The Future of Payment Systems (Routledge International Studies in Money and Banking)

The Future of Payment Systems

Payment systems are the circulation system for modern monetary economies, ensuring money for spending and saving moves to the right person in the right place at the right time. When these circulation systems break down, monetary economies themselves run the risk of seizing up. This is why robust payment systems are considered so important by central banks and policymakers throughout the world. This volume draws on wide-ranging contributions from prominent international experts, discussing some of the most pressing issues facing policymakers and practitioners in the field of payment systems today. Because payment systems have been with us for at least as long as money itself, many of the questions raised in this book are timeless. Improvements in information technology mean, however, that answers to these questions are unlikely to be timeless. This book tackles issues regarding the form payment systems might take in the future, the risks associated with this evolution, the techniques being deployed to assess these risks and the implications these risks have for the respective roles of the public and private sector. Based on a conference, ‘The Future of Payment Systems’, organised by the Bank of England, this book will make fascinating reading for practitioners and policymakers in the field of payment systems, as well as students and researchers engaged with the economics of payments and central banking policy. Andrew G. Haldane is Head of Systemic Risk Assessment at the Bank of England. Stephen Millard is a Senior Economist at the Bank of England. Victoria Saporta is a Senior Economist at the Bank of England.

Routledge international studies in money and banking

1 Private Banking in Europe Lynn Bicker 2 Bank Deregulation and Monetary Order George Selgin 3 Money in Islam A study in Islamic political economy Masudul Alam Choudhury 4 The Future of European Financial Centres Kirsten Bindemann 5 Payment Systems in Global Perspective Maxwell J. Fry, Isaak Kilato, Sandra Roger, Krzysztof Senderowicz, David Sheppard, Francisco Solis and John Trundle 6 What is Money? John Smithin 7 Finance A characteristics approach Edited by David Blake

8 Organisational Change and Retail Finance An ethnographic perspective Richard Harper, Dave Randall and Mark Rouncefield 9 The History of the Bundesbank Lessons for the European Central Bank Jakob de Haan 10 The Euro A challenge and opportunity for financial markets Published on behalf of Société Universitaire Européenne de Recherches Financières (SUERF) Edited by Michael Artis, Axel Weber and Elizabeth Hennessy 11 Central Banking in Eastern Europe Edited by Nigel Healey and Barry Harrison 12 Money, Credit and Prices Stability Paul Dalziel

13 Monetary Policy, Capital Flows and Exchange Rates Essays in memory of Maxwell Fry Edited by William Allen and David Dickinson 14 Adapting to Financial Globalisation Published on behalf of Société Universitaire Européenne de Recherches Financières (SUERF) Edited by Morten Balling, Eduard H. Hochreiter and Elizabeth Hennessy 15 Monetary Macroeconomics A new approach Alvaro Cencini 16 Monetary Stability in Europe Stefan Collignon 17 Technology and Finance Challenges for financial markets, business strategies and policy makers Published on behalf of Société Universitaire Européenne de Recherches Financières (SUERF) Edited by Morten Balling, Frank Lierman, and Andrew Mullineux 18 Monetary Unions Theory, history, public choice Edited by Forrest H. Capie and Geoffrey E. Wood 19 HRM and Occupational Health and Safety Carol Boyd

20 Central Banking Systems Compared The ECB, the pre-Euro Bundesbank and the Federal Reserve System Emmanuel Apel 21 A History of Monetary Unions John Chown 22 Dollarization Lessons from Europe and the Americas Edited by Louis-Philippe Rochon and Mario Seccareccia 23 Islamic Economics and Finance: A Glossary, 2nd Edition Muhammad Akram Khan 24 Financial Market Risk Measurement and analysis Cornelis A. Los 25 Financial Geography A banker’s view Risto Laulajainen 26 Money Doctors The experience of international financial advising 1850–2000 Edited by Marc Flandreau 27 Exchange Rate Dynamics A new open economy macroeconomics perspective Edited by Jean-Oliver Hairault and Thepthida Sopraseuth 28 Fixing Financial Crises in the 21st Century Edited by Andrew G. Haldane

29 Monetary Policy and Unemployment The U.S., Euro-area and Japan Edited by Willi Semmler

37 The Structure of Financial Regulation Edited by David G. Mayes and Geoffrey E. Wood

30 Exchange Rates, Capital Flows and Policy Edited by Peter Sinclair, Rebecca Driver and Christoph Thoenissen

38 Monetary Policy in Central Europe Miroslav Beblav´y

31 Great Architects of International Finance The Bretton Woods era Anthony M. Endres 32 The Means to Prosperity Fiscal policy reconsidered Edited by Per Gunnar Berglund and Matias Vernengo 33 Competition and Profitability in European Financial Services Strategic, systemic and policy issues Edited by Morten Balling, Frank Lierman and Andy Mullineux 34 Tax Systems and Tax Reforms in South and East Asia Edited by Luigi Bernardi, Angela Fraschini and Parthasarathi Shome 35 Institutional Change in the Payments System and Monetary Policy Edited by Stefan W. Schmitz and Geoffrey E. Wood 36 The Lender of Last Resort Edited by F.H. Capie and G.E. Wood

39 Money and Payments in Theory and Practice Sergio Rossi 40 Open Market Operations and Financial Markets Edited by David G. Mayes and Jan Toporowski 41 Banking in Central and Eastern Europe 1980–2006 A comprehensive analysis of banking sector transformation in the former Soviet Union, Czechoslovakia, East Germany, Yugoslavia, Belarus, Bulgaria, Croatia, the Czech Republic, Hungary, Kazakhstan, Poland, Romania, the Russian Federation, Serbia and Montenegro, Slovakia, Ukraine and Uzbekistan Stephan Barisitz 42 Debt, Risk and Liquidity in Futures Markets Edited by Barry A. Goss 43 The Future of Payment Systems Edited by Andrew G. Haldane, Stephen Millard and Victoria Saporta

The Future of Payment Systems

Edited by Andrew G. Haldane, Stephen Millard and Victoria Saporta

First published 2008 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Simultaneously published in the USA and Canada by Routledge 270 Madison Ave, New York, NY 10016 Routledge is an imprint of the Taylor & Francis Group, an informa business This edition published in the Taylor & Francis e-Library, 2008. “To purchase your own copy of this or any of Taylor & Francis or Routledge’s collection of thousands of eBooks please go to www.eBookstore.tandf.co.uk.” © 2008 Selection and editorial matter, The Governor and Company of the Bank of England; individual chapters, the contributors All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data A catalog record for this book has been requested ISBN 0-203-94014-8 Master e-book ISBN

ISBN10: 0-415-43860-8 (hbk) ISBN10: 0-203-94014-8 (ebk) ISBN13: 978-0-415-43860-5 (hbk) ISBN13: 978-0-203-94014-3 (ebk)

Contents

List of figures List of tables Notes on contributors Foreword Acknowledgements General introduction: the future of payment systems

x xii xiii xv xvi 1

ANDREW G. HALDANE, STEPHEN MILLARD AND VICTORIA SAPORTA

PART I

Payment systems and public policy 1 Central banks and payment systems: past, present and future

13

15

STEPHEN MILLARD AND VICTORIA SAPORTA

2 The role of a central bank in payment systems

45

EDWARD J. GREEN

3 Some challenges for research in payments

57

EDWARD J. GREEN

4 Payment economics and the role of central banks

68

JEFFREY LACKER

PART II

New approaches to modelling payments 5 New models of old (?) payment questions RICARDO CAVALCANTI AND NEIL WALLACE

73 75

viii

Contents

6 Optimal settlement rules for payment systems

87

BENJAMIN LESTER, STEPHEN MILLARD AND MATTHEW WILLISON

7 The microstructure of money

100

JAMES McANDREWS

PART III

Current payment policy issues 8 Wholesale payments: questioning the market-failure hypothesis

117

119

GEORGE SELGIN

9 Central bank intraday collateral policy and implications for tiering in RTGS payment systems

138

JOHN P. JACKSON AND MARK J. MANNING

10 Central banks’ interest calculating conventions: deviating from the intraday/overnight status quo

160

GEORGE SPEIGHT, MATTHEW WILLISON, MORTEN BECH AND JING YANG

11 How should we regulate banks’ liquidity?

175

JEAN-CHARLES ROCHET

PART IV

Policy perspectives on the future of payments

187

12 The diffusion of real-time gross settlement

189

MORTEN L. BECH

13 E-settlement: soon a reality?

206

HARRY LEINONEN

14 Real-time liquidity management in a globally-connected market

230

RICHARD PATTINSON

15 Will central banking survive electronic money? STEFAN W. SCHMITZ

233

Contents 16 Payment systems and central banks: where are we now and where will e-payments take us?

ix 255

CHARLES FREEDMAN

Index

262

Figures

1.1 6.1 6.2 6.3 6.4 6.5 6.6 7.1 7.2 9.1 9.2 9.3 9.4 9.5 9.6 9.7 10.1 10.2 12.1 12.2 12.3 12.4 12.5 12.6 12.7 12.8 12.9 13.1 13.2

Stylised models of intervention in payment systems DNS equilibrium RTGS equilibrium ( = 0) RTGS equilibrium ( = 0.005) RTGS equilibrium ( = 0.05) Existence of DNS and RTGS equilibria Welfare Intraday pattern of activity: total value and volume of federal funds traded Fifth and 95th percentiles of the average federal funds rate minus the target rate Payment flows when C accesses the system via bank A A time-line for actions High degree of internalisation Low default probability The impact of imperfect monitoring The impact of tiering risk The impact of tiering risk with high payment value Change in shape of yield curve An alternative change in the shape of the yield curve Value of transfers originated on Fedwire Adoption of RTGS in Europe – 1995 Adoption of RTGS in Europe – 2005 Adoption of RTGS in Asia – 2005 Adoption of RTGS in Africa – 2005 Adoption of RTGS in South America Adoption of RTGS in Central America – 2005 S-curve and adopter groups Adoption of RTGS in central banking Direct interbank communication in a network-based infrastructure The common account number space

28 94 94 95 95 97 98 106 108 143 144 148 149 152 153 155 162 171 191 192 193 194 196 197 197 199 199 208 209

Figures xi 13.3 E-settlement is part of the credit transfer circle, which provides efficient electronic communications between participants in a payment 13.4 The digital e-settlement stamp is part of the payment message 13.5 The digital encrypted stamp with central bank cover will follow the payment message through the network 13.6 E-settlement stamps are produced by e-settlement modules, which are closely integrated with banks’ payment systems 13.7 A dedicated interbank network connects all banks and the central bank with each other for payment processing 13.8 The increasing market failure gap 14.1 Links between world market infrastructure

210 213 213 214 215 222 230

Tables

1.1 1.A1 1.A2 1.A3 7.1 10.1 10.2 10.3 10.4 12.1 12.2

Ranking models of intervention 31 G10 models for intervention in large-value payment systems 37 G10 models for intervention in ACHs 38 G10 models for intervention in the embedded payment systems of securities settlement systems 39 Regression results 109–10 Pay-offs when central bank charges and remunerates at end of day only 167 Equilibrium strategies when the central bank charges and remunerates at end of day only 167 Pay-offs when central bank charges and remunerates at midday and at end of day 169 Equilibrium strategies when the central bank charges and remunerates at midday and at end of day 169 Imported RTGS systems 201 Liquidity savings features 203

Contributors

Morten Bech. Economist, Federal Reserve Bank of New York. Prior to that, Dr Bech worked for the Danish Central Bank where he helped design the KRONOS RTGS system. Dr Bech has written extensively on payment system issues in central bank publications and academic journals. Dr Bech was a visitor at the BoE in the summer of 2006. Ricardo de Cavalcanti. Associate Professor of Economics, Graduate School of Economics, Getulio Vargas Foundation, Rio de Janeiro, Brazil. Charles Freedman. Currently Scholar in Residence in the Department of Economics at Carleton University and consultant to the IMF and central banks. He worked at the Bank of Canada from 1974 to 2003, serving as Deputy Governor from 1988 to 2003. Edward J. Green. Professor of Economics, The Pennsylvania State University. John Jackson. Economist, Systemic Risk Reduction Division, Bank of England. Jeffrey Lacker. President of the Federal Reserve Bank of Richmond. Harry Leinonen. Adviser to the Board of the Bank of Finland and the Finnish representative in the Eurosystem Payment and Settlement System Committee; Finnish representative in the EU Commission; Government Expert Group and Market Group on Payment Issues. Benjamin Lester. University of Pennsylvania. Benjamin is currently finishing his PhD in Economics at the University of Pennsylvania. His work focuses on the macroeconomic implications of issues in money and banking. Mark Manning. Senior manager, Systemic Risk Reduction Division, Bank of England. James J. McAndrews. Vice President and Head of the Money and Payment Studies Function in the Research and Statistics Group of the Federal Reserve Bank of New York. He received a PhD in economics from the University of Iowa.

xiv

Contributors

Richard Pattinson. Head of Regulatory and Industry Issues, Global Payments, Barclays Bank. Holds a number of external positions in the industry primarily concerned with payments and settlements and the management of payment system liquidity including: Chairman, CHAPS Clearing Company Limited (UK); Deputy Chairman, SWIFT (UK) Limited (UK); Director, Voca Limited (UK); Director, EBA Association (France); Director, CLS Group Holding AG (Switzerland); Director, CLS Bank International (USA); Member Bank of England Money Market Liaison Group; Member UK Market Advisory Committee. Jean-Charles Rochet. Professor of Economics and Mathematics at Toulouse School of Economics (Toulouse University) and Research Director at Institut D’Economie Industrielle, Toulouse, France. George Selgin. Professor of Economics at the University of Georgia’s Terry College of Business. His latest book, Good Money: Private Enterprise and the Beginnings of Modern Coinage, is forthcoming from the University of Michigan Press. Stefan W. Smitz. Currently at Oesterreichische Nationalbank. Co-editor of Institutional Change in the Payments System and Monetary Policy, Routledge (with Geoffrey E. Wood) and Carl Menger and the Evolution of Payment Systems: From Barter to Electronic Money, Edward Elgar (with M. Latzer). George Speight. Senior Manager, Systemic Risk Reduction Division, Bank of England. Neil Wallace. Professor of Economics, The Pennsylvania State University. Matthew Willison. Economist, Systemic Risk Assessment Division, Bank of England. Jing Yang. Senior Economist, International Finance Division, Bank of England.

Foreword

The Bank of England has two core purposes – monetary stability and financial stability. Payment systems are the mechanism by which money is transferred to enact both real transactions (such as buying bread) and financial transactions (such as buying bonds). So robust payment systems are integral to both of the Bank’s core purposes. The same is true in central banks around the world. Indeed, in many central banks their payment system role predated the formalisation of their financial and monetary stability objectives. The role of central banks in payment systems has, however, changed significantly over the past decade. Doubtless it is set to change further over the next decade, not least due to the impact of advances in information technology. But what form will this change take? And what are the risks – for policymakers, for payment system operators, for the public at large – associated with this change? These are among the most topical and involved questions facing central banks today. With these questions in mind, the Bank of England hosted a two-day conference on 19 and 20 May 2005 with the title ‘The Future of Payment Systems’. The conference aimed to draw together the views of academics, payment system practitioners and policymakers on payment system issues. All too rarely have attempts been made to integrate the distinct perspectives of these three parties. This volume brings together in one place these contributions, as a first step towards such a synthesis. You will not be surprised to hear that the volume is long on questions and short on answers. That is in the nature of conferences, perhaps especially successful ones. But I hope, nonetheless, you find it useful as a contribution towards understanding the likely future course of payment systems, which has important implications for us all. Sir John Gieve Deputy Governor for Financial Stability Bank of England

Acknowledgements

This volume brings together the papers from a conference on ‘The Future of Payment Systems’ that was held at the Bank of England on 19–20 May 2005. We would first like to thank all the contributors to this volume for their help in preparing papers on topics of our choosing rather than theirs, presenting them at our conference and then revising them for this volume. We would like also to thank everyone who attended and contributed to the conference, as well as all those who made it happen. In particular, we would like to thank the discussants at this conference whose comments led to substantial improvements in all of the papers: Martin Andersson, Morten Bech, Xavier Freixas, Charles Goodhart, Charles Kahn, Nobuhiro Kiyotaki, Thorsten Koeppl, John Mohr, John Moore, Erlend Nier, Will Roberds, Matthew Willison and Randall Wright. And, in particular, we would also like to thank Francesca Desquesnes whose efforts ensured the event ran without a hitch. Many people have commented on the various chapters in this book and we thank them all. Roy Clive, Raxita Dodia, Elizabeth Hughes, Sandra Mills and Julie Pickering have done sterling work in helping pull the manuscript together; and the help of Thomas Sutton and Terry Clague at Routledge has been invaluable at various stages of the project. Of course, all remaining errors and omissions are ours. The views expressed in the chapters in this book are those of the authors and do not necessarily reflect those of the Bank of England, the Federal Reserve System, the Federal Reserve Bank of New York, the Federal Reserve Bank of Richmond, the Oesterreichische Nationalbank, Barclays Bank or the Bank of Finland. Finally, the authors and publishers would like to thank the following for granting permission to reproduce material in this work: the Bank of Finland, Barclays Bank, the Federal Reserve Banks of New York and Richmond, Charles Freedman, Ed Green, Jean-Charles Rochet, Stefan Schmitz and Neil Wallace and Ricardo Cavalcanti. The chapter by George Selgin was reprinted from the International Review of Law and Economics, Vol. 24, No. 3, Selgin, G, ‘Wholesale payments: questioning the market-failure hypothesis’, pages 333–350, Copyright (2004) with permission from Elsevier to whom we give our thanks. Every effort has been made to contact copyright holders for their permission to reprint material in this book. The publishers would be grateful to hear from

Acknowledgements

xvii

any copyright holder who is not here acknowledged and will undertake to rectify any errors or omissions in future editions of the book. Andrew G. Haldane Stephen Millard Victoria Saporta Bank of England

General introduction The future of payment systems Andrew G. Haldane, Stephen Millard and Victoria Saporta

A conundrum At the heart of the study of payment systems lies a contradiction. Mere mention of the words ‘payment systems’ to an economist tends to conjure up images of an obscure and rather technical sub-discipline – or perhaps even backwater – of the profession. This backwater is believed to be inhabited by a small and rather reclusive set of fanatics. This tribe uses tools and a language of their own and spends its time studying issues that are well outside the mainstream. Yet, for the public at large, ‘payment systems’ are part and parcel of their everyday lives. The use of cash, credit and debit cards and electronic money transfers to enact payments and transfers is a practical and straightforward task. Payment systems are unwittingly used by almost everyone, probably several times a day, every day of the week. The tools and the language used to describe these instruments, while distinct, are well understood by almost everyone. So payment systems are obscure yet commonplace, highly technical yet understood by everyone. How do we resolve this conundrum? It was this question which prompted the Bank of England to host an international conference on ‘The Future of Payment Systems’ on 19 and 20 May 2005 in London. This volume collates together the main contributions from that conference. Unlike the conference itself, the volume is organised into four blocks. Part I considers the intersection between payment systems and public policy. The chapters trace the anthropological origins of payment systems: How and why they came into being and how their evolution has been, and is being, shaped by public policy? Historically, central banks and payment systems have been inextricably linked. But technology is reshaping those historical relationships in important ways. Part II of the volume considers some of the methodological advances which have recently been made in the study of payment systems: What models and empirical methods have been used to analyse payment system behaviour? This is a rapidly evolving – though at present rather diffuse – area of the economics profession. Part III of the volume illustrates how some of these approaches can be used to address a number of topical payment system issues, for example, the role of central bank intraday liquidity policy and regulatory liquidity requirements in reducing payment system risks.

2

A.G. Haldane et al.

Finally, having traced the origins of the species, in Part IV of the volume we plot the possible course of payment systems in the future – hence the title of this volume, The Future of Payment Systems. Parts I, II and III of the volume are natural precursors to Part IV because an understanding of the fundamentals of payment instruments is essential when predicting how technology might shape those future fundamentals. The chapters in Part IV sketch some alternative – and in some cases quite radical – visions of payment systems of the future. This volume does not profess to contain all of the answers, nor to solve completely the conundrum. But it may provide, we hope, some clues on what the key issues and questions might be, now and in the future. To that end, the remainder of this introduction considers a few key generic payment system themes which emerge from the volume; it then places the chapters from the volume in the context of these themes and the wider literature on payments.

The economics of payments and payment systems To develop and understand the economics of payments, we first need some definitions and methodology. Perhaps a good – if not entirely uncontroversial – place to start is with a working definition of ‘payments’ and ‘payment systems’. A ‘payment’ is a transfer of monetary value. So a ‘payment system’ is no more than an organized arrangement for transferring value between its participants. So defined, it is clear that payment systems are fundamental to the functioning of all monetary economies. If money is the lifeblood of modern monetary economies, payment systems are the circulation system. Failures in this circulation system risk a seizing up in the real and financial transactions they support, with potentially significant welfare costs. Because value need not be embodied in monetary assets but also in real goods, it could reasonably be argued that payment systems predate the existence of money itself. Certainly, payment systems predate the emergence of central banks, the latter which in many cases emerged during the twentieth century. Some barter economies had quite sophisticated and hence organized exchange of value systems in place, which legitimately could be termed a primitive payment system. But as first commodity money replaced goods, fiat money displaced commodity money, and finally commercial bank (or ‘inside’) money replaced central bank (or ‘outside’) money as the media of exchange, payment systems have increasingly involved monetary transfers routed through financial institutions. And this monetary evolution, in turn, resulted in central and commercial banks increasingly taking centre stage in the design and operation of payment systems (see Part I). These definitions, and this evolution, of payment systems make clear why the study of payments is many-faceted. In particular, the economics of payments can be thought to cover at least the following sub-disciplines and questions: •

Foundations of money and payment systems: Why have payments or money in the first place? What fundamental frictions in the economy do money and

General introduction

•

•

3

payment systems help mitigate? And what can this welfare-theoretic approach to money and payments tell us about the future evolution of payment systems? Payments and the macroeconomy: Where do payments fit within the wider macroeconomy? In particular, what are the macroeconomic benefits of wellfunctioning payment systems, measured in terms of output, employment and inflation? The industrial organization of payments: What is a suitable industrial structure for payments from a societal perspective? In particular, what and how large are the market failures embedded in the payments industry? And how best should these be tackled? How do we design payment systems to provide incentives for appropriate behaviour by system participants from a social welfare perspective?

These questions are distinct and so too have been the analytical frameworks used to tackle them. The chapters in the volume seek to address some of these different questions using often quite different analytical approaches (see, for example, Part II). To date, there has been no grand synthesis of these different approaches. Green’s chapters in this volume argue compellingly against us expecting such a grand synthesis any time soon, in part because the questions being posed of payment systems are so deep and broad, ranging from the microeconomics of money, through the macroeconomics of payments to the industrial organization of banking and finance. The chapters in the volume hopefully give a flavour for such depth and breadth. And, as such, they may hopefully serve as a staging post towards a synthesis of these various strands. The foundations of money and payment systems Kahn and Roberds (2006) suggest that the foundations of payment systems rest on their ability to deal with two fundamental problems inherent in an economy: a mismatch between the times that different agents wish to trade and limited enforcement of privately-issued debt obligations (‘inside money’, such as deposits with commercial banks). They point to the existence of two types of payment system: ‘store-of-value’ systems based on the ability of agents to verify the asset being used to effect payment, and ‘account-based’ systems based on the ability of agents to verify the identities of the account holders. Most payment systems that we observe in the real world feature elements of store-of-value and account-based systems and nearly all rely on the transfer of inside money. The chapter by Cavalcanti and Wallace in this volume uses a model based on these imperfections to investigate the usefulness of inside money as a medium of exchange. In particular, they assume that individuals cannot commit to future actions and to some extent their histories are not known (there is ‘imperfect monitoring’). Specifically, some people are not monitored at all and others are perfectly monitored. The perfectly-monitored agents are able to issue inside money. Cavalcanti and Wallace use this model to show that private money can

4

A.G. Haldane et al.

be useful and that, in the absence of private money, central bank money and central bank lending through the discount window can deliver welfare benefits over and above what can be accomplished with a private interbank money market. They argue that the deep insights gained from this framework can be used to help shape answers to questions about the future design of payment systems, though their existing framework is perhaps as yet too primitive to deliver sharp and robust policy insights. He et al. (2005) develop a search-theoretic model of a payment system in which the introduction of cash benefits the economy for the same reasons as in the chapter by Wallace and Cavalcanti; that is, it solves the problems of lack of commitment and the inability to monitor your trading partners. But they note that the use of cash itself carries costs; it needs to be verified and can be stolen. In their model, they allow agents to guard against the risk of theft by depositing their cash in banks and making payments from their bank account to other agents’ bank accounts by a payment system they call ‘cheques’. They find that the introduction of a payment system expands the range of parameter values consistent with there being an equilibrium in which money is accepted as a medium of exchange – in other words, that the presence of a payment system enables money to solve the imperfections discussed by Kahn and Roberds (2006). But the payment system in He et al. (2005) is risk free. Two more imperfections that need to be considered in a model of payment systems are the risk that the members of the system default on their obligations – credit risk – and the risk that the system itself breaks down – operational risk. Lester (2005) extends the He et al. model to allow for credit risk while Millard and Willison (2006) extend the model to allow for operational risk in payment systems. In both cases, the authors show how one might begin to quantify the welfare benefits of reducing the risks brought about by these frictions. Payments and the macroeconomy The approaches discussed above begin to introduce inside money – and payment systems that enable the transfer of inside money – into a model of the economy in a way that is intellectually rigorous, showing that it delivers welfare gains by mitigating some of the fundamental frictions that may exist in the real world, such as the lack of a double coincidence of wants. This work is essentially grounded in the microeconomics of money. Relatively little work has, however, been done to illustrate the potential links between the design and operation of payment systems and the macroeconomy – for example, how payment systems can improve welfare through their effects on standard macroeconomic variables such as output and inflation. Cifuentes and Willison (2006) suggest that payment systems can benefit the economy by reducing the liquidity needed to make payments, thereby allowing banks to reallocate their resources to less liquid assets which have a higher expected return. Millard et al. (2006) use a standard macroeconomic modelling

General introduction

5

approach to argue that, if banks require liquid assets to make payments through a payment system, then central banks can improve welfare by creating such liquid assets at the beginning of the day and withdrawing them at the end of the day. If this activity is near-costless – as would be expected in a pure fiat money world – it follows that the operation of the payment system will only affect the intraday money stock and not macroeconomic magnitudes, such as output and inflation. The chapters by Green in this volume make essentially the same point, arguing that central banks are uniquely placed to offer free, short-term credit against illiquid collateral to other financial intermediaries and that doing so improves welfare in the economy, without adversely affecting monetary control and hence inflation and output. Such a liquidity policy by the central bank may, of course, have implications for the efficiency with which banks manage their own liquidity. The chapter by Bech et al. in this volume argues that, if the central bank were to charge interest based on settlement banks’ balances with them more frequently than once a day, the settlement banks would be given an incentive to monitor more closely their customers’ use of intraday credit; this, in turn, would be positive for financial stability since intraday credit risk would be reduced, or at least internalized to some degree in banks’ decision-making.1 But if banks’ customers did not insist their payments be made by a particular time, the banks would tend to delay payments so they could earn interest on positive balances for longer; this would be negative for financial stability since an operational incident would then result in more payments not having been made. Either way, there would be no implications for the ability of the central bank to carry out monetary policy. The industrial organization of payments Perhaps the largest body of literature on the economics of payments has focussed on its industrial organization. Within this, analysis has typically centred on the potential externalities involved in the payments industry and costeffective ways of mitigating the market failures associated with them. Payment systems are an example of a network industry in which the welfare of existing members increases each time a new member joins. In addition, there are likely to be economies of scale in the provision of payment services. Both these features are likely to imply a tendency towards concentration and a lack of competition. These potential externalities are discussed extensively and rigorously in the chapters by Millard and Saporta, Green and Lacker. But the scale of these problems, and hence the appropriate solution to them, remain vexed questions. The chapters by Green and Lacker argue that the importance of many of these externalities may have been overstated historically. Hence they favour minimalist interventions by the public sector to offset these frictions. Allen et al. (2006) and Schanz (2006) discuss one such non-interventionist strategy. If a payment system is mutually-owned by its users, the monopoly problem can, to an extent, be mitigated.

6

A.G. Haldane et al.

A second set of externalities arise from the way in which payment systems can generate risks for the wider financial system – so-called systemic risks. These can arise either because payment systems can act as a channel through which financial contagion is propagated between institutions or because such systems act as a potential single point of failure. The mitigation of these systemic risks is the focus of a number of the chapters in the volume. One of the key tools typically used to mitigate such systemic risks is the design and operation of payment systems including, crucially, the rules of the game for settlement among participants in these systems. Payment systems have historically settled using what is called deferred net settlement (DNS). Payments are collected together over the course of a day, the net amount that each bank owes or is owed is calculated, and then net amounts are settled between participants. As the volume and value of payments increased dramatically throughout the 1980s and 1990s, central banks became worried about the risks inherent in such netting systems. In particular, if one participant failed during the day, all processed payment orders could be unwound with the consequent risk of other participants failing to be able to meet their obligations, thereby generating contagious default. In response, over the past decade there has been a large increase in the number of large-value payment systems in the world that employ real-time gross settlement (RTGS), where payments are paid in full as soon as they are submitted to the system. The diffusion of RTGS systems across the world’s economies is discussed by Bech in his chapter. But RTGS requires banks to hold a much larger amount of the settlement asset to make these gross payments. As Bech and Garratt (2003) show using a simple game-theoretic model, this creates incentives for banks to delay payments in the hope of obtaining liquidity from incoming payments which they can then use to make their outgoing payments. The net result is a ‘bad’ equilibrium where all banks may delay payments and no banks save liquidity by sodoing. McAndrews and Rajan (2000) present some empirical evidence that suggests this may happen in practice. In his chapter for this volume, McAndrews calls this approach to understanding the effects of payment system design on payment outcomes the ‘market microstructure of money’. More generally, he discusses how the methodological insights from the market microstructure literature on securities markets can be used and adapted to address a range of topical payment system issues.

Key issues in payments today Armed with this toolkit, how best might we apply it to key policy issues in payments (Part III)? As Millard and Saporta discuss, historically, central banks’ relationship with payment systems has extended beyond providing their liabilities as the ultimate settlement asset, to include owning, operating and regulating certain payment systems. But which of these roles, if any, should central banks play?

General introduction

7

Norman et al. (2006) show how the need for an ultimate settlement asset for payments is inextricably linked to the evolution of central banking. Indeed, Millard (2006) argues that providing the ultimate settlement asset is what leads a central bank to care about monetary and financial stability. This foundational approach does not necessarily suggest, however, a clear additional role for the central bank in the ownership, operation and/or regulation of payment systems. The nature and extent of such intervention depends critically on how large systemic risk externalities are believed to be. This, in turn, requires an appreciation of the ways in which different risks within a system – credit, liquidity, operational, etc. – might interact in the face of changes in payment system design.2 There appears to be relatively little consensus, however, on either the precise scale or the source of systemic risks within many payment systems. This is an issue taken up by several chapters in the volume. For example, Selgin argues that a bank is only exposed to the risk of another bank failing to meet a net obligation in a DNS system if it credits customer accounts before settlement occurs. But that is not something banks are required to do. Indeed, such customer credits can typically be reversed in the event of settlement not taking place. In this way, Selgin argues that the credit risk in DNS systems is not as severe as is often assumed and hence that RTGS systems might be a sledgehammer approach to cracking a nut. Moreover, although RTGS systems all but eliminate credit risk, they do so at the expense of requiring much higher levels of liquidity than DNS systems, and in this way potentially aggravate liquidity risk. The chapter by Lester et al. in this volume presents a search-theoretic framework to examine the trade-off between cost and risk inherent in the choice between RTGS and DNS payment systems. It shows that when the costs of settling payments on an RTGS basis are high, only a DNS equilibrium can exist. For intermediate values of costs either settlement rule may hold in equilibrium and there is nothing to suggest that agents in the economy will necessarily use the welfare superior mode of settlement. The results from the chapter thus support a role for central banks in encouraging the move from DNS to RTGS, as they did during the 1990s and as documented by Bech in his chapter in this volume. As stressed in the chapters by Green and Bech et al. in this volume, one design feature that can help to reduce the liquidity costs and hence improve the trade-off is for central banks to provide intraday liquidity at low cost. For example, in the United Kingdom and the Euro Area (as well as in a number of other countries) the central banks provide liquidity to the members of their large-value payment systems at a zero intraday interest rate against high-quality collateral. An alternative (or additional) way of improving the risk-efficiency trade-off is by designing payment systems that combine the liquidity-savings features of DNS with the finality offered by RTGS: so-called ‘hybrid’ systems. McAndrews and Trundle (2001) discuss the development of these systems and how they improve the risk-efficiency trade-off. One way the banks themselves can economize on liquidity costs is through the use of correspondent banks, which process payments on behalf of indirect

8

A.G. Haldane et al.

system participants. The chapter by Jackson and Manning in this volume argues that such ‘tiering’ may bring welfare benefits for the economy as a whole for three reasons: correspondent banks are likely to monitor their customers better than the central bank; correspondent banks can internalize payments, reducing further the collateral needed by the second-tier banks; and the opportunity cost of collateral is likely to be lower for the correspondent banks than it is for the second-tier banks. But they argue that tiering may also introduce costs into the economy since correspondent banks may not have sufficient incentive to monitor their customers, internalized payments might be subject to greater legal risk, and operational or financial problems at a settlement bank will disrupt not only its own payments, but also those of its customers. More generally, the move to RTGS has increased the demand for liquidity by financial firms and hence the potential for liquidity risk materializing in the banking system. This is one of the factors that has resulted in liquidity risk becoming an issue of increased importance for banking regulators over recent years.3 The chapter by Rochet in this volume looks at why it is necessary to regulate banks’ liquidity holdings and how regulators might go about doing this. He argues that liquidity regulation for banks can be justified by two different motives: limiting the risk and the impact of individual bank failures, and limiting the need for massive liquidity injections by the central bank in case of a macroeconomic shock. In his view, a simple form of ‘stock liquidity requirement’ can cover the objective of protecting small depositors. But he suggests that there is a need additionally for a second type of liquidity requirement, based on some ex ante indices of exposure to macroeconomic shocks by individual banks. This systemic liquidity requirement would limit the need for an ex post liquidity injection by the central bank in response to a systemic shock. This would be a radical departure from existing regulatory practices, but perhaps not an implausible one should liquidity risks continue to escalate.

The future of payment systems This book is called The Future of Payment Systems so this introduction would not be complete without some discussion of where payment systems might be headed. Several of the chapters in the volume address this issue (see Part IV). Drawing on Kahn and Roberds (2006), it seems likely that payment systems in the future will still fall into either ‘account-based’ or ‘store-of-value-based’ categories. Rosenblat (1999) argues that future retail payments are likely to be made using one or other of two things: a debit card (which would act in the same way as a credit card when the user was overdrawn) and a cash card (an anonymous stored-value card). The former represents an account-based system; the latter a store-of-value-based system. One interesting open question is whether cash cards will replace cash completely. Rosenblat argues that they will because if such cards require user authorization, there would be no incentive for anyone to steal them (unlike cash itself). One alternative form that the ‘cash substitute’ could take is units of mobile phone airtime; the technology already exists for

General introduction

9

these to be transferred anonymously from one mobile phone to another and mobile phones are ubiquitous, helping deal with the network externality problem that issuers of cash cards have faced in the past. The only remaining issue is whether non-banks (i.e. the mobile phone companies) would be allowed by the public authorities to enable this: that is, issue their own money. In terms of wholesale payments, we can again consider the possibilities of different ‘store-of-value-based’ and ‘account-based’ systems. The e-settlement vision put forward by Leinonen in his chapter in this volume is one particularly radical version of a ‘store-of-value-based’ system. Under this system, wholesale payments are made directly between banks using encrypted bytes of information – central bank e-money – with no central processing. Leinonen argues that the efficiency gains from moving to such a system are large, but that the network externalities and increasing returns to scale operating in the payments industry and the lack of competition among banks act as barriers against the adoption of this more efficient technology. Nonetheless, he maintains that the new, and more efficient, technology of network-based e-settlement will eventually triumph over the centralized systems that we have today. In terms of ‘account-based’ systems, wholesale payments are already made in this way across accounts held at the central bank. One question here is whether these systems could be merged into one large payment system in which wholesale payments can be made between banks anywhere in the world. In his chapter in this volume, Pattinson suggests that the introduction of the Continuous Linked Settlement (CLS) system has taken us much closer to this position. There are already a cat’s cradle of links between the world’s largest wholesale payment systems, with CLS at its epicentre. Whether payment systems and regulators have fully taken into account the systemic risk implications of this increased interconnection between systems is, however, a moot point. In particular, liquidity risk – and the accompanying need for effective cross-border liquidity management – may have increased inadvertently as a result of this increased interconnection between infrastructures. However wholesale payment systems develop, there are likely to be implications for the demand for central bank money and, it could be argued, for the ability of the central bank to carry out monetary policy. The chapter by Millard and Saporta in this book surveys some of the literature on this. Woodford (2004) notes that as long as central bank money is used to settle payments there will always be some demand for it, even in the absence of central-bank-issued cash. This line of argument is discussed further in the chapter by Schmitz in this volume. Schmitz argues that there are strong economies of scale reasons for having a single unit of account. Given that the current unit of account corresponds to units of central bank money, and that changing to an alternative unit of account would be costly, it is efficient for central bank money to remain as the ultimate settlement asset. As long as this remains the case, the central bank can continue to carry out monetary policy, more or less, as it does now. And if the demand for central bank money were to fall to zero, the central bank could still carry out monetary policy provided it could control the supply of, or demand

10

A.G. Haldane et al.

for, the ultimate settlement asset (whatever it was) via reserve requirements in, and an interest rate corridor on central bank borrowing and lending of the ultimate settlement asset. Freedman makes a similar point in his chapter in this volume. He suggests, though, that how a central bank might act to control the supply of a settlement asset that were not its own money is unlikely to become a practical issue in the foreseeable future. As long as central banks maintain a stable value for their money, it is likely that it will remain the settlement asset of choice.

Conclusion We have not solved the conundrum with which we began this introduction. We hope, however, that this volume is some small step towards beginning to close the gap between the perceptions of practitioners, academics and the general public on the role and importance of payment systems. In particular, we hope this volume helps raise the interest of those who had previously thought payment systems obscure and demystified, at least to a degree, the technical language built up around them. With luck, readers might even feel spurred to tackle some of these fascinating and important issues for themselves.

Notes 1 Lacker (2006) makes a similar point, arguing that the interest rate should be related to the rate at which agents discount the utility they obtain from consumption if incentives – in this case, banks’ payment incentives during the day – are not to be adversely affected. 2 Definitions all taken from Bank for International Settlements (2003). 3 In his chapter in this volume, Rochet also mentions two other factors that have led to an increase in liquidity risk: increases in banking sector concentration, as well as in the complexity and size of financial markets, and the increased use of derivative products, which generate a large demand for liquidity.

References Allen, H., Christodoulou, G. and Millard, S.P. (2006) ‘Financial infrastructure and corporate governance’, Bank of England Working Paper No. 316. Bank for International Settlements (2003) A glossary of terms used in payments and settlement systems. Bech, M. and Garratt, R. (2003) ‘The intraday liquidity management game’, Journal of Economic Theory, 109: 198–219. Cifuentes, R. and Willison, M. (2006) ‘Why payment systems matter: measuring their benefits for the economy’, unpublished thesis, Bank of England. He, P., Huang, L. and Wright, R. (2005) ‘Money and banking in search equilibrium’, International Economic Review, 46: 637–70. Kahn, C.M. and Roberds, W. (2006) ‘An introduction to payments economics’, unpublished thesis, University of Illinois. Lacker, J. (2006) ‘Central Bank credit in the theory of money and payments’, speech

General introduction

11

given at Federal Reserve Bank of New York Conference on Economics of Payments II, 29 March. Lester, B. (2005) ‘A model of interbank settlement’, unpublished thesis, University of Pennsylvania. McAndrews, J. and Rajan, S. (2000) ‘The timing and funding of Fedwire funds transfers’, Federal Reserve Bank of New York Economic Policy Review. McAndrews, J. and Trundle, J. (2001) ‘New payment system designs: causes and consequences’, Bank of England Financial Stability Review, 11: 127–36. Millard, S.P. (2006) ‘The foundations of money, payments and central banking: a review’, unpublished thesis, Bank of England. Millard, S.P. and Willison, M. (2006) ‘The welfare benefits of stable and efficient payment systems’, Bank of England Working Paper No. 301. Millard, S.P., Speight, G.E. and Willison, M. (2006) ‘Why do central banks observe a distinction between intraday and overnight interest rates?’, unpublished thesis, Bank of England. Norman, B., Shaw, R. and Speight, G.E. (2006) ‘The history of interbank settlement arrangements: exploring central banks’ role in the payment system’, unpublished thesis, Bank of England. Rosenblat, T.S. (1999) ‘What makes the money go round?’, PLD thesis, Massachusetts Institute of Technology. Schanz, J. (2006) ‘Innovation and ownership structure in payment systems’, unpublished thesis, Bank of England. Woodford, M. (2004) Interest and Prices: Foundation of a Theory of Monetary Policy, Princeton, NJ: Princeton University Press.

Part I

Payment systems and public policy

1

Central banks and payment systems Past, present and future Stephen Millard and Victoria Saporta1

Introduction Central banking and payment systems – mechanisms that enable the transfer of monetary value – are inextricably linked. In the past, institutions that developed into modern central banks stood at the top of the inter-bank payments hierarchy, providing the ultimate settlement asset exchanged by commercial banks when settling payments with each other. At present, modern central banks devote a considerable proportion of their resources to operating, overseeing and influencing developments in payment systems. In the future, innovations in payment system technology might permanently change the role of central banks, possibly even leading to their demise. And yet, the economics literature in the field is surprisingly scarce. With some honourable exceptions (including papers by the contributors to this volume), mainstream monetary economics has largely ignored the mechanics of how payments are actually made and banking theory has largely ignored the management of liquidity intraday. Even within central banks, payment systems are often treated as simply ‘the plumbing’ and left to technocrats. The aim of this chapter is to paint a broad-brush picture of the economic links between central banks and payments in the past, the present and the future. The purpose is ambitious and impossible to cover comprehensively in a single chapter – hence ‘broad-brush’. In particular, we start by arguing that the modern roles of central banks can be seen as natural outgrowths of their historical role in the inter-bank payments hierarchy. We then proceed to ask what are the characteristics of payment systems modern monetary authorities should be interested in? And how should this interest be made operational? Should central banks own, operate and/or oversee payment systems? We conclude with some tentative thoughts on how the payments landscape may evolve in the future and what that may mean for the future role of central banks. The chapter is organised as follows. We first provide background on the development of payment systems and central banking, arguing that historically they have been closely linked – the past. We then go on to analyse the role of modern-day central banks in the payment system, in particular in which systems should they be interested and how should they exercise this interest – the

16

S. Millard and V. Saporta

present. Finally, we offer ideas about the future direction of payment systems and of central bank involvement therein – the future.

Payment systems and central banking – the past Natural pyramiding Historically, the evolution of central banking can be traced back to the market’s natural demand for an efficient way to make payments. This natural demand can lead to the development of a hierarchy or pyramid in payments with the liabilities of a proto central bank at its apex, as the ‘settlement asset’ of choice. In other words, where institutions could provide a safe settlement asset that other banks use to settle obligations ultimately between themselves, they often developed the characteristics that in the twentieth century we came to associate with modern central banks. Payment systems form the means by which monetary value is transferred. Agents have a natural demand for a safe and verifiable asset – money – that they can use to transfer value in exchange for goods. This demand is derived from the low probability of the ‘double coincidence of wants’ necessary for trade in a barter economy (Jevons (1875) and, in a modern context, Kiyotaki and Wright (1989, 1993)). Given this asset, agents will eventually wish to find a way of being able to make payments – transfers of this asset – without having to carry it. There are at least two reasons for this. First, as suggested inter alia by He et al. (2005), money is susceptible to theft. Banks developed as places where people could deposit their gold for safekeeping. The banks would then issue their customers with receipts. These receipts represented a form of debt and, eventually, this debt became ‘transferable’ in the sense that it became possible for a merchant who wished to make a purchase to transfer the debt to the seller as payment for his goods. Final settlement occurred when the sellers went back to the bank to call in the debt. Second, as suggested inter alia by Kohn (1999), it was hard to verify the true value of different coins (the predominant form of money at this stage). Banks developed as places where agents could have their money counted and valued by money changers. As it was efficient for this process to only happen once, agents would leave their money – once counted and valued – with the money changers who would issue them with receipts. Payments were made with both payer and payee present at the bank. Where the payee did not hold an account at the payer’s bank, he either opened one or could ask the bank to transfer the money to his own bank. Since banks were close to each other, this was done by the payer’s banker walking over to the payee’s banker with the money. But, in an economy with many banks, it is inefficient for every agent to have an account with each and every bank and the banks themselves might be a long distance from each other. One solution is for each bank in the economy to have an account with all other banks and net obligations bilaterally with them. In a world with many banks this will tend to result in an inefficiently large number of

Central banks and payment systems

17

inter-bank accounts. A more efficient solution is for a hierarchy – or pyramid – of banks to develop, with banks at the bottom of the pyramid having accounts with correspondent banks in its upper tier which in turn have accounts with banks at the apex of the pyramid. Indeed, there is plenty of historical evidence that such pyramiding evolved naturally in a free-banking environment without the need for the state to superimpose and/or guarantee a ‘settlement institution’ at the apex of the pyramid.2 One example is the case of England (and, later, the United Kingdom). The Bank of England was founded in 1694 and was granted a number of privileges by the British Government, in return for its services in raising finance and managing the Government’s accounts.3 Due to these privileges, the Bank has been the largest and best capitalised bank in the United Kingdom for most of its history. Its large capital base and creditworthiness meant that it became the ‘custodian’ of choice – other banks naturally felt that it was the safest institution in which to hold their gold reserves, which they exchanged against Bank of England notes. Consequently, Bank of England notes (and later deposits) became the ultimate settlement asset for making payments, placing the Bank at the top of the payments pyramid in the United Kingdom. But, for most of its history, despite being the Government’s banker, the Bank did not enjoy an explicit government guarantee, nor was there an explicit or implicit acceptance that if the Bank chose to put the capital of its shareholders at risk the Government would step in to cover any resulting loss. For example, in 1890 the Chancellor of the Exchequer refused a request by the Governor of the Bank to guarantee its shareholders against loss if it were to support Barings Bank. A second example of natural pyramiding is the development of the Suffolk Bank system in Boston in the early nineteenth century. The development of this system is discussed in Goodhart (1988), Trivoli (1979) and Calomiris and Kahn (1996). At the time, ceteris paribus, Boston banks could issue fewer notes than their New England country competitors because the probability of a note being presented for payment varied negatively with the difficulty of travelling to the bank that issued it. This put the Boston banks at a competitive disadvantage to country banks and encouraged them to develop secure and systematic ways to redeem the various note issues that were circulating freely around the city. The Suffolk Bank ran the most successful system – it undertook to redeem at par the notes of country banks as long as they maintained sufficiently large deposits, topped up as necessary so as to make redemption at par possible. Moreover, the Suffolk Bank refused entry to its clearing system to banks it deemed not to have the requisite degree of integrity. In effect, it undertook an early form of supervision of banks. A third example relates to the arrangements for inter-bank payments in the United States during the period 1837–1913 (when there was no central bank in the country). Green and Todd (2001) explain that a hierarchy of correspondent bank relationships developed. Each small city had one or more correspondent banks and New York City had a number of banks that facilitated interregional

18

S. Millard and V. Saporta

payments; that is, there was essentially a ‘mutualised cooperative’ at the top of the pyramid. Put in the words of Smith (1936): The conspicuous position held by the banks of New York City in this respect – in 1912 six or seven of them held about three-quarters of all banks’ balances – seemed to point to the existence of spontaneous tendencies to the pyramiding and centralisation of reserves and the natural development of a quasi-central banking agency, even if one is not superimposed. (our italics)4 There is, of course, the issue of whether natural pyramiding is socially optimal or whether the government may wish to intervene by creating an institution that sits at the summit. On one view – referred to as the ‘jaundiced view’ in Calomiris and Kahn (1996) – private systems such as the Suffolk Bank system are driven by large banks seeking to limit the supply of money and engage in monopoly pricing. Any gains are at the expense of the smaller banks and the public as a whole. An alternative view – the ‘sanguine view’ – is that such arrangements increase efficiency and reduce risk in the banking system. Calomiris and Kahn (1996) suggest that empirical evidence backs the sanguine view in the case of the Suffolk Bank system (see also Selgin and White (1994) for a similar view).5 But regardless of whether such natural pyramiding is socially optimal, the fact that it seems to occur raises the question of how many banks would naturally take this role at the top of the hierarchy? Does the market, in each currency, tend to one proto central bank or more? The relative standing of different banks and the structure of capital market flows in a country are important factors – as in the case of the Bank of England. Another important factor is the structure of the banking market.6 In an oligopolistic ‘free banking’ market with few banks, it may still be efficient for banks to hold bilateral correspondent accounts with each other, settling in each others’ monies, rather than in an outside settlement asset. According to Green and Todd (2001), in Canada banks did just this, until recently. In consequence, markets with a few large banks dominating the system may tend to develop flatter upper-tier structures. In contrast, in a unitbank system – that is, a system consisting of a large number of small independent units – efficiency considerations will lead the smaller units to seek an arrangement that would decrease the number of inter-bank relationships. In such systems ‘proto central banking agencies’ may develop naturally.7 Features of the settlement institution What are the financial features that such proto central banks need to display to enjoy a comparative advantage in performing the functions of the settlement institution at the apex of the pyramid? First, if a central bank is commercially-oriented – in practice, if it is privately-owned – it needs to find ways of overcoming various conflicts of interest. Banks whose payments are made via the settlement asset of the proto central

Central banks and payment systems

19

bank would need assurances that it is not using the information in their accounts to compete unfairly with them. Further, there may be a tension between a bank’s pursuit of profit and its role as a central bank. There is plenty of historical evidence that suggests that when the provider of the ultimate settlement asset is also a commercial bank, conflicts of interest ensue, especially during periods of financial crisis. For example, in 1793, the Bank of England was asked to aid some large country banks – it refused to do so and some important failures occurred that spilled over to the London market. Henry Thornton (1802) writes that ‘a sense of unfairness of the burden cast on the Bank by the large and sudden demands of the banking establishments in the country, probably contributed to an unwillingness to grant them relief’. Goodhart (1988), inter alia, describes a number of other examples, involving commercial rivalries among the Suffolk Bank and the New England country banks, the Bank of England and the London bill brokers and the Banque de France and potential commercial competitors. One way of overcoming the conflict of interest problem is for the central bank to stop competing with the other banks for non-bank business. In effect, this was how the Bank of England overcame the problem, withdrawing from all (new) commercial activity with non-bank entities between around 1880 and 1910 (Goodhart, 2004). Second, the banks whose payments are made using the central bank’s liabilities as settlement asset need to be confident in the credit quality and liquidity of this asset – where liquidity should be taken to mean the acceptability of this asset as a means of payment by others. At least one of three features recur in the development of institutions as central banks and help explain other banks’ willingness to use these assets: (a) the provider’s bank notes and deposits are backed by a commodity with intrinsic value (such as gold); (b) the provider has a very large capital base such that the probability of failing to realise its obligations is very small; and (c) the provider holds an explicit or implicit government guarantee. During different times in its history, the Bank of England had each of the features (a), (b) and (c). The fact that the Bank had the largest capital base of any bank in the United Kingdom well into the nineteenth century – feature (b) – was the key factor in explaining why the Bank’s liabilities became the settlement asset of choice. During this time, the Bank’s relative standing as the banker to the Government might have created the impression that it had an implicit government guarantee – feature (c) – but any such impression was certainly less firmly held than today. The Barings episode in 1890 is a concrete example of the Government refusing to underwrite the capital of the Bank. It was not until 1844 that the Banking Act placed restrictions on the Bank’s ability to print notes that were not backed by gold. It stipulated that the Bank had to hold gold reserves against all the notes it issued in excess of a fiduciary issue of £14 million – essentially forcing the Bank to display feature (a). However, the regulation was suspended during subsequent periods; in particular, during the liquidity crises that occurred in 1847, 1857 and 1866, the government allowed the Bank to issue additional notes not backed by gold. Again, however, there was no indication that the government would underwrite any losses the

20

S. Millard and V. Saporta

Bank made as a result of intervention. In 1946, the Bank was nationalised – effectively giving it feature (c).8 But the Bank of England’s history may be unusual. In many cases, more than one bank exhibited features (a) and (b). Either you would expect to see these banks ‘jockeying for position’ as the central bank or you would expect to see flatter upper tier structures as in Canada or the New York Clearing House system with ultimate settlement (where necessary) carried out in gold or government bonds.9 In such cases, eventually central banks tended to be superimposed by the state to provide the ultimate settlement asset – as, for example, in the cases of the Bank of Canada, the Federal Reserve, the Reichsbank and the Swiss National Bank, among others. Often, but not always, this happened in response to banking crises that were perceived to result from the lack of a central bank being able to provide lender-of-last-resort assistance. For example, the Federal Reserve System was set up in 1914 in response to the banking panic of 1907 as a direct result of a perceived need for a government-backed lender of last resort in such circumstances. This is discussed in more detail below. Proto central banks, financial and monetary stability At present, central banks all around the world generally share two closely related core purposes – monetary and financial stability. In this sub-section we ask why central banks evolved as the natural candidates for taking on these two responsibilities and argue that the answer lies in the key role proto central banks played in the payment system. That is, these core purposes can be seen as natural outgrowths of a central bank’s role in payments. In Section 3, we reverse the question and ask, given that modern central banks are currently the public institutions commonly charged with the preservation of monetary and financial stability, what do these core functions imply for their modern interest, and active involvement, in payment systems? Historically, privately-owned settlement institutions that supplied the settlement asset at the top of the payments pyramid had a natural interest in ensuring the ability of their client base – the banking sector as a whole – to meet the public’s demand for liquidity. The reason for this is that if it allowed a solvent commercial bank to fail as a result of a run, it would only aggravate the situation and this could ultimately result in a run on itself. Also, assuming the commercial bank stayed in business, the central bank would make a high return on this lending (given that lender-of-last-resort assistance would typically be given at high rates of interest). Put differently, profit maximisation is consistent with Bagehot’s (1873) rule that a central bank should always lend to liquid but solvent institutions against collateral – that is, be a ‘lender of last resort’. Historical evidence backs this assertion. For example, the Bank of England provided lender-of-last-resort assistance during the financial crises of 1857 and 1866. Equally, the status of a proto central bank at the top of the payments pyramid derived from the fact that it was perceived to be ‘safe’ – that is, an institution with a large capital base, holding high quality assets. So a commerciallyoriented central bank would also need to be concerned about its own soundness.

Central banks and payment systems

21

This would give it incentives to be careful about to whom it should provide settlement accounts and to monitor these banks; one can think of this as an early form of banking supervision. In addition, it also had to weigh carefully the advantages of providing lender-of-last-resort assistance to the banking system to avoid a drop in its revenue stream against the risk of lending to an insolvent institution and making a loss that could decrease its capital base and threaten its reputation as the supplier of the ultimate settlement asset. As a result, proto central banks were more likely to let healthy banks go down than risk lending to unhealthy banks by mistake. Hence, in a fractional reserve system, central banks without a government guarantee have incentives to maintain financial stability – that is, grow their balance sheet to avoid crises – but at a sub-optimally low level. This is the typical justification for a role for the public sector in providing financial stability. Indeed, the full title of the Federal Reserve Act of 1913 reads, ‘An act to provide for the establishment of Federal reserve banks, to furnish an elastic currency, to afford means of rediscounting commercial paper, to establish a more effective supervision of banking in the United States, and for other purposes’ (our italics). Privately-owned providers of the ultimate settlement asset also have incentives to maintain the value of their liabilities. In particular, if the proto central bank printed more and more of its notes without a corresponding increase in the demand for them, the notes would fall in value relative to those of other banks. Eventually, the proto central bank would no longer be seen as ‘safe’ and it would lose the revenue it obtained from acting as the settlement institution. This is why these proto central banks emerged as natural candidates for ultimately being charged with maintaining ‘integrity’ and ‘confidence’ in the currency in modern fiat monetary systems.

Payment systems and central banking – the present Payment systems and central banks have evolved in tandem. In addition, the ultimate development of the core functions of modern central banks – monetary and financial stability – has been closely linked to their role in the provision of the ultimate settlement asset in the payment system. But, apart from providing the ultimate settlement asset, central banks have played a number of other roles in the provision of payment services including the ownership and operation of some payment systems, both retail and wholesale, and overseeing systems which they do not own or operate themselves. This raises a current pertinent policy issue for central banks. History aside and given their core functions of monetary and financial stability, to what extent, and why, should central banks play these other roles in payment systems in modern times? In what follows, we first tackle the ‘why’ question, asking specifically in which payment systems a modern central bank should be interested. We then tackle the question of how this interest might best be made operational.

22

S. Millard and V. Saporta

In which payment systems should a central bank be interested? Modern central banks are typically charged with the provision of monetary and financial stability. Any attempt therefore to answer the question of why central banks should have an interest in payment systems should start with an analysis of how disruptions in payment systems could affect monetary and financial stability. And this, in turn, should start with a definition of what we mean by monetary and financial stability. To define ‘monetary stability’ it is useful to start from a definition of ‘money’. Money is normally defined by its four functions: unit of account, store of value, medium of exchange and means of deferred payment (settlement). By controlling inflation – that is, helping to ensure stable prices – the central bank enables money to perform the first two of its functions, though its roles as a medium of exchange and means of deferred payment are likely to be severely compromised in a situation of high inflation. But it is possible to think of situations in which, although inflation is low, money is still not able to perform its roles as a medium of exchange and means of deferred payment. Historically, shortages of coin have meant that agents in the economy have been unable to use money as a medium of exchange and have had to resort to barter. Alternatively, and in a modern context, the failure of a payment system might mean that agents are unable to use money held in bank accounts to make payments, leading to a loss of confidence in money more generally. This could happen regardless of the rate of inflation.10 This suggests a broad – monetary stability – objective for a central bank of ensuring that money can perform its functions of unit of account, store of value, medium of exchange and means of deferred payment in all states of the world. Financial instability is normally thought of as a situation in which shocks to the financial system – institutions, markets and financial infrastructure (including payment systems) – have contagious external effects elsewhere within the financial system with consequences for social welfare.11 It is worth noting that the financial stability responsibility of central banks can be derived from the broad monetary stability responsibility as we have defined it above, as problems in financial institutions and markets would clearly disrupt the ability of money to perform its functions of medium of exchange, store of value and means of deferred payment. Having defined the objectives of a central bank, we now consider what this implies for its interest in payment systems. To understand the characteristics of payment systems – means of transferring monetary value – one first needs to understand the economic purposes of money and of banks (given that commercial bank deposits form the bulk of the stock of money in modern economies). Money developed as a way of overcoming the problem of a lack of ‘double coincidence of wants’ in a barter economy, i.e. as a medium of exchange. The most basic form of money is ‘cash’. Where a central bank is the monopoly issuer of banknotes, a key element of its broad monetary stability remit will be ensuring that its banknotes can act as a reliable medium of exchange in all cir-

Central banks and payment systems

23

cumstances – in effect, to ensure that cash can always act as the payment medium of last resort. But cash forms only a small part of the total money supply; the bulk of money in the economy today exists in the form of bank deposits. ‘Banks’ originally developed to deal with two specific problems with cash: susceptibility to theft and the difficulty in verifying its true value. Because of these problems, cash was not always able to perform its functions of medium of exchange, unit of account, store of value and means of deferred payment. Banks enabled payments to be made with ‘bank money’ – banknotes, bills of exchange and payments ‘in bank’ – rather than with cash. In essence, banks and bank money enable ‘money’ to perform its functions more effectively. Where economic agents wish to make a payment to someone who has an account at the same bank, they can make the payment ‘in bank’. One can think of the internal accounting systems of banks over which such payments are made as a ‘payment system’ since they are allowing agents to make effective use of ‘bank money’. Where the payer and payee do not hold accounts with the same bank, then the payer will need to transfer a claim on his bank to the payee. Eventually, this claim will be deposited by the payee, leaving his bank with a claim on the payer’s bank. As this process will be happening for many agents banking with different banks, all the banks will be building up claims on each other, which need to be settled at some time. ‘Final settlement’ takes place via the transfer of an asset that the creditor bank is happy to accept. Although some banks will be happy to accept settlement in the money of other banks, there will be some level at which the banks will not accept settlement in each other’s money; settlement between such banks can only then take place through an ultimate settlement asset. This whole process – from the point at which a payer transfers a claim to a payee through to final settlement – defines an inter-bank payment system. Such systems enable money to perform its roles of a store of value, medium of exchange and means of deferred payment. Different payment systems have evolved over time to handle different types of payments. Some distinctions can be made between payments that support particular financial market transactions (e.g. unsecured inter-bank loans, securities purchases), payments that are made electronically or by paper, regular payments that can be preset in a bank’s systems, payments that are made for retail purchases in shops, etc. Disruptions to the systems used for making certain types of payment are likely to matter more to a central bank than to other systems. But in deriving the key set of characteristics for a central bank, we need to assess what matters for the ability of money – and bank money in particular – to perform its functions. This ability can be disrupted if a system that processes large values and/or volumes of payments is disrupted and the participants in the system cannot divert their payments to another system. Further, if disruptions to a single system led to a general loss of confidence in all payment systems, then bank money would again be prevented from effectively fulfilling its functions of a medium of exchange and means of deferred payment.12 Indeed, in this case, it is likely that agents would turn to cash to make their payments. This possibility

24

S. Millard and V. Saporta

again emphasises the importance to a note-issuing central bank of ensuring confidence in its banknotes. The above analysis suggests that payment systems can be characterised in terms of three features: •

•

•

Size – other things equal, problems in a system handling a large value of payments are more likely to lead to contagious losses than problems in a system handling only a small value of payments, since the inter-bank exposures within them are likely to be higher relative to banking system capital. Problems in a system handling a large volume of payments are also likely to lead to the disruption of more transactions than problems in a system handling a small volume of payments. Types of payment – payment systems that support particular financial markets within which problems could result in financial instability; or systems that enable the payment of non-discretionary payments (such as wages, salaries and bills) as opposed to discretionary payments (such as retail purchases); or handle payments that are more or less urgent than payments handled by other systems. Availability of substitutes – that is, if a system were to develop problems, would agents still be able to make their payments by switching easily/ costlessly to an alternative system?

We can now attempt to map these characteristics into the broad monetary and financial stability objectives of a central bank. Agents’ demand to make payments is what creates a demand on the part of the banks for the ultimate settlement asset – central bank money (whether held as cash or accounts at the central bank). And it is precisely this demand for its liabilities that enables the central bank to carry out monetary policy. So, in order to carry out monetary policy so as to maintain stable prices, a central bank will need to take an interest in the payment systems in which agents use the ultimate settlement asset. In most developed economies, these will almost certainly include the large value payment systems where unsecured money market transactions are settled and the security settlement systems for government bonds and other central bank-eligible securities.13 In addition, with the broad objective of ensuring that money can perform its functions in all states of the world, a central bank should, in principle, also take an interest in disruptions to systems that could lead to contagious losses among banks and other financial institutions, disruptions to other systems and/or problems elsewhere in the financial system more generally – that is, financial instability – since these problems could, in turn, prevent money from fulfilling its functions. In most developed economies, the set of relevant systems is likely to comprise the large-value payment and clearing and settlement systems that support financial market transactions in money, securities and derivatives. A central bank should also take an interest, in conjunction, where relevant, with the prudential supervisory authority, in disruptions to the internal systems of

Central banks and payment systems

25

banks where such disruptions prevented customers of that bank from making payments with contagious effects on the rest of the financial system. A recent example is the disruption of Bank of New York (BoNY) following the 11 September 2001, terrorist attacks.14 Beyond these systems, the central bank will need to consider what would be the minimum necessary involvement to enable bank money to carry out its functions in both normal and crisis states of the world. One answer would be to rely on cash. But in a situation where there were no other means for agents to make payments, banks could potentially run into liquidity problems with systemic consequences. In such situations, the monetary authority would need to engineer a huge increase in the supply of cash, which it may not be able to do with adequate speed (depending on the amounts involved). Moreover, reverting to a ‘cash economy’ would reintroduce the welfare costs that bank money eliminates and will be welfare-reducing in itself. A better answer would be to ensure that, in each state of the world, there was at least one payment system available to a sufficient proportion of the population. There would be different options available to a central bank to make this objective operational. The central bank could provide such a system itself or ensure that it was in place via regulation or provide a back-up system ready to come into operation in times of crisis. Alternatively, it could ensure – via regulation, oversight, encouraging competition, etc. – that there were many substitute systems available for agents to use. How should central banks’ interest in payment systems be made operational? Central banks have an interest in ensuring that the payment systems that are important to the functioning of the monetary and financial system remain open in both normal and crisis states. But this interest does not necessarily imply that a central bank or any other public authority should intervene in the payment system. Public sector intervention can only be justified in the presence of market failures and only then if cost-effective instruments that can mitigate these failures exist. In this section, we first set out the externalities that may justify public sector intervention in payment systems before going on to describe and evaluate a set of stylised models of intervention that could mitigate such externalities. Market failures and payment systems Payment systems can give rise to the following externalities that may justify public sector intervention: •

As discussed in Bank of England (2005), payment systems create systemic risk externalities that private sector owners and operators of payment systems may not internalise, in the absence of intervention. Although there is no universally agreed definition of ‘systemic risk’, it is generally accepted

26

•

•

•

S. Millard and V. Saporta that it includes the following four contagion-related risks, all of which may be mitigated by the public sector influencing the way payment systems are designed and operate: (i) the risk that the failure of one member of a system will lead to other members suffering losses and potentially also failing; (ii) the risk that operational disruptions to one member, or features of the design of the payment system, may lead to ‘liquidity sinks’, whereby a member may receive large pay-ins but for technical reasons cannot, or for other reasons does not, make pay-outs, which in turn exposes other members of the system to liquidity pressures; (iii) the risk that the operational or financial failure of a payment system itself may disrupt financial markets with knock-on consequences for banks and other financial intermediaries; and (iv) the risk that news about the failure of one payment system leads to failures of other payment systems that were not subject to the original shock through ‘loss of confidence’ effects operating through agents’ expectations. Systemic risk externalities fall squarely within the remit of a central bank. That said, to the extent that some of the systemic risks can be mitigated through strengthening the credit and operational risk management of individual members of the payment system, the prudential regulatory authority – in countries where it is separate from the central bank – should also be involved. Moreover, for the case of the internal payment systems of individual banks, where these are judged to be of systemic importance to the payment and settlement system as a whole, the interest would clearly lie with the prudential regulator. As discussed in Bolt and Humphrey (2005), payment systems are characterised by network externalities and increasing returns to scale which in turn imply a strong tendency to monopoly. While relevant to the central bank, one could argue that regulating against this type of market failure would more naturally fall to competition authorities.15 A number of payment systems are organised as member-owned cooperatives and are typically beset by collective action problems.16 Such problems can potentially lead to an ‘inefficient’ set of payment systems being available to consumers. Again, while relevant to the central banks, it is not clear that the central bank is best placed in dealing with such problems; again, the competition authorities may be better suited. Finally, the informational asymmetries between banks and their customers, which lead to ‘consumer protection/conduct of business issues’ within the banking industry, may give rise to a need for regulating the information provided by banks on their provision of payment services.17 Clearly, the government agency charged with regulating the conduct of business of banks has the natural locus here.

Models of intervention No form of public sector intervention to correct market failures will be effective without appropriate instruments for monitoring performance and appropriate

Central banks and payment systems

27

powers of enforcement; that is, some lever to ensure compliance with a set of standards or principles. Together, these determine the degree of control over the payment system that the public sector is able to apply in mitigating the social costs of the market failures set out above. While the central bank will be in a position to exercise a degree of control through its role as the provider of the ultimate settlement asset and its influence as banker to the banks, it may be desirable/necessary for the public sector at large to supplement this by taking additional actions. In practice, these will be drawn from choices taken along three dimensions: (i) public sector ownership of the system; (ii) public sector operation of the infrastructure; and (iii) public sector oversight or regulation of the system. Before going any further, it is useful to define each of these terms. By ownership of the system, we refer here to the case in which the public sector, typically the central bank, has an ownership stake in the entity governing the payment system, or a role in its governance. A controlling stake (greater than 50 per cent) affords the pubic sector the ability to design the system in accordance with its own objectives (preferences) and enforce continued compliance. By operation of the infrastructure, we refer to active public sector engagement in the design, implementation and operation of all, or a sub-set, of the elements such as software, hardware, communication networks, data centres and contingency sites that underpin modern-day payment infrastructures. In practice, the central bank may be the authority best-placed to take on this role, as it can leverage the systems it maintains to hold, monitor and control accounts for financial institutions that bank with it. It is worth noting, however, that operation of the payments infrastructure is a separable activity from the provision of the ultimate settlement asset. While acting as the settlement agent gives the central bank direct access to information on the payment flows of system members that settle in central bank money, operation of the infrastructure provides a complementary, yet distinct, instrument for exercising direct control over operational capacity and performance. By oversight of the system we refer to day-to-day regulatory activity that ensures continued compliance with a set of minimum standards and design principles (e.g. the Core principles for systemically important payment systems set by the Bank of International Settlements (2001)). In practice, this activity is invariably carried out by the central bank but, both conceptually and in practice, it need not be. Oversight, with adequate powers/influence, may be a substitute vehicle for enforcement in the absence of public ownership of the system. Each activity can, in practice, be carried out with varying degrees of formality and intensity: ownership can range from no role in governance, through a seat on the Board, through to a controlling stake; operation can range from complete outsourcing with appropriate control/monitoring procedures through to design, maintenance and operation of all key components of the infrastructure; and oversight, if carried out at all, can be carried out with limited and informal powers through to extensive and formal powers of direction and enforcement. For illustrative purposes, however, we can think of a spectrum of possible models of intervention, in which these three activities are somehow combined.

28

S. Millard and V. Saporta

Model 2

Model 1

OWNS

Model 4

Model 5

OVERSEES Model 6

Model 3 OPERATES

Figure 1.1 Stylised models of intervention in payment systems.

These models can be shown to lie along the surfaces and in the interior of a three-dimensional cube (as shown in Figure 1.1). An evaluation of the effectiveness of the various alternative models of intervention can then start with an assessment based around the six relevant corner models identified in the diagram: model 1 (‘owner/operator’) lies at one corner of the cube, with the central bank assuming a controlling ownership stake and operating the key components of the infrastructure; model 2 (‘owner’) involves a controlling stake in the ownership of relevant payment systems, but with the infrastructure operated by a private sector provider; model 3 (‘operator’) involves the central bank operating key components of the infrastructure on behalf of a private sector owner; model 4 (‘overseer’) involves pure oversight with powers of enforcement (without excluding the role of central banks in providing the ultimate settlement asset for systems that settle in central bank money); model 5 (‘overseer/operator’) combines oversight and operation of the infrastructure, without ownership; and, finally, in model 6 (‘laissez-faire’), there is no active public sector intervention. The two further corner models, not identified in the diagram, combine both ownership and oversight. While this might be feasible – i.e. central banks may wish to perform an internal audit function on their provision of payment services – it is difficult to imagine that, in such circumstances, oversight offers incremental value in achieving central bank objectives. Hence, we do not consider these in our analysis. Evaluation of corner models of intervention Models in which the central bank assumes ownership (i.e. models tending towards 1 and 2 in Figure 1.1) may be highly effective in achieving the central

Central banks and payment systems

29

bank’s objective of ensuring that payment systems that are critical in the operation of the monetary and financial system are free from systemic risk externalities. With complete power of enforcement and access to information for monitoring purposes, these models can give the central bank power to determine the level of resilience and systemic risk mitigation built into payment systems. A counter argument here, however, is that public ownership can undermine the continued engagement of members of the payment system. Without a stake in how the system is run, the banking system may be more inclined to divert flows to competing, and perhaps more risky, systems and payments vehicles. Indeed, it is important to recognise that powers of direction over the characteristics of the system do not translate into powers of direction over its usage: a central bank may be able to force every bank to have an account, but cannot force them to use it. An additional consideration might be that models incorporating ownership, especially those combining this with operation, will entail a cost to the central bank and ultimately the taxpayer. If full-cost recovery could not be achieved, the central bank under these models would, in effect, be subsidising the provision of the payments infrastructure. And although a subsidy might be justified if central bank intervention was directed towards addressing the private sector’s underprovision of a public good in the absence of negative externalities, a priori it appears less appropriate when private sector solutions are plagued by negative externalities. In addition, a central bank might not be indifferent between models with and without operation. First, and most critically, outsourced operation of the infrastructure raises the question of control over access to central bank money settlement. Ensuring balance sheet integrity is an important element of a central bank’s monetary stability objective and systems settling in central bank money, particularly those settling in real time, can potentially expose the central bank to losses arising from system-error or mismanagement. Where settlement is outsourced, the central bank can only control such exposure at arms’ length. Different central banks may have different risk tolerances in this regard. Second, the higher the degree of operational control exercised, the greater the central bank’s access to information and data and the greater its ability to both monitor and enforce operational standards applied to the system. Third, outsourced operation potentially introduces principal-agent frictions, if the private firm’s incentives cannot be aligned perfectly with those of the central bank. Finally, the cost considerations outlined above are likely to be an order of magnitude greater when the central bank is also the operator. Under models tending towards model 3 (‘operator’), the central bank has a high degree of operational control, but no formal powers of enforcement afforded by either ownership or oversight. Here, the central bank’s monitoring capability may be enhanced and, subject to its ability to recover costs (or its willingness to subsidise), the central bank can assume a certain level of control over initiatives to address operational risk. It may also be that regular and close engagement with the users may afford the central bank additional influence and

30

S. Millard and V. Saporta

control over other aspects of the owner’s strategy. However, this model in itself affords no legal means to enforce risk-mitigating actions over which it has no direct control – e.g. over the settlement model adopted by the payment system, the rules covering direct credit caps, membership rules, etc. Under an appropriately designed pure-oversight model (one tending towards model 4), on the other hand, the overseer would have powers to force a private system owner to comply with any desired risk-mitigating action, which may be as effective as under an ownership model. However, being a step-removed from the operation of the system, it is unlikely that the overseer would be able to obtain the same quality of information on operational processes and performance. That said, with adequate powers, the overseer would be able to prescribe the requisite level of reporting/information provision, and hence the gap here might not be so large. Also, with any specific actions imposed by the overseer implemented at the members’ own expense, there is no issue around cost recovery: oversight essentially constitutes taxation of the members. Finally, as noted above, a model with mutual ownership might encourage the continued engagement of the members. Model 5 is essentially a hybrid of models 3 and 4, whereby the public sector operates the payment system and also has regulatory powers to impose penalties to enforce performance. With close operational control supported by powers of enforcement, and no issues around disengagement of members, such a model would perhaps be the most effective in achieving the central bank’s objectives. A caveat here is that the combination of operation of the infrastructure and oversight might expose the central bank to conflicts of interest, either apparent or real. In the presence of market failures and cost-effective instruments allowing their mitigation, model 6 would clearly be the least appropriate in achieving public sector objectives. So far we have focussed on evaluating models of intervention in terms of their effectiveness in mitigating systemic risk externalities that fall clearly within the remit of modern central banks. But as we noted earlier, there are several other market failures that may raise distinct but equally valid public sector concerns that may or may not fall within a central bank’s official remit. In seeking to eliminate these market failures, the public sector at large will have an interest in facilitating the development of an infrastructure that promotes welfare through providing the optimal set of payment instruments (direct debits, credits, cards, cash and the like), infrastructural platforms (the IT networks that ensure that payment messages are sent, processed and settled) and settlement assets (combination of bank and central bank money), at the optimal price-quality combinations.18 The models considered in Figure 1.1 might equally be applied in the pursuit of these public policy objectives. Something close to a pure oversight model (model 4) is likely to be most effective in this regard. A regulator, if appropriately empowered, can catalyse infrastructural development beneficial to welfare by resolving coordination problems in the industry under the threat of regulation

Central banks and payment systems

31

– this is effectively the line the Office of Fair Trading in the United Kingdom pursued in persuading the industry to build a Faster Payments Service.19 A regulator is also well-placed in taking steps to eliminate excessive rents in a sector characterised by network externalities and scale economies (not unlike in other network industries, such as telecoms or electricity generators). And by addressing market failures through regulation rather than subsidy, oversight would seem to be the appropriate public policy instrument to mitigate the frictions that lead to the innovation and consumer protection market failures outlined above. Of course, if different authorities have different interests and multiple overlapping oversight regimes emerge, conflicts or duplication of effort can arise. In this regard, clarity as to each authority’s responsibility is essential. If this strategy does not work, the public sector may seek to solve the problem by building and operating the service itself, for a period at least.20 This is the approach many central banks have taken historically. But permanent public sector operation of the payments infrastructure is unlikely to be as beneficial to this objective as private sector ownership subject to regulation because the public sector is unlikely to be as innovative as a private sector provider in developing efficient, high quality and cheap IT network solutions. Model 2, in which the central bank owns but does not operate the system, will rank higher than models that also involve operation because, despite owning the system, private sector providers of infrastructure are free to compete for the central bank’s business. Table 1.1 ranks the stylised models of intervention against their effectiveness in mitigating the systemic risk externalities that fall squarely within most central banks’ remits and the other frictions set out on pages 25–6. It is clear from the table that, while models involving ownership may be effective in delivering on central banks’ monetary and financial stability objectives, these models could stifle competition and innovation in the payments sphere and hence potentially compromise other public sector objectives. If combined with outsourced operation, however, there may be a lower efficiency cost to public ownership. Taking all public sector objectives together, an oversight-only model might be preferred, as long as sufficient powers of enforcement could be conferred upon the overseer whose responsibilities were clearly defined. Table 1.1 Ranking models of intervention Model

Degree of effectiveness in mitigating systemic risk externalities

Degree of effectiveness in achieving additional public sector objectives

Model 1: owner/operator Model 2: owner Model 3: operator Model 4: overseer Model 5: overseer/operator Model 6: laissez-faire

High/medium High/medium Medium/low High/medium High Low

Low Medium Medium/low High Medium/low Low

32

S. Millard and V. Saporta

Practical considerations Whether such an oversight-only solution will be chosen will depend on a number of practical considerations, including the risk preferences and budget sets of the public authorities involved (which in turn may depend on political economy considerations) and a weighing of the costs of introducing a statutorybased oversight regime relative to the expected welfare benefits. One factor affecting costs would be the prevailing institutional structure for financial system regulation and oversight in the country concerned. In countries where multiple authorities have overlapping interests in systemically important payment systems – e.g. in countries where the prudential regulator is not the same agency as the central bank – the net (of costs) social benefits of introducing formal legislation delineating responsibilities and powers between agencies may turn out to be lower than in an arrangement whereby the central bank and the prudential regulator agree on how they should co-operate in discharging their respective roles, leveraging on their existing powers.21 Moreover, in an increasingly globalised financial and monetary system, payment systems that are of systemic importance to a particular country may not be located in that country at all – e.g. Euroclear, which owns the UK securities settlement system CREST, is incorporated in Belgium and CLS Bank, which settles wholesale foreign exchange transactions between sterling and all major currencies and which is incorporated in the United States. For these systems, an oversight arrangement based on domestic-only statutory powers is unlikely to be enforceable; instead, international co-operative oversight arrangements between ‘host’ and ‘home’ authorities become necessary. There is also the question of whether a uniform approach to intervention is necessary. Indeed, on an examination of models of intervention across various countries, consistency of approach seems to be the exception rather than the rule. A series of tables in the Annex maps a set of international payment systems to the corner models shown in Figure 1.1. It is immediately clear that, of the countries included in the tables, Canada is the only one adopting the same model of intervention for all of the payment systems covered in the tables. Elsewhere, central banks have tended to adopt a mix of models, with models involving ownership and/or operation of the system more prevalent in the case of largevalue payment systems or payment systems embedded in securities settlement systems. Historical (or path) dependence appears to be one factor behind the cross-country variation in models of intervention.22 That said, the different models applied may also reflect a judgement on the extent of each system’s contribution to the market failures the intervention is intended to address, or the balance of central bank interests versus those of other public authorities. For example, a ‘lighter central bank touch’ might be envisaged for retail versus wholesale payment systems, where competition, innovation and consumer protection concerns may weigh more heavily.

Central banks and payment systems

33

Payment systems and central banking – the future In this section we offer some tentative ideas about the future direction of payments and central bank involvement in payment systems. We first examine whether cash will endure and we then move on to discuss how wholesale financial transactions might be made in the future. The chapters by Freedman, Leinonen, Pattinson and Schmitz in Part IV of this volume discuss these issues in more detail. We also tentatively explore the implications for the ability of central banks to carry out their core monetary and financial stability functions in the future. How will consumers pay for goods and services in the future? When thinking about the question of how consumers will pay for goods in the future, one question to ask is whether cash will endure or whether it will be replaced by some form of electronic money. Cash transactions, though declining, still form about 75 per cent of personal payments in the United Kingdom, although their value tends to be small (abstracting from illegal black market transactions). In thinking about why cash is so enduring, it is important to note that any replacement would have to offer its user the same level of anonymity, universal acceptability and recognisability; no current alternative has ever done this. Another advantage of cash over, say, credit and debit cards is the fact that final payment takes place simultaneously with the provision of the good or service; the seller is not exposed to settlement risk. Agents like anonymity. Indeed, the lack of anonymity in credit cards, for example, has led to the large and increasing problem of identity theft; in the United Kingdom, for instance, credit card fraud in 2004 totalled $966 million.23 An alternative view – expounded by Buiter (2005) – is that since anonymity is of most use to criminals, there may be a case for the government to do away with legal tender currency issue by the state, while making sure that private note issuance continued to be banned. In that case, all payments would have to be made using a medium – such as credit cards – in which the purchaser of any good could always be identified. Drehmann et al. (2002) point out that such a policy ‘would be appallingly illiberal’ and so unlikely to be contemplated. But the fact that no current alternative to cash is able to match its attributes does not mean that such an alternative will not exist in the future. One might think that eventually, in place of cash, some form of e-money will exist that offers the same complete anonymity, universal acceptability and recognisability as cash but will not be useable by anyone other than the holder of the e-money. This will reduce the incentive of others to steal the e-money and so make it a ‘safer’ asset to hold than cash. But it may well be that cash and e-money can coexist. A key issue here is anonymity. In practice, regulators have forced e-money transactions to be auditable (that is, limited anonymity); if they continue to do so, cash would always have the advantage of anonymity. E-money will always be subject to ‘operational risk’ where cash is not – at least once it has been taken out of the

34

S. Millard and V. Saporta

ATM.24 Finally, unless the central bank were to issue the e-money (or, equivalently, the government were to underwrite the issuing companies), then it would always be subject to credit risk. Dowd (1998) and Friedman (1999) suggest that the demand for central bank money – and cash in particular – has fallen dramatically over recent years and that it will, possibly, fall to zero eventually. Given this, they argue that changes in the supply of central bank money – that is, monetary policy – will increasingly have less impact on the wider economy, in the limit having no impact at all. In effect, central bank money is just one of a number of competing monies; the price level itself, at that point, would need to be tied to a commodity or, alternatively, a bundle of financial assets. But Woodford (2004) notes that as long as central bank money is the ultimate settlement asset – that is, there is a need for it in order that banks can make payments to each other – there would always be some demand for it even in the absence of central bank notes and this would mean that central banks could carry out monetary policy exactly as before. Buiter (2005) argues along similar lines, noting that even if there were no demand for central bank money in normal times, there would still be a need for central banks to supply liquidity to the banking sector in times of stress. Hence, central bank deposits are likely to be replaced by overdraft facilities, lines of credit or other contingent claims on central bank money; if the demand for such claims were sufficiently stable, the monetary authority could still set short-term interest rates. How will wholesale payments be made in the future? With respect to wholesale payments, the trends seem to point to two opposing corner outcomes: one integrated payment system perhaps covering the whole world (a mega-version of the Continuous Linked Settlement (CLS) system that settles foreign exchange transactions for the major world currencies) or a large number of competing private payment systems. The benefits of the first corner outcome would be large savings in collateral, IT communications and other costs; the downside would be the ‘single point of failure’ problem associated with a massive concentration of risk in one system and the general inefficiencies usually associated with monopoly providers. In terms of implications for central banks, the net effect on monetary and financial stability would appear to depend upon the level of systemic risk in the system (or systems), together with the degree of control that central banks can exercise over this system(s). Aligning incentives of different national overseers-regulators will become an even more important determinant of which outcome prevails. A future of competitive private systems granting agents the ability to transfer financial assets to pay for goods in real time forms the basis of the vision of the future put forward by King (1999).25 In a world of competitive systems, the systems would need to be linked, but this need not be via a central bank provided there was general agreement on acceptable settlement assets. In such a world, central banks would become pure regulators operating monetary policy by tightly defining the unit of account – like ‘weights and measures’ inspectors.

Central banks and payment systems

35

The unit of account could be one of the settlement assets but need not be. Greenfield and Yeager (1983), for example, analyse an economy in which the unit of account is a bundle of standardised commodities (conceivably financial assets) and is not used as a medium of exchange. In addition, accounts held in it cannot be redeemed for the ‘bundle’ but instead are redeemed for something of intrinsic value. In this world, a central bank could define this unit of account and offer a continuous quote of its value against all the financial assets used as media of exchange. But, as pointed out by Selgin and White (1994), the information costs associated with doing this are large; information and transactions costs for agents in the economy can be minimised by everyone quoting prices in the dominant medium of exchange. Financial stability policy in this world would involve ensuring the operational integrity of the payment systems and the appropriate level of systemic risk in the ‘custodian sector’ – as this is what the banking sector would effectively become. Whether it would make sense for the central bank to do this rather than another government body is, of course, an open question. But there is still the issue of what the acceptable settlement asset would be. Kiyotaki and Moore (2004) argue that payments must be made in terms of money rather than other assets because there is a commitment problem with repaying loans with returns from investment. Kocherlakota (2004), reviewing this paper, says that it fails to address why assets that could pay a higher return than money and that have no enforcement problems are not used: e.g. government debt.26 Wallace (1988) makes the same point by noting that the acceptability of central bank money today hinges on the fact that it is ‘legal tender’ and the fact that legal restrictions stop other banks from issuing interest-yielding bearer bonds that could act as ‘money’ while offering a higher return than central bank money. Selgin and White (2005) make the point that central bank money is likely to continue being preferred to other assets for two reasons: (a) because it defines the unit of account, it will not be subject to bid-ask spreads, and (b) payment in it is final given its ‘legal tender’ status. If central bank money remained, then central banks would remain at the top of the payments pyramid. This is the vision presented by Harry Leinonen in his chapter in this volume. He envisages a situation in which individuals can settle payments between each other in real time and in central bank money over a network without a central payments-processing infrastructure. For this to work, the central bank will need to provide liquidity in the form of ‘bytes of encrypted information’ (i.e. central bank e-money!) that it can distribute to banks over this same network. Once these have been distributed, the central bank would not need to keep a tab on them until the end of the day (or any other point at which they may charge interest or remunerate balances). On the other hand, if government bonds became ‘money’, then the top of the payments pyramid would instead be represented by a Central Securities Depository (CSD) that would be responsible for transferring securities from one account to another across its books. Essentially, this is exactly like a payment system run by a proto central bank, as we described in the first section of this chapter. In this case, the CSD would have become the central bank.

36

S. Millard and V. Saporta

As the rate of inflation would be determined by government borrowing – which would fix the supply of government bonds – monetary policy would become indistinguishable from fiscal policy. Guarding against systemic risk externalities would involve ensuring that the CSD were operationally robust and reliable. The rationale would be the same as for payment systems today: if the CSD were to go down, agents would have to find an alternative – and expensive – way of making payments such as transferring a valuable commodity or, at the very worst, resorting to barter. If financial assets could be transferred from hand to hand there would be no need for a settlement institution at the ‘apex’ of the pyramid. Again, monetary policy would involve ensuring that these financial assets were what agents said they were – just as goldsmiths and money changers assured the quality of money in the past – and ensuring that the supply of these financial assets did not grow quickly. Likewise, guarding against systemic risk externalities would involve ensuring that the process of transferring financial assets from hand to hand was operationally secure and that the custodian sector (which would presumably intermediate in these assets) was appropriately regulated. The likelihood and desirability of these scenarios is open to question. And even if market forces were to drive the financial system in one direction, governments could still enforce, to some degree, settlement in certain assets via certain institutions through legal restrictions.

Concluding remarks Based on an eclectic choice of raw material, this chapter has painted a broadbrush picture of the economic links between central banks and payments in the past, the present and the future. In particular, we argued that the core functions of monetary and financial stability of modern-day central banks can be traced back to their payments role as providers of the ultimate settlement asset. We then argued that central banks charged with the preservation of monetary and financial stability should have an interest in payment systems that process large values/volumes and/or payment types, disruptions to which can give rise to significant social welfare costs. We then evaluated different models of public intervention, concluding that an oversight model backed with appropriate enforcement powers might provide the best balance between central bank objectives and other broader public policy objectives. Whether such a model is chosen in practice will depend on a number of practical considerations, including the risk preferences and budget sets of public authorities and the weighing of social benefits versus the costs of introducing a statute-based regime; this, in turn, will depend on country-specific institutional arrangements for the regulation and oversight of the financial system at large. Finally, we ended with some tentative thoughts about how retail and wholesale payments may be made in the future and how this might affect central banks’ ability to conduct monetary policy and to guard against systemic risk.

Central banks and payment systems

37

Annex: International models of intervention in payment systems Tables 1.A1 to 1.A3 explore existing models for intervention in selected countries’ large-value payments systems, automated clearing houses (ACHs) and embedded payment systems of securities settlement systems. These are mapped to the first five corner models illustrated in Figure 1.1. (There are currently, in the countries considered, no existing examples of a laissez-faire (model 6) approach for these categories of payment system, so this model has been excluded. To re-cap, the models considered are: Model 1: own and operate Model 2: own only Model 3: operate only Model 4: oversee only Model 5: operate and oversee.

Table 1.A1 G10 models for intervention in large-value payment systems Country

Domestic LVPS(s)

Australia Belgium Canada France

HVCS ELLIPS LVTS TBF PNS RTGSplus BI-REL BOJ-NET TOP NICS SCP RIX SIC CHAPS* Fedwire CHIPS TARGET EURO1

Germany Italy Japan Netherlands Norway New Zealand Sweden Switzerland UK US ECB

Model 1

Model 2

✓

Model 5 ✓ ✓

✓

✓ ✓ ✓ ✓

✓

Model 4

✓

✓

✓

Model 3

✓ ✓

✓ ✓

✓ ✓

Note *It should be noted that, although operated and overseen by the Bank of England, CHAPS does not map perfectly to model 5 as oversight is conducted without formal powers of enforcement.

38

S. Millard and V. Saporta

Table 1.A2 G10 models for intervention in ACHs Country

Domestic ACH(s)

Australia

BECS CECS CEC ACSS SIT RPS BI-COMP* Zengin Interpay NICS Retail ISL BGC DTA LSV BACS EPN ACH

Belgium Canada France Germany Italy Japan Netherlands Norway New Zealand Sweden Switzerland UK US

Model 1

Model 2

Model 3

Model 4 ✓ ✓ ✓ ✓

✓ ✓

✓

✓

✓ ✓ ✓ ✓

Model 5

✓

✓

✓ ✓ ✓

Note *Operation of one of the two sub-systems settling across BI-COMP is carried out by SIA on behalf of the Bank of Italy

In the case of large-value systems, the majority of central banks own and operate (model 1). The most notable exceptions here are the United Kingdom, Australia, New Zealand and Belgium, where the infrastructure for the largevalue system is operated by the central bank, but the system is privately owned; and Switzerland, where the central bank owns but does not operate the system. There are some other examples of pure-oversight models for large-value systems, but these occur where private and public large-value systems co-exist (CHIPS in the United States; PNS in France; and Euro-1 in the Euro area). The pure oversight model is most prevalent in the case of ACHs. Only in the cases of the United States and Germany does the central bank own and operate an ACH (in the United States, this is in competition with a private sector provider, EPN), while in Italy the central bank owns the system but operates only part of the infrastructure. In Belgium, oversight is combined with a role in operating the infrastructure. Finally, in the case of securities settlement systems, most central banks have adopted either model 1, 4 or 5. As for its large-value system, Switzerland is an outlier here, having adopted model 2 (ownership with outsourced operation). The examples of model 1 tend to be where either a separate system exists for settlement of government securities (United States, Japan and Belgium) or where the cash leg of securities transactions occurs via a central bank owned and operated large-value system. Otherwise, most central banks have adopted an oversight model, with some also assuming some operational involvement, depending on whether an interfaced or integrated settlement model is applied.

US

Netherlands Norway New Zealand Switzerland UK

Fedwire Securities DTC

Austraclear NBB Clearing CIK CDSX VPC Euroclear France Clearstream Banking AG Monte Titoli BOJ-NET JGB Services JASDEC Necigef VPO Austraclear NZ SWX CREST

Australia Belgium

Canada Sweden France Germany Italy Japan

Domestic SSS(s)

Country

✓

✓

✓ ✓

✓

✓

Model 1

✓

Model 2

Model 3

✓

✓

✓

✓

Model 4

Interfaced model Settles through LVPS Settles through LVPS Integrated model Interfaced model Interfaced model Settles through LVPS Settles through LVPS Interfaced model Interfaced model

✓

✓ ✓

Cash transfers onto DTC accounts take place through Fedwire

Settles through LVPS A hybrid-integrated model, sitting between models 4 and 5. DvP occurs with finality in CREST across memo. accounts mirrored by cash movements in the Bank’s RTGS system

Interfaced model

✓

✓ ✓

Notes

Model 5

Table 1.A3 G10 models for intervention in the embedded payment systems of securities settlement systems

40

S. Millard and V. Saporta

Notes 1 The views expressed are those of the authors and do not necessarily reflect those of the Bank of England. The chapter has benefited from comments from Helen Allen, Andrew Bailey, Morten Bech, Paul Bedford, Alastair Clark, Charles Goodhart, Andrew Gracie, John Jackson, Nigel Jenkinson, Charles Kahn, Ana Lasaosa, Harry Leinonen, Chris Mann, Mark Manning, Adrian Penalver, David Rule, George Speight, Matthew Willison, Al Wilson and Jing Yang. A number of the arguments presented were refined following discussions with participants at the June 2006 Bank of England workshop on ‘payment economics’. Any errors are entirely our own. 2 Here we follow BIS (2001) and define as a ‘settlement institution’ the institution across whose books settlement takes place, or, put differently, the institution that supplies the asset which members of the system have agreed to accept for settlement of their obligations – the ‘settlement asset’. 3 For example, until 1826, the Bank was the only joint stock bank – other banks’ capital was constrained to the fortunes of a maximum of six partners. And early in the eighteenth century, forging Bank of England notes was punishable by death (Collins, 1988). 4 Cited in Goodhart (1988), page 35. The term ‘quasi-central banking agency’ in the quote includes both private banks – as in the Suffolk Bank example – and mutuallyowned clearing houses – as in the New York Clearing House example. 5 Even if tendencies that lead to the centralisation of reserves do not result in monopoly pricing, there might still be a rationale for government intervention in ‘superimposing’ a central bank. In particular, during financial crises, conflicts of interest might prevent commercially-oriented central banking agencies from behaving in a socially optimal manner. 6 There is a substantial literature that investigates the economic reasons why different firms adopt different internal management ‘hierarchies’ – see, for example, Rajan and Zingales (2001). It is possible that this literature can shed light on the development of different payment hierarchies. 7 The clearing arrangements developed by the Suffolk Bank and by the New York Clearing House in the nineteenth century in the United States fall into this category. Another example is provided by the Schulze-Delitsch savings banks in Germany which in the nineteenth Century sought to clear through Dresdner Bank which in turn developed into a proto central bank (Goodhart, 1988). 8 The Act of 1833, gave other banks the right to redeem their own liabilities with Bank of England notes rather than gold – that is, Bank of England notes became legal tender. The impact of the Act was far from startling, however, because Bank of England notes had been de facto (rather than de jure) legal tender for a long time (Collins, 1988). 9 At times when liquidity was particularly tight (that is, when gold had been leaving the system), the New York Clearing House (CH) issued liabilities backed by commercial bank notes that banks could use to settle their multilateral net obligations. It issued them to member banks against holdings of the notes of other banks; so, when a bank received another bank’s notes as a deposit, it could post these with the CH in exchange for a clearing house loan certificate. On such occasions, the CH would operate as a proto central bank in a fractional reserve system – that is, it would supply a settlement asset (the clearing house loan securities) that was not fully supported by either a commodity with intrinsic value or a government guarantee (Timberlake, 1984). 10 The Irish banking strike of 1966 is a good example where agents were unable to use bank accounts to make payments and were unable to withdraw cash. In this case, money was completely unable to fulfil its role as a medium of exchange at a time when inflation was low. See Murphy, 1978.

Central banks and payment systems

41

11 For a broader definition of financial stability see Haldane et al. (2004). 12 For example, an ‘informational externality’ can be based on imprecise information about fundamentals – e.g. all affected systems depend on the same telecoms infrastructure which agents believe with some probability to have been the source of the disruption of the original system – or conceivably be purely ‘sunspot’ driven – that is, agents stop using another payment system based on the belief that others will stop using it and their beliefs turn out to be self-fulfilling) 13 As an aside, it is worth noting that for ‘balance sheet’ control reasons, there is also a clear need for a central bank to be interested in the system for getting ‘cash’ out into the economy (e.g. the Note Circulation Scheme in the United Kingdom) as well as any payment system whose workings lead to the injection/withdrawal of central bank money into/out of the economy (e.g. in the United Kingdom, the large value payment of CHAPS and the security settlement system of CREST). 14 BoNY and JP Morgan Chase – the so-called ‘clearing banks’ – settle across their books over 75 per cent of transfer volume in US government securities, placing them at two critical nodes in the network of inter-bank payment flows (McAndrews and Potter, 2002). Following the attacks on 11 September, BoNY became a liquidity sink as payments were made to it which it was temporarily unable to recycle by making pay-outs. At some point BoNY was reported to be overdue of $100 billion in payments and unable to make payments or lend funds. Failures to deliver US government securities rose from $1.7 billion per day the week of 5 September to $190 billion the week ending 19 September. Absent massive intervention by the Federal Reserve, money and securities markets are likely to have been severely disrupted (Lacker, 2004). 15 It is possible of course that the central bank is formally charged with guarding against both systemic risk and competition-related externalities. This is the case, for example, in Australia. International comparisons, however, reveal that the Reserve Bank of Australia is the exception rather than the rule in this respect. See Bank of England, 2005: Table 1. 16 The classic reference on collective action problem in a mutual is Hart and Moore (1995). 17 This suggestion was put forward by Cruickshank (2000). 18 By ‘quality’ we mean features such as robust anti-fraud security, speedy clearing cycles and ease of use. 19 See www.oft.gov.uk/News/Press+releases/2005/94–05.htm for the press announcement by the OFT. Further information on how this position was reached can be found by following the links on this page. There are numerous other examples. In the late 1970s, for example, the Securities Exchange Commission (SEC) pursued a similar strategy in catalysing the establishment of DTC (the user-owned company that settles equities and corporate bonds in the US) and the European Commission is currently pursuing a similar approach for catalysing the creation of a Single Euro Payments Area (SEPA) in retail payments. 20 This is effectively what the Bank of England did when it built CREST, the central securities depository and large-value securities settlement system in the United Kingdom, in the mid-1990s, following a failed private sector initiative to build an electronic settlement service for equities. And it is, in effect, what the Eurosystem is suggesting in its recent proposal to build a central-bank owned and operated securities settlement system for Euro-denominated securities (TARGET2-Securities). See the press release issued by the European Central Bank on 7 July 2006 (www.ecb.int/ press/pr/activities/paym/html/index.en.html). 21 This is the current arrangement in the United Kingdom. A publicly available memorandum of understanding between Her Majesty’s Treasury, the Bank of England and the Financial Services Authority allocates responsibility for oversight of payment systems that are systemically significant to the United Kingdom to the Bank of

42

22

23 24 25

26

S. Millard and V. Saporta England. That said, the Bank of England appears to be an outlier among central banks, internationally, in not having such an arrangement backed by statutory powers. See Bank of England, 2005: Table 1. In countries with fragmented banking systems, one of the purposes of central banking institutions when they were founded was to solve the coordination problems that prevented the establishment of a single inter-bank payments network (see Norman et al., 2006). The large fixed costs involved in setting up payment systems from scratch explains why, once a reasonably efficient model of ownership and operation is established (regardless of whether it is central bank owned or privately owned), shifts to different models are observed infrequently. This is the path-dependence element of the historical explanation. Kahn and Roberds (2005) analyse this issue in the context of a model of a simple economy with credit arrangements. Of course, cash can be lost or stolen; indeed, the papers by He et al. (2005) and Millard and Willison (2006) suggest this possibility as a reason why payment systems developed in the first place. Capie et al. (2003) interpret King’s (1999) vision as one of electronic barter where goods are exchanged for goods in real time using electronic means. They argue that the transaction and information costs that led to the development of money in the first place cannot ever be removed by technology; hence, there will always be the need for (central bank) money. Of course, if another financial asset became an ‘acceptable’ asset in the sense we go on to describe, such an asset would realise the advantages of money described in their paper; we would be back at the conclusion that central bank money would vanish. Of course, there is nothing to stop the central bank paying interest on its money, at which point the distinction between central bank money and government debt essentially vanishes and fiscal and monetary policy would be, literally, the same thing.

References Bagehot, W. (1873) Lombard Street: A description of the money market, London: Henry S. King. Bank for International Settlements (BIS) (2001) Core principals for systemically important payment systems, Committee on payment and settlement systems. Bank of England (2005) Payment System Oversight Report 2004, January. Bolt, W. and Humphrey, D. (2005) ‘Public good issues in TARGET: natural monopoly, scale economies, network effects and cost allocation’, European Central Bank Working Paper No. 505. Buiter, W.H. (2005) ‘New developments in monetary economics: two ghosts, two eccentricities, a fallacy, a mirage and a mythos’, Economic Journal, Vol. 115, pages C1–31. Calomiris, C.W. and Kahn, C.M. (1996) ‘The efficiency of self-regulated payments systems: learning from the Suffolk system’, Journal of Money, Credit and Banking, Vol. 28, No. 4, pages 766–97. Capie, F.H., Tsomocos, D.P. and Wood, G.E. (2003) ‘E-barter versus fiat money: will central banks survive?’, Bank of England Working Paper No. 197. Collins, M. (1988) Money and banking in the UK: A history, London: Croom Helm. Cruickshank, D. (2000) Competition in UK banking: A report to the Chancellor of the Exchequer, Norwich: Her Majesty’s Stationary Office. Dowd, K. (1998) ‘Monetary policy in the 21st century: an impossible task?’, CATO Journal, Vol. 17, pages 327–31. Drehmann, M., Goodhart, C.A.E. and Krueger, M. (2002) ‘The challenges facing foreign

Central banks and payment systems

43

currency usage: will traditional transactions medium be able to resist competition from new technologies?’, Economic Policy, Vol. 17, pages 193–227. Friedman, B.M. (1999) ‘The future of monetary policy: the central bank as an army with only a signal corps?’, International Finance, Vol. 2, No. 3, pages 321–38. Goodhart, C.A.E. (1988) The evolution of central banks, Cambridge, Massachusetts: MIT Press. Goodhart, C.A.E. (2004) ‘Financial supervision from an historical perspective’, Paper given to Bank of Finland Conference on ‘The structure of financial regulation’. Green, E.J. and Todd, R. (2001) ‘Thoughts on the Fed’s role in the payment system’, Federal Reserve of Minneapolis Quarterly Review, Winter, pages 12–27. Greenfield, R.L. and Yeager, L.B. (1993) ‘A laissez-faire approach to monetary stability’, in White, L (ed.), Free banking. Volume 3: Modern theory and policy, Aldershot: Elgar. Haldane, A., Hall, S., Saporta, V. and Tanaka, M. (2004) ‘Financial stability and macroeconomic models’, Bank of England Financial Stability Review, June, pages 80–8. Hart, O. and Moore, J. (1995) ‘The Governance of Exchanges: members’ co-operatives versus outside ownership’, LSE Financial Markets Group Discussion Paper No. 229. He, P., Huang, L. and Wright, R. (2005) ‘Money and banking in search equilibrium’, International Economic Review, Vol. 46, pages 637–70. Jevons, W.S. (1875) Money and the mechanisms of exchange, New York: D. Appleton and Company. Kahn, C.M. and Roberds, W. (2005) ‘Identity, identity theft and credit’, University of Illinois, mimeo. King, M.A. (1999) ‘Challenges for monetary policy: old and new’, Bank of England Quarterly Bulletin, November, pages 428–38. Kiyotaki, N. and Moore, J. (2004) ‘Liquidity and asset prices’, London School of Economics, mimeo. Kiyotaki, N. and Wright, R. (1989) ‘On money as a medium of exchange’, Journal of Political Economy, Vol. 97, No. 4, pages 927–54. Kiyotaki, N. and Wright, R. (1993) ‘A search-theoretic approach to monetary economics’, American Economic Review, Vol. 83, No. 1, pages 63–77. Kocherlakota, N.R. (2004) ‘Optimal monetary policy: what we know and what we don’t know’, University of Minnesota, mimeo. Kohn, M. (1999) ‘Early deposit banking’, Dartmouth University Working Paper No. 99–03. Lacker, J. (2004) ‘Payment system disruptions and the federal reserve, following September 11, 2001’, Journal of Monetary Economics, Vol. 51, pages 935–65. McAndrews, J. and Potter, S. (2002) ‘Liquidity effects of the events of September 11, 2001’, Federal Reserve Bank of New York, Economic Policy Review, Vol. 8, No. 2, pages 59–79. Millard, S.P. and Willison, M. (2006) ‘The welfare benefits of stable and efficient payment systems’, Bank of England Working Paper No. 301. Murphy, A.E. (1978) ‘Money in an economy without banks: the case of Ireland’, Manchester School of Economic and Social Studies, Vol. 46, No. 1, pages 41–50. Norman, B., Shaw, R. and Speight, G. (2006) ‘The history of interbank settlement arrangements: exploring central banks’ role in the payment system’, Bank of England, mimeo. Rajan, R. and Zingales, L. (2001) ‘The firm as a dedicated hierarchy: a theory of the origins and growth of firms’, Quarterly Journal of Economics, Vol. 116, pages 805–51.

44

S. Millard and V. Saporta

Selgin, G.A. and White, L.H. (1994) ‘How would the invisible hand handle money?’, Journal of Economic Literature, Vol. 32, No. 4, pages 1718–49. Selgin, G.A. and White, L.H. (2005) ‘The future of fiat money: a Mengerian perspective’, University of Georgia, mimeo. Smith, V. (1936) The rationale for central banking, London: P&S King Ltd. Thornton, H. (1802) An enquiry into the nature and effects of paper credit of Great Britain, London: Hatchard. Timberlake Jr, R.H. (1984) ‘The central banking role of clearinghouse associations’, Journal of Money, Credit and Banking, Vol. 16, No. 1, pages 1–15. Trivoli, G. (1979) The Suffolk Bank: a study of a free-enterprise clearing system, London: Adam Smith Institute. Wallace, N. (1988) ‘A suggestion for oversimplifying the theory of money’, Economic Journal, Vol. 390, pages 25–36. Woodford, M. (2004) Interest and prices: foundation of a theory of monetary policy, Princeton, NJ: Princeton University Press.

2

The role of a central bank in payment systems Edward J. Green

Introduction The central banks of all industrialised countries specialise to some extent in what roles they play in their respective payment systems. Each defers to other entities in its respective economy, both private and public, to assume roles that it elects not to fill. Nevertheless, there is considerable variation across industrialised economies in how broad or focused a role the central bank assumes. Does the body of economic learning about central banking and payment systems have anything to say about what is the preferred point along this spectrum, or about the range of acceptable points? I suggest here that this learning does have an implication. Specifically, the central bank is a specialised organisation that is uniquely able to offer free, short-term credit on illiquid collateral to other financial intermediaries. It provides this service via a set of settlement accounts on its books for those intermediaries. A central bank does well, and arguably does best, by specialising in providing these services while leaving other roles (including clearing of retail and commercial transactions, transmission of payment messages and commercial regulation of payment intermediaries) to different entities that can specialise in those tasks.

Commencing the argument Let me commence this argument by defining some terms, and also by indicating briefly how some of its premises are justified. A payment system is a comprehensive system for settling the obligations of purchasers of goods, services and financial assets, and of their intermediaries. Examples are large-value payment systems that settle obligations between banks, and payment-card networks through which many consumer transactions are settled. Sometimes an individual component of such a comprehensive system is also called a payment system, but that is not relevant here. Rather, the discussion will concern systems in which numerous entities, including providers of such specialised components, interact to provide an overall service. The institutional arrangement for meshing these various contributions functions essentially as a market. As in other markets, comparative advantage determines which

46

E.J. Green

entity should best do what. So the topic of this chapter might be restated as: What is the comparative advantage of a central bank with respect to the other participants in a payment system? Define a central bank to be an institution that • •

•

has both the government and private financial intermediaries (that will generically be called ‘banks’) as account holders; is therefore in a position to influence overall interbank credit market conditions through its credit policies towards account-holding banks and its intermediation on behalf of the government; and has been given lead public policy responsibility for achieving credit market conditions that foster prosperity and economic stability, and price stability in particular.

In view of the position and responsibility just described, the central bank is in a privileged position to • • •

manage a system of accounts for interbank settlement; provide short-term credit to facilitate settlements; and accept illiquid collateral for that credit.

These are the three activities in which a central bank should specialise. The argument for this proposal rests on three premises. The first premise is that it is economically efficient for financial intermediaries to have access to credit at a cost (either a direct interest price or an indirect collateral cost) just high enough to balance the small risk that the central bank assumes by extending it. This assertion is supported by a model of payment system credit due to Freeman (1996a, 1996b). Freeman derived the optimality of such essentially free credit along with a zero inflation-rate policy, and Zhou (2000) has further shown in that framework that free credit is conditionally optimal even when inflation is positive. I provide an informal discussion of Freeman’s model and its logic, and of why I consider the model to be a convincing economic theory, in a separate chapter (Green, 2006) elsewhere in this volume. The second premise is that a central bank is uniquely able to make short-term loans on illiquid collateral. A collateral asset is illiquid if it will possibly take time considerably beyond the maturity date of the loan to sell the asset for its full value, although the full value (including a market rate of appreciation) can ultimately be obtained. A private-sector intermediary that lends subject to a balance-sheet constraint cannot wait a long time to recoup the value of loans in default, so it cannot afford to take illiquid assets for collateral. In contrast, a central bank in a fiat-money regime issues credit by creating outside money. Suppose that the only constraint on such money creation, imposed by the requirement of price stability, is that the money so emitted must be reabsorbed eventually by sale of the collateral, or through repayment of the loan. This constraint does not rule out acceptance of illiquid assets as collateral. This premise

The role of a central bank in payment system

47

can be derived formally in a suitably parameterised overlapping-generations model of money, and is a consequence more generally of formal or informal theories in which inflation is a function of the long-run rate of money growth. It is not a consequence of a ‘fiscal theory’ of the price level, such as the theory of Woodford (2003) discussed in Green (2006), nor of a generic overlappinggenerations model in which the price level is indeterminate (cf. Brock and Scheinkman, 1980.) However, neither of those models (or any other model of which I am aware) implies that a central bank cannot accept illiquid collateral or that doing so would make it more difficult or less likely than otherwise to achieve price stability. The third premise is that not only does a central bank have a comparative advantage relative to most other organisations in supplying credit to banks, but there are likely to be some diseconomies of scope when a central bank attempts to play other roles as well. Because of those diseconomies, at least prima facie, a central bank has a comparative disadvantage at playing such other roles. Richard Todd and I have argued in some detail for this premise in a recent paper (Green and Todd, 2001) concerning the role of the US Federal Reserve System in consumer and commercial payments.

Must a monetary authority be involved with payment systems? The topic of this chapter has been framed as a question about the industrial organisation of the payment system: What should a central bank do, or refrain from doing, in order to enhance the performance of the market for payment services? However, a central bank is the natural candidate to play the macroeconomic roles of issuing base money and regulating the price level. If performing those macroeconomic functions is a central bank’s primary role, then perhaps the question framed above is of secondary importance. The primary question should be, what level or kind of involvement with payment systems must the central bank maintain, in order to function effectively as a monetary authority? Let me briefly examine this question, before returning to issues of industrial organisation. Stipulate, for a moment, that a central bank must issue money in order to be an effective monetary authority. There is a long-standing issue in monetary economics, regarding whether a central bank needs to monopolise the provision of money. To a large extent, arguments for a central bank monopoly have assumed that money is being used to exact a seignorage tax. If issuing money is profitable for the government, the argument goes, then it is profitable for everyone, so there will be a glut of money in the absence of monopoly. Today, though, seignorage is recognised to be an inefficient tax. It is universally thought that, except perhaps during a temporary public-finance emergency, a central bank can better contribute to public welfare by avoiding inflation than by raising seignorage that results in inflation. Issuing money according to a low-inflation policy is not a profitable activity, so a glut of money from private issuance need not be

48

E.J. Green

feared. That is, the traditional argument for making the central bank a monopolist in money issuance is moot, according to current understanding of how monetary policy should be conducted. There are other, more recent arguments regarding the potential benefits and risks of permitting private-sector intermediaries to issue money. Research by Cavalcanti and Wallace (1999) and Wallace (2005) suggests a potential benefit, essentially that private issuers can target more precisely than the central bank the recipients of the money that is created. Research by Smith (1988) suggests a potential cost – exposure to the possibility of sunspot equilibrium. These are representative of economists’ arguments about this topic. It is noteworthy that the arguments on all sides of the topic assume that central bank money and private monies are perfect substitutes. This assumption contrasts with the discussion among central bankers and bank supervisors, regarding whether or not central bank money is an inherently superior settlement asset. (For example, one of the CPSS ‘Core Principles for Systemically Important Payment Systems’ (2001) seems to suppose that central bank money is superior. Some scepticism towards that position is expressed in a comment letter on the draft core principles that was submitted to the CPSS by the Federal Reserve Bank of Chicago (2000). This contrast evidences a more fundamental distinction. The policymakers’ discussion focuses on whether or not some degree of central bank monopoly in money issuance would improve the stability of payment systems, while the economists’ discussion focuses on whether or not monopoly would improve the stability of the monetary system and of the real economy. It is the economist’s discussion that is relevant to my preliminary question here. Lately there has been some discussion of whether or not, in principle, a central bank really does need to issue money, or indeed will be able to issue money in a future environment where advances in payment technology have taken full effect. One view, expressed informally by King (1999), is that money might become difficult to issue, and that such difficulty might significantly impair the effectiveness of monetary policy. An alternative view, taken by Woodford (2003) and others, is that the effectiveness of monetary policy is based on the central bank’s ability to set a nominal interest rate, and that such rate setting can be accomplished with zero net issuance of money in equilibrium. This ‘fiscal theory’ view implies that a central bank can be an effective monetary authority without being involved in the payment system in any way whatsoever. The considerations just discussed have to do with the possibility that some form of involvement with payment systems may intrinsically benefit the conduct of monetary policy or the pursuit of financial stability. One also hears arguments, at least in the United States, that involvement with payment systems can provide some extrinsic benefit to the central bank. One such argument is that information obtained from involvement with payment systems can profitably be factored into monetary policy decision making. There is scant evidence for that supposition, though. The minutes of the FOMC, the Fed’s policy-making committee, have seldom, if ever, mentioned such information. Another argument,

The role of a central bank in payment system

49

usually privately expressed, is that involvement with payment systems garners valuable political support for the central bank. Setting aside the serious question about whether or not it is proper for (or in long-term interest of) the central bank to engage in interest group politics, consider whether or not involvement with payment systems is politically effective in the short run. In the US situation, this claim is superficially plausible because the US banking system includes thousands of small ‘community banks’, scattered throughout every Congressional district. The owners of these banks are a cohesive, national, political lobby. However, they mobilise to support Fed policies that they perceive to give them subsidy or regulatory advantage over their competitors. They are not known for supporting the Fed when it is necessary to raise interest rates, which is the occasion when political support might be helpful to monetary policy. In summary, the argument that involvement with payment systems confers extrinsic, informational or political advantages for monetary policy making is not convincing. In general, the idea that involvement with payment systems confers any extrinsic advantage on monetary policy making has no support, in the context either of the US or other countries. Economists differ among themselves about whether or not some form of involvement in payment systems confers an intrinsic monetary-policy advantage, but the intrinsically beneficial involvement, if any, does not extend beyond the three roles that I identified in the introduction. Even if a central bank becomes a better monetary authority on account of performing these functions, there is not a consensus among economists that it must have a monopoly position. In particular, there is no consensus among economists that private clearinghouses should be prevented or discouraged from playing the same settlement roles as a central bank plays.

The market structure of payments Now I return to the main investigation, regarding the benefit that some form of central bank involvement may confer on payment systems. This discussion draws heavily on Green and Todd (2001), which pertains specifically to the Federal Reserve’s involvement in US consumer and commercial payment systems. Recall that a payment system can be viewed as a market in which various types of entity provide complementary services to payors and payees, as well as intermediate-good services to the payment intermediaries with which those end-users deal directly. Which services should be provided by which form of entities involves questions of market structure and comparative advantage. Regarding market structure, on the production side, many payment technologies have high fixed costs and exhibit declining average costs throughout a very wide quantity range. In addition there are economies of scope, such as the opportunity for several payment systems to share a common informationtransmission infrastructure. On the demand side, there are ‘network externalities’ – the preference of payors and payees to co-ordinate with one another on which payment system they use. These features distinguish markets for payment services from the neoclassical model of a competitive market. However, the

50

E.J. Green

assumptions of the neoclassical model constitute sufficient – not necessary – conditions for competitive equilibrium to exist and to be efficient. Several decades of experience with ‘deregulation’ largely corroborate the theory of contestable markets, according to which fairly nonintrusive regulation supports efficient outcomes in industries having economies of scale and scope on the production side. Winston (1993) estimates that ‘deregulation’ has achieved a 7–9 per cent improvement in contribution to GNP from the US sectors affected by regulatory reform, without significant distributional side-effects. With respect to network externalities, the situation is less clear-cut. One analysis, based on decentralised, myopic decision making by consumers, suggests that this demand side phenomenon may lead to inefficient outcomes. A different analysis, based on a cooperative solution concept for markets with network externalities, predicts efficient outcomes. (Weinberg (1997) discusses these two theories.) The conflict between these two theories was essentially the crux of the contentious Microsoft case in competition law. The preceding paragraph has dealt with competition among payment system operators. Another issue is strategic interaction among participants in payment systems. Bech and Garrett (2003) formalise one theory. They study strategic interaction within a single payment date, and find that ‘gridlock’ can be an inefficient outcome of interaction. Each bank desires to be in a credit position versus the payment system as much of the time as possible, but this is a zero-sum game among the participants. Participants non-cooperatively delay their payments in order to retain their clearing balances for as long as possible, so in equilibrium payments are not made until the end of the settlement day. In Bech and Garrett’s model, there is a public cost attached to this, and thus an economic inefficiency. McAndrews and Rajan (2000) document an intraday pattern of Fedwire payments that they attribute to strategic interaction. They emphasise the very high level of payment activity in the late afternoon, close to the time when the CHIPS payment system settles over Fedwire. They suspect that even payment orders received in the morning are often not executed until this time. This phenomenon plausibly reflects banks’ desire to avoid being in a debit position vis-à-vis Fedwire, a motivation that is easy to understand because the Fed charges interest on intraday credit. Many Fedwire payments are not time sensitive for the transactors, as long as they are settled by the end of the day of submission to the payment system. Thus, while banks seem to be playing the zero-sum game modelled by Bech and Garrett, the feature of their model that early settlement is a public good does not plausibly hold for Fedwire. Of course, there is a public good problem if time sensitive payments are being delayed, but McAndrews and Rajan do not provide evidence of that. If a payor informs its bank that a payment is time sensitive, then the bank’s cost from alienating that customer by holding back the payment would credibly exceed the gain from minimising interest on an intraday debt by doing so. Moreover, the private gain to a bank from delaying a payment is certainly exacerbated by, and may be nothing but an artifact of, the Fed’s pricing of intraday credit. If it is efficient to offer free intraday credit, as Freeman (1996a) and Zhou (2000) argue, then any inefficiency in the timing of

The role of a central bank in payment system

51

Fedwire transactions is arguably a result of the Fed’s daylight credit policy, rather than being a market failure requiring pricing of daylight credit to remedy.

The comparative advantage of a central bank I explained earlier that a payment system is, in effect, a market where various specialised entities collaborate to produce services. These entities notably include commercial banks, correspondent banks, clearinghouses and the firms that provide technology infrastructure, as well as central banks. The division of labour among these entities is determined by comparative advantage, and by economies of scope internal to each of the entities. The general presumption in economics as a whole is that the operation of these forces produces efficient outcomes. The gist of my preceding discussion of market structure is that this presumption applies to payment systems in particular. The comparative advantage of the central bank consists in maintaining deposit accounts for banks and in providing short-term credit to, and effecting transfers of balances among, those accounts as a means of settling interbank obligations. This characterisation reflects both economic history and economic theory. Historically, central banks have been chartered to perform two functions. One is to be an intermediary between the government and its lenders, enabling the government to obtain credit by ensuring that implicit default through inflation will occur only in genuine national emergencies. The other is to serve broad public interest as the trustworthy and neutral apex of a hierarchy of banks that, in turn, provide the nonbank public with accounts used to settle financial business and personal payments by transfer balances. Indeed, there is an economy of scope between these functions that gives the central bank comparative advantage in performing the latter. Since almost all banks need to transfer funds from their customers to the government to pay taxes, the government’s bank is in a natural position to serve as apex of the correspondent-banking hierarchy. This role as apex puts the central bank in a unique and distinguished position in the payments business. Its role with respect to banks is closely analogous to the role the banks play with respect to their nonbank customers – including netting, extension of credit and concomitant monitoring of creditworthiness. Moreover, just as banks are often structured to avoid conflicts of interest with their own nonbank customers, central banks evolved in part to avoid conflicts of interest with banks. Market demand rose for a special-purpose intermediary (that is, one that does not do business with nonbank customers) that could play the roles just mentioned without the incentive conflicts to which a commercial bank serving its peers would be exposed. The most immediate incentive conflict – the temptation to steal customers’ profitable nonbank customers – is ruled out (except in emergency conditions) explicitly by the central bank charter, which prohibits lending to nonbank entities under normal circumstances. Nonprofit status and other features of governance are designed further to control potential conflict of interest that might arise through the central bank’s discharge of its payment system functions.

52

E.J. Green

This historically-oriented description of the role of the central bank in the payment system is consonant with the body of recent economic theory regarding central banks described earlier. Together, history and theory suggest that there are two payment system functions that a central bank is better able than other institutions (except, perhaps, a clearinghouse) to perform for banks. First, the central bank can manage, in the broad public interest, a system of accounts that all banks are eligible to own, and that the banks can use to settle interbank transactions. Second, by extending credit to banks, the central bank can provide the benefits of netting and immediate finality of payments. Its ability to perform these functions, and particularly its position of neutrality and trust among the public and the institutions that it serves, is the unique strength of the central bank in the payment system. From this finding, together with the general principle that the public is best served when each institution in the economy focuses its resources in its area of unique strength. I conclude that central banks should restrict their involvement with payment systems principally to the accountmanagement and lending roles just discussed.

Do economies of scope justify broader involvement? The restrictive conclusion that has just been stated might be relaxed to some extent by taking the possibility of economies of scope into account. An economy of scope is a situation in which an entity is able to perform one task more efficiently than otherwise, because it is also performing a related task. Thus, while the central bank might not have a comparative advantage in performing some task on a stand-alone basis, the relationship of the task to account management and lending activities might give the central bank an advantage over other entities to perform it. The aftermath of the terrorist attack in the United States that took place in 2001 was a notable instance in which officials asserted that an economy of scope had been demonstrated between operation of a payment system by a country’s central bank and the core financial stability objective of that central bank. Among its effects, the attack impaired the interbank collection of cheques for the better part of a week. During that interval, the Federal Reserve agreed to accept legal possession of cheques that had not been transferred physically from its customer banks, and to give those banks immediate credit for the value of the cheques. This credit for cheques in the process of collection was a significant source of central bank credit beyond what was given overtly through the discount window. Numerous Federal Reserve officials characterised this extension of credit, during a crisis in which financial stability was at risk, as something that the Fed would not have been able to accomplish if it had not been the operator of a retail payment system. In fact, the extension of credit exemplified lending on illiquid collateral (in this case, cheques in the process of collection), which is one of the activities in which a central bank possesses a comparative advantage over other institutions. The Fed allows borrowers from the discount window to retain custody of collateral assets under appropriate safeguards, so

The role of a central bank in payment system

53

the fact that customer banks’ cheques could not be brought to Federal Reserve premises did not automatically disqualify them as collateral assets for a discount loan. While the fine print of the Fed’s discount regulations might have disqualified their acceptance, an ad hoc arrangement to accept them during the crisis could have been made very easily and immediately. Moreover, it seems likely that cheques in the process of collection were not contemplated as discount collateral precisely because they already serve routinely as collateral for credit extended in the course of the Fed’s cheque collection business. This oversight would surely be corrected in the course of an exit by the Fed from that business. In conclusion, because the Federal Reserve’s operation of a retail payment system does not enable it to lend on illiquid collateral that could not otherwise be accepted, the perceived economy of scope between operating a payment system and responding to a financial-stability crisis is illusory. More generally, in practice there are almost always pros and cons associated with prima facie economies of scope between central bank functions. For example, historically account management and lending to banks were argued to have had an economy of scope with bank supervision, but many countries have established separate supervisory authorities in recognition that any economy of scope is outweighed by the costs of giving that responsibility to the central bank. The careful econometric research required to substantiate that an economy of scope exists has seldom, if ever, been conducted for payment systems. One claimed economy of scope, between maintaining a system of settlement accounts and operating the ‘backbone’ system for making large-value transfers between those accounts, has sufficient intuitive plausibility to be widely accepted in the absence of such research. Despite this exception, it would be difficult to argue that those central banks that have taken on diffuse missions in their country’s payment systems on the basis of such arguments are actually serving the public well. One role for central banks that is conspicuously missing from the very short list enumerated here is regulation. There are at least three types of regulation to which payment systems might be made subject: prudential (or ‘safety and soundness’) regulation, regulation of competitive conduct, and consumer protection regulation. Central banks seem to have no comparative advantage in regulation of competition or consumer protection. Nevertheless, there is a recent tendency (exemplified in CPSS advice) to take on these responsibilities. This seems to me particularly ill-advised in the case of competition regulation. In terms of both cost structure on the producer side and the presence of network externalities on the demand side, payment systems closely resemble the telecommunications industry, on which national competition authorities expand very significant resources to maintain expertise that bears little or no relation to central banking. Given that another authority possesses such expertise, and that the central bank faces a dilemma of either regulating incompetently or else duplicating public investment in expertise, the wisest course would be to let the competition authority regulate payment systems in that regard. Prudential regulation is the type of regulation that people intuitively feel most

54

E.J. Green

strongly should be in the province of the central bank. There is a view that is widely held, even in countries where a specialised agency regulates and examines individual banks, that the central bank should regulate payment systems. This view is recommended on the grounds that the payment system is closely connected with the ‘systemic contagion’ of bank failures and other financial shocks, and the authority to regulate this transmission channel ought to go along with the central bank’s responsibility to deal with widespread financial troubles when they occur. That rationale is not an assertion that there is an economy of scope in the precise sense of economic theory, however. If the payment system is regulated by a separate agency, and the central bank believes that new regulations or a different regulatory stance would be conducive to financial stability, then the central bank can make its case to the agency and can use its political capital, if necessary, to exert pressure for change. A consideration against having the central bank regulate payment systems directly, on the other hand, is that the arrangement puts one payment system participant in the position of regulating others. There is a potential for conflict of interest. This potential is illustrated well by the preference expressed by the CPSS (2001) for the use of central bank money rather than inside money as a settlement asset. If we consider this issue in light of Freeman’s model of payments, a good settlement asset is one that can be emitted and reabsorbed (in a manner analogous to open market operations) by the entity with which the settlement accounts are held. The model depicts a central bank and a clearinghouse as being equally capable in this regard. Indeed, in the nineteenth century, US clearinghouses did emit and reabsorb inside settlement assets that they used for crisis management. In the context of this theoretical conclusion and of economic history, a regulatory presumption by central banks against creation by a clearinghouse of an inside settlement asset has the outward appearance of abuse of regulatory authority to disadvantage competition with the central bank from clearinghouses. If there is a cogent and creditable reason to discourage use of inside settlement assets, then a separate agency might be in a better position than the central bank to solicit public support for such a policy.

Conclusion A central bank does several things that are of immense value to the payment system, and it is better for the central bank to focus resources on doing those crucial things excellently than to dissipate resources by taking on additional tasks. Such a policy of deliberate focus is a difficult one for any organisation to adopt and maintain. Indeed, many central bankers probably believe that it would be difficult in practice to adopt such a policy. There are too many demands from the banking industry, the legislature, and various other stakeholders to embark on additional tasks, and those demands cannot be ignored. To rebut those sceptical views, I would point out that several OECD central banks actually do approach the high degree of focus recommended here. The closest of all is the Bank of Canada. In the early 1990s, the Bank and the Canadian banking indus-

The role of a central bank in payment system

55

try embarked on an initiative to consider deliberately what kind of payment system would serve Canada well. As this initiative progressed, the Bank of Canada concurrently revised its monetary-policy operating procedures to take full advantage of the improved payment environment. The new regime went into effect in 1999. Currently the Bank of Canada plays the three roles (maintaining settlement accounts, providing short-term credit and accepting illiquid collateral) enumerated at the beginning of this chapter. The Bank neither owns nor operates either Canada’s Large Value Transfer System (LVTS) or other retail components of the Canadian payment system. The Bank does serve as the final guarantor of LVTS obligations, a role that is considered not to create a material problem of moral hazard because the LVTS is well capitalised. The Bank regulates several payment systems that are deemed to be systemically important, but shares regulatory authority with Canada’s Department of Finance. One might argue that Canada is more fortunately situated than some other countries with respect to its ability to adopt such a highly focused stance. However, anyone making this argument should bear the burden of proof for it, and that burden is especially heavy because several other, highly regarded and successful, central banks have also chosen payment system roles that approximate the Bank of Canada’s focused role. In summary, a central bank can best contribute to the payment system by maintaining a system of accounts for interbank settlement, denominated in a money of stable value, and by offering short-term credit on illiquid collateral to the holders of those accounts. These services are of immense value to the payment system, and a central bank has a comparative advantage over most other institutions in providing them. In a developed economy where other institutions are able to provide services complementary to these, the social consequence of broader central bank involvement in the payment system is likely to be a marginal gain, at best.

References Bech, M.L. and Garrett, R. (2003) ‘The intraday liquidity management game’, Journal of Economic Theory, 109, pages 198–209. Cavalcanti, R. and Wallace, N. (1999) ‘Inside and outside money as alternative media of exchange’, Journal of Money, Credit and Banking, 31, pages 443–57. Committee on Payment and Settlement Systems (CPSS) (2001) ‘Core principles for systemically important payment systems’, Bank for International Settlements. Federal Reserve Bank of Chicago. (2000) ‘Comments regarding implementation of the core principles for systemically important payment systems’, www.chicagofed. org/bankwide_public_policy/files/\FinalStaffComCPVI09080.pdf (accessed 26 January 2006). Freeman, S. (1996a) ‘The payments system, liquidity, and rediscounting’, American Economic Review, 86, pages 1126–38. Freeman, S. (1996b) ‘Clearinghouse banks and banknote over-issue’, Journal of Monetary Economics, 38, pages 101–15. Green, E.J. (2006) ‘Some challenges for research in payments’, this volume.

56

E.J. Green

Green, E.J. and Todd, R.M. (2001) ‘Thoughts on the Fed’s role in the payment system’, Federal Reserve Bank of Minneapolis Quarterly Review, pages 12–27. King, M.A. (1999) ‘Challenges for monetary policy: new and old’, Bank of England Quarterly Bulletin, 39, pages 397–415. McAndrews, J. and Rajan, S. (2000) ‘The timing and funding of Fedwire funds transfers’, Federal Reserve Bank of New York Economic Policy Review, pages 17–32. Smith, B. (1988) ‘Legal restrictions, “sunspots” and Peel’s Bank Act: the real bills doctrine versus the quantity theory reconsidered’, Journal of Political Economy, 96, pages 3–19. Wallace, N. (2005) ‘From private banking to central banking: ingredients of a welfare analysis’, International Economic Review, 46, pages 619–31. Weinberg, J. (1997) ‘The organization of private payment networks’, Federal Reserve Bank of Richmond Economic Quarterly, 83, pages 25–44. Winston, C. (1993) ‘Economic deregulation: days of reckoning for microeconomists’, Journal of Economic Literature, 31, pages 1263–89. Woodford, M. (2003) ‘Interest and prices: foundations of a theory of monetary policy’, Princeton, NJ: Princeton University Press. Zhou, R. (2000) ‘Understanding intraday credit in large-value payment systems’, Federal Reserve Bank of Chicago Economic Perspectives, 24, pages 29–44.

3

Some challenges for research in payments Edward J. Green

In this chapter I discuss four directions in payment research that provide particular challenges in both pure and applied economics, chosen from among the many important topics in this active field. They are: 1 2 3 4

Formulating better basic models. Making market-microstructure data publicly available. Providing sound advice about payment systems risk. Understanding the relationship between payments and other business processes.

Challenge 1: to formulate better basic models As a working definition, suppose that payment economics comprises the topics that pertain to both monetary economics and industrial organisation. Monetary economics is the study of economic environments where: • • •

•

limited trading opportunities do not exhaust economy-wide gains to trade; institutions such as monetary and banking regimes can link these opportunities to enlarge the feasible gains; centralised, command-and-control institutions or all-encompassing contingent contracts that would provide prior resolution of all decisions are infeasible; and good outcomes require trust based on self-fulfilling expectations of favourable institutional performance.

Industrial organisation is an approach to economic analysis that recognises that: • •

economic agents are strategic players, not passive price takers; and economic activity involves increasing returns and externalities.

These definitions help define what a good economic model of payments should contain. One should begin by specifying an environment in terms of the agents who populate it, their technological opportunities, their preferences, the information that

58

E.J. Green

they possess, the protocols for communication among them, and so forth. From this specification, it should be clear what are the institutions that would be feasible to operate in this community. Then an equilibrium concept – specifically, one that recognises agents’ strategic incentives – should be set forth, as well as a welfare criterion that ranks the allocations that potential equilibria would implement. The equilibria of various institutional frameworks can then be studied, and recommendations about optimal institutions can be made in a way that is forthright about the combination of analytic and normative assumptions on which they rest. If these are the foundations for models in payment economics, then what facts should be explained according to it? The most basic fact is that virtually all trade utilises one of two institutions that coexist in the economy. One is the transfer tokens of stored value. Historically, these tokens have generally been coins or pieces of paper currency, either publicly or privately issued. Recently, there are also electronic implementations of stored-value transfer such as ‘smart cards’. The other institutional framework for trading is the recording of a pair of offsetting ledger entries in accounts of the parties to the transaction on the books of intermediaries, supported by another pair of offsetting entries in intermediaries’ accounts at a higher level intermediary such as a correspondent bank or central bank if the transactors have accounts at two different intermediaries. There was no economic theory, or model, of payments that could be regarded as even a serious attempt to explain this fact until about 15 years ago. Today there are two such models. One is an overlapping-generations model with settlement frictions, due to Freeman (1996a, 1996b). A second is a model of bilateral trade that formalises the neoclassical idea of ‘lack of double coincidence of wants’, the prototype of which was developed by Kiyotaki and Wright (1989), that Cavalcanti et al. (1999), Cavalcanti and Wallace (1999) and Wallace (2005) have augmented by a representation of financial intermediaries. Each of these models is now examined, to see how they fit the principles outlined above and also to suggest the respects in which they are not entirely successful. The overlapping-generations model with settlement frictions, which might be called the settlement-friction model for short, is a descendant of Sargent and Wallace’s (1982) model of the coexistence of transactions using outside money and those using debt. It adopts the modelling strategy of that paper in positing two types of trader within each generational cohort, one of which must acquire goods from the other at the beginning of life in order to fully exploit potential gains to trade, and also in positing a fixed sequence of trade meetings with limited participation at each date. The various traders must utilise these opportunities to accomplish their transactions. As this model is specified by Freeman (1996a, 1996b), an agent whose trade meeting to acquire first-period consumption from a contemporary comes before he has had a chance to meet someone who wants to purchase his endowment does not yet have money, so the only way to make his purchase is on credit. Fortunately, since both sides of this transaction are contemporaries, the two will

Some challenges for research in payments

59

have a subsequent trading meeting during the second (and final) period of their lives, at which the debt can be settled. All endowments are perishable and are received in the first period of life, so the repayment must be monetary. That is, a nominal asset must be used for settlement. That is not a problem, if the secondperiod meeting comes in time for the creditor to purchase consumption from young members of the next cohort later in the period. In that case, the traders who are creditors in equilibrium value money for the same reason as the traders in Samuelson’s classic model. They are also willing to accept debt issued by the traders who are debtors in equilibrium because it is safe debt that settles in outside money, and is therefore a perfect substitute for the outside money that they would alternatively have got by selling their endowments to members of the preceding cohort. The problem is that the second-period trade meetings between contemporaries are not so nicely timed for everyone. If debt could not be intermediated, then creditors would be unwilling to sell to debtors who could not make timely repayments. Or, if those specific debtors could not be identified in advance, then creditors would be unwilling to sell to debtors at all. However, if there are appropriately timed trade meetings among agents in their second period of life, then creditors who cannot receive timely repayment are able to sell their claims to others who can meet those debts later, and those others recover the purchase price of the claims by receiving payment from the debtors. That is, payment debt is intermediated by a subset of the creditors. What this structure accomplishes is to motivate three features of actual payment systems: 1

2 3

Some trade (between members of adjacent cohorts) has to be conducted using outside money, while other trade (between creditors and debtors in the same cohort) has to be conducted using trade credit. The trade credit has to be nominal – that is, settled in fiat money rather than in a commodity. In general, an intermediary must be active in order for all debt to be settled.

There will be inefficiency if an intermediary is active, but is unable to hold enough outside money to purchase at face value all of the debt offered for sale. The situation when this problem is severe, so that debt claims have to be sold at far below par, is one of a high interest rate prevailing in the intermediation market. There are two solutions that turn out to be equivalent in this model. One is to allow the intermediary to emit outside money but require him to reabsorb it by the end of the period. This is tantamount to making the intermediary a central bank that conducts open market operations. The other is to allow the intermediary to issue debt of his own in return for the debt claims that he acquires from other agents for settlement. The intermediary must settle these debt claims later, of course. This debt issued by the intermediary closely resembles the securities that clearinghouses issue to members in return for their claims on the members’ debtors. In summary, the model implies that:

60

E.J. Green

•

an interest rate on debt created in the payment system that is above the level warranted by the riskiness of that debt is inefficient; a central bank and a clearinghouse operate, although having different asset structures and payment flows among payment intermediaries, support identical real allocations.

•

The first implication of the settlement-friction model mentioned above – the coexistence of money and credit transactions – is just the basic fact that was identified earlier as a challenge for a theory of payments to explain. The second and third implications – that payment-system debt is nominal and requires intermediation – provide further, accurate details about this basic fact that were mentioned in framing the challenge. The fourth implication shows that the settlement-friction model has the sort of welfare implication that one expects from a good theory in economics: it identifies a significant efficiency issue, and it also provides an observable, potentially decidable criterion (the interest-rate level relative to the level of risk) for how well the payment system is working with respect to that issue. The criterion reflects a classical view that the main contribution of monetary policy to economic efficiency is to facilitate the operation of the payment system by regulating conditions in the interbank market for the short-term credit generated in the payment process. This is a cogent view, well aligned with what a central bank is able directly to affect. Certainly, central banks have focused their attention on this matter for weeks or months during periods of financial stress, such as 1987, 1998 and 2001 in the United States. If one were to emphasise the settlementfriction model very heavily as a model of the economic role of central banks, though, the upshot might be a view of optimal monetary policy that is more focused on this role, and correspondingly less focused on attempting to control broad real or nominal aggregates, than most central banks embrace today. The fifth implication of the model is a feature that is a hallmark of good scientific theories – the ability to explain facts beyond the handful for which it was explicitly designed. For, besides the coexistence of cash and trade-credit transactions, it is also a basic fact that central banks and clearinghouses have long coexisted in most of the world’s mature economies, with neither institution seeming to have so pronounced an advantage over the other from participants’ perspective to drive it out of business. Indeed, factual questions about coexistence of, and competition among, institutional forms, and normative questions about whether the formation and survival of efficient institutions is an outcome of competition, are challenging and important. The value of the settlementfriction model in this regard is further established by the research of Fujiki (2003) concerning institutional structures of foreign exchange settlement. The foregoing discussion establishes that the settlement-friction model is, on the whole, a scientifically successful economic theory, the policy implications of which should be taken seriously. Nevertheless it has three limitations in common with all overlapping-generations models. First, monetary equilibrium would not survive if the model were enriched and made more realistic by adding

Some challenges for research in payments

61

a real asset with a positive rate of return. Second, while the model as typically formulated has a unique equilibrium, uniqueness is an artifact of the restrictive assumption that agents have zero endowments in the second period of their lives. If that assumption is relaxed, then the resulting model will have a continuum of equilibria with deterministic price paths, and will also be replete with ‘sunspot’ equilibria. Because an eminently reasonable generalisation of the model makes equilibrium indeterminate, welfare analysis and policy advice based on the special version of the model are fragile. (Note, however, that fragility takes the form that there is simply nothing to say if the special assumption is relaxed, rather than that a contradictory conclusion must be drawn if the assumption is relaxed.) Third, while the model provides qualitative insight about how the payment process works and why it is structured as it is, it does not seem to provide a helpful framework for the quantitative study of data. For those who think that empirical work should be organised around the estimation of explicit, coherent theoretical models, the unsuitability of this model for that purpose is a significant limitation. To further consider the difficulty of using the model as a framework for empirical research, recall that one of Samuelson’s motivations to formulate the original overlapping-generations model was to study social security systems. Recent advances in computing power have made it possible to formulate versions of the model – adapted in ways such as specifying 30 or 40 periods (interpreted as years) of working life and 20 periods or so of retirement, instead of one period of each – that do provide suitable frameworks for quantitative studies. An analogous strategy of building a larger-dimension version of the model and analysing it computationally probably cannot work for the settlement-friction model. The reason is that the specification of trading opportunities in the model, while judicious and fruitful for explanation, is ad hoc. Consider a bank’s operational problem of making payments, for instance. Settlements at different times are distinct services, and the resources for making those payments – such as staff with specialised competence in the bank’s operations centre – are production inputs with economies of scope in their application. It would be costly to bring in part-time ‘bubble staff’ for only an hour or two each day to handle the spike in demand for Fedwire (the United States Federal Reserve RTGS system) payments that precedes the closing time of CHIPS (clearinghouse owned/operated net settlement system for US dollar payments) that McAndrews and Rajan (2000) have documented, for example. Exogenously or endogenously, banks may be heterogeneous in their respective customers’ levels of intraday demand variability or in their ability to accommodate that variability. These are the sort of issues that make it difficult for banks to coordinate their settlements directly, and that therefore make it socially valuable for Fedwire to grant intraday credit to the banks that use it. To analyse convincingly a question like how the development of automated, ‘straight-through’ payment processing will affect the level of demand for Fedwire credit, one would need to derive the sequencing of the meeting opportunities envisioned in the settlement-friction model under both the manual and the automated

62

E.J. Green

technologies. Providing such a derivation is part of the challenge to formulate a more satisfactory payment model. The second noteworthy model of an environment in which use of outside money coexists in equilibrium with use of the services of payment intermediaries focuses on bilateral trade between pairs of agents who lack a double coincidence of wants, although there would be economy-wide gains to trade if everyone could trade simultaneously. A formal setting with this feature, and in which the agents are explicitly modelled as having a high degree of anonymity that makes multilateral trading agreements infeasible to negotiate or enforce, is the random-matching model introduced by Kiyotaki and Wright (1989). The population is a continuum of infinite-lived traders who maximise expected utility, who have preferences to consume one another’s endowments in a pattern that excludes double coincidence of wants, and who are randomly matched into trading pairs at each date. Kiyotaki and Wright assume that no trader can learn any other’s trading history, so an agreement to participate in a pattern of multilateral trades over time could not be enforced. They (and subsequent researchers who study the use of fiat money) show how coordination on the use of either commodity money or fiat money can partially compensate for traders’ anonymity. Payment intermediation has been introduced into a random-matching model by Cavalcanti et al. (1999), and also Cavalcanti and Wallace (1999) and Wallace (2005). Payment intermediaries are modelled as traders who possess the same, random, meeting of technology as other traders, but whose trading histories are publicly known. This publicity makes it possible for intermediaries to subject themselves to self-enforcing agreements to repay their own debts and also the debts of other intermediaries. In fact, the probability that an intermediary will have its own notes returned to it is zero. What is important is that each intermediary commits itself to provide valuable goods in return for the notes of others, and to limit its net emission of notes – that is, the excess of its cumulative note issuance over its cumulative acceptance of other intermediaries’ notes. Claims on intermediaries – broadly resembling private banknotes – thus become acceptable to other traders. In fact, they become a circulating medium of exchange that is accepted by one non-intermediate trader from another, as well as when offered directly by an intermediary. Such inside money complements the outside money in the economy so that a higher level of welfare is reached. Let us compare this sort of random-matching model of payments with the settlement-friction model. Both models posit constraints on traders’ ability to deal directly with one another at mutually convenient times, and both succeed in explaining the coexistence of transactions made directly with outside money and those made via financial intermediaries that issue inside-money debt claims. The settlement-friction model has a richer set of implications, but the additional implications are fruits of a style of modelling that allows assumptions to be made ad hoc to produce an equilibrium with preconceived characteristics. In contrast, the random-matching model is highly stylised, but parsimonious. Traders’ anonymity and the lack of double coincidence of wants (or the more

Some challenges for research in payments

63

general idea that immediate gains to trade in small coalitions do not exhaust the gains that economy-wide trade would afford) are modelled in a conceptually straightforward way. This is also a model in which the acceptability of a medium of payment is significantly a matter of social convention, so that a security that everyone agrees to treat as money can be useful as such, despite being intrinsically less appropriate (for example, lower in rate of growth) than another asset.

Challenge 2: to make market-microstructure data publicly available A second challenge for payments research is to gather, and make available to the research community a body of statistical data from which models of payments could be assessed and estimated. From the preceding discussion, it should be clear that data about what finance researchers call ‘market microstructure’ is particularly important. Consider large value payments among banks and other financial intermediaries, for example. What is the pattern of these payments (including value, volume and concentration of payors and payees) through the course of a typical day? How is that pattern affected by various special circumstances, ranging from ‘triple witching days’ in markets for cash-settled derivatives to episodes of malfunction of critical payment-technology infrastructure? Having this sort of information is basic to understanding how large value payment systems work and to proposing and evaluating policy towards them. Obviously this is a shared challenge for researchers, the payments industry, and especially central banks. There is the usual problem of assembling and maintaining very large data sets: finding someone to do it and to finance it. There are also issues such as data confidentiality that are somewhat specific, although by no means unique, to payments data. Owners of other data that present such problems have shown determination and creativity in making it available to researchers. For example, the US Census Bureau operates a number of facilities around the country where researchers can submit their computer programs to be run by Bureau staff. Since researchers are almost exclusively interested in the overall statistical measures that are computed from the data in this way, rather than in the names of individuals whose privacy the delegated-computation arrangement safeguards. This way of providing access is very successful. Another strategy that might be considered for the types of data that I have mentioned would be to create synthetic data sets, using a computer program for random number generation, that would have the same statistical moments or other relevant statistical features as the actual data, but that would not contain actual, individual transactions. If the statistical verisimilitude of the synthetic data set were credibly vouched for, then journal editors and other pivotal members of the scientific community would be disposed to accept its use as a proxy for the actual data. However, researchers’ access to payment data might be provided, the returns to providing it would be large. Today, lack of data access is a factor that

64

E.J. Green

discourages empirical researchers from working on payment-system issues. And that fact tends to discourage economic theorists, in turn, from investing the extra effort to turn insightful, schematic theories into estimable and testable ones. The costs of data not being publicly accessible is a silent one of researchers, who look for good problems to address and simply find problems in other areas, and perhaps not even being aware that they would have chosen payment topics if they had had the opportunity. Nevertheless, the cost to payment economics is substantial. A broad-based, scientific community provides rapid and relatively straightforward growth of knowledge, and correction of mistaken belief. Without a good infrastructure of data conveniently available to independent researchers, analogous to the published national income statistics that researchers on other issues of interest to central banks take for granted, the scale of research in payment economics is limited and the benefits of more intensive research are missed.

Challenge 3: to provide sound advice about payment risk A third challenge is to understand better the role of payment-system credit, and correspondingly to be able to provide better founded advice about the risk management issues that this credit raises. Here is the issue: there is broad consensus (although it is not quite unanimous) among policy makers that an RTGS system is the gold standard for a systemically important large value payment system. There is also consensus, supported by some recent contributions in paymenteconomics theory (such as Freeman and Hernandez-Verme, 2004), that an RTGS system that does not provide credit to its participants can be at least as risky as a net settlement system. Essentially, what is done in virtually all RTGS systems operated by central banks is exactly what the settlement-friction model recommends. The central bank effects RTGS by becoming the central counterparty to all participants, providing intraday bridge loans for the difference between the payments that they would have to pay in an RTGS system without credit and the smaller payments that they would have to make in a net settlement system. As a result, at the peak of its exposure during the day, the central bank holds a huge portfolio of short-term loans that are exceptionally safe on the whole, but never absolutely safe. A first question is, how risky is this loan portfolio? To answer this question from a public perspective is subtle. Simply to say, for example, that the central bank has never lost a penny on its short-term credit to the payment system. But that may not be relevant if the RTGS design creates incentives for participants to become more exposed to losses than they would otherwise be, and if there are other creditors of those participants who would be junior to the central bank in an insolvency. If so, then would that situation be an externality that the RTGS operator is unwisely imposing and that those other creditors cannot control? That is the sort of assumption that Lagunoff and Schreft (2001) make in their modelling of systemic risk, for example. A polar assumption, analogous to a strong-form efficient market hypothesis in asset-pricing research, would be that

Some challenges for research in payments

65

junior counterparties have reasonably current and accurate knowledge of how an RTGS participant uses intraday credit from the operator, and that those counterparties price their own credit extensions to the participant. The participant thus has the right incentive to economise on RTGS credit, since the operational cost of doing so will result in savings from more favourable pricing of other credit that it receives. How would researchers distinguish between these hypotheses? A threshold question is, are RTGS participants heterogeneous in their levels and patterns of use of intraday credit? With the sort of market-microdata for payments that were described above, this question would be answerable. If the answer is affirmative, as seems likely, then it could be asked whether participants’ strategies of RTGS credit use are correlated with, for example, the terms that they receive in the overnight money market (see Furfine, 2000). Even if it is determined that RTGS credit does not impose an uncontrollable externality on participants’ counterparties, there is the second question of how an RTGS operator should manage the credit risk that it takes on directly. One, decision-theoretic, aspect of this issue is how the very small risk of an extremely large loss to the operator ought to be quantified. Once it is quantified, one approach to managing it is to price it, as is done by the US Federal reserve. An alternate approach, taken by numerous other central banks, is to require collateral. Participants tend to view collateral as being costly, since the assets eligible to be used as collateral have lower rates of return than other investments. Thinking about collateral in terms of the Diamond–Dybvig theorem (along lines suggested by Wallace (1996)) suggests that there is also a social cost if RTGS collateral requirements force the banking system to hold a portfolio more heavily weighted towards low-return assets than would otherwise be necessary. An estimate of the magnitude of this cost would be useful. Moreover, if one thinks of central banks – the RTGS operators – as having much less need for liquidity than other market participants because they can create it themselves, then there is a welfare-economic question of whether they should accept collateral that is less liquid than other secured lenders would accept. Finally, if so, there are financial-economics questions of how such collateral should be valued, and there are other questions – partly in the domain of law and economics – about how it should be managed. A third aspect of the challenge regarding the welfare economics of RTGS credit is the issue of whether or not such credit, when offered cheaply or for free, is a subsidy. A traditional view is that payment credit is indeed a subsidy if it is priced below the intertemporal rate of substitution or the intertemporal marginal rate of return of agents in the economy. The settlement-friction model suggests a contrasting view – that payment credit is a unique economic institution that, despite a superficial resemblance to investment credit, has a completely different rationale. In fact, Zhou (2000) has analysed a version of the model implying that (abstracting from risk) payment credit should be priced at zero, even when there is a positive rate of inflation. In the settlement-friction model in its usual form and as she uses it, however, there is no arbitrage opportunity for the diversion of

66

E.J. Green

intraday credit. In the actual economy, there is a possibility that an RTGS participant with very low transaction costs might be able to profit from doing something like using RTGS credit every morning to purchase government securities, and then selling the securities at the end of the day to fund the RTGS payment from the proceeds. If there were a tendency for government securities to be priced higher at the end of the day than at the beginning, then this daylighttrading strategy might be profitable on average over time for a participant that is large enough to take huge positions. Whether or not there actually does exist any such arbitrage possibility is worthy of study.

Challenge 4: to understand the relationship between payments and other business processes A fourth challenge is substantially different than the others, which have all been towards the monetary-economics end of the payment-economics spectrum. This last challenge is better to integrate models of payments and related business processes. Much emphasis has been placed in payment economics on understanding the benefits of net settlement and on the transaction costs of payment float. Yet, for moderate-size commercial payments as well as for household payments, the costs and benefits of these for a single transaction must be fairly minor. In contrast, US bankers and corporate treasurers tend to quote numbers in a range of something like $5–$20 as the all-in cost of making a payment when related costs, such as invoicing and reconcilement, are included. Besides these variable costs, there are substantial fixed costs of payment-related information technology – particularly of comprehensive enterprise-management software that structures how payments and other business processes are related. There are other sorts of complementarity as well, such as retailers’ use for marketing purposes of information generated by customer payments. Models that treat payment explicitly as a component of a more comprehensive transaction, and that incorporate these various considerations, will likely be essential for understanding what drives the choices that transactors and their intermediaries make among payment options.

References Cavalcanti, Ricardo and Neil Wallace (1999) ‘Inside and outside money as alternative media of exchange’, Journal of Money, Credit, and Banking, 31, pages 443–457. Cavalcanti, Ricardo, Andres Erosa and Ted Temzilides (1999) ‘Private money and reserve management in a random-matching model’, Journal of Political Economy, 107, pages 929–945. Freeman, Scott (1996a) ‘The payments system, liquidity, and rediscounting’, American Economic Review, 86, pages 1126–1138. Freeman, Scott (1996b) ‘Clearinghouse banks and banknote over-issue’, Journal of Monetary Economics, 38, pages 101–115. Freeman, Scott and Paula Hernandez-Verme (2004) ‘Default and fragility in the payments system’, unpublished thesis Texas A&M University.

Some challenges for research in payments

67

Fujiki, Hiroshi (2003) ‘A model of the Federal Reserve Act under the international gold standard system’, Journal of Monetary Economics, 50, pages 1333–1350. Furfine, Craig (2000) ‘Interbank payments and the daily federal funds rate’, Journal of Monetary Economics, 46, pages 535–553. Kiyotaki, N. and R. Wright (1989) ‘On money on a medium of exchange’, Journal of Political Economy, 97, pages 927–954. Lagunoff, Roger and Stacey L. Schreft (2001) ‘A model of financial fragility’, Journal of Economic Theory, 99, pages 220–264. McAndrews, James and Samira Rajan (2000) ‘The timing and funding of Fedwire funds transfers’, Federal Reserve Bank of New York Economic Policy Review, pages 17–32. Sargent, Thomas and Neil Wallace (1982) ‘The real-bills doctrine versus the quantity theory: a reconsideration’, Journal of Political Economy, 90, pages 1212–1236. Wallace, Neil (1996) ‘Narrow banking meets the Diamond–Dybvig model’, Federal Reserve Bank of Minneapolis Quarterly Review, 20, pages 3–13. Wallace, Neil (2005) ‘From private banking to central banking: ingredients of a welfare analysis’, International Economic Review, 46, pages 619–631. Zhou, Ruilin (2000) ‘Understanding intraday credit in large-value payment systems,’ Federal Reserve Bank of Chicago Economic Perspectives, 24, pages 29–44.

4

Payment economics and the role of central banks Jeffrey Lacker

The role of payment economics A distinct and coherent field of payment economics appears to be emerging, and it deserves some attention, especially among central bank policymakers (Lacker and Weinberg, 2003). In this chapter, I will say a few words about payment economics, and then discuss the role of the central bank in the payment system and implications of that role for several current issues. At the core of payment economics are systems of exchange financed by private and/or public liabilities and the institutions that facilitate the clearing and settlement of these instruments. In other words, payment economics can be defined as the study of the mechanics of exchange. It is based on the core insight of monetary economics that the instruments that people use to pay for goods and services serve to communicate reliably (that is, in an incentive-compatible way) about the buyer’s past actions (Townsend, 1989; Kocherlakota, 1998). Payment economics extends banking theory to encompass the role of banks as private issuers of payment instruments, and reflects the observation that virtually all institutions usually thought of as banks are significantly involved in payments. Indeed, the defining feature of banks appears to be their issuance of payment instrument liabilities, as opposed to their role as balance sheet intermediaries between savers and borrowers. Banks, from this perspective, are specialized institutions for facilitating the transmission and recording of relevant payment information. The industrial organization of the banking system therefore affects the characteristics of the monetary system. Payment economics thus lies at the intersection of monetary and banking economics with industrial organization. The fact that payment instruments and specialized institutions are at the core of the economics of payment arrangements has important methodological implications. It means that the choice of payment instruments, and the structure of the institutional arrangements that support them, should be viewed as endogenous. This defines an approach known as mechanism design – the cornerstone of modern monetary theory. Under this approach, payment instruments are seen as messages that embody contingent contracts, and one can model the information and risk allocation characteristics of alternative payment arrangements in a way that takes into account the limitations imposed by real world payment technologies

Payment economics and the role of central banks

69

– for example, the costliness and falsifiability of communication, verification, and authentication. For example, the mid-1980s presumption that paper cheques were socially inefficient because of their higher processing costs ignored other, apparently consumer-relevant, characteristics of cheques. The central role of communication in payments instruments and institutions has implications for the organization of payments activities. Communication technologies invariably are characterized by such features as economies of scale, common costs, and joint production. These features often take the form of network effects, in which much of the benefits and costs of network activities are shared among multiple participants. Private organizations that deal effectively with such characteristics can be described as clubs, in which terms of membership are just as important as the pricing and terms of service provision in inducing efficient participation. There is a tradition in industrial organization of questioning the extent to which competition ensures efficient performance in markets with these characteristics. This hinges on the extent to which markets are contestable, as has been emphasized by Ed Green and Dick Todd in their essay for the Minneapolis Fed’s 2000 Annual Report (Green and Todd, 2001).

The role of central banks in payments I would like to sketch out a tentative view that seems consistent with the emerging lessons of payment economics. It is not the only possible view one could take, but it strikes me as compelling. Until a better one comes along, I view it as a logical benchmark model. This view is built on two core ideas. First, central banks have more or less nationalized the clearinghouses at the ‘apex’ of the payment system. One can debate whether this was efficiency enhancing, as Goodhart (1988) argues, or whether it arose instead to re-allocate the costs and benefits of clearing and settlement activities. Clearinghouse activities appear to have aspects of club goods, as I noted earlier, and for club goods there is often a range of allocations consistent with efficiency – that is, with Pareto optimality. Central bank intervention sometimes alters the distribution of net benefits among payment system participants. For example, the Fed’s entry into cheque clearing seems to have been less about efficiency improvements than it was about shifting the costs of clearing cheques drawn on country banks. In any event, legal restrictions nowadays more or less compel many banks to settle at least some transactions through the transfer of central bank account balances. The second core idea is that many, if not most, of the private sector institutions that are the major players in the payment systems benefit from a substantial public sector safety net. In many cases explicit deposit insurance provides the most visible government support. But in addition, significant support is provided in conjunction with central bank payment operations. Central banks generally supply credit, both intraday and overnight, to key payment system participants. (The Swiss, until recently, were notable exceptions.) Moreover, there is a widespread perception among private payment system participants that central bank

70

J. Lacker

credit will be made available, perhaps even overnight, to facilitate the resolution of operational problems or other settlement disruptions. As Marvin Goodfriend and I have emphasized in a joint paper, this constitutes a backstop line of credit provided by the central bank (Goodfriend and Lacker, 1999). Indeed, operational protocols and the routine provision of daylight credit in some cases leave the central bank with no other choice but to lend. For example, in the case of the disruption at the Bank of New York in November 1985, the extension of overnight credit was a fait accompli (Lacker, 2004).

Moral hazard Taking the terms on which central banks clear, settle, and lend as given, the usual presumption is that competitive pressures will drive private sector institutions towards second-best efficiency. Underpriced access to central bank credit will of course distort private sector choices. Absent countermeasures, banks will take excessive risks and central bank credit will be overused, a distortion often referred to as moral hazard. It is in the nature of lines of credit, however, that they are underpriced at the point in time at which they are utilized. Credit lines provide guaranteed access to funds at a prespecified rate that does not vary with the borrower’s ex post creditworthiness. Thus borrowers essentially obtain insurance against adverse shocks to their creditworthiness. Private line of credit lenders are generally compensated for this insurance provision through up-front fees. Other features of typical credit lines act to constrain moral hazard. Lenders limit the extent of their liability through loan covenants that let them deny credit if certain financial conditions are not satisfied. In addition, lenders generally monitor borrower financial conditions on a regular basis, and often reserve the right to audit borrowers. The potential for moral hazard due to a public sector safety net, and in particular the provision of central bank credit in connection with payment operations, is to my mind the central rationale for central bank oversight of payment system participants. Such oversight should be aimed at measuring and efficiently constraining private risk taking that could affect the extension of central bank credit or the provision of public sector support. Much central bank payment system supervisory activity obviously fits this description well. Having said that, it is my sense that central banks have not come close to offsetting fully the safety net’s moral hazard distortion, although I would be hard pressed to document that claim, except to note the extent to which access to central bank settlement seems to be highly prized by financial institutions. This description of central bank payment activities implies a minimal service provision role – basically just offering clearing accounts that are used to settle interbank obligations. And this role is a byproduct of having de facto monopolized interbank settlement. In this, I find Green and Todd (2001) persuasive when they argue that the rationale for more extensive central bank service provision depends on the extent to which there are economies of scope between additional activities and the basic clearing account function. A focus on payment

Payment economics and the role of central banks

71

systems as communications mechanisms suggests the importance for this question of the relative effectiveness of alternative configurations of communications architectures, and potential economies in verifying messages and safeguarding information. My sense, however, is that there are far less by way of economies of scope than would be needed to justify, on economic efficiency grounds, the current scale of Federal Reserve service provision, particularly in clearing ‘retail’ payments such as cheques and automated clearing houses (ACHs). In fact, I have argued elsewhere that the evidence suggests that the Fed’s role in clearing retail payments rests on altering the allocation of clearing costs that would result from purely private provision. The imminent transition away from paper cheque clearing makes the Green and Todd question a live issue right now in the United States. Notice that I have made no use of the notion of ‘market failure’. My own working hypothesis is that market failures are largely absent from the payment system. After all, participants in any given payment arrangement are all linked by voluntary contractual relationships. Thus I find it hard to see how an externality, in the classic sense, could possibly arise. (The only genuine payment system externality I know of occurred when the Federal Reserve incinerated worn paper currency, thus polluting the air.) Note that the lack of an observed market does not mean market failure. For example, large banks do not clear cheques for rural banks in the United States. Surely this is due to the terms on which the Fed provides the same service. After all, there was an active market before the Fed did it. But as I argued earlier, we do not need a market failure to motivate central bank supervision of private payment system activities. To me, central banks’ policy interest is amply motivated by the presence of a substantial public sector safety net to payment system participants, and the central bank’s role in providing and setting the terms and conditions of important elements of that safety net.

Questions about the role of central banks in payments The perspective I have just outlined implies that the terms and operational conditions on which central banks extend daylight and overnight credit are of central importance. Years ago, when many aspects of current arrangements were put in place, operational considerations made it costly to implement systems that did not automatically extend daylight central bank credit in one form or another. New technologies may have significantly altered this cost–benefit trade-off, and in my opinion a re-examination of daylight credit policies is in order. For example, many banks monitor and control the extension of daylight credit to many of their corporate customers, and indeed, supervisors expect them to. It would be ironic for central bank risk management to lag behind private sector practices in this regard. A focus on central bank credit also makes clear that paying interest on reserves is more important than is commonly appreciated. The prohibition of interest on central bank deposits, as in the United States, greatly enhances the demand for daylight credit, in the sense that larger overnight balances act as

72

J. Lacker

substitutes for daylight overdrafts. As a result, limitations on central bank credit extension would be less costly if reserves earned interest. More broadly, it seems plausible that a huge fraction of settlement activity originates in transactions whose main purpose is to allow entities to evade interest prohibitions, and thus to some extent are socially wasteful. The relationship between central bank credit and the broader public safety net has implications that are sometimes overlooked. For example, the collateralization of central bank credit extension may reduce risks to the central bank, but it can increase risk to the deposit insurance fund. Therefore, the central bank ought to consider more than just its own balance sheet risk in making lending decisions. This is especially important because, as the lender of last resort, the central bank can often force an institution’s closure by refusing credit. Notwithstanding several seemingly strong policy pronouncements, I do not believe that our understanding of the economics of intraday credit is at this point sufficient to provide quantitative guidance on the optimal pricing of daylight credit, even apart from moral hazard considerations. In that light, volumes such as those devoted to the advance of payment economics are among the most noble uses of central bank resources.

References Goodfriend, M. and Lacker, J.M. (1999) ‘Limited Commitment and Central Bank Lending’, Federal Reserve Bank of Richmond Economic Quarterly, 85, 4, pages 1–27. Goodhart, C.A.E. (1988) The Evolution of Central Banks, Cambridge: The MIT Press. Green, E.J. and Todd, R.M. (2001) ‘Thoughts on the Fed’s Role in the Payments System’, Federal Reserve Bank of Minneapolis Quarterly Review, 25, pages 12–27. Kocherlakota, N.R. (1998) ‘Money is Memory’, Journal of Economic Theory, 81, pages 232–251. Lacker, J.M. (2004) ‘Payment System Disruptions and the Federal Reserve After September 11, 2001’, Journal of Monetary Economics, 5, 1, pages 935–965. Lacker, J.M. and Weinberg, J.A. (2003) ‘Payment Economics: Studying the Mechanics of Exchange’, Journal of Monetary Economics, 50, pages 381–387. Townsend, R.M. (1989) ‘Currency and Credit in a Private Information Economy’, Journal of Political Economy, 97, pages 1323–1344.

Part II

New approaches to modelling payments

5

New models of old (?) payment questions Ricardo Cavalcanti and Neil Wallace1

Introduction In both the United States and the United Kingdom, a monopoly on ‘currency’ issue grew out of a system in which there were many issuers of banknotes. In the United Kingdom, that monopoly was created in 1844, and was accompanied by a 100 per cent specie marginal reserve requirement against banknote issue. The 1844 law, Peel’s Act, was a victory for the currency school, whose members advocated some version of hard money, or what much later came to be called monetarism. The 1844 law was opposed by members of the banking school: those who advocated some versions of laissez-faire in intermediation. Among the questions alluded to in the debates were: Was the private note-issuing system accomplishing anything? If it was, then would it be desirable to have the Bank of England manage its monopoly so as to emulate what the private note system was accomplishing? In this chapter, we revisit those questions and do so for at least three reasons. First, one test of progress in monetary theory is its ability to provide new insights about old questions that have never been satisfactorily resolved. Second, those old questions have modern analogues: should central banks operate lending facilities and, if so, how? Should stored value, and other modern analogues of private note-issue, be regulated and, if so, how? Third, the modelling ideas that throw light on those questions have implications for seemingly unrelated questions: for example, how best to model cashless economies. Why do we assert that the nineteenth-century debates were never satisfactorily resolved? At the beginning of the twentieth century, the dominant monetary theory consisted of the classical dichotomy. While that theory could accommodate private credit instruments that to some extent substitute for outside or base money, either by treating such substitutes as part of the stock of a broader concept of money or by treating them as increasing the velocity of outside or base money, neither treatment could say anything about the welfare consequences of different monetary systems or, for that matter, the welfare consequences of money. At the beginning of the twenty-first century, the dominant monetary theory consists of descendents of the classical dichotomy: models with real balances in utility or production functions or models with cash-in-advance constraints. These descendents were designed to overcome the blatant inconsistencies of the classical dichotomy:

76

R. Cavalcanti and N. Wallace

the kind of inconsistency that Patinkin (1951) pointed out. They were not designed to and cannot address the questions raised in the nineteenth-century debates any better than could the classical dichotomy. In this chapter, we set out some ideas about how such questions might be approached. Our goal is to convince readers that the ideas are fruitful: both for the nineteenth-century questions about good monetary systems and for other questions concerning monetary systems. However, one warning is in order; we have essentially no results about the implications of the modelling ideas we set out.

Some general ideas One challenge to any model of money is: why is trade being modelled using money when trade could conceivably be accomplished with some version of borrowing and lending between people? There is, by now, a well-known answer. Individuals cannot commit to future actions and to some extent their histories are not known. This answer goes back at least to Ostroy (1973). (See also Townsend (1989) and Kocherlakota (1998).) Neither part is controversial. The inability to commit, although inconsistent with the Arrow–Debreu model, is a standard and plausible assumption of game theory. The problem of partially unknown histories is, in modern game theory, labelled imperfect monitoring. It means that previous actions of some people are not common knowledge. It is the assumption in moral-hazard models and is implicit in the idea that money is used in trade among strangers and the related idea that money is evidence of past actions that are otherwise imperfectly known. We like the answer, and, therefore, build a model that rests on it. Throughout we maintain the assumption that people cannot commit to future actions. As regards monitoring, we assume that some people are not monitored at all and others are perfectly monitored. The unmonitored people are the demanders of tangible media of exchange; the monitored people are the potential issuers of private money and, in most respects, are the focus of our discussion. The kind of private money we analyse is best thought of as payable-to-thebearer bills of exchange that have only the issuer’s name on them. The private money has this form, a form which bypasses banks as we ordinarily describe them, because this form more easily gets us to a model in which the welfare consequences of different systems can be analysed. (Something like this is done in the Diamond–Dybvig model (Diamond and Dybvig, 1983) of banking in which what is described as a banking system is best thought of as a mechanism in a model consisting only of consumers, who are the owners of a consolidated banking-business sector.) Throughout, we work against the background of a model in which each person, including any issuer of private money, is individually a small part of the total economy and in which for purposes of production and consumption people meet in pairs. In our model, people do not choose to meet in pairs and a pair need not be viewed as a natural production unit, as it is in models of marriage and seems to be in many search models of labour. (In our setting, larger

New models of old (?) payment questions

77

production-consumption meetings, if they could occur, would enhance welfare.) In our model, one pairwise meeting per date for the purposes of production and consumption is free and any other kind of meeting for that purpose is infinitely costly. The pairwise meeting structure helps us in several respects; it is consistent with absence-of-double-coincidence difficulties, with imperfect monitoring (each person may know only what the person has seen in the meetings in which the person has been a participant), and with non-trivial float (although float will not play a significant role here). As in any model of private money, potential over-issue has to be prevented. In our model, it is prevented by threatened punishment. The punishment we use here, although rather mild, always includes the loss of the ability to issue valuable money. To make that feasible, we assume throughout that the private money issued by one person is potentially distinguishable from that issued by anyone else. This is a strong recognizability assumption. We suspect that weakening it, by permitting some sort of counterfeiting, would matter a lot.

The model The model is almost identical to that in Wallace (2005), which, in turn, builds closely on our previous work (Cavalcanti and Wallace, 1999a, 1999b). In particular, the sense in which there is imperfect monitoring is carried over from the specification in our earlier papers. A background specialization environment We use the familiar specialization setting of Shi (1995) and Trejos and Wright (1995). Time is discrete. There is a unit measure of each of K  3 specialization types of infinitely lived people and there are K distinct, produced, and perishable goods at each date. A specialization-type k person, k {1,2,...K}, produces only good k and consumes only good k + 1 (modulo K). Each person maximizes expected discounted utility with discount factor  (0,1). For a specializationtype k person, utility in a period is u(qk+1) – qk, where qk+1 ℜ+ is consumption of good k + 1 and qk ℜ+ is production of good k. The function u : ℜ+ → ℜ is strictly concave, strictly increasing, differentiable, and satisfies u(0) = 0 and u'(∞) = 0. In addition, u'(0) is sufficiently large. A word is in order about the assumption that the number of people is uncountable. So far as we can see, this assumption plays only one role. It implies that a person’s action in a two-person meeting does not influence his or her future trading opportunities except by way of what happens to the person: not by way of what happens to the person’s trading partner. That should hold approximately for a sufficiently large finite number of people. In other words, we suspect that the outcomes we describe resemble those of the comparable model with a sufficiently large finite number of people, provided there is discounting that is held fixed as the number of people is allowed to get large.

78

R. Cavalcanti and N. Wallace

Imperfect monitoring and the sequence of actions We make one other distinction among people. We assume that the set of each specialization type is partitioned in an exogenous way into two parts. Throughout, the fraction n, is not monitored at all. The history of each such person, except as may be revealed by the person’s money holdings, is private to the person. The rest, the fraction m = 1 – n (m for monitored) are perfectly monitored. That is, the history of each monitored person is common knowledge. It is as if each monitored person wears a computer chip that transmits actions of the person to everyone else. In this model, m represents the economy’s monitoring capacity. As part of not being monitored, each unmonitored person can hide money. We use the following sequence of actions in discrete time. At the start of each date, each person has a state consisting of the person’s type, history, and money holding. A person’s type, specialization type, and whether monitored or not, is assumed to be common knowledge and is permanent. Money holding, a scalar, is defined to be the sum of outside money plus private money acquired from others. As this suggests, we only consider allocations in which all valuable monies, all private monies and outside money, are perfect substitutes. (Richer allocations that distinguish among valuable monies, both subsets of private money issuers and between private money and outside money, could be considered.2) Then, there are pairwise meetings at random, during which there is production and consumption which gives payoffs according to the preferences and technologies described above. After those meetings conclude, monitored people simultaneously meet the planner and, to be consistent, are all together. However, by assumption, there is no production or consumption that goes on after the pairwise meetings. At best, there are transfers of money. (The planner can be thought of as a benevolent central bank running a discount window and having unlimited access to outside money, while trades among the monitored people after the pairwise meetings can be thought of as being something like a federal funds market.3) Weakly implementable allocations and welfare We will be doing a limited kind of mechanism design analysis. We start by defining a set of allocations. Then, we describe a simple coordination game in which people choose individually either to cooperate or defect. If everyone cooperates and unmonitored people, who can hide money, choose to truthfully self-select, then the allocation is weakly implementable; otherwise not. Our goal is to describe the best weakly implementable allocation, where best is defined below. An allocation describes what happens in pairwise meetings and what happens when monitored people meet the planner, all conditioned on the date and on the states of the people in the pairwise meeting and the state of each monitored person when meeting the planner. Then, given an initial condition in the form of

New models of old (?) payment questions

79

a distribution over money holdings and histories, such a description of what happens in meetings at each date is sufficient to describe the evolution of the economy. Given a suggested allocation, the following game is played. Consider a pairwise meeting. The allocation includes a suggested trade in the meeting. Both parties simultaneously choose between cooperate and defect. If both cooperate, then the suggested trade is carried out. If either says defect, then they leave the meeting without trading. If an unmonitored person defects, then there are no further consequences. The person goes on to the next date with what the person has. If a monitored person defects, then there are further consequences to be described momentarily. In the meeting with the planner, each monitored person again chooses between cooperate and defect. As regards the consequences of defection for a monitored person, we assume that the person can at any time join the ranks of the unmonitored people and suffer no additional punishment except that the person’s private money is no longer accepted. In describing the consequences of defections, we are explicitly ruling out punishment of a large segment of the economy in response to individual defections. For example, we are ruling out permanent autarky for the entire economy as a response to an individual defection. Notice that our defection scheme permits free exit from the set of monitored people. However, we do not permit free entry into that set. Definition 1: An allocation is weakly implementable if there is a sub-game perfect Nash equilibrium in which each person cooperates and each unmonitored person also self-selects the trade intended for people with the person’s actual holding. Two comments are in order about this definition. First, it only requires that there is some equilibrium that implements the allocation. Second, it permits only individual defections, not group defections. In particular, it does not permit cooperative defection by the pair in a meeting. There are several obvious consequences of the definition. It is weakly implementable to have any recognizable money be worthless and for the usual reason: if a person thinks that others will not accept an intrinsically useless object in the future, then the person will not accept it now. Thus, it is weakly implementable to have all private monies be worthless, to have outside money be worthless (it is important that we are assuming outside money to be uniform), and to have all money be worthless. In particular, allocations in which there is no valuable private money are special cases of more general allocations that include valuable private money. (According to our model, the Peel’s Act monetary system could arise without a law.) Therefore, in order not to dwell on the completely obvious, our focus will be on describing as carefully as we can what is sacrificed by not having valuable private money. The simplest welfare criterion for the model is an ex ante representativeagent criterion: one that treats people as identical before the assignment of types and states. In particular, according to this criterion, the probability of being in the monitored set is m and the probability of being in the unmonitored set is n.

80

R. Cavalcanti and N. Wallace

Finally, because there is no capital in this model, the initial condition, the distribution of money holdings and histories, can be treated as something that the planner chooses along with the allocation. The ex ante welfare criterion can be expressed as the expected discounted value of the gains from trade over all single-coincidence pairwise meetings: gains from trade in the sense of the magnitude of g(q)  u(q) – q. Obviously, maximum ex ante welfare is achieved by production and consumption equal to q* = arg maxq[u(q) – q] in every single-coincidence meeting. As we will see, the constraints on monitoring and punishments rule out that allocation. However, the maximum nicely summarizes the economic problem represented by the model; good arrangements will tend to weaken the tie between what happens in single-coincidence meetings and the individual histories of the participants in the meeting.

The role of inside (private) money: an example with {0,1} money holdings We present a simple example that shows that private money can actually play a role. Although we do this in the context of individual money holdings in the set {0,1}, the forces at work are general. We describe stationary and symmetric allocations that are weakly implementable with valuable private money, but that are not in its absence. Consider an allocation in which the same output level, some y (0,q*], is produced in all single-coincidence meetings except in two circumstances: when an unmonitored producer has money or when an unmonitored consumer does not have money. Suppose that nothing is produced in those single-coincidence meetings. (The exception for unmonitored producers is implied by the bound on money holdings and their participation constraint; that on unmonitored consumers is an arbitrary part of the allocation and will be discussed below.) Moreover, suppose that unmonitored consumers surrender money when they consume y and that unmonitored producers receive money when they produce y. In single-coincidence meetings between monitored and unmonitored people, the monitored consumer provides (newly issued private) money to the unmonitored producer and the monitored producer collects (outstanding private) money from the unmonitored consumer, which is then turned over to the planner. In meetings between monitored people, no money changes hands. Suppose further that half the unmonitored people start without money and half with money and that all monitored people start without money. The above trades imply that that distribution persists (is a steady state). In order to express the participation constraints, it is a helpful short-hand to compute the discounted values implied by this allocation. Let v n(z) be the discounted value for an unmonitored person at the beginning of a date with money holdings z {0,1}. These values satisfy K(1 – )v n(0) = (–y + )

(1)

New models of old (?) payment questions

81

and K(1 – )v n(1) = (u(y) – )

(2)

where  = v n(1) – v n(0) and  = m + n/2. (These linear equations have a unique solution that implies  = u(y) + y, where  = 2 + K(1 – )/.) And let v m be the discounted value for a monitored person at the beginning of a date without money. It satisfies K(1 – )v m = (u(y) – y)

(3)

We do not need to express the discounted value for a monitored person of starting a period with money, money issued by another monitored person, because (i) there are no such people in equilibrium, and (ii) a defection does not give rise to such a person. For incentive feasibility, there are three relevant constraints. One is the participation constraint for an unmonitored producer: v n(1) – y  v n(0)

(4)

The others are two constraints for monitored people: v m – y  v n(0) and v m  v n(1)

(5)

The first is the participation constraint for a monitored producer (the pay-off for a monitored producer who defects is that of an unmonitored person without money because the defector’s printing press becomes worthless); and the second says that a monitored person is willing to surrender to the planner the money received in a trade. Because v m = v n(0) + v n(1) (see equations (1) to (3)), participation constraint (equation (4)) implies participation constraints (equation (5)). Next, we describe necessary conditions for duplicating the above consumption and production pattern without private money. In order to duplicate the pattern, each monitored person must begin a period with outside money. Otherwise, when a monitored person is a consumer in a meeting with an unmonitored producer without money, the producer cannot be induced to produce y. In the simpler set-up of our earlier paper (Cavalcanti and Wallace, 1999b), there was nothing like a discount window or a federal funds market, and the stock of money was constant. Hence, it was simply impossible to have the spending described in the allocation: the monitored people who spent money in the previous period would not have money at the start of the next date. Now, that argument does not apply because the planner could give money to those monitored people who spent money and could collect money from those who have acquired money. If that is done and the trades are as described by the allocation, then vm as given by equation (3) again describes the discounted value for any monitored person. In addition, the v n(z) are unaffected. However, the

82

R. Cavalcanti and N. Wallace

constraints are now different. In place of the constraints on v m in equation (5), there is just one relevant constraint: v m – y  v n(1)

(6)

The constraint says that there will be no defection when a monitored person with money is called on to produce y. The new constraint is tighter than the two it replaces and is not implied by equation (4). In fact, it is easy to describe magnitudes of y and the other parameters for which equation (4) holds, but for which equation (6) does not. For example, if y =  > 0, as is implied if y is the outcome of a take-it-or-leave-it offer by an unmonitored consumer to an unmonitored producer, then v n(0) = 0 and v n(0) = v m. Therefore, equation (4) holds, but equation (6) does not. Hence, implementabilty can fail without private money. The greater temptation to defect when a transfer of outside money replaces private money issue does not seem to depend on the special assumption about money holdings. The result does, however, depend on two features of the model. One is the assumption that outside money is uniform. If each unit of outside money were unique, then a defection could render worthless the particular unit held in the same way as the person’s printing press is rendered worthless. And it depends on the uncertainty about spending. If future spending were known when the monitored person meets the planner, then without private money the planner’s transfer could be made just sufficient to support that spending. Although the above comparison is suggestive, it is not decisive even about this simple setting with the special {0,1} money holdings. The example does not establish that private money is necessary for an optimum, even among stationary allocations. Even with y = q*, the allocation described above does not maximize welfare. In that allocation, a monitored producer does not produce for an unmonitored consumer who has no money. But some production in such meetings – even if offset by lower production in other meetings in order to satisfy participation constraints – would almost certainly increase welfare because u is strictly concave. Given {0,1} money holdings, an upper bound on welfare is given by y = q* in all single-coincidence meetings except those in which the unmonitored producer has money and y = 0 in those meetings. However, it is immediate that any allocation with the same positive output in all meetings except those in which the unmonitored producer has money is not implementable. Given such an allocation, in a meeting with a monitored producer, an unmonitored consumer with money will envy the trade of an unmonitored consumer without money unless the former is not asked to turn over money. But, if not, then money never flows from the set of unmonitored people to the monitored, which, in turn, implies that money cannot flow the other way. But that contradicts the presumed spending of monitored consumers in meetings with unmonitored producers. This immediately tells us that the optimum will have some binding truth-telling or participa-

New models of old (?) payment questions

83

tion constraints. That, in turn, makes it challenging to describe optima even in the highly special case of money holdings in the set {0,1}.

The planner or a market as a potential substitute for private money The above discussion points to the potential gain from private money. Suppose, however, that we are stuck with only outside money. Is there a presumption that there is a role for an active planner? For this question, the simple case of money holdings in the set {0,1} is misleading. So let’s think about general money holdings. It should be evident that dispersion of money holdings is not a good thing in this model. In general, if the consumer has small money holdings, then it will be impossible to get an unmonitored producer to produce much for such a consumer. And, as we have seen in the example above, it will be difficult to get even monitored producers with large money holdings to produce much. Hence, it would seem desirable for the planner to transfer money from those monitored people with large holdings and to transfer money to those with small holdings. Of course, those who are asked to give up money have to be willing to do so because they have the option to defect.4 A scheme of such transfers is an insurance arrangement. One of the things sacrificed by a monitored person who defects is the right to continue in it. Another is participation as a monitored consumer in meetings with monitored producers; in such meetings, an optimal arrangement will tend to have output be less dependent on the consumer’s money holdings if the consumer is monitored than if the consumer is not monitored. It, too, is a kind of insurance. Obviously, the binding constraints for transfer schemes arise when taking money from monitored people. One way to avoid those constraints is to inflate. An extreme is to give only non-negative transfers to monitored people and to make them a decreasing function of the wealth of monitored people. That will shift purchasing power towards the monitored people with little money. Of course, that will also produce a falling value of money, which, itself, tends to have undesirable effects because it tightens participation constraints. It should be emphasized, by the way, as in our earlier papers, that inflation and deflation are not the only ways to produce non-zero returns on money in this model. Even in the simple case of {0,1} money holdings, there is no reason why output in meetings should not depend on the monitoring status of the participants. In particular, a positive average return on money for unmonitored people can be achieved by having a monitored consumer get less in a meeting with an unmonitored producer than does an unmonitored consumer in any single-coincidence meeting. Is a market among monitored people a perfect substitute for activity by a planner? This question seems particularly relevant in our model because there are no aggregate shocks in the model. In the model, the market would be one in which people are insured against

84

R. Cavalcanti and N. Wallace

the kind of pairwise meetings they experience. Moreover, the market would have to be subject to participation constraints because individuals can defect. One possible specification is a special case of the formulation in Kehoe and Levine (1993), a competitive formulation in which each person faces a budget set of the usual sort and the person’s own participation constraints: constraints that are common knowledge. And, obviously, the market would be subject to a feasibility constraint on total money holdings, a constraint that the planner does not have to satisfy. (By the way, the scheme of transfers of outside money described above in the case of {0,1} money holdings could be accomplished by a market with the following trades: each monitored person who ends up after pairwise trade with two units of money willingly surrenders one unit in the market (the second unit would violate the bound) and each monitored person who ends up with zero acquires one unit.) One way to think of a market is as a constraint on what the planner can accomplish (Hammond, 1987). This view of a market seems to be the same as imposing the stronger requirement on allocations that a group does not want to defect to anything that is feasible for the group. A surmise is such, that under a more stringent notion of implementability, the only advantage of a planner over a market is the planner’s freedom to change the total amount of outside money.

Generalizations of the model Given that we have done little but pose questions of the simple model that we have set out, it seems gratuitous for us to suggest generalizations of the model. However, showing that the model lends itself in a straightforward way to various generalizations is part of its attractiveness. Imperfect monitoring is, of course, consistent with having people experience private-information shocks to preferences. One extreme version of such shocks was described in Cavalcanti and Wallace (1999a). There, we assumed that people receive at each date a private-information realization that determines whether or not they can produce at that date. The presence of such a shock has essentially no consequences for how we describe the unmonitored people because they cannot gain by misrepresenting their realization. For monitored people, in contrast, such shocks introduce into the model the kind of truth-telling constraints in Green (1987). One of the consequences is to make the planner’s dealings with monitored people dependent on individual histories. The model above has the simplest timing consistent with uncertainty about spending opportunities. Obviously, there are many alternatives that would retain that feature. And nothing in the model is inconsistent with aggregate shocks or with something like a deterministic seasonal. The imperfect monitoring we have assumed is very special. A troubling aspect of imperfect monitoring is that there are innumerable ways of specifying it. A lag in updating each person’s history is adopted in Kocherlakota and Wallace (1998). Such a lag is applied to the monitored people of the model above in Mills (2001). And, although they do not attempt a mechanism-design

New models of old (?) payment questions

85

analysis, implicit in Cavalcanti et al. (1999) is the assumption that the planner’s only information about issuers of inside money comes from the money that shows up in a clearing house run by the planner. In some respects, the crucial assumptions we have made are about recognizability. We have assumed that outside money is uniform, but that private monies can be distinguished according to the issuer. Missing from the model is the notion that uniformity of money is desirable.

Concluding remarks This volume is about the future of payments and the challenges that that future poses for central banks. We have focused on seemingly old questions: is private money useful? In the absence of private money, is there a role for a central bank discount window over and above what a federal funds market could accomplish? Our model hints at affirmative answers to both questions. Moreover, the model seems relevant for some new questions. Is management of central-bank, intra-day credit a new question or is it a version of the question about a role for a discount window as we have posed it? That depends in part on whether it is sensible to think of intra-day credit as being extended to perfectly monitored agents who have a demand for it because of their dealings with strangers. And what sort of model of a cashless economy should we focus on? Presumably, the relevant cashless economy should be a limit of a cash economy as cash becomes less important. Because we like the ideas we described at the outset that explain why cash rather than IOUs are used, we are inclined to use such a model as our model of a cash economy. But what sort of limit should we take? In such a model, we can get a cashless economy in one of two ways: we can let the ability of individuals to commit to future actions get perfect or we can let monitoring get perfect. To us, the choice is clear. We should let monitoring get perfect; after all, that is what improved information technology makes possible. This has an immediate implication: the limiting cashless economy is not an Arrow–Debreu economy. We have suggested some ideas about how to deal with a fundamental issue in monetary theory: the margin between money and credit. And we think that those ideas are fruitful both for old questions about monetary systems and for new ones related to the future of payment systems.

Notes 1 We are indebted to Stacey Schreft of the Federal Bank of Kansas City and to John Moore of the University of Edinburgh for helpful comments on an earlier draft. 2 Distinctions among the money issued by subsets of monitored people are discussed in Wallace (2003). 3 In a sense, excluding the unmonitored people from meeting the planner and others after pairwise meetings is without loss of generality. Because unmonitored people can hide money, the planner can at best give non-negative transfers to them that are weakly

86

R. Cavalcanti and N. Wallace

increasing in their money holdings. And even that can be regarded as problematic. How does the planner prevent the same unmonitored person from showing up many times at a date for a transfer? 4 There is a literature on matching models that avoids heterogeneity of money holdings. One route is the so-called large family model (Shi, 1997). Another is the device introduced by Lagos and Wright (2005): quasi-linear preferences in a good that is traded in a centralized market. These models have two limitations. First, the assumptions that do away with the heterogeneity are special. Second, the possible role of policy in dealing with heterogeneity is lost.

References Cavalcanti, R. and Wallace, N. (1999a) ‘A model of private banknote issue’, Review of Economic Dynamics, 2: 104–36. Cavalcanti, R. and Wallace, N. (1999b) ‘Inside and outside money as alternative media of exchange’, Journal of Money Credit and Banking, 31(2): 443–57. Cavalcanti, R., Erosa, A., and Temzelides, T. (1999) ‘Private money and reserve management in a random matching model’, Journal of Political Economy, 107: 929–45. Diamond, D. and Dybvig, P. (1983) ‘Banks runs, deposit insurance, and liquidity’, Journal of Political Economy, 91: 401–19. Green, E.J. (1987) ‘Lending and the smoothing of uninsurable income’, in E.C. Prescott and N. Wallace (eds) Contractual Arrangements for Intertemporal Trade, Minneapolis, MN: University of Minnesota Press. Hammond, P. (1987) ‘Markets as constraints: multilateral incentive compatibility in continuum economies’, Review of Economic Studies, 54: 399–412. Kehoe, T. and Levine, D. (1993) ‘Debt-constrained asset markets’, Review of Economic Studies, 60: 865–88. Kocherlakota, N. (1998) ‘Money is memory’, Journal of Economic Theory, 81: 232–51. Kocherlakota, N. and Wallace, N. (1998) ‘Optimal allocations with incomplete recordkeeping and no commitment’, Journal of Economic Theory, 81: 272–89. Lagos, R. and Wright, R. (2005) ‘A unified framework for monetary theory and policy analysis’, Journal of Political Economy, 113: 463–84. Mills, D.C. (2001) Outside and Inside Money: A Mechanism Design Approach, PhD dissertation, Pennsylvania State University. Ostroy, J. (1973) ‘The informational efficiency of monetary exchange’, American Economic Review, 63: 597–610. Patinkin, D. (1951) ‘The invalidity of classical monetary theory’, Econometrica, 19: 134–51. Shi, S. (1995) ‘Money and prices: a model of search and bargaining’, Journal of Economic Theory, 67: 467–98. Shi, S. (1997) ‘A divisible search model of money’, Econometrica, 65: 75–102. Trejos, A. and Wright, R. (1995) ‘Search, bargaining, money and prices’, Journal of Political Economy, 103: 118–41. Townsend, R. (1989) ‘Currency and credit in a private information economy’, Journal of Political Economy, 97: 1323–44. Wallace, N. (2003) ‘Commentary’ in D. Altig and B. Smith (eds) Evolution and Procedures in Central Banking, Cambridge, England: Cambridge University Press. Wallace, N. (2005) ‘From private banking to central banking: ingredients of a welfare analysis’, International Economic Review, 46: 619–36.

6

Optimal settlement rules for payment systems Benjamin Lester, Stephen Millard and Matthew Willison1

Introduction Payments are transfers of value between agents. For all payments that are not made in cash, finalization of payment occurs separately to the exchange of goods and will involve a payment system: a specification for when and how the actual funds are delivered consisting of a settlement asset, credit arrangements, infrastructure and rules. Indeed, Zhou (2000) defines a payment system as a ‘contractual and operational arrangement that banks and other financial institutions use to transfer funds to each other’. Such systems support a vast amount of economic activity. For example, on an average day in 2005 CHAPS Sterling, the United Kingdom’s largevalue payment system, processed about 120,000 transactions with a total value of around £210 billion, about 20 per cent of the United Kingdom’s annual gross domestic product.2 Given this, problems in a payment system could affect the functioning of the financial system and in turn the wider economy. As part of their role in ensuring the stability of their financial systems, central banks ‘oversee’ a number of payment systems with the goal of assessing and, if necessary, reducing the amount of risk that they bring to the financial system.3 Historically, interbank payments have been settled via end-of-day deferred net settlement (DNS) systems. As the volume and value of interbank payments passing through such systems increased rapidly in the 1980s and 1990s, central banks became increasingly concerned about the risk that stemmed from such systems. In particular, where payments are credited to customer accounts before being finally settled, credit exposures can build up and a failure of one participant in the system can then lead to the failure of other participants in the system. Fry et al. (1999) report that, at the same time as these exposures were becoming larger, advances in IT meant that it became increasingly technologically feasible to settle payments gross and in real time. Since doing this eliminates credit risk from a payment system, central banks increasingly favoured real-time gross settlement (RTGS) as the settlement rule within their countries’ large-value payment systems. In particular, in 1995 Switzerland and the United States were the only major countries relying on RTGS systems for their large-value payments. The Bank of Japan, which had offered both DNS and RTGS systems, switched to only offering RTGS in the late 1990s; in the United Kingdom,

88

B. Lester et al.

CHAPS switched to settling payments on an RTGS basis in 1996; and Euro area central-bank-administered, wholesale, systems have operated as RTGS systems since 1997.4 But, RTGS systems can be more costly than DNS systems. In addition, to the higher IT costs involved in setting up and running such systems, RTGS systems are ‘liquidity hungry’ relative to DNS systems. That is, participant banks require more liquidity to settle their payments in an RTGS system than in a DNS system. In turn, this liquidity is costly as when banks do not have it to hand, they will need to borrow. To a degree, central banks can mitigate this cost by providing intraday liquidity at low cost – typically free, so long as it is collateralized – but, even then, this liquidity will still carry an opportunity cost. So, it is not at all clear that moving from a DNS to an RTGS payment system is necessarily welfare improving. Indeed, in a comparison of the costs of secured net settlement on CHIPS – at the time a DNS system – to those of an otherwise equivalent RTGS system, Schoenmaker (1995) concludes that ‘the estimated extra cost of RTGS exceeds the estimated reduction in settlement risk’.5 Furthermore, George Selgin, in his chapter in this volume, argues that the ‘credit risk’ between banks in DNS systems used as a justification for imposing RTGS does not exist. He argues that this is because a bank is only exposed to the risk of another bank failing to meet a net obligation in a DNS system if it credits customer accounts before settlement occurs, something it does not actually have to do. In addition, such customer credits can typically be reversed in the event of settlement not taking place. Therefore, Selgin argues, all agents involved face the right incentives to manage these risks and there is no market failure in payment systems. As a result, he suggests that the imposition of RTGS, where the market had settled on a DNS system, must be welfare reducing. The purpose of this chapter is to construct a model within which we can begin to explore the trade-off between cost and risk in payment systems. The model shows that under certain assumptions, DNS and RTGS can both be a payment system’s settlement rule in equilibrium. The presence of multiple equilibria opens up the possibility that private agents fail to coordinate on the optimal equilibrium settlement rule. However, we stop short of claiming that there can be a case for government intervention to help coordinate private agents to choose the optimal rule. Rather, the purpose of the chapter is to highlight some of the features of the economy that play an important part in determining a payment system’s settlement rule. Relaxing some of the restrictive assumptions we make remains avenues for future research. We consider a banking economy in the spirit of He et al. (2005), ignoring theft as a motive for banking and instead focussing on the case of interestbearing deposits. Moreover, we introduce into the He et al. model two possible payment systems, and endogenize the choice of payment system for both buyers and sellers. Incorporating this creates a framework within which we can analyse the given payment system as an equilibrium outcome of the economy. In particular, in our model the end recipients of payments placed through a DNS system are exposed directly to the possibility of default by the banks of the

Optimal settlement rules for payment systems

89

payees. We make three key assumptions in the model. The first is that prices are exogenously fixed. One good always transfers for one unit of money. The implication of this is that settlement rules can have different effects on buyers and sellers. The second assumption is that banks are unable to charge buyers for the cost of RTGS. Third, sellers cannot offer incentives to buyers to switch settlement rules. This assumption partly derives from the first – sellers cannot use prices to induce buyers to switch settlement rules – but goes further by prohibiting sellers from using other strategies to get buyers to switch; e.g. sellers cannot base the quality of the good on the settlement rule. We start by considering an economy in which a DNS equilibrium exists and first show that if the costs of using an RTGS system are too high, then there will not be an equilibrium in which agents use an RTGS system. We think of this economy as representative of a time when IT had not developed to the extent to allow RTGS to occur at an economical price. We also show that if these costs become low enough an RTGS equilibrium will exist, in addition to the DNS equilibrium. We think of this economy as representative of developed economies in the early 1990s. Finally, we show that if the costs of RTGS are low enough relative to the costs of DNS, under the assumptions we make, the equilibrium in which all agents use the RTGS system generates a higher value of social welfare than that in which all agents use a DNS system.6 Since our model cannot handle dynamics and, in particular, the endogenous decision to move from one equilibrium to another, we cannot handle the question of which equilibrium – RTGS or DNS – will be selected by agents in the economy. We would argue that historically, DNS was the only system that could be used in equilibrium and that we have now moved to a situation in which either a DNS or an RTGS system could be used in equilibrium. An analysis of how private agents coordinate on one of the equilibria and whether they can overcome potential coordination failures is left to subsequent research. This research would help us better understand the role of public authorities in the move towards RTGS that occurred in the early to late 1990s in many countries. The chapter is structured as follows. We first outline our model before discussing equilibria within it. We then compare welfare over regions of the parameter space within which both DNS and RTGS equilibria exist showing that, under the assumptions we make, we can find a critical value for the costs of the RTGS system below which the RTGS equilibrium welfare dominates the DNS equilibrium. Finally, we conclude with some suggestions for future work.

The model We begin by describing a simple random matching model of money that we use as a platform for our analysis. In the economy there is a unit continuum of infinitely-lived agents. A proportion M [0,1] of agents are each endowed with one indivisible unit of fiat money. Agents produce and consume indivisible goods. In each period, agents are randomly and anonymously matched with one another. In any pair-wise meeting, a double coincidence of wants occurs with

90

B. Lester et al.

zero probability. Single coincidence meetings, where one agent wants the other’s good but not vice versa, occur with probability, x. Given the absence of any double coincidence of wants, the only feasible trades involve the exchange of one unit of money for one good. In any single coincidence meeting, the buyer receives utility, u while the seller incurs cost of production, c. Banks We build on this simple random matching model of money by allowing agents either to hold money in the form of cash or to deposit it in a bank. Banks are modelled in the same way as in He et al. (2005) and, in particular, are perfectly competitive. An agent deposits his money at a bank with probability,  and pays a fee, , which is derived from the setup of the banking sector. Banks face a fixed cost, a, for managing each account. Banks also make loans, L, to agents without money but must retain a fraction  of deposits (D) as reserves. We denote the measure of agents holding their money as cash as M0 and the measure of agents holding either cash or having a bank account as M1 (total money supply). So L + M = M1 and D + M0 = M1. Banks charge an upfront fee, , for loans. Settlement Whether an agent chooses to hold his money in the form of cash or a bank deposit it has implications for when money is transferred between buyers and sellers. To capture this we assume that each period is divided up into two subperiods. Trade between agents takes place during the first sub-period, which we refer to as the morning. If a buyer uses cash, money is transferred between the buyer and the seller in the morning since cash changes hands at the point of sale. If a buyer makes a payment to a seller from his bank account, the money has to be transferred via an inter-bank payment system.7 There is a single inter-bank payment system in the economy. When inter-bank payments are received depends on the rule in place governing how payments are settled. Payments are settled in the morning if they are made through a real-time gross settlement (RTGS) payment system and are settled in the afternoon if they are made through a deferred net settlement (DNS) payment system. The timing of the settlement process would be irrelevant if there is perfect commitment among banks, as there is in He et al. (2005). Agents would simply choose the cheaper of the two settlement rules as money is transferred from the buyer’s bank to the seller’s bank before the next trading sub-period with certainty. However, the timing of the settlement process is crucial if there is a possibility of bank insolvency between when trading occurs and when the transfers of money are completed. Sellers may fail to receive funds altogether or receive them only at a cost if a bank becomes insolvent before the settlement of payment occurs. We introduce the possibility of bank default into the model by assuming that there is an exogenous risk that each bank could become insolvent and consequently default on any outstanding payment obligations it may have. Banks

Optimal settlement rules for payment systems

91

can become insolvent between the morning and the afternoon in every period. We assume that there is full deposit insurance implying that depositors face no costs as a result of bank default. In the following morning, their money is transferred to accounts at new banks that replace the insolvent institutions. Any default situation is resolved and sellers receive their money but only at a cost. The expected per payment cost incurred by a seller is . This cost represents a deadweight loss to the economy and can be thought of as being the real resource cost of winding up insolvent institutions and sorting out the residual claims on it (i.e. ‘bankruptcy’ costs). Sellers are not exposed to default risk when they receive an RTGS payment because they receive a payment before buyers’ banks can become insolvent. But reducing the lag between trade and settlement comes at a cost. First, there exist bureaucratic costs to settling every transaction more promptly. Moreover, a higher frequency of settlements requires banks to hold larger amounts of idle reserves, thereby decreasing the revenue earned per unit of money deposited.8 Finally, some banks will also be forced to borrow from the monetary authority or enter the interbank market in order to obtain sufficient reserves to complete settlement of all of their customers’ transactions. These additional costs will result in customers facing higher fees in a perfectly competitive banking sector. The cost of making RTGS payments is captured by a per-period cost  that is levied on all bank accounts in proportion to the amount of RTGS payments made. In a single coincidence meeting, the buyer proposes whether he wishes to pay in cash or in the form of a payment from his bank account. The seller then chooses between accepting a payment through these means and not trading. Trade does not take place if they cannot agree. Trade always occurs when a buyer offers to pay in cash because a seller incurs no cost and is exposed to no risk when receiving cash. If a buyer offers to make a payment from his bank account and the interbank payment system is RTGS, the seller accepts with probability one since there is no cost or risk to him from receiving a payment this way. If the interbank payment system is DNS, a seller accepts a payment with probability . Let = 1(0) if the interbank payment system is DNS (RTGS). Therefore, trade takes place in a single coincidence meeting, when the buyer wishes to make a payment from his bank account, with probability P( ,) = (1 – ) + . It also follows that the cost to a bank of providing payment services is K( ) = (1 – ). This implies that the per-period fee that each (perfectly competitive) bank charges a customer for providing deposit services is ( ) = a – (1 – )rp + (1 – ) where r is the discount rate. Bellman equations We can now derive the Bellman equations for the value of being a buyer holding cash, V1m, the value of being a buyer who has a bank deposit, V1d, and the value of being a seller, V0. These are shown in equations (1)–(3).

92

B. Lester et al. rV0 = (M0x + (M1 – M0)xP( ,))(V1 – V0 – c) – (M1 – M0)x 

(1)

rV1m = (1 – M1)x(u + V0 – V1) + V1 – V1m

(2)

rV1d = (1 – M1)xP( ,)(u + V0 – V1) + V1 – V1d – (1 + r)( )

(3)

where V1 = max {V1m,V1d} is the value of being a buyer prior to the decision of whether or not to deposit his money in the bank. The first component of equation (1) is the gains from trade enjoyed by a seller multiplied by the probability of meeting someone to trade with in any given period. The second component is the expected cost of failing to receive a DNS payment times the probability of accepting a DNS payment. The first part of equation (2) is the gains from trade enjoyed by a buyer times the probability of trading in a given period when he offers to make payments in cash. The first component of equation (3) is the same except that it represents the gains from trade times the probability of trading when a buyer deposits his money in a bank and makes payments from his account. The second component of equation (3) is the fee paid by a buyer for depositing his money in the bank.

Equilibrium analysis In this section we examine pure-strategy equilibria in our model in which all money is deposited in banks; i.e. V1 = V1d, M0 = 0 and M1 = M/. Therefore, a buyer’s Bellman equation is: rV1 = (1 – M1)xP( ,)(u + V0 – V1) – (1 + r)( )

(4)

We derive conditions for an equilibrium in which trade occurs with only a DNS payment system and one in which trade occurs with only an RTGS payment system. We then compare the conditions under which each type of equilibrium exists. Such equilibria will exist if the following conditions hold: Individual rationality: the value of being a seller or a buyer is at least as good as leaving the market and living in autarky: V0 0 and V1 0. Incentive compatibility: a seller has an incentive to produce his good in exchange for one unit of money and a buyer has an incentive to trade his unit of money for one unit of the good: V1 – V0 c +  and u V1 – V0. Banking constraint: agents choose to deposit money in a bank: ( )  0. The probability that trade occurs equals one in any single coincidence meeting when we restrict attention to pure-strategy equilibria. Thus, the only difference between holding money as cash and depositing it in the bank is the banking fee. Buyers have incentives to deposit money in banks as long as they receive a return from doing so; that is, the banking fee is negative. Hence, ( )  0 is both necessary and sufficient for agents to choose to deposit money. The buyer’s individual rationality constraint holds if the seller’s individual

Optimal settlement rules for payment systems

93

rationality constraint and the buyer’s incentive compatibility constraints are both satisfied. Thus, we only need to derive the conditions under which V0 0, V1 – V0 c + , u V1 – V0 and ( )  0 hold. Now, it is not feasible for all agents without money to borrow from a bank because then there would be no sellers. So, in equilibrium, the loan market clears at a rate at which sellers are indifferent between remaining a seller and borrowing. As stated in He et al. (2005), they will be indifferent when the cost of borrowing, , is equal to the difference between the value of being a buyer and the value of being a seller, V1 – V0. Using this result, equations (1) and (4) then imply the following expressions for the values of being a buyer and of being a seller, respectively: M1xP( ,)[(1 – M1)xP( ,)u – (1 + r)(a + (1 – ))] rV0 = r((1 + r) – r) + xP( ,) M1x[(1 – M1)xP( ,) + r((1 + r) – r)][P( ,)c + ] – r((1 + r) – r) + xP( ,)

(5)

[r + M1xP( ,)][(1 – M1)xP( ,)u – (1 + r)(a + (1 – ))] rV1 = r((1 + r) – r) + xP( ,) M1x[(1 – M1)xP( ,) – r(1 + r)(1 – )][P( ,)c + ] – r((1 + r) – r) + xP( ,)

(6)

We use these equations to derive ranges of parameter values for which there exists an equilibrium in which there is trade when there is only a DNS payment system ( =  = 1) and when there is only an RTGS payment system ( = 0). The ranges of values of x and a for which there exist such equilibria are illustrated in Figures 6.1–6.4. In each figure, the dashed line represents the seller’s individual rationality and incentive compatibility conditions. These conditions hold for values of x that are sufficiently high, relative to the value of a, that (x,a) lies to the right of the dashed line. This is because the value of being a seller is increasing in the probability of meeting an agent with whom to trade (x) but decreasing in the cost of operating a banking account (a). The dashed line is upward sloping since when a rises, sellers must be compensated by an increase in x for their individual rationality and incentive compatibility conditions to still hold. The solid line represents the banking constraint. This condition is met for values of x that are sufficiently high relative to the value of a because the banking fee is only negative if the cost of operating an account (a) is low enough compared with the loan market rate (). The loan market rate is increasing in the gain from moving from being a seller to a buyer, which is obviously increasing in the probability of trading (x). It follows that the solid line is upward sloping.9 The grey, highlighted, areas show possible equilibria.

94

B. Lester et al.

Figure 6.1 shows the DNS equilibrium. RTGS equilibria are shown in Figures 6.2–6.4. The figures show that the range of parameter values for which there exists an RTGS equilibrium grows smaller the more costly is the RTGS (i.e. the higher is ). When  is high enough no RTGS equilibrium will exist (see Figure 6.4). Historically,  was very high and hence, trade was only possible if payments settled on a deferred net basis. It is likely that the value of  has fallen over time 0.03 a

0.02

0.01

0

0.1

x

0 0.2

Figure 6.1 DNS equilibrium.

0.03 a

0.02

0.01

0

0.1

Figure 6.2 RTGS equilibrium ( = 0).

x

0 0.2

Optimal settlement rules for payment systems

95

0.03 a

0.02

0.01

0

0.1

x

0 0.2

Figure 6.3 RTGS equilibrium ( = 0.005).

0.03 a

0.02

0.01

0

0.1

x

0 0.2

Figure 6.4 RTGS equilibrium. Note: Figures 6.1 to 6.4 are drawn for M = 0.5,  = 0.7, u = 5, c = 2,  = 0.003, r = 0.02.

(say, with improvements in information technology). This means that an equilibrium in which payments settle on an RTGS basis is now possible, in addition to one in which they settle on a DNS basis (Figures 6.2 and 6.3). Further reductions in  would expand the range of parameter values for which there exist both RTGS and DNS equilibria.

96

B. Lester et al.

Welfare In this section we will analyse the level of social welfare under the two settlement rules. We identify the (positive) values of  and  under which one of the settlement rules has superior welfare properties under the particular assumptions we have made about fixed prices and the difference between buyers and sellers in terms of how they incur the costs of DNS. Social welfare, W, is defined as the average of the value of being a buyer and the value of being a seller: W = M1V1 + (1 – M1)V0

(7)

The social planner chooses whether payments settle on a DNS or RTGS basis to maximize W subject to agents’ individual rationality constraints, incentive compatibility constraints and banking constraint. Using equations (5) and (6) and setting both and  to unity, we can show a DNS equilibrium will exist if  [1,2], where  1 ensures that the banking constraint holds and  = 2 ensures that the seller’s individual rationality and incentive compatibility conditions hold. The values of 1 and 2 are (r + x)a – rx(1 – )[(1 – M1)u + M1c] 1 = rM1x(1 – ) and (1 – M1)xu – (1 + r)a 2 = – c. (1 – M1)x + r((1 + r) – r) Similarly, using equations (5) and (6) and setting to zero, we can show that an RTGS equilibrium will exist if   min {1,2}, where   1 ensures that the banking constraint holds and   2 ensures that the seller’s individual rationality and incentive compatibility conditions hold. The values of 1 and 2 are rx(1 – )[(1 – M1)u + M1c] 1 = – a r+x and (1 – M1)xu – [(1 – M1)x + r((1 + r) – r]c 2 = –a. 1+r It is straightforward to show that (1 – M1)x 1 (<)2 if u  (>)c. r + (1 – M1)x

Optimal settlement rules for payment systems

97



RTGS

Autarchy

DNS and RTGS

DNS

2

Min {1, 2}



Figure 6.5 Existence of DNS and RTGS equilibria.

When c is relatively low, the relevant constraint on  is the banking constraint; the cost of settlement must be sufficiently small so that banks can continue to pay interest on deposits. Alternatively, when c is relatively high, the relevant constraint are the seller’s incentive constraints because  must be sufficiently small that a seller finds it profitable to produce and trade. Before comparing welfare levels under either settlement rule, note that we can ignore the constraint  1 because of the fact that 1 0 implies 1  0. In other words, if the banking constraint can be satisfied under RTGS for a positive value of  it will definitely be satisfied (i.e. not bind) under DNS. The banking constraint is satisfied under RTGS as long as the gains from trade enjoyed by a seller are sufficiently high relative to the cost of RTGS. Under DNS, there is no cost from using this settlement rule and there is also an additional benefit to the seller of not facing the cost  in the following period. It follows that if the banking constraint can be satisfied under RTGS the banking constraint will not bind under DNS. Figure 6.5 depicts the different equilibria that can occur for different values of  and . Both the DNS and the RTGS equilibria are possible when the costs of default and the costs of making payments on an RTGS basis are low enough that   2 and   min {1,2}. When  exceeds 2, but   min {1,2}, there exists an RTGS equilibrium but no DNS equilibrium. There is a DNS equilibrium but no RTGS equilibrium when   2 and  > min {1,2}. Finally, if both costs are sufficiently high, no DNS or RTGS equilibrium will exist. When either a DNS equilibrium or an RTGS equilibrium could exist, welfare would not be unambiguously higher in one case than the other. We can use equations (5), (6) and (7) to calculate welfare under each of the two settlement rules, again setting both and  to unity in the DNS case and setting to zero in the RTGS case. Welfare is the same under either settlement rule if

98

B. Lester et al.  L

2 A B

Min {1, 2}



Figure 6.6 Welfare.

(1 + r)(r + x)  = . x[r((1 + r) – r) + (1 – M1)x – rM1] When  = 0 (that is, RTGS is costless) welfare is the same under both settlement rules only if DNS is also free of cost ( = 0). Thus, the locus values of  and  for which welfare is the same under either settlement rule passes through the origin and divides the area in which both equilibria exist into two.10 Welfare is higher under RTGS above the locus (region A) and is higher under DNS below the locus (region B). The locus is depicted in Figure 6.6. In summary, when trade could occur under either settlement rule in equilibrium, it may be possible to rank the equilibria in terms of welfare. But without modelling how agents select settlement rules, the analysis does not indicate whether they could coordinate on the equilibrium providing the highest welfare.

Conclusions and future work In this chapter we have constructed a model for examining the trade-off between cost and risk in DNS and RTGS payment systems. The model showed that when the costs of settling payments on an RTGS basis are high, only a DNS equilibrium can exist. The opposite is true when the costs of default are high: only an RTGS equilibrium can exist. Either settlement rule may hold in equilibrium for intermediate values of costs. The model as it stands makes several restrictive assumptions and so is not general enough to answer the question of whether agents can coordinate on a welfare-superior settlement rule and any role for public policy in preventing coordination failures. Nonetheless the model offers a starting point for developing a general model with which these questions can be tackled.

Optimal settlement rules for payment systems

99

Notes 1 The views expressed in this chapter are those of the authors and not necessarily those of the Bank of England. 2 Bank of England (2006). 3 Bank for International Settlements (2005). For more on the Bank of England’s roles and responsibilities in the area of payment systems, see Bank of England (2005). 4 The chapter by Morten Bech in this volume discusses the diffusion of RTGS systems across the world. 5 The Clearing House Interbank Payment System (CHIPS) is a private payment system operated by the New York Clearing House. At the time of Schoenmaker’s work (1995) it was a DNS system; as of 2001 settlement happens almost continuously. 6 While in this chapter we assume the existence of a cost–risk trade-off between RTGS and DNS, this is derived from first principles in Lester (2005). 7 We assume that although a seller may not already have a bank account when a buyer offers to make a payment from his bank account, he will open an account if he agrees to receive money in this way. The seller pays a fee for holding a bank account in the following period when he is a buyer. 8 Taken literally, this would suggest that  would be higher if the payment system in operation were RTGS than if it were DNS. But, to keep the model simple while making our point, we lump this effect in  and leave  exogenous. 9 For the parameter values we use, the buyer’s incentive compatibility condition holds for all values of x and a. 10 The locus is upward sloping because the value of being a buyer (V1) is decreasing in the production cost (c) and the cost of default ().

References Bank for International Settlements (2005) New developments in large-value payment systems, Committee on Payment and Settlement Systems Publication No. 67. Bank of England (2005) Payment systems oversight report 2004, London: Bank of England. Bank of England (2006) Payment systems oversight report 2005, London: Bank of England. Fry, M.J., Kilato, I., Roger, S., Senderowicz, K., Sheppard, D., Solis, F. and Trundle, J. (1999) Payment systems in global perspective, London: Routledge. He, P., Huang, L. and Wright, R. (2005) ‘Money and banking in search equilibrium’, International Economic Review, 46: 637–70. Lester, B. (2005) ‘A model of interbank settlement’, unpublished thesis, University of Pennsylvania. Schoenmaker, D. (1995) ‘A comparison of alternative interbank settlement systems’, London School of Economics Financial Markets Group Special Paper, No. 204. Zhou, R (2000), ‘Understanding intraday credit in large-value payment systems’, Federal Reserve Bank of Chicago Economic Perspectives, 24(3): 29–44.

7

The microstructure of money James McAndrews1

Introduction In his chapter in this volume, Ed Green suggested that ‘payment economics comprises the topics that pertain to both monetary economics and industrial organization’: loosely paraphrasing, the study of the industrial organization of money. In this chapter, I approach the study of payment systems as the study of the market microstructure of money. In doing so, I will also use the lens of market microstructure to address issues relating to trends in payment systems. The study of the microstructure of financial markets has focused on the institutional arrangements for the exchange of financial instruments. This study has examined the efficiency of different market structures. Various measures of efficiency including the size of the bid–ask spread and the speed of execution have been explored. In addition, the conditions under which one market structure or another is more appropriate, given the underlying economic environment, has been a question of interest. Payment systems are analogous to financial markets. In this chapter I will explore this analogy in depth. I will examine first the analogies between alternative arrangements for the exchange of financial contracts and for money. The primary focus is on payment systems in which only money (or deposit balances) is exchanged. A key difference between payment systems and financial exchanges is that the price is held fixed in payment systems (at least at first glance). Nonetheless, the organization of payment systems has been examined in theory by many authors who argue that the behavior of participants differs in payment systems of differing designs. Given these analogies we can apply the microstructure literature of financial markets to that of payment systems. This application yields various insights into the conditions in which one payment system design is more appropriate than another based on the volatilities of payment sizes, arrival rates, and likelihood of offsetting other payments. A second line of application of microstructure literature focuses on the priced aspects of payment systems. This line of thought suggests that large-value payment systems provide the settlement system for the overnight money market. As a result, the microstructure of the payment system and the money market are

The microstructure of money

101

linked. I explore this application by presenting empirical estimates of an inventory model of dealer behavior in the fed funds market to a sample of large US banks. The null hypothesis of no inventory effects, as would be the case in a frictionless market, would suggest that the current balance of the bank would not affect the price quotes of the bank. Such a hypothesis would be consistent with a risk neutral dealer that had no limits to the balance it could hold in the central bank. I find some evidence of inventory effects in the US federal funds market. The examination of the empirical results of the inventory model leads to a discussion of how the design of the payment system affects the overnight money market. A review of the features of payment systems suggest that daylight credit policies, the design of the payment system as an RTGS, a DNS, or a queueaugmented system, the importance of links to ancillary payment systems, discount window policies, and reserve accounting rules all affect both the precautionary demand for overnight balances and the tightness, or elasticities of demand, of the money market. This discussion suggests new methods of examining payment system policy and its effects in the money market. The chapter concludes with a discussion of the role of payment systems in the money market. In addition, I discuss recent trends in payment systems, how the discussion of microstructure can assist us in understanding those trends and what they portend for payment system developments.

Designs for payment systems and security markets The design of large-value payment systems has been closely examined in Bank for International Settlements (1989, 1990, 1997, 2005). The alternative designs for large-value payment systems can be roughly characterized as follows. A netting system, often referred to as a designated-time net settlement system (DNS), cumulates payment orders until a specified time. At the designated time a settlement agent calculates the multilateral net amount that each participant owes or is owed. The net debtors are required to send funds in the amount of their net debts, and upon receipt of these funds, the net creditors are paid out the amounts owed to them.2 A real-time gross settlement system (RTGS), in contrast, settles each payment order against balances on account of the settlement institution. Each payment order is settled, if possible, on arrival and in the full amount of the payment order.3 Recently, a number of systems that are not easily classified as either DNS or RTGS have been put into operation. These systems, such as the ‘new CHIPS’ system in the United States and ‘RTGS plus’ in Germany, are not easily characterized. One aspect of these systems, perhaps easiest to understand in the RTGS plus system, is that participants can specify a priority for each payment order. An express payment is meant to be settled as in an RTGS system: immediately against balances. A limit payment is intended to be settled in a liquidity savings mode. The participants can establish the limits and they can be limits on the total of payments to be settled in that mode, and the bilateral or the multilateral net amount settled in that mode. The purpose of the limits is

102

J. McAndrews to reserve liquidity exclusively for the execution of urgent payments (Express payments); to prevent one-sided dissipation of liquidity in comparison with individual participants; to synchronise the payment flow with other participants; and to promote its early submission.4

There are numerous possible variations on the possible designs of payment systems of this type, that is, payment systems that allow prioritizing payments by criteria other than the time of submission, and that allow payments to be settled out of a queue against incoming balances. This type of payment system has been called a ‘hybrid system’ but in this chapter I will refer to it as a ‘limit payment order’ system.5 Each of the designs for payments system has an analogue in the designs of security markets. A RTGS system is analogous to a continuous auction market, in which trades can be made throughout the market day. In continuous auction markets, trades arrive throughout the day, and specialists or dealers match buy and sell orders, concluding transactions quickly. In a continuous auction for a security, the price moves based on market conditions, while in a payment system only the quantity of deposit account balances moves.6 Payment systems typically settle trades for securities, debts, or commercial goods or services that are delivered outside of the payment system. For example, the title to a building or security is transferred in an office, and the banks representing the buyer and seller transfer the funds in an RTGS system. A DNS system is analogous to a call market, such as occurs at the opening of the New York Stock Exchange, in which orders are accumulated and trade occurs at a price that clears the market. In a DNS, payment orders are accumulated and netting of trades occurs. In call markets, liquidity is concentrated and various authors have posited reasons why call markets might be efficient in certain environments. Similarly, a DNS system concentrates funding liquidity, as offsetting payments reduce the amount of account balances that need to be transferred between parties. Finally, limit-order books and crossing markets accumulate trade orders on either side of the market at particular prices, and settle them when a market order or a new limit order enters the queue and ‘crosses’ or satisfies the limit price of the queued order. The analogy to payment systems is found in limit payment order systems in which payment orders are queued pending the satisfaction of the limit condition established by the participant. For example, the limit condition might be the arrival of funds into the participant’s account, such as an offsetting payment order from a bilateral counterparty. In this analogy, an express payment is like a market order, crossing the limit payment order, and allowing it to settle without pre-establishing account balances, as would be necessary in a pure RTGS system. Large-value payments are usually intended to settle debts of fixed nominal value. As such, much of the market microstructure literature of securities markets, which studies the efficiency of price-setting in markets of various designs, would not apply directly to payment systems. However, several aspects of the market microstructure literature do apply to payment systems.

The microstructure of money

103

The participants in payment systems are typically banks. In most payment systems, banks either have to post collateral or pay fees if they borrow during the day (from the central bank) to acquire more account balances. By the end of the day, the bank has to repay their intraday borrowing, borrow from the discount window (usually at a rate that exceeds the overnight rate in the market), or incur an overnight overdraft on which the bank pays a penalty rate. As a result, banks treat the balances that they transfer in payment systems as scarce. A number of market microstructure concerns are applicable to the analysis of payment systems. Before delving into the areas which have been investigated in the literature, I will contrast some of the fundamental assumptions in the market microstructure literature. Generally, in securities markets, traders are often modeled as being risk averse, or being subject to convex costs of trading; the basic risk facing participants in securities markets is market risk of inventory: the risk that the price of the security will change during the planning period; and securities markets are subject to relatively high bid–ask spreads. In contrast, in the money market, banks are usually modeled as risk-neutral, but subject to a convex cost function related to penalties to missing their reserve requirement (although even these might be only slightly convex on most days because of a high degree of substitutability of reserves across the days of a maintenance period); the basic risk they are subject to is liquidity risk, rather than market risk; and money markets typically display low bid–ask spreads. The market microstructure literature contrasts the outcomes of continuous and call markets, suggesting that call markets can be useful when liquidity is particularly scarce. Admanti and Pfleiderer (1988) and Pagano (1989) show that clustering of trades occurs even in continuous auctions. Because multiple equilibria can occur, however, call markets can assist in selecting a particular clustering equilibrium, minimizing the costs of trade. In the case of payment systems, this would translate into minimizing the costs of balances needed to settle the day’s payments. McAndrews and Rajan (2000) conduct an empirical investigation of clustering of payments in the US Fedwire RTGS system. Other studies of continuous versus call auctions focus on adverse selection effects of market design. Vayanos (1999) models a situation in which market participants with private information about their own risk sharing demands limit trades in a continuous auction to limit the market impact. This motive, and other similar motives, might result in payment system participants limiting the amount they wish to pay at one time in an RTGS, instead preferring to wait on the arrival of expected payments from their counterparties. Such motives are discussed in Bank for International Settlements (2005) and in Kahn et al. (2003), and may result in situations in which netting is preferable to RTGS. Freixas and Parigi (1998) consider interbank contagion effects as depositors seek to avoid bank failure in a gross or net payment system. Kahn and Roberds (1998) consider a rich model that reveals that the choice of payment system design creates alternative adverse selection and moral hazard incentives for banks. Another issue discussed by market microstructure theorists is the value of transparency in securities markets. Transparency has been shown theoretically to

104

J. McAndrews

reduce the problems of adverse selection in securities markets, as in Pagano and Roell (1996), although empirical results suggest mixed effects. The effect of greater transparency in payment systems has been examined in the paper by Willison (2005). A number of other issues, including the strategic supply of liquidity (Angelini 1998; Bech and Garratt 2003), intermarket competition and liquidity fragmentation are examined in the market microstructure literature, all of which can be applied to the study of payment systems. How does a bank manage to pay back its intraday borrowings from the central bank if it experiences a net outflow from its account as a result of payments its customers request to be made, or that it makes for itself? The bank can enter the overnight market and borrow funds from another bank. We now turn to a discussion of the market microstructure of the overnight market.

Payment systems and the money market Large-value payment systems provide the settlement infrastructure for the overnight money market in most countries. In the United States, federal funds market sales are settled by the seller of funds making a funds transfer on Fedwire, the Federal Reserve System’s large-value, RTGS payment system. The money market assists banks as they make payments. As Skeie (2004) points out in a model of banking and money, as banks make net outflows from their payment activity, they experience a demand to borrow on the overnight market to avoid overnight negative positions that would otherwise result. The recognition that payments activities of banks and their overnight borrowing are so closely linked suggests that the microstructure of the payment system influences activity on the money market. Consider first an RTGS payment system. In such a system, information about a bank’s balance evolves over the day as payments flow into its account and customers request that payments are made on their behalf. In this case, a bank may not have uncertainty about its balance resolved until very near the close of the payment system. The reserve accounting rules that govern the reserve requirements in the United States allow banks to meet their requirements by averaging reserve levels over a two-week reserve maintenance period.7 The ability to average shortages on one day with excess amounts on another day tends to increase the elasticity of a bank’s demand for balances late in the day (on an average day). This lessens the bank’s need to be extremely precise in its reserve management and maintain its reserves in an extremely tight band. However, because reserve balances do not earn interest, banks have an incentive to avoid holding excess reserves, and as shortfalls in reserve holdings incur penalties, banks wish to avoid shortfalls in reserves as well. In addition, banks’ balances should not fall below zero on any day. These requirements suggest that banks take care that their balances come close to the level that they are targeting. As they make and receive payments in the late afternoon, they manage their account balances to achieve their desired

The microstructure of money

105

end-of-day balance. These considerations suggest that inventories might play a role in the federal funds market. Modern microstructure theory has focused on informational differences in traders, but I will focus on inventory effects in the federal funds market as an example of the microstructure approach to payment systems and the money market. The federal funds market has been well described in the papers by Demilralp et al. (2004), Furfine (1999, 2003), Goodfriend and Whelpley (1993), Griffiths and Winters (1995), Hamilton (1996, 1997), Lee (2003), and Stigum (1990). These papers have established many regularities of the federal funds market. Inventory models One of the most basic models of market microstructure suggests that there are inventory effects in dealer markets. Garman (1976) examines a risk-neutral dealer market. In his model the stochastic flow of buy and sell orders is pricedependent. The dealer has an obligation to maintain continuous trading, and because the orders arrive stochastically, the dealer is motivated to carry an inventory. Garman goes on to examine an inventory-independent pricing policy. Amihud and Mendelson (1980) extend Garman’s result to derive the optimal inventory-dependent pricing policy by the monopoly dealer when it has constraints on its short and long position. They show that prices, both bid and ask, are monotonically decreasing in inventory. This result, and similar ones for risk-averse dealers (as in Stoll (1978) and Ho and Stoll (1981, 1983)) have led to various empirical tests of the theory. Hasbrouck and Sofianos (1993) and Madhavan and Smidt (1993) show mean reversion in specialist inventories; Lyons (1995) applies the inventory model to the foreign exchange market; Manaster and Mann (1996) find that market makers in the Chicago Mercantile Exchange with long positions tend to sell. Biais et al. (2004) review many other studies as well. The result that bid and ask prices are monotonically decreasing in inventories is often implemented by showing that the midpoint of the bid–ask spread of dealers is decreasing in the inventory of dealers. In what follows I apply this to the federal funds market. Sales in the federal funds market (again, excellent descriptions of the market are available in Demilralp et al. (2004) and in Furfine (1999)) are delivered on Fedwire. Federal funds trades can either be brokered or direct. Furfine (1999) examines the patterns of participant, timing, and concentration in the federal funds market. He finds that the largest five banks by asset size in 1998 accounted for 24 percent of industry assets, but an even greater share of both federal fund purchases and sales. Those banks accounted for 38 percent of federal funds sold and purchased 34 percent of federal funds bought (by value). Similarly the top ten banks in asset size sold 47 percent of the federal funds transactions, and bought 48 percent of the federal funds by value in the period. Taking advantage of this concentration in federal funds markets we collect a sample of likely federal funds transactions from the Fedwire transactions journal.8 Figure 7.1 displays the intraday pattern of identified federal funds

106

J. McAndrews

1,800 1,600

14 Volume Value

12

US$ (billions)

10 1,200 1,000

8

800

6

600 4

Volume (thousands)

1,400

400 2

200 0 12:30 a.m.

2:30 a.m.

4:30 a.m.

6:30 a.m.

8:30 a.m.

10:30 a.m.

12:30 p.m.

2:30 p.m.

4:30 p.m.

0 6:30 p.m.

Time

Figure 7.1 Intraday pattern of activity: total value and volume of federal funds traded (2 October 2001–31 September 2004) (source: Federal Reserve Bank of New York: author’s calculations). Note Values shown are the average values in a calendar minute calculated over the days of the sample period.

activity averaged over the sample period, from October 2001 through September 2004. The levels of the lines in the chart are per-minute averages for each minute of the Fedwire operating day, averaged across the sample period.9 There is a great deal of clustering in federal funds trading activity, as there is in payments activity generally in Fedwire. Our test of inventory effects focuses on banks that are active on both sides of the market for Fedwire during the period of heaviest trade in federal funds. Recognizing that large banks are active on both sides of the market, we treat the banks as dealers. There may well be economies of scope between making and receiving large amounts of payments and buying and selling federal funds. It may be that these economies of liquidity generation can explain the concentration of banks actively buying and selling federal funds. We treat our sample of banks as dealers, and test whether their balances have an effect on the transaction prices of their purchases and sales. While the perfect market hypothesis suggests that dealers face no inventory constraints, a finding of inventory effects on prices does not imply that the market for federal funds is inefficient. As Amihud and Mendelson (1980) point out, there can be inventory effects in an efficient market, in that no one can profit from knowledge of the market-maker’s inventory position and its pricing policy.

The microstructure of money

107

Data and tests The data consists of the federal funds transactions of several US commercial banks that are regularly active on both sides of the market. In addition, I supplement this with the bank’s balances throughout the day (calculated from the transaction journal) and the bank’s transactions deposits from the Call Reports. The measure of the balance we use is the bank’s actual balance divided by that quarter’s level of transactions deposits from the Call Reports; this provides a useful way to scale the level of a bank’s balance in relation to a rough gauge of its reserve demands. I will call this variable the bank’s transactions balance. Using this data, I first calculate the value-weighted spread of interest rates on a bank’s sales of federal funds and its purchases of federal funds over the period between 4.00 p.m. and 4.15 p.m. (as well as between 4.30 p.m. and 4.45 p.m.). Then I calculate the midpoint of the value-weighted average federal funds bidask spread over the period and subtract the target federal funds rate. I’ll call this variable the bank’s midpoint minus target rate. The test we construct is to test whether the midpoint of the excess of the bidask spread over the target rate of dealers is decreasing in the inventory of dealers. I first regress the midpoint minus target rate between 4.00 and 4.15 on the bank’s transactions balance (its inventory of balances) at 4.00 p.m. I choose 4.00 p.m. because the market is so active at that time. The estimation tests whether a bank’s balance at 4.00 p.m. influences its federal funds activity over the following 15 minutes. This might be considered a rigorous test, as banks may not pay close attention to the balance at any one moment, and they may not adjust their federal funds activity in response to a particular balance. Instead, banks may have a better estimate of the average balance they wish to achieve over some time period late in the day. If any inventory effects are found, it might be useful to test these less restrictive approaches to identifying inventory effects. The null hypothesis is that there are no effects on the prices at which the bank transacts. The interpretation of this hypothesis is that a finding of no inventory effects is consistent with an efficient market: banks can substitute expected funds for current balances and make loans even with low current balances late in the day. This suggests that banks have strong expectations for their ability to borrow later in the day. As shown in Figure 7.2, which plots the 5th and 95th percentile of the excess of the average federal funds rate (calculated as the average federal funds rate in the particular minute across the days of the sample period) over the target federal funds rate, it is clear that the very sparse trading in the early morning hours leads to greater variability in observed trades during those periods. The regression also includes a number of calendar and event effects. In particular each regression we report has dummies for the month, for days that precede or follow a holiday, for Good Friday (a day of low activity), for the days of the reserve maintenance period, for the first and last banking days of the month, and for the end of the quarter. Finally, we include dummies for days of

108

J. McAndrews 4.00 5th percentile 95th percentile 3.00

1.00

0.00

Average midpoint target (%)

2.00

1.00

12:30 a.m.

2:30 a.m.

4:30 a.m.

6:30 a.m.

8:30 a.m.

10:30 a.m.

12:30 p.m.

2:30 p.m.

4:30 p.m.

2.00 6:30 p.m.

Time

Figure 7.2 Fifth and 95th percentile of the average federal funds rate minus the target rate (2 October 2001–30 September 2004) (source: Federal Reserve Bank of New York: author’s calculations). Note Values shown are the 5th and 95th percentile of the distribution of the average federal funds rate minus the target rate for each minute calculated over the days of the sample period.

federal funds target rate increases and, separately, one for decreases in the target rate.10 I perform two other tests. First, I conduct the same test described above for the 4.30 to 4.45 time period. As the 4.00 time was chosen somewhat arbitrarily among those times in which federal funds trading is active, I wanted to test whether any findings were robust to changes in the time considered. Second, I also regress the deviations in the midpoint minus target rate on the bank’s deviations in its transactions balance, both at 4.00 and 4.30 p.m. Results The results of these tests are shown in Table 7.1 (all coefficients and standard errors in Table 7.1 have been multiplied by 10,000 to reduce the number of zeros in the Table). First, we find weak results in favor of inventory effects, with the coefficient on the transactions balance variables always negative and significant at the 10 percent level for the 4.00 p.m. estimations. The estimated coefficient is negative but smaller in absolute value, and statistically insignificant, for the

The microstructure of money

109

Table 7.1 Regression results Dependent variable

Midpoint – target rate (4:00 – 4:15)

Transaction –0.456 balances (0.242)* Deviation from mean transaction balances January February March April May June July August September October November December Pre-holiday Post-holiday Good Friday 1st Thursday 1st Friday 1st Monday 1st Tuesday 1st Wednesday

0.613 (0.73) 0.407 (0.731) –0.116 (0.729) 0.844 (0.739) 1.232 (0.739)* 0 (0) 0.137 (0.749) 0.598 (0.78) 0.787 (0.775) 0.784 (0.729) 1.87 (0.759)** –0.817 (0.741) –3.57 (0.79)*** 4.939 (0.822)*** 0.134 (2.967) 0.435 (0.667) –4.364 (0.68)*** 0.466 (0.685) –3.87 (0.667)*** –2.617 (0.677)***

Deviation Midpoint – from midpoint – target rate target rate (4:30 – 4:45) (4:00 – 4:15)

Deviation from midpoint – target rate (4:30 – 4:45)

–0.075 (0.146) –0.426 (0.248)* 0.633 (0.731) 0.427 (0.731) –0.092 (0.729) 0.846 (0.739) 1.234 (0.739)* 0 (0) 0.13 (0.749) 0.587 (0.78) 0.777 (0.775) 0.779 (0.729) 1.874 (0.759)** –0.817 (0.741) –3.559 (0.79)*** 4.939 (0.822)*** 0.081 (2.967) 0.437 (0.667) –4.366 (0.68)*** 0.473 (0.685) –3.869 (0.667)*** –2.613 (0.677)***

–0.068 (0.146) –1.179 (0.509)** –0.38 (0.518) –1.463 (0.507)*** –0.396 (0.513) –0.826 (0.508) 0.398 (0.521) 0.435 (0.513) –0.286 (0.521) 0.046 (0.524) –0.754 (0.504) 0 (0) –1.031 (0.513)** –3.712 (0.589)*** 3.794 (0.579)*** 5.672 (2.639)** –0.295 (0.481) –4.361 (0.471)*** –1.021 (0.472)** –3.578 (0.469)*** –3.683 (0.476)***

–1.179 (0.509)** –0.379 (0.518) –1.461 (0.507)*** –0.394 (0.512) –0.825 (0.508) 0.4 (0.521) 0.435 (0.513) –0.286 (0.521) 0.047 (0.524) –0.755 (0.504) 0 (0) –1.031 (0.513)** –3.713 (0.589)*** 3.793 (0.579)*** 5.671 (2.639)** –0.295 (0.481) –4.361 (0.471)*** –1.021 (0.472)** –3.577 (0.469)*** –3.682 (0.476)*** continued

110

J. McAndrews

Table 7.1 continued Dependent variable

2nd Thursday

Midpoint – target rate (4:00 – 4:15)

–1.556 (0.691)** 2nd Friday –2.22 (0.669)*** 2nd Monday 0 (0) 2nd Tuesday –2.153 (0.686)*** 2nd Wednesday –0.468 (0.675) First of month 3.987 (0.806)*** Last of month 9.207 (0.818)*** End-of-quarter –3.398 (1.617)** Fed funds target –12.078 increase (2.978)*** Fed funds target 11.512 decrease (1.75)*** Constant 1.525 (0.73)** Observations 1640 Adjusted R2 0.21

Deviation Midpoint – from midpoint – target rate target rate (4:30 – 4:45) (4:00 – 4:15) –1.555 (0.691)** –2.217 (0.67)*** 0 (0) –2.157 (0.686)*** –0.466 (0.675) 3.986 (0.806)*** 9.214 (0.819)*** –3.401 (1.617)** –12.101 (2.979)*** 11.533 (1.75)*** –145.181 (0.722)*** 1640 0.21

–1.381 (0.468)*** –1.934 (0.488)*** 0 (0) –2.927 (0.479)*** –2.101 (0.475)*** 6.458 (0.526)*** 8.382 (0.581)*** –0.276 (1.082) –2.746 (1.879) 16.586 (1.868)*** 3.108 (0.507)*** 2648 0.23

Deviation from midpoint – target rate (4:30 – 4:45) –1.381 (0.468)*** –1.933 (0.488)*** 0 (0) –2.927 (0.479)*** –2.101 (0.475)*** 6.459 (0.526)*** 8.384 (0.581)*** –0.275 (1.082) –2.746 (1.879) 16.588 (1.868)*** –139.205 (0.506)*** 2648 0.23

Notes Standard errors in parentheses: * significant at 10%; ** significant at 5%; *** significant at 1%. All coefficients and standard errors reported at e + 4 level.

4.30 p.m. regressions. The estimated coefficients are not particularly significant in an economic sense for the 4.00 p.m. estimate. Column 1 suggests that a doubling of the average transactions balance (in absolute value) would lower the midpoint minus target rate by about 5–18 percent (using estimated coefficients from Column 1 or Column 2, respectively). But the average midpoint minus target rate is itself very small, with the constant in the regression (from Column 1) being 0.015 basis points. These results then suggest that the federal funds market appears quite efficient on average at the 4.00 and 4.30 p.m. times. The result on the calendar and event variables also point to some interesting microstructure stylized facts. As found in Griffiths and Winters (1995), Hamilton (1996), Lee (2003) and Demilralp et al. (2004), most of the reserve maintenance period days are significant. On both Fridays, and the last Tuesday of the maintenance period, the banks in the sample post lower midpoints over target

The microstructure of money

111

rates. As the other authors have suggested, especially for the last Tuesday, that may reflect a desire to avoid being ‘locked-in’ with excess reserves at the end of the maintenance period. None of the monthly effects are consistently significant. Good Friday is insignificant. The pre-holiday and post-holiday dates, and the first and end-of-the-month days are significant. The pre-holiday is correlated with lower rates, but the other days, all days of typically high payment volume are correlated with higher rates. The end-of-quarter days are significant at the 5 percent level, and although they are typically high payment days, they are associated with marginally lower rates. Most notable, perhaps, are the estimated effects of increases and decreases in the federal funds target rate. On days on which the target rate increased the midpoint spread over the target rate was approximately three basis points lower for an increase in the target rate of 25 basis points. Similarly a decrease in the target rate was correlated with a similar size increase in the spread over the target rate. These variables are the largest in terms of economic significance in the estimation. It appears that the market, after controlling for all of the calendar effects and balances, is slightly softer after adjusting to a new, higher, target rate. These results suggest that the market for federal funds displays relatively slight calendar and inventory effects, in terms of economic significance. Of somewhat more economic significance are the effects that changes in the target rates have on behavior in the market, with transaction rate spread midpoints actually moving downward slightly, relative to the target rate, on days of rate increases. These slight effects suggest that on average the market is quite efficient in that it would be hard to profit from these spreads. These tests are weak in the sense that they measure current balances only during a short time period during the day. Other possible tests would be to measure the change in a bank’s balance (excluding all of its fed funds transactions) over the course of the day, or over some lengthy period during the day. However, the test reported here gives a view of the efficiency of the market at a particular time late in the trading day; this might be a good measure of the view that traders have of their ability to obtain funds in the market later in the day.

Discussion: market microstructure of payments and current trends Market microstructure gives us a lens through which we can view payment systems and developments in payment systems. Consider the federal funds market. This market is decentralized and operates through brokers and directly as banks use telephones to contact counterparties. It settles and convenes in the context of an RTGS system. A large number of participants can enter the market if they wish.11 The reserve accounting procedures appear to have slight, but significant, effects on participants’ behavior. Now imagine how the market would differ if trades were settled only in a net settlement system (DNS). One could imagine many different outcomes, depending on how early participants were informed of their net obligations. Suppose

112

J. McAndrews

that participants were informed of their net obligations in the payment system only shortly before the designated time of settlement. Furthermore, suppose that positions of all parties were transparent to all participants. Then, depending on the distribution of the positions, those on one side of the market might face a credit squeeze if the other side of the market is quite concentrated. Alternatively, the presence of a net settlement payment system might encourage the market participants to develop a more centralized call market for the money market. These remarks are intended to focus attention on the money market effects of payment system design. Whether a payment system is a DNS, an RTGS, or a limit payment order system has implications for the money market that utilizes the payment system as its settlement system. Continuous auction money markets are naturally associated with RTGS systems, although it is not necessary that the settlement must be continuous. In fact, it is likely that the causation runs the other way in that RTGS systems, with binding constraints on intraday overdrafts, may require continuous trading in money markets to assist banks in maintaining the level of their balances in a desirable range. In an RTGS then, it is not surprising that many banks both buy and sell money market loans, as they utilize the overnight money market to assist in their intraday money management. This realization in turn suggests that, as technology improves, net settlement of the next day’s money market loans might be possible and desirable, to reduce the outstanding dues to and dues from in the banking system. Looking at current trends in the payment system, the development of limit payment order systems is certainly a notable trend, as discussed in Bank for International Settlements (2005). The improvements in the technology for communication and computation that make these technologies feasible and economical for payment system participants continue. It is likely therefore that these systems will continue to develop. How will the money market be affected? As in other forms of trading, one can imagine the development of electronic limit order books for overnight money markets.12 It is possible to imagine that the liquidity suppliers post interest rates, perhaps through brokers, and banks that need to make payments then ‘cross’ or ‘hit’ the offer of funds. Such a market development could assist banks in planning their liquidity needs at different times, and may lead to a smoother method for providing liquidity for payments. The development in other securities markets, including stock markets, of electronic limit order book systems has led to a voluminous debate over the merits of floor-based trading systems and electronic limit order books and between the benefits of transparency of the ‘book’ of limit orders, and lack of transparency. As summarized by Biais et al. (2004) most theoretical analyses of transparency suggest that it reduces adverse selection in markets. However, empirical analyses suggest some mixed results. Biais et al. (2004) suggest this may be because transparency may make it more difficult for large traders to supply liquidity. The analysis of Willison (2005) suggests that transparency in limit payment order systems is not as crucial as the common knowledge of the limit order priority system itself.

The microstructure of money

113

Payment systems have also become somewhat more tied together in recent years, both within a currency area, as the development of the Target system in Europe shows, and across currencies, as the development of CLS Bank International shows. These links have assisted banks in managing their positions in different currencies and in different countries. It is likely that, as banks continue to pursue international opportunities, payment system participants – banks and central banks – will continue to work to improve and develop these links, similarly to the way that securities markets in both Europe and the United States have done so.

Summary The microstructure approach to the study of payment systems views payment systems as analogous to security settlement systems. Payment system designs have analogues in the designs of securities settlement systems. Similarities in the economics of payment systems and security settlement systems suggest that further study of payment systems, borrowing from the financial market microstructure literature, would be fruitful. The tests for inventory, calendar, and event effects in the US federal funds market suggested that there are only slight effects of these variables on the midpoint of the spreads of federal funds loans above the target rate. Furthermore the spread above the target rate was very small, indicating that the market is quite efficient on average, and that the open market desk is quite accurate in its operations. Notable developments in payment systems include the recent adoption of ‘limit payment order’ systems, and the linking of payment systems through Target and CLS Bank International. These developments are similar to developments in securities markets, and are likely to be pursued further. The study of market microstructure in financial markets has been greatly affected by the increasing presence of high-quality, high-frequency data. Recent work by central banks in simulating the performance of payment systems has revealed that central banks have been maintaining data on payment system activity. This is a first step in organizing data that can be quite useful in studying the market microstructure of money.

Notes 1 I wish to thank Adam Ashcraft, Bruno Biais, and Xavier Freixas, and the participants at the Future of Payments Conference at the Bank of England for helpful comments. I thank Kurt Johnson for excellent research assistance. The views expressed in this chapter are those of the author and do not necessarily reflect the views of the Federal Reserve Bank of New York or the Federal Reserve System. 2 This rough characterization does not fully specify a DNS as it does not describe the rules of the DNS in the event of a default. See Bank for International Settlements (1989, 1990) for more complete descriptions of DNS systems. 3 If there are not sufficient balances on account, a payment order may be queued, pending the arrival of additional balances, or rejected. See Bank for International Settlements (1997) for a full discussion of various alternatives and implementations of RTGS systems.

114

J. McAndrews

4 From the RTGS plus website of the Deutsche Bundesbank (surveyed April 20, 2005): www.rtgsplus.de/en/leistungsumfang/limitsteuerung/inhalt_e.htm. 5 See McAndrews and Trundle (2001) and Bank for International Settlements (2005) for more detailed descriptions of the range of alternatives for these payment systems. 6 Again, I confine myself here to ‘pure’ payment systems, and ignore delivery-versuspayment systems in which both a security and account balances are exchanged simultaneously. 7 See Regulation D, 12 CFR 204, Board of Governors of the Federal Reserve System. 8 The method for identifying federal funds purchases from the Fedwire transactions journal was pioneered by Furfine (1999), and applied by Demilralp et al. (2004), as well. The method I use is very similar to Demilralp et al. (2004) ‘N-to-N’ method. We search for payments from a bank to its counterparty on day t that is in round values of $100,000, and a matching return payment on day t + 1 that is slightly larger than the first payments, and within a reasonable range given federal funds rates reported by brokers for that day. Specific details are available from the author. 9 On May 22, 2004, the Fedwire operating day lengthened from its previous hours of 12.30 a.m. to 6.30 p.m. to 9.30 p.m. the previous day to 6.30 p.m. We do not include the hours of 9.30 p.m. to 12.30 a.m. in this chart. 10 We note that open market operations are usually conducted in the late morning, well before the 4.00 p.m. time of our tests. 11 Demilralp et al. (2004) have a good discussion of who is eligible to participate in federal funds trading. 12 See www.e-mid.it/index.php for a description of the screen-based money market platform.

References Admanti, A. and Pfleiderer, P. (1988) ‘A theory of intraday patterns: volume and price variability’, Review of Financial Studies, 1: 3–40. Amihud, Y. and Mendelson, H. (1980) ‘Dealership market: market-making with inventory’, Journal of Financial Economics, 8: 32–53. Angelini, P. (1998) ‘An analysis of competitive externalities in gross settlement systems’, Journal of Banking and Finance, 22: 1–18. Bank for International Settlements (1989) Report on Netting Schemes (Angell Report), Committee on Payment and Settlement Systems Publication No. 2. Bank for International Settlements (1990) Report of the Committee on Interbank Netting Schemes of the Central Banks of the Group of Ten Countries (Lamfalussy Report), Committee on Payment and Settlement Systems Publication No. 4. Bank for International Settlements (1997) Real-time Gross Settlement Systems, Committee on Payment and Settlement Systems Publication No. 22. Bank for International Settlements (2005) New Developments in Large-value Payment Systems, Committee on Payment and Settlement Systems Publication No. 67. Bech, M. and Garratt, R. (2003) ‘The intraday liquidity management game’, Journal of Economic Theory, 109: 198–219. Biais, B., Glosten, L., and Spatt, C. (2004) ‘Market microstructure: a survey of microfoundations, empirical results, and policy implications’, IDEI Working Paper. Demiralp, S., Preslopsky. B., and Whitesell, W. (2004) ‘Overnight interbank loan markets’, Board of Governors of the Federal Reserve System Finance and Economics Discussion Series. Freixas, X. and Parigi, B. (1998) ‘Contagion and efficiency in gross and net interbank payment systems’, Journal of Financial Intermediation, 10: 3–31.

The microstructure of money

115

Furfine, C. (1999) ‘The microstructure of the federal funds market’, Financial Markets, Institutions and Instruments, 8(5): 24–44. Furfine, C. (2003) ‘Interbank exposures: quantifying the risk of contagion’, Journal of Money Credit and Banking, 35: 111–28. Garman, M. (1976) ‘Market microstructure’, Journal of Financial Economics, 3: 257–75. Goodfriend, M. and Whelpley, W. (1993) ‘Federal funds’, in T.Q. Cook and R.K. Laroche (eds) Instruments of the Money Market, Richmond, VA: Federal Reserve Bank of Richmond. Griffiths, M. and Winters, D. (1995) ‘Day-of-the-week effects in federal funds rates: further empirical findings’, Journal of Banking and Finance, 19: 1265–84. Hamilton, J. (1996) ‘The daily market for federal funds’, Journal of Political Economy, 104: 26–56. Hamilton, J. (1997) ‘Measuring the liquidity effect’, American Economic Review, 87: 80–97. Hasbrouck, J. and Sofianos, G. (1993) ‘The trades of market makers: an empirical analysis of NYSE specialists’, Journal of Finance, 48: 1565–93. Ho, T. and Stoll, H. (1981) ‘Optimal dealer pricing under transactions and return uncertainty’, Journal of Financial Economics, 9: 47–73. Ho, T. and Stoll, H. (1983) ‘The dynamics of dealer markets under competition’, Journal of Finance, 38: 1053–74. Kahn, C., McAndrews, J., and Roberds, W. (2003) ‘Settlement risk under gross and net settlement’, Journal of Money Credit and Banking, 35(4): 591–608. Kahn, C. and Roberds, W. (1998) ‘Payments system settlement and bank incentives’, Review of Financial Studies, 11: 845–70. Lee, Y.S. (2003) ‘The federal funds market and the overnight eurodollar market’, Journal of Banking and Finance, 27: 749–71. Lyons, R. (1995) ‘Tests of microstructural hypotheses in the foreign exchange market’, Journal of Financial Economics, 39: 321–51. Madhavan, A. and Smidt, S. (1993) ‘An analysis of changes in specialist inventories and quotations’, Journal of Financial Economics, 30: 99–134. Manaster, S. and Mann, S. (1996) ‘Life in the pits: competitive market making and inventory control’, Review of Financial Studies, 9: 953–75. McAndrews, J. and Rajan, S. (2000) ‘The timing and funding of Fedwire funds transfers’, Federal Reserve Bank of New York Economic Policy Review. McAndrews, J. and Trundle, J. (2001) ‘New payment system design: causes and consequences’, Bank of England Financial Stability Review. Pagano, M. (1989). ‘Trading volume and asset liquidity’, Quarterly Journal of Economics, 104: 255–76. Pagano, M. and Roell, A. (1996) ‘Transparency and liquidity: a comparison of auction and dealer markets with informed trading’, Journal of Finance 51: 579–611. Skeie, D. (2004) ‘Money and modern bank runs’, unpublished thesis, Federal Reserve Bank of New York. Stigum, M. (1990) The Money Market, Homewood, IL: Dow Jones-Irwin. Stoll, H. (1978) ‘The supply of dealer services in securities markets’, Journal of Finance, 33: 1133–51. Vayanos, D. (1999) ‘Strategic trading and welfare in a dynamic market’, Review of Economic Studies, 66(2): 219–54. Willison, M. (2005) ‘Real-time gross settlement and hybrid payment systems: a comparison’, Bank of England Working Paper No. 252.

Part III

Current payment policy issues

8

Wholesale payments Questioning the market-failure hypothesis George Selgin1

The kind of situation which economists are prone to consider as requiring corrective governmental action is, in fact, often the result of governmental action. Coase (1988: 133)

Introduction When a bank, acting on its own behalf or that of a client, needs to transfer a large sum, say $10 million, to another bank, the transfer will almost certainly be made through a ‘wholesale’ payments system. Wholesale payments systems receive payments messages from sending banks, relay them to receiving banks, and oversee the final settlement of interbank accounts. Two basic kinds of wholesale payments systems exist. A Deferred Net Settlement (DNS) system gathers payment orders throughout the business day (or some other preset period), calculates end-of-day multilateral net obligations, and then arranges for the transfer of reserves (‘good funds’) from net senders to net recipients of funds. A Real-Time Gross Settlement (RTGS) system executes payment orders as they arrive, at once transferring reserve credits representing the gross value of individual payments. The strictest Real-Time Gross Settlement (RTGS-GF) systems require their members to possess clearing balances or ‘good funds’ sufficient to cover all payments, while less strict versions (RTGSDO) allow their participants to rely on intraday credits or ‘daylight overdrafts’ to cover payments in excess of their available balances, on the understanding that the credits must be repaid at day’s end. An RTGS-DO arrangement combines the continuous settlement feature of RTGS-GF with at least some of the reserve efficiencies found in DNS systems. However, unless intraday credits are provided free of charge, a sequence of wholesale payments administered by either sort of RTGS system will give rise to a greater demand for bank reserves than an identical sequence of payments using DNS. This is one reason why DNS has historically been the preferred means for handling both retail (small value) and wholesale transfers. The 1990s, however, witnessed a remarkable change in wholesale payments arrangements, with traditional DNS arrangements giving way to RTGS systems,

120

G. Selgin

and especially to RTGS-DO systems. Government monetary authorities, including the Fed, have actively promoted the change both by recommending or mandating changes in private wholesale payment systems and by competing against private networks with their own RTGS systems. Government promotion of RTGS has typically been defended on the grounds that traditional net settlement systems are beset by serious externalities that could have catastrophic consequences. Here, I critically assess market failure arguments for reforming or abolishing traditional DNS systems, showing that these arguments have been based, not on empirical evidence or on careful consideration of actual DNS procedures, but on a spurious analogy drawn between DNS systems and Fedwire, the Federal Reserve’s own RTGS-DO system. While there are good reasons for thinking that Fedwire does indeed harbor a serious externality problem, or at least that it did so prior to reforms initiated in 1994, the frequently made claim that unregulated DNS systems suffered from an analogous but independent externality problem overlooks crucial differences between Fedwire and traditional DNS arrangements, including CHIPS prior to the 1990s, and especially the different ways in which each generates and assigns intraday credit risk. My position is not that traditional DNS arrangements were trouble-free. Nor do I deny that such arrangements may have permitted excessive risk taking. However, I argue that, to the extent that risks were excessive, the cause was not market failure but implicit guarantees extended by regulatory authorities themselves, which tended to corrupt otherwise sound market-based arrangements. My modest aim is to show how the literature on wholesale payments has tended to confuse regulatory failure with market failure, thereby diverting attention from potential first-best solutions to alternatives that may not even qualify as second-best.

The trouble with Fedwire At the end of the 1990s, the United States was the only industrialized nation that relied heavily upon both DNS and RTGS arrangements to handle wholesale payments. The Clearing House Interbank Payment System (CHIPS) was a private DNS system operated by the New York Clearing House.2 Fedwire was (and remains) a RTGS-DO system operated by the Federal Reserve. Each system handled about $1.5 trillion in wholesale transfers annually. The coexistence of these two arrangements, despite superior reserve-holding economies generally realized through net settlement, was largely due to the Fed’s willingness to supply sending banks with low-cost daylight overdrafts. Prior to 1994, the Fed supplied intraday credit routinely and free of charge, effectively matching the reserve economies of net settlement. While the Fed allowed sending banks to overdraw their accounts to execute payments, it also provided for immediate settlement by crediting the accounts of receiving banks by the full amount of gross payments. The Fed guaranteed the ‘finality’ of these payments, meaning that it could not have recourse to credited

Wholesale payments

121

accounts if a sending bank with an overdrawn account failed to settle up with it by the end of the business day. The risks associated with Fed extensions of intraday credit, including the risk of settlement failure, were therefore borne approximately by the Federal Reserve System (which might find itself holding Fedwire-generated claims on a failed institution) and ultimately by the general public. Thus, until 1994 at least, Fedwire suffered from a potentially serious flaw, in that it deprived both receiving banks and their customers (the ultimate recipients of payments) of any strong incentive to monitor sending banks or to limit their acceptance of payments orders sent to them. It also relieved senders of any reason to fear that, in selecting unsound banks as their agents, they would remain liable for promised payments in the event that their banks failed to settle with the Fed. The result was a serious moral hazard problem, with private agents initiating transactions that exposed third parties to credit risk. One way out of this moral hazard was to have the Fed charge a fee on its intraday credits sufficient to cover the risk of non-payment. Although the Fed acknowledged the need for some such remedy during the 1980s, when the volume of Fedwire transactions was increasing rapidly, only in April 1994 did it begin charging a marginal daylight overdraft fee of 24 basis points (annual rate), which was increased to 36 basis points a year later.3 This reform led to a significant reduction in measured Fedwire overdrafts (Richards, 1995). Nonetheless, some experts maintain that the Fed’s current intraday lending rates, which are levied only on overdrafts in excess of 10 percent of a bank’s risk-based capital, and which are waived if they sum to less than $25 in two weeks, may still involve some under-pricing of settlement risk.4 In fact, the Federal Reserve Board had originally intended to raise the Fed’s overdraft fee above 36 basis points, but changed its mind in part because it feared that such a move might prompt a substantial shift of wholesale payments volume to private payments networks, including CHIPS (Richards, 1995: 1068).

DNS: guilt by association Fed officials hesitated to implement any Fedwire reforms that might have sponsored a substantial shift in wholesale payments activity from Fedwire to CHIPS. Although bureaucratic turf-preserving and budget-maximizing behavior might account for this hesitation, it appears to have been due at least in part to officials’ belief that CHIPS’ traditional operating procedures were no less in need of reform than Fedwire’s had been. Their reasoning, which is shared by many monetary authorities across the globe as well as by some economists, goes as follows. In a traditional, ‘unsecured’ DNS system, like CHIPS before 1990, payment orders are, essentially, binding IOUs that come due at settlement time.5 Therefore, if at any time of the day system participant A has received payment orders from participant B exceeding by $X the value of such orders B has received from A, A may be said to have granted B $X of ‘intraday credit.’ (Some authorities even claim that sending banks in a traditional DNS system

122

G. Selgin

routinely rely on ‘daylight overdrafts’ on their clearinghouse accounts, implying that intraday credit is granted, not by receiving banks, but by the clearinghouse itself.) DNS-system intraday credit is, moreover, provided free of any nominal or explicit charge: just like Fedwire’s pre-1994 intraday credits. DNS-system participants have no other (non-price) means for controlling the volume of intraday credits they grant one another (or, as some would have it, obtain from the clearinghouse), except by rejecting payments orders altogether. A free-rider problem makes it less than worthwhile for receiving banks to go to the trouble of finding out whether a sending bank might be unworthy of an intraday loan. DNS systems therefore suffer from the same moral hazard problem present in Fedwire. The tendency of regulators and economists to treat traditional or ‘unsecured’ DNS systems as close cousins of Fedwire, with its reliance upon free or underpriced intraday credits or ‘daylight overdrafts,’ is evident throughout the literature on wholesale payments: •

•

•

•

•

•

Board of Governors of the Federal Reserve System (1988: 7): ‘CHIPS participants [that] have initiated transfers with a total dollar value greater than that of the transfers they have received . . . are essentially receiving intraday credit from the participants that have received transfers with a total value higher than that of transfers they have sent.’ Rochet and Tirole (1996: 840): ‘[I]n net payment systems, intraday overdrafts do not appear explicitly and are therefore necessarily free’; ‘Since its inception, CHIPS has operated with explicit [sic] intraday overdrafts.’ Folkerts-Landau (1997: 5): ‘Netting arrangements . . . expose the participants to credit risks as they extend large volumes of payments-related intraday credit to each other.’ Kahn and Roberds (1999: 30): Large-value payment systems ‘have traditionally enjoyed access to significant amounts of virtually free intraday credit. . . . In net settlement systems, such credit is granted when a bank accumulates a large net debit position vis-à-vis other banks in the system.’ Roberds (1999: 2): ‘The principal disadvantage of a net settlement system is that the central counterparty . . . ends up bearing most of the credit risk and liquidity risk associated with the settlement of payment obligations.’ Mengle et al. (1987: 7): ‘On private net settlement networks such as CHIPS, the [intraday credit] supply curve faced by system participants diverges from the supply curve reflecting risks to society the same as occurs on Fedwire.’

Using such arguments regulators have claimed, not only that unsecured DNS arrangements involve externalities comparable to those found in Fedwire, but also that DNS-system externalities pose a different and perhaps greater hazard. This last conclusion stems from beliefs concerning the different consequences of a bank’s failure to settle in the two kinds of systems. Under Fedwire, as we have seen, if a bank fails to pay-off its intraday debts to the Fed, settlement proceeds

Wholesale payments

123

regardless, with the Fed footing the bill. (Meanwhile the defaulting bank may be granted a discount-window loan, which unlike other central bank loans may have to be fully collateralized.) In a traditional DNS system, in contrast, contractual agreements would generally call for a payments ‘unwind.’ An unwind commences with cancellation of all of the previous settlement period’s payment orders to and from the failed institution. Remaining multilateral net positions are then re-calculated, and settlement is once again attempted. Should one or more other banks then find themselves unable to settle owing to a deterioration of their net positions, payments messages to and from those banks are also cancelled. The process continues in this fashion until surviving banks are able to settle. In principle, the failure of a single DNS-system participant could cause many other participants to default. Just how great was the risk of a DNS-system participant failing unexpectedly, and how extensive would the fallout from such a failure have been in practice? The empirical record supplies only negative evidence: there had never been an actual settlement failure on CHIPS or any other important DNS wholesalepayments network.6 Economists and policy-makers therefore had to rely on simulations to assess the likely consequences of a DNS-system unwind. According to Humphrey’s (1986) influential simulation, the failure of a major CHIPS participant, given rules in place at the time, might have caused dozens of large banks to fail, triggering a system-wide crisis. Fear of such a catastrophe led regulators to treat traditional DNS systems as being especially in need of regulation, or (an increasingly popular option) of replacement by some form of RTGS. The Bank of Japan, which for many years offered both DNS and RTGS services to Japanese banks, switched to offering RTGS only in the late 1990s, despite Japanese financial firms’ apparent preference for the DNS alternative: just prior to the change, only about 3 percent of Japan’s wholesale payments, measured in value terms, were handled by the Bank of Japan’s RTGS system. The United Kingdom converted CHAPS – its counterpart to CHIPS – into an RTGS system in 1996. Central bank administered wholesale payments systems throughout the rest of the EU have, in response to BIS recommendations, operated on a real-time gross basis since 1997. The Reserve Bank of New Zealand switched to RTGS in March 1998, and in 2000 the Canadian Payments Association established, at the Bank of Canada’s urging, a hybrid Large Value Payment System that relies on the Bank of Canada as a source of collateralized intraday loans and as an ultimate guarantor of end-of-day settlement. CHIPS itself, finally, was converted into a hybrid system offering near continuous and irrevocable settlement for executed payments in January 2001.7 Traditional DNS systems have thus been largely abolished, despite having long been generally favored by parties engaged in wholesale payments.8

‘Mirage’ externalities The wholesale payments system reforms of the 1990s are supposed to have addressed excessive risk taking, and excessive risk of payment unwinds in

124

G. Selgin

particular, stemming from externalities inherent in unregulated DNS systems. However, a close look at traditional DNS-system contractual arrangements suggests that the externalities that are supposed to have given rise to an inefficient risk–return trade-off were not genuine market failures at all, but market failure ‘mirages’ that tend to appear when DNS systems are viewed through Fedwirecoloured glasses. Putting those glasses aside allows us to get an undistorted picture of traditional, voluntary contractual arrangements of commercial banks and private clearinghouses, where by ‘voluntary’, I mean free from pressure, edicts, or (subsidized) competition from government agencies. As noted earlier, the overwhelming historical tendency has been for financial firms to establish clearinghouses that rely on unsecured DNS to handle both small- and largevalue payments. In the absence of government involvement, clearinghouses have generally been bankers’ associations or ‘clubs’ (Dowd, 1994). Among their organizational features, the following are especially relevant: •

•

•

•

Because the clearinghouse is not a bank, it receives payment orders and calculates net multilateral positions, but (unlike a central bank) does not maintain account balances for its members. Because clearinghouse members do not maintain accounts with the clearinghouse, ‘daylight overdrafts’ in the strict meaning of the term play no part in traditional DNS systems, or play only an indirect part (as when net settlement transfers are administered through a RTGS-DO system). Nor do private clearinghouses normally provide intraday credit in any form to their members. Such clearinghouses therefore do not usually assume any payments-related risk.9 Intraday DNS payments have traditionally been provisional payments only. Private agents contract for ‘check’ rather than ‘receiver’ finality in both retail (check) and wholesale (wire) payments. Check finality means that a bank receiving a payment order is under no obligation to release funds to the payee until end-of-day settlement is complete. Moreover, should the bank grant the payee access to funds prior to settlement, the funds in question generally remain revocable in the event of a settlement failure. In legal terms, the ‘acceptance’ of payment orders by receiving banks, which renders them liable to payees for the amount of the orders (while in turn obliging sending banks to pay them) is conditional upon settlement.10 Because acceptance of payments orders in a traditional, unsecured DNS system is conditional upon settlement, a settlement failure in such a system renders orders sent by a failed participant null and void. The effect is as if the orders in question had never been sent. This is the essence of the ‘unwinding’ process. It is therefore misleading to speak of receiving banks in DNS systems providing sending banks with ‘intraday credits’: in a traditional DNS system, payment orders are not legally binding IOUs; they are properly regarded as non-binding pledges only.11

Wholesale payments

125

Taken together these points imply that the only ‘intraday credits’ actually extended in traditional DNS systems are credits granted by receiving banks to their own customers when they grant beneficiaries immediate access to funds. No basis exists, therefore, for claiming that receiving banks in DNS systems are – in the absence of government interference with private contracts – unable to effectively regulate their exposure to intraday credit risk, or that such risk is an external or third-party consequence of transactions undertaken, not by receiving banks, but by sending banks and their clients. Finally, because the intraday credits to beneficiaries can generally be revoked in the event of a settlement failure (the one circumstance in which credits might conceivably need to be repaid) risk is borne not primarily by receiving banks but by the beneficiaries themselves and also by payment originators who remain obliged to the beneficiaries even though their banks’ failure may have deprived them of access to their accounts.12 This bearing of risk by payment senders and beneficiaries, as opposed to some externality, may account for the extension of intraday credits free of charge in unsecured DNS systems. To insist that unsecured DNS systems involved underpriced intraday credit ‘the same as occurs [or occurred] on Fedwire,’ while simultaneously finding fault with such systems because they expose payment senders and beneficiaries to the possibility of a payments unwind, is self-contradictory.13 The use of terminology taken from Fedwire (or other RTGS-DO systems) to describe DNS arrangements has thus caused economists and regulators to overlook features of deferred settlement aimed at containing and controlling payment-related risks in a cost-effective way. By making the acceptance of payment orders conditional only, pending successful settlement in good funds, receiving banks in private DNS systems avoid having to monitor sending banks – a burden that would be great enough in a small DNS system such as CHIPS, and impossible in a network with as many participants as Fedwire. Receiving banks can instead concentrate on keeping informed of the credit worthiness of their own customers – the beneficiaries of wire transfers – in order to decide whether to release funds to them prior to settlement. Bankers are, presumably, capable of assessing the credit-worthiness of their own account holders. In practice, receiving banks in traditional DNS systems often grant immediate credit to customers. This tendency reflected customers’ apparent creditworthiness as well as receiving banks’ low estimate of the probability of a settlement failure. The low perceived risk of a settlement failure was no doubt informed by the extreme rarity of actual settlement failures in traditional DNS systems, which was partly a consequence of careful up-front selection (by means of capital and liquidity standards and the like) and monitoring of clearinghouse members. The presumption that a settlement failure was highly unlikely may also have been informed at least in part by the belief that regulators would intervene rather than allow such a failure: a possibility considered below. Payment beneficiaries nevertheless retained an incentive to limit their use of revocable funds, and to insist that funds be sent to them through banks that they considered sound. Finally, payment originators had reason to select sending banks carefully,

126

G. Selgin

knowing that a sending bank’s failure prior to settlement would leave beneficiaries’ claims against them intact, while depriving them of access to most, if not all, of their undelivered bank balances. As Mengle (1990: 158) observed: In wholesale wire transfers, the sender is most likely a corporation, possibly a bank. Given the size of the transfer, it is plausible that senders are of sufficient sophistication to monitor the soundness of the banks with which they do business. Failure of a sending bank is something against which a prudent sender can protect itself.14 That unsecured DNS systems did not involve any inherent externality problem does not necessarily mean that such systems were efficient. Economists’ standard practice is, nonetheless, one of assuming that, unless special reasons can be found for thinking otherwise, ‘parties in privity will contract for the most efficient allocation of risk’ (Scott, 1990: 182). There was nothing to stop receiving banks in private DNS systems from voluntarily agreeing to accept payment orders unconditionally, making irrevocable payments to beneficiaries in anticipation of settlement. But then these banks, unlike receiving banks in Fedwire, would have had reason to adjust their fees to reflect any perceived risk of a settlement failure. Faced with the choice of having to pay receiving banks a fee sufficient to compensate them for bearing such risk, and bearing the risk themselves, payment senders and beneficiaries apparently preferred the latter option, and did so presumably, because they believed themselves capable of controlling or absorbing such risks for less than what receiving banks would charge them for performing those same services. Indeed, the sophistication of large payment originators and beneficiaries makes check finality appear to have been even better suited to satisfying the ‘least-cost avoider’ principle for optimal risk assignment with respect to large-value wire transfers than it has been with respect to check payments.

The systemic risk problem Although individual participants in a traditional DNS system might have been fully in control of credit risks associated with their own intraday lending activities, they may also have been exposed to losses resulting from other participants’ imprudence, which could have exposed them to the diffuse adverse effects of a payments unwind. Did the existence of such ‘systemic’ risk itself justify restricting activity in DNS systems or replacing DNS with RTGS? Strictly speaking, it did not. Although it is true that a bank’s multilateral net position may be adversely affected by failure of another bank even if it had no bilateral transactions at all with the failed bank, such interdependence does not imply a market failure. As Scott (1990: 185) observes: a bank that joins a voluntary payments network enters a contractual arrangement with the network operator and, indirectly, with all other network

Wholesale payments

127

participants. Network members have an incentive to establish rules designed to limit systemic risk generated within the network without having to be encouraged or forced to do so by regulatory authorities.15 Systemic risk is, moreover, not a problem unique to DNS systems. All banks, regardless of the payment systems they rely upon, depend to some extent on payments from other banks to finance their own payments. Participants in a DNS system expose themselves to systemic risk when, upon receipt of a payment order, they credit their customers’ accounts in anticipation of settlement. In an RTGS-GF system, payment orders are settled immediately, or are rejected, so that the orders cannot be said to be the basis for payment expectations that may ultimately be disappointed. But this does not mean that RTGS-GF system participants cannot act upon expectations that are later falsified regarding the flow of good funds through the banking system. For example, a bank might promise to make a payment to a second bank at 16:00 h, anticipating the 15:00 h arrival of a payment promised to it by a third bank. But the 15:00 h payment may never be sent. The same thing can of course happen in an RTGS-DO system. The claim that participants in RTGS systems ‘cannot respond to payments that have not been received’ (Van den Bergh and Veale, 1994: 103) is therefore invalid. It follows that, with regard to systemic risk, the difference between RTGS and DNS is a difference of degree rather than a difference in kind; and this difference may be warranted in view of the much higher liquidity or collateral costs (or more frequent payment-order rejections and delays) that real-time gross settlement entails.16 Comparing the costs of secured net settlement on CHIPS to those of RTGS, Schoenmaker (1995: 26) concludes that ‘the estimated extra cost of RTGS exceeds the estimated reduction in settlement and settlement risk.’17 What about the potentially catastrophic consequences of a DNS unwind? Are there not sufficient grounds for preferring RTGS to DNS even if the former arrangement is much more costly and does not completely eliminate systemic risk? One reply to this question asks, ‘What catastrophic consequences?’ To repeat: there has never been a settlement failure and consequent unwind on CHIPS or any other major DNS wholesale-payments system; and the chances of a DNS-system participant failing suddenly enough to precipitate a settlement crisis (as can happen if other participants have no inkling of troubles at a participating bank even on the very day on which it fails) can only be judged remote.18 Of course, were an unwind to have truly catastrophic consequences, even a tiny risk might be considered unacceptable, and especially so in light of the possibility that market participants might underestimate or choose to ignore risks that they are scarcely able to contemplate. Research in cognitive psychology and behavioral economics suggests that, contrary to the predictions of orthodox rational choice theory, individual inferences concerning extreme events can exhibit systematic biases. However, the same research offers no consistent prediction concerning the direction of the bias, and therefore supplies little guidance for the formulation of public policy (Gerson, 2001). Private markets do

128

G. Selgin

supply catastrophe insurance, after all. Moreover, the assumption that an unwind is likely to prove catastrophic is itself open to serious doubts. That assumption has been based largely on the results of Humphrey’s (1986) simulation, which itself assumed that CHIPS participants could have had no recourse at all to beneficiaries’ account balances following cancellation of failed banks’ payment messages. This is tantamount to assuming that every beneficiary of an intraday DNS payment draws its account balance down to zero prior to settlement time, which is unrealistic as well as question-begging. Intraday payment beneficiaries in DNS systems function like lenders of last resort, except that, instead of augmenting their banks’ liquidity by supplying them with last-minute reserves, they allow the banks to make last minute, negative adjustments to the beneficiaries’ deposit credits. If beneficiaries’ end-of-day account balances were always at least equal to their provisional intraday credits (the opposite of Humphrey’s assumption), a payments unwind could never cause a bank to fail even in a banking system bereft of capital. Reality falls between the two extremes, with end-of-day beneficiary balances offering some protection against settlement failure.19 Provisional payment clauses would never have arisen in the first place had they not been capable of re-directing risk, if only to a limited extent. Nor would they still exist, despite regulations aimed at making unwinds less probable than ever, if banks could not bring themselves to contemplate the possibility of a settlement failure. In general, the more frequent the clearing-and-settlement sessions in a DNS system, the less likely it is that any participant will fail before its payment orders are rejected by the system. RTGS can thus be viewed as a limiting case in which settlement occurs with each and every payment. The predominance of daily clearings in past DNS systems suggests that such clearings typically sufficed to reduce the odds of an unwind to tolerably low levels. This interior solution to the ‘optimal frequency of settlement’ problem – a solution informed by decades of experience – stands in stark contrast with the ad hoc corner solution of realtime gross settlement favored by regulatory authorities. Besides exaggerating the likelihood and adverse consequences of a DNS unwind, regulators have also tended to overlook certain advantages that an unwind rule offers relative to an alternative rule guaranteeing intraday finality of payments. Consider, for example, how a large (uninsured) deposit holder might respond under each rule to a rumor that his bank is going to fail later that day. Under guaranteed finality, the depositor can rescue most of his balance by wiring funds to another bank, thereby shifting default risk from himself to the payment system. Under provisionality, in contrast, the same depositor would have no choice but to run on his bank for cash, knowing that the failure of his bank before the end of the day will result in the cancellation of any payment messages sent by it. Although a run to currency may seem more disorderly than a wire transfer ‘run’ to other banks, it does not necessarily expose third parties to default risk.20 In this respect at least, an unsecured DNS arrangement is more incentive-compatible than a secured one, and also more incentive-compatible than RTGS systems with intraday credit, including Fedwire.21

Wholesale payments

129

Government, not market, failure Although no convincing grounds exist for the claim that unregulated DNS systems suffer from inherent market-based imperfections, good reasons exist for holding regulatory authorities themselves responsible for undermining the safety of wholesale payments arrangements that might otherwise be expected to achieve an efficient and acceptable risk–return trade-off. The most important source of ‘government failure’ in past DNS systems consisted of central-bank-provided finality guarantees.22 Central bankers’ empirically unsupported views concerning the likelihood of systemic failures, and bureaucratic or political considerations as well, have inclined them to bail out banks – and big banks especially – that might otherwise have been unable to meet their net settlement obligations. In the past such guarantees, if they existed at all, tended to be implicit only. But some recent DNS-system reforms, based upon the misguided policy desideratum of zero unwind risk, have made them explicit. Whether implicit or explicit, government-based finality guarantees undermine private incentives to monitor and control payments-related risks in DNS arrangements, including receiving banks’ incentive to limit beneficiaries’ access to unsettled payments and beneficiaries’ incentive to resist employing advanced funds prior to settlement. The extent to which the routine practice, in CHIPS and other DNS systems, of immediately releasing intraday funds to payment beneficiaries was encouraged by the presence of central bank guarantees (or by the presence of alternative central-bank administered payments arrangements in which intraday finality was provided for less than its true social cost) remains a crucial but as yet unexamined empirical question.23 If excessive risk taking in DNS systems has been due, not to market failure, but to the presence of central-bank guarantees, then DNS systems might be made ‘perfectly’ safe by ending the guarantees. In principle, this requires nothing beyond enforcement of the generally approved but frequently broken ‘classical’ rule limiting last-resort assistance to illiquid but solvent banks.24 The 1991 FDICIA reform limited Fed lending to undercapitalized banks to no more than 60 days within any 120-day period, and sanctions stricter than those included in FDICIA could further discourage Fed lending to insolvent institutions (Kaufman, 1999: 5–6). Alternatively, the Fed and other central banks could be altogether prohibited from providing extended credit to banks. As Kaufman (1999: 6–9) and several other economists have observed, well-organized modern markets for both government securities (or commercial paper) and bank reserves make direct central bank lending to troubled banks unnecessary in most industrialized economies. Central banks need only provide adequate supplies of base money, which they can do by means of open-market operations, leaving to the private market the task of reallocating reserves among solvent banks, perhaps through pre-established lines of credit. In traditional DNS arrangements, solvent banks that found themselves short of funds for settlement routinely relied on interbank loans to tide them over until they were able to replenish their reserves

130

G. Selgin

by liquidating non-reserve assets. CHIPS participants, for example, had up to 1 hour after clearing to acquire needed settlement funds.25 Most monetary authorities are of course unwilling to give up all or part of their power to extend ‘last resort’ loans, a power they regard as essential for containing systemic risk. But where deposits are largely insured, systemic risk has come increasingly to be identified with the risk of a wholesale payments crisis. Monetary authorities have tended, in other words, to argue in a circle, appealing to systemic risk problems originating in their own implicit guarantees as reason for providing those guarantees, while disguising the circularity of their arguments by misrepresenting DNS problems as market failures. In fact, the persistence and prominence of provisionality clauses in bank deposit contracts suggests that participants in DNS systems are, after all, not entirely convinced that their central banks will intervene to prevent a payments unwind.26 Although any positive probability of central-bank support does presumably reduce private parties’ incentives to monitor and control settlement risk, recent reforms, including modified DNS risk-sharing arrangements aimed at ‘securing’ those arrangements while converting provisional intraday payment messages into ‘final’ payments and (in the US) measures aimed at favoring Fedwire over CHIPS, only worsen the implied moral hazard, by replacing uncertain ex post finality guarantees with certain ex ante ones.27 Another source of government failure in DNS systems is the failure of courts to enforce private contracts. In the United States both the Federal Reserve and the courts can override private payment contracts. Prior to 1990 the rights and liabilities of parties to wholesale payment transactions were unclear, particularly with regard to the consequences of a settlement failure, owing to the lack of any case or statutory law addressing this problem (Mengle et al., 1987: 4). The very rarity of unwinds had the ironic effect of generating uncertainty regarding the enforceability of standard unwinding rules and provisionality clauses. Section 4A of the Uniform Commercial Code – the law that governed wire transfers in most of the United States throughout the 1990s – clarifies matters, but did so in part by declaring bank contracts providing for provisionality of wholesale payments to be generally unenforceable! The Code allowed for two exceptions, one to accommodate CHIPS, the other to accommodate the Automated Clearing House (ACH): a government-operated DNS system used mainly for smaller payments. The CHIPS exception, however, allowed its members to revoke payments in the event of a settlement failure only so long as they also took part in a special Lamfalussy-type loss-sharing arrangement that was not a traditional DNS feature.

‘Secured’ net settlement: chipping away at CHIPS Attempts to ‘secure’ DNS systems through regulations aimed at reducing their exposure to systemic risk have been a third source of government failure in wholesale payments. The special risk-sharing provisions first adopted by CHIPS are one of three means now employed on a worldwide basis to achieve such

Wholesale payments

131

security, the other two being ‘caps’ on DNS participants’ net bilateral and multilateral intraday clearinghouse ‘debits.’ Were such special arrangements and restrictions a response to genuine market failures, serving to ‘internalize the cost of third party risk’ (Folkerts-Landau, 1997: 6), they might contribute to the overall efficiency of payments. But when adopted in response to market-failure ‘mirages’ the efforts can undermine efficiency instead of enhancing it. Consider, for example, the effects of bilateral and multilateral debit caps. If it were really true that receiving banks in DNS systems, by failing to reject payment orders, became creditors to banks sending those orders and thereby passively exposed themselves to credit risk, then such restrictions might be the only reliable means (apart from switching to RTGS) for keeping intraday credit exposures within efficient bounds. But we have seen that receiving banks in DNS systems actually expose themselves to intraday credit risk only when they grant payment beneficiaries pre-settlement access to anticipated funds, and then only to the extent that they may be unable to recover funds advanced to those beneficiaries following an actual settlement failure. It follows that bilateral and multilateral caps may serve no purpose other than to alter and restrict artificially the flow of wholesale payments, presumably in a manner that makes such payments less efficient. Nor does the fact that many DNS systems, including CHIPS, who have adopted CAPS voluntarily, necessarily contradict this conclusion. Many ‘voluntary’ restrictions have been adopted only in response to encouragement by central bankers, whose advice may be coupled with implicit threats, including (for example) the threat to deny private DNS systems access to central bank final settlement facilities. Much the same may be said regarding special risk-sharing arrangements that CHIPS and other DNS systems were encouraged to adopt as a substitute for the standard practice of making all payments conditional upon settlement. Under the Lamfalussy provisions for secured DNS, if any DNS participant defaults, surviving participants are required to share responsibility for its net settlement obligations. Ironically, such special risk-sharing provisions can create precisely the sort of externality problem that was originally supposed to justify regulation of DNS systems, but which may not have been present in such systems before they became objects of regulators’ scrutiny (Rochet and Tirole, 1996). Market-failure mirages have thus served as the basis for lifting wholesale payments out of the unwind frying pan and into the moral-hazard fire.28 If attempts to ‘secure’ DNS systems may only have served to render them less efficient (if not more risky) than before, reforms that have abolished DNS systems altogether in favor of RTGS may involve even larger welfare losses. As noted previously, unless supplemented by underpriced intraday credits (which introduce a moral-hazard problem), RTGS imposes much higher liquidity or collateral costs on banks, forcing them to maintain higher average reserve balances, or to incur more liquidity risk, than would be the case under deferred net settlement. Wherever RTGS reforms have been motivated, not by private agents’ voluntary quest for efficiency but by regulators’ desire to correct non-existent market failures, reason exists for suspecting that the reforms have not been efficient.

132

G. Selgin

Conclusion: private contracts versus government policy The widespread view that unregulated private wholesale payment arrangements, and deferred net settlement in particular, promote excessive risk taking has been based upon regulators’ misunderstanding of (1) the manner in which private DNS arrangements generate and assign intraday credit risk and (2) the role that government guarantees have played in undermining otherwise sound privatemarket wholesale payments arrangements. Unlike Fedwire, CHIPS and other traditional DNS arrangements do not rely upon underpriced intraday credits, or rely upon them only to the extent that they must participate in an RTGS-DO system to effect final settlements. The only credits that are directly connected to intraday DNS payments are credits granted by receiving banks to their own customers; and these credits shift settlement risks to third parties only in so far as they are backed by extra-market finality guarantees. Recent reforms doing away with DNS systems, or subjecting them to restrictions or loss-sharing provisions that may run counter to what participants would have contracted for, have been inefficient if not counterproductive substitutes for reforms aimed at eliminating uncalled for finality guarantees extended by regulatory authorities. Although it may not be possible to reverse the wave of central bank sponsored reforms aimed at abolishing or at least ‘securing’ DNS systems, it is not too late for economists to strive for greater clarity concerning what such reforms have and have not accomplished. The reforms have succeeded in reducing or eliminating unwind risk in wholesale payments. In some cases, the reforms may have countered inefficiencies stemming from the presence of central bank guarantees or from the failure of courts to enforce private contracts. What the reforms cannot be shown to have achieved is the correction of any genuine market failure.

Notes 1 The author thanks Bill Bergman, Sandra Haasis, David Humphrey, George Kaufman, David Mustard, Harold Nitsch, Will Roberds, Art Snow, Larry Wall, Ron Warren, Lawrence H. White, several anonymous referees, and participants in the University of Georgia Economics Department workshop for their helpful comments and suggestions. 2 In January 2001, CHIPS switched from deferred net settlement to near-continuous settlement. 3 Since 1986 the Fed has also imposed limits, known as ‘net debit caps,’ on the maximum overdraft individual banks could obtain from it without being subject to special administrative actions. These caps appear, however, to have had only a very limited effect on the overall value of overdrafts (Hancock and Wilcox, 1996). 4 According to Zhou (1999), allowing for deductibles Fedwire’s ‘imputed average annual rate [was] only around 11 basis points’ after 1995. Zhou’s study is also valuable for its excellent critique of models (e.g. Freeman, 1999) that claim to demonstrate the optimality of a zero intraday lending rate. The essence of Freeman’s argument is that the assumption of aggregate default risk by central banks transforms that risk into less problematic inflation risk. Because Freeman’s model assumes an exogenous default rate, it cannot allow for the moral hazard problem induced by

Wholesale payments

5

6

7

8 9

133

having the costs of intraday borrowing borne largely by third parties. A model aimed at addressing the question of the optimal intraday borrowing rate should at least be capable of representing the problem of moral hazard! By a traditional or ‘unsecured’ DNS system I mean a system in which there are no caps or limits imposed on the volume of payment flows and where the failure of any participant to settle results in the cancellation of payment messages to and from that participant. The oft-cited Bank of New York computer glitch of December 1985 affected the government securities market only and did not result in any CHIPS unwind. The Fed automatically debited the Bank of New York’s reserve account to pay for securities it received on behalf of that bank’s clients, but because the glitch kept the Fed from knowing where to deliver the securities, the Bank of New York ended up with a $32 billion overdraft, which lasted for 90 minutes. CHIPS’s conversion to continuous settlement coincided with the establishment of a new private-sector network for foreign-exchange settlement: the Continuous Linked Settlement (CLS) Bank, aimed at avoiding Herstatt risk. Herstatt was a small German bank that failed in 1974, after having received irrevocable mark payments, but before the dollar leg of its transactions had been settled. Note that this was not a settlement failure in the sense used throughout the paper: Herstatt was closed at 08:30 h New York time, with no outstanding CHIPS payment orders. In 1995, wholesale payments in most nations were processed by DNS systems, and Switzerland was the only major country that relied upon real-time gross settlement for all of its large-value payments. Referring specifically to the New York Clearing House Association Cannon (1910: 209) observes that: The association is in no way responsible for the balances, except in so far as they are actually paid into the hands of the manager, and then its responsibility is strictly limited to the faithful distribution by him among the creditor banks of the amounts which he has received.

CHIPS today likewise disclaims responsibility for any obligations incurred by its members. 10 Even today most US banks provide their customers with ‘deposit account agreement and disclosure’ statements that include language like the following (from the Seaway National Bank of Chicago): All non-cash items (for example, checks) deposited to your Account are posted subject to our receipt of final payment by the payor bank. If final payment is not received . . . you authorize us to charge any of your Accounts, without prior notice and at any time, for the amount of the returned item, our returned item fee, any interest paid on the item, and any other fee we pay or incur . . .. With respect to wire transfers . . . you agree to enter into and comply with our wire transfer agreement. . . . Credit given by us to you with respect to [a] wholesale (wire) funds transfer entry is provisional until we receive final settlement for such entry through a Federal Reserve Bank. If we do not receive final settlement, you are hereby notified and agree that we are entitled to a refund of the amount credited to your account in connection with such entry, and the party (the originator of the entry) making payment to you via such entry shall not be deemed to have paid you the amount of the entry. 11 The practice of ‘unwinding’ transactions of failed clearinghouse participants, although rarely resorted to, goes back to the early days of clearinghouses. According to Cannon (1910: 277–278), members of the Chicago Clearing House (founded in 1865)

134

G. Selgin who have been so unfortunate as to fail have seldom attempted to clear on the day of their failure. . . . When, however, a failing bank has attempted to make its regular clearings and has not been able to pay the balance against it . . . all the exchanges presented by and against it have been returned and a new settlement made, the same as if it had not participated in the exchanges of the day. In this way, those who had presented checks against the failing member were enabled to return them to their customers, thereby intrenching themselves against loss.

Cannon goes on to observe that, although [n]o formal provision was made for such action at the onset, . . . as time passed on this course was found to be the best means by which to avoid serious complications. Accordingly, the rule has been embodied in the constitution. It is similar in its provisions to that existing in the constitutions of nearly all the clearinghouse associations of the country. 12 I abstract here from any possible influence of government guarantees, which are discussed below. 13 Some writers (e.g. Dale, 1998: 231–232) recognize that receiving banks might avoid intraday credit exposure in traditional DNS systems simply by refraining from releasing funds prior to settlement, but insist nevertheless that, insofar as funds are released during the day, resulting credit exposures ‘are the result, not of credit judgments made by the participant banks, but of customer transactions over which the participants have little or no control.’ The argument begs the question: what is it that compels the banks to release funds at once? 14 Mengle goes on to observe that this argument no longer holds if a sending bank participates in a DNS network indirectly, via a correspondent, because it is not reasonable to expect senders to ‘know’ their banks’ correspondents. This is true enough; however, the problem in question reflects, not any inherent shortcoming of DNS or check finality, but yet another regrettable consequence of historical restrictions on nationwide and international branch banking. In places (like Canada) where DNS arrangements have developed along with unrestricted branch banking, senders generally have the option of doing business directly with clearing banks. 15 In 1990, for example, CHIPS, anticipating the Bank of International Settlements’ Lamfalussy Report (which recommended that participants in DNS systems be required to collectively post enough collateral to cover default by any single participant, thereby further limiting the potential for a payments unwind), adopted rules and fees especially designed to limit its members’ exposure to systemic risk. It is not clear, however, that CHIPS considered these changes worthwhile except as a means for pre-empting more stringent government regulation. 16 Shen (1997: 51–53) offers an excellent discussion of systemic (‘liquidity’) risk in RTGS systems. He notes (Shen, 1997: 53) that, in the Swiss Interbank Clearing System (in which banks do not have access to central-bank supplied intraday credit), on an average day c.1997 ‘at least 45% of payments experience[d] some delay in their execution due to the lack of liquidity.’ See also Kahn et al. (2003). The reality of frequent payment order rejections in some actual RTGS systems is to be contrasted with the (so far) purely hypothetical possibility of a DNS-system unwind. The Fed, of course, avoids the problem of liquidity risk in Fedwire only by exposing itself to credit risk. 17 An anonymous referee, while agreeing that ‘the choice between RTGS versus DNS is a matter of relative costs,’ observes that ‘which system is more cost-effective is by no means a settled issue,’ and observes that I fail to supply ‘any new evidence’ of the cost-effectiveness of traditional DNS. That is true enough. However, the referee appears to forget that standard practice, informed by the welfare theorems of neoclas-

Wholesale payments

18

19

20 21 22

23

24

25

26 27

28

135

sical economics, places the burden of proof not on those who deny the presence of a market failure, but on those who insist that such a failure is present. The likelihood that either conventional accounting ratio analysis or CAMEL (capital, assets, management, earnings, and liquidity) ratings will fail to identify banks that are about to fail diminishes as the frequency of monitoring increases. Thus, of more than 1600 banks that failed between 1980 and 1994, only 16 percent had high (1 or 2) CAMEL ratings based on examinations made within a year of the banks’ failure, whereas 36 percent had ratings of 1 or 2 two years prior to failing. More frequent financial monitoring and public disclosure of CAMEL ratings and such could presumably go far in further reducing the (already low) probability of a DNS settlement failure. Angelini et al. (1996) performed a Humphrey-style simulation for Italy’s DNS system, which differed from CHIPS mainly in having many more (288) direct participants. They found that only 4 percent of these participants had the potential to trigger a systemic ‘crisis,’ and that in no case would the ‘crisis’ have caused more than seven other participants to default. The probability of no chain of defaults occurring following a single participant’s failure was 96 percent. This was again assuming, following Humphrey, zero recovery of provisional (intraday) credit to account holders. This is not to deny that a run to currency would, ceteris paribus, reduce the money multiplier, and would therefore necessitate some expansion of the monetary base to maintain the money stock. For a more general discussion of the incentive advantages of net settlement see Roberds (1999) and Kahn et al. (2003). In personal correspondence Roberds acknowledges the inherent risk-control advantages of net settlement, observing that ‘The problem with more modern net settlement systems is that the incentive to undertake [traditional] risk-mitigating activities has necessarily been diminished by the presence of central banks.’ That CHIPS participant practices were influenced by Fed guarantees of one sort or another is strongly suggested by the fact, reported to me by Humphrey, that those participants persistently refused to follow regulators’ suggestion that they purchase ‘settlement insurance’ on the grounds that ‘the risks [meaning, presumably, the premiums] were too great.’ During the 1980s, for example, approximately 90 percent of 418 banks that received extended emergency credit from the Fed failed subsequently, and most of them were known or at least suspected by regulators to have been insolvent when assistance was granted to them (Kaufman, 1999: 4). From 1857 to 1907, when interbank lending markets were poorly developed, members of CHIPS’ predecessor, the New York Clearing House Association, operated an effective coinsurance scheme by agreeing to accept fully collateralized clearinghouse ‘loan certificates’ in lieu of gold for settlements during financial emergencies (Timberlake, 1984). Were banks and their customers truly certain that last-resort loans would be employed to prevent a payments unwind, they would be indifferent between receiver finality and check finality rules (Scott, 1990: 186). Dale (1998: 229) points out that the very involvement of central banks ‘in the settlement process can lead to market expectations of official support in the event of any threatened disruption’ and that for this reason settlement risk exposure is best contained by means of settlement arrangements in which central bank involvement ‘is least evident.’ In CHIPS this particular externality problem was partly avoided by linking each participant’s share of losses to the bilateral credit line the participant has set for a defaulting bank. A reduced moral hazard is thus achieved at the expense of more rigidly constrained payments flows.

136

G. Selgin

References Angelini, P., Maresca, G., and Russo, D. (1996) ‘Systemic risk in the netting system,’ Journal of Banking and Finance, 20(5): 853–868. Board of Governors of the Federal Reserve System (1988) Controlling risk in the payments system, Washington, DC: Board of Governors of the Federal Reserve System. Cannon, J.G. (1910) Clearing houses, Washington, DC: Government Printing Office. Coase, R.H. (1988) The firm, the market and the law, Chicago, IL: University of Chicago Press. Dale, R. (1998) ‘Risk management and public policy in payment, clearing and settlement systems,’ International Finance, 1(2): 229–259. Dowd, K. (1994) ‘Competitive banking, bankers’ clubs, and bank regulation,’ Journal of Money Credit, and Banking, 26(2): 289–308. Folkerts-Landau, D. (1997) ‘Wholesale payments and financial discipline, efficiency, and liquidity,’ International Monetary Fund Working Paper No. 154. Freeman, S. (1999) ‘Rediscounting under aggregate risk,’ Journal of Monetary Economics, 43: 197–216. Gerson, J.E. (2001) ‘Theoretical approaches to catastrophic risk,’ unpublished thesis, University of Chicago. Hancock, D. and Wilcox, J.A. (1996) ‘Intraday management of bank reserves: the effects of caps and fees on daylight overdrafts,’ Journal of Money, Credit and Banking, 28: 870–908. Humphrey, D.B. (1986) ‘Payments finality and the risk of settlement failure,’ in A. Saunders and L.J. White (eds) Technology and the regulation of financial markets, Lexington, KY: Lexington Books. Kahn, C.M. and Roberds, W. (1999) ‘The design of wholesale payments networks: the importance of incentives,’ Federal Reserve Bank of Atlanta Economic Review, 84(3): 30–39. Kahn, C.M., McAndrews, J. and Roberds, W. (2003) ‘Settlement risks under gross and net settlement,’ Journal of Money Credit and Banking, 35: 591–608. Kaufman, G. (1999) ‘Do lender of last resort operations require bank regulation?,’ unpublished thesis, Loyola University Chicago. Mengle, D.L. (1990) ‘Legal and regulatory reform in electronic payments: an evaluation of payment finality rules,’ in D.B. Humphrey (ed.) The US payment system: efficiency, risk and the role of the Federal Reserve, Boston, MA: Kluwer. Mengle, D.L., Humphrey, D.B., and Summers, B.J. (1987) ‘Intraday credit: risk, value, and pricing,’ Federal Reserve Bank of Richmond Economic Review, 73(1): 3–14. Richards, H.W. (1995) ‘Daylight overdraft fees and the Federal Reserve’s payment system risk policy,’ Federal Reserve Bulletin, 81(12): 1065–1077. Roberds, W. (1999) ‘The incentive effects of settlement systems: a comparison of gross settlement, net settlement, and gross settlement with queuing,’ Bank of Japan Institute for Monetary and Economic Studies Discussion Paper No. 99-E-25. Rochet, J.-C. and Tirole, J.T. (1996) ‘Controlling risk in payment systems,’ Journal of Money Credit and Banking, 28: 832–862. Schoenmaker, D. (1995) ‘A comparison of alternative interbank settlement systems,’ London School of Economics Financial Markets Group Discussion Paper No. 204. Scott, H.S. (1990) ‘Commentary,’ in D.B. Humphrey (ed.) The US payment system: efficiency, risk and the role of the Federal Reserve, Boston, MA: Kluwer.

Wholesale payments

137

Shen, P. (1997) ‘Settlement risk in large-value payment systems,’ Federal Reserve Bank of Kansas City Economic Review 82(2): 45–62. Timberlake, R.H. (1984) ‘The central banking role of clearinghouse associations,’ Journal of Money Credit and Banking, 16: 1–15. Van den Bergh, P. and Veale, J.M. (1994) ‘Payment system risk and risk management,’ in B.J. Summers (ed.) The payment system: design, management, and supervision, Washington, DC: International Monetary Fund. Zhou, R. (1999) ‘Understanding intraday credit in large-value payment systems,’ unpublished thesis, Federal Reserve Bank of Chicago.

9

Central bank intraday collateral policy and implications for tiering in RTGS payment systems John P. Jackson and Mark J. Manning1

Introduction In this chapter we present a model of a Real-Time Gross Settlement (RTGS) payment system with tiered membership where settlement is facilitated by intraday credit extensions from the central bank. RTGS systems process and settle payment instructions individually in real time, ensuring intraday finality. Furthermore, central banks typically provide the settlement accounts across which payments are processed; hence, settlement is typically effected in central bank money, thereby eliminating counterparty risks between members once settlement has taken place. The model allows us to examine the key factors that influence both an agent’s decision over whether to participate directly in an RTGS payment system, and a central bank’s decision as to whether to require collateralization of intraday credit extensions to payment system participants. The design of all payment arrangements must reflect a trade-off between cost and risk. As noted in BIS (2005), ‘if a system was so costly or burdensome that no one used it, the system would have no effect on risks no matter how extensive its risk controls’. This applies as much to RTGS systems as to any other system design, for while addressing the counterparty credit risks associated with Deferred Net Settlement (DNS) systems, RTGS of payments can be a significant, and costly, drain on a bank’s liquidity (Kahn and Roberds, 2001). To alleviate this burden, central banks also typically offer intraday credit to payment system participants. In the absence of such credit, users would have to pre-fund their settlement accounts at the central bank, thereby incurring a substantial opportunity cost of holding liquidity.2 Chakravorti (2000), Kahn and Roberds (2001) and Bech and Garratt (2003) as well as the chapter by Bech et al. in this volume all highlight the behavioural implications of costly liquidity needs. They note that system participants might then seek to reduce these costs by delaying the submission of payments, with potentially adverse consequences for liquidity recycling in the system, operational risk, and, to the extent that obligations are not settled as expected, ultimately social welfare. Central bank provision of intraday credit to payment system participants entails a potential credit exposure. Several risk-mitigating measures might be taken, but these impose costs on payment system participants, and potentially

Central bank intraday collateral policy

139

also on society. Many central banks, including the Bank of England, require full collateralization of intraday credit exposures; this greatly reduces credit risk, but imposes an opportunity cost of posting collateral. By contrast, the US Federal Reserve does not require collateralization, but instead charges an interest rate on all intraday overdrafts and imposes credit limits on agents’ usage of intraday credit. Costly intraday credit can have similar behavioural consequences to prefunding requirements: prompting banks to economize on their usage of liquidity in the system by delaying payments. The trade-off faced by central banks, between assuming greater credit exposure and imposing costs on participants, has been subject to considerable scrutiny in recent years. Furfine and Stehm (1998), for instance, highlight the deadweight welfare losses associated with costly collateral requirements. They conclude that, from a social welfare perspective, a policy of free liquidity provision would be preferred to full collateralization unless the opportunity cost of collateral tended to zero. However, more recent work by Mills (2005) suggests that models of this type may not have adequately accounted for the credit risks faced by the central bank under zero collateralization. A related strand of literature focuses on how central banks, if they do require that intraday credit be collateralized, can reduce the opportunity costs incurred by system participants in posting such collateral. Manning and Willison (2005) show that allowing cross-border usage of collateral enables agents to economize on their total collateral holdings, while the chapter by Green in this volume suggests that central banks could accept less liquid (and hence lower cost) collateral than might other secured lenders. Alternatively, Willison (2005) considers recourse to more liquidity-efficient payment system designs, so-called hybrid systems, to reduce the amount of intraday credit needed to settle a given set of payments.3 Where significant costs of obtaining intraday credit remain, agents might choose not to participate directly in an RTGS system at all. The chapter by Rochet in this volume argues that to the extent that an agent chooses to by-pass an RTGS system, by entering into bilateral agreements with other agents or by shifting flows to a competing DNS system, systemic risk may be increased. Such alternative arrangements typically include recourse to a correspondent bank, who processes payments on behalf of indirect system participants. This phenomenon, known as ‘tiering’, is a commonly observed feature in many RTGS systems internationally. CHAPS Sterling, with only 13 out of around 350 commercial deposit-taking banks operating in the United Kingdom participating directly, is particularly highly tiered (Harrison et al., 2005). In this chapter we explore the implications tiering can have on the welfare costs associated with risk mitigation in payment arrangements. Our starting point is an insight from Kahn and Roberds (2006), who identify delegated monitoring as an alternative enforcement device to collateralization. They show that, in the presence of private information about the reliability of agents, delegated monitoring can economize on the need for agents to post collateral to guarantee repayment of intraday credit. This monitoring is achieved

140

J.P. Jackson and M.J. Manning

through a tiered structure, whereby a direct participant of a payment system absorbs the risk associated with credit extensions to its customer banks, hence maintaining good incentives to monitor. When default states occur only a fraction of the time, a full collateralization policy can achieve only a second-best outcome; any collateral posted to the central bank in non-default states is a deadweight loss to society, arising as a result of the central bank’s imperfect information about settlement banks’ credit quality and their wish to minimize their own credit losses (and hence costs to the taxpayer). Under such a scenario, if monitoring is sufficiently accurate, and monitoring costs sufficiently low, delegated monitoring can achieve a smaller deviation from the first-best. In this chapter we highlight two additional channels by which tiering in payment systems might lead to a reduction in the deadweight social costs associated with full collateralization of intraday credit extensions by the central bank. First, we consider internalization of payments. This refers to a situation where payments made between customers of the same correspondent bank are settled internally across the correspondent’s books, without being processed through the payment system. Internalization allows payments to be made without recourse to intraday credit from the central bank, thereby avoiding any costs associated with collateral posting requirements. Therefore, to the extent that tiering facilitates the internalization of payments, it can reduce the costs imposed by a central bank’s full collateralization policy. Furthermore, to the extent that payment flows from a correspondent’s customers to other first-tier participants are likely to be spread out through the day, there may be a diversification, or ‘collateral-pooling’ benefit. That is, unless customers’ payment flows are perfectly correlated, the total pool of collateral required to generate intraday credit on behalf of several customer banks, will be smaller than that required were each customer’s collateral needs served from segregated pools of collateral. Where payments made by second-tier participants are not internalized, but rather are effected by the correspondent over central bank settlement accounts, any collateral-posting requirements at the central bank would apply. But here too, tiering may reduce the deadweight social costs of collateral if agents with high opportunity costs of posting collateral are able to take advantage of lower collateral-posting costs enjoyed by their correspondent. The opportunity cost of posting collateral may be proxied by the reverse-repo spread (i.e. the spread between secured and unsecured borrowing costs). This may vary across agents, according to differences in credit-worthiness which affect the unsecured cost of borrowing. Also, market imperfections and regulatory policy may influence agents’ relative opportunity cost of collateral. For example, in the United Kingdom, banks subject to the Stock Liquidity Requirement (SLR), a prudential liquidity regime, are able to meet intraday collateral requirements using assets that they have to hold, in any case, to meet their prudential requirement at the end of the day; hence they have a very low opportunity cost of posting collateral intraday.4

Central bank intraday collateral policy

141

The foregoing discussion highlights the potential benefits of tiering. However, tiering can also introduce additional risks to the system. First, a correspondent bank might not have sufficient incentive to monitor because it does not fully internalize the potential systemic consequences of a customer’s default which triggers liquidity or solvency problems of its own. Even if sufficient incentives were to exist, monitoring of second-tier agents by a correspondent might not be accurate, causing the correspondent to incur credit and liquidity exposures that might in turn lead to wider contagion. Harrison et al. (2005) analyse the credit risk implications of the highly tiered structure of the UK large value payment system, concluding that this channel might not impose significant risks on the system as a whole, except in extreme circumstances. However, other risks exist. For instance, internalized payments might be subject to greater legal risks as they are not likely to be covered by provisions providing protection against bankruptcy law (such as the Settlement Finality Directive in the European Union). Tiering increases the risk that operational or financial problems at a settlement bank lead to disruption of payments in a large part of the system. Another important concern is that, in response to liquidity problems among second-tier banks, settlement banks might decide to cut or restrict intraday credit, further exacerbating these liquidity problems. In this chapter, we apply a simple model of an RTGS payment system, in which agents rely on the central bank to provide intraday credit to facilitate settlement of a single payment obligation. We allow the paying agent to choose whether to be a direct participant of the payment system, or to settle its obligation via an existing direct participant. With complete information as to the determinants of the agent’s choice, the central bank chooses whether or not intraday credit should be fully collateralized.5 Using this framework, we are able to show that, when the central bank requires full collateralization, it may be optimal for an agent to become an indirect participant, so as to take advantage of cost-efficiency benefits arising from monitoring as a substitute for collateralization. These benefits are increased to the extent that payments can be internalized and that agents can take advantage of their correspondent’s lower collateral posting costs.6 Furthermore, in the absence of spillover risks from tiering, private and social costs are aligned when the central bank opts for full collateralization. Therefore, any private costefficiency benefit derived translates directly into a social welfare improvement. We do show, however, that a wedge between social and private costs is likely to exist under zero collateralization and that, although welfare might be maximized if the central bank requested zero collateral and the agent chose to access the system indirectly, this outcome is not achievable: it will always be in the agent’s interest to access directly under zero collateralization. Unless the probability of default is very low, this will rarely be optimal for the central bank. With imperfect monitoring and the potential for systemic spillovers from tiering, a wedge will also emerge between social and private costs when the central bank opts for full collateralization, with this wedge increasing in the degree of monitoring imperfection, the value of payments to be settled and the

142

J.P. Jackson and M.J. Manning

degree of spillover per unit of exposure. In this case, policy intervention might be desirable to address the risks introduced. The chapter is organized, as follows. We first outline our analytical framework. We then apply this framework to analyse agents’ decisions under alternative scenarios for the quality of monitoring and the existence of tiering externalities before offering some conclusions.

Analytical framework In this section, we present a simple model of payment arrangements to explore two key decisions: a bank’s decision as to whether to access an RTGS payment system directly, or via a correspondent banking arrangement and a central bank’s decision as to whether to collateralize intraday credit extensions. Our analysis draws on the framework presented in Kahn and Roberds (2006), but applies this in a much-simplified, stylized and reduced-form fashion.7 We first provide an overview of the model set-up, going on to describe in greater detail the actions taken and costs incurred under each alternative arrangement. The model set-up and timeline for actions The essence of the model is a game of complete information, with actions taken sequentially by two players: the central bank; and a commercial bank, C. There are two further agents in the game, banks A and B, both of whom are direct settlement members of the payment system, with A also a potential provider of correspondent banking services to C. Bank B never provides payment services. All agents are assumed to be risk-neutral. Neither bank A nor bank B take any direct actions in the game, although we do establish the terms on which A provides correspondent banking services if called upon to do so, ensuring that it would be rational for A to offer such services. Bank C makes a single payment, of value unity.8 No other payments are made. Time consists of a single day, divided into four periods. In period 0 the central bank sets its collateral policy with respect to bank C, choosing actions from the set {F,Z}, where F = full collateralization; and Z = zero collateralization, so as to minimize expected social costs.9 In period 1, observing (with certainty) the central bank’s policy choice, C minimizes its expected costs with respect to its decision as to whether to fulfil a single payment obligation directly in the RTGS payment system, or via correspondent banking services provided by direct payment system participant, A. Its set of potential actions is then {D,I}, where D = direct participation; and I = indirect participation. We assume that, if indifferent, C will always choose to participate directly. The state of the world is characterized by {e,}. Parameter e [0,1] describes the possible orientation of payment flows in the system; with probability (1 – e), C is obliged to make a payment to A, whereas with probability e, C has an obligation to B. As will be discussed below, the orientation of payment flows has

Central bank intraday collateral policy

143

implications for the degree of internalization possible when C’s payments are settled indirectly via A. The orientation of payment flows is realized by a draw from nature in period 2 and revealed immediately to the agents. Parameter  [0,1] is the probability that C suffers an exogenous default shock which prevents it from repaying intraday credit extended by either A or the central bank.10 The incidence of a default shock is also realized by a draw from nature in period 2. The outcome is not revealed to agents until period 3, although a signal as to whether or not a shock has occurred can be obtained by monitoring in period 2. Settlement of C’s payment obligation occurs in the payment system in period 2. It is assumed that C has no endowment of the settlement asset at the start of the period and hence always requires an intraday credit extension, either from the central bank or from A, before settlement can be effected. Equally, we assume that A requires intraday credit from the central bank before it can make a payment on C’s behalf. All intraday credit extended in period 2 is to be repaid by the end of period 3, by which time C expects to have received a sufficient quantity of the settlement asset (from a maturing investment). Parameter  may be interpreted as the probability that this investment fails and returns nothing. If settling directly in the system under full collateralization, C also posts collateral in period 2 to support its request for intraday liquidity. C then settles its obligation with finality in central bank money.11 If bank A is settling on behalf of C, A first obtains intraday credit from the central bank. Any cost to A of posting collateral to the central bank is passed on directly to C.12 This collateral requirement in respect of C’s payments under an indirect arrangement can, however, be reduced by internalization. When C has a payment obligation to A, this can be settled directly on A’s books. Only if C has to make a payment to B will settlement occur in central bank money. This is shown in Figure 9.1 below. Whether settling C’s obligation in central bank money, or internalizing across its own books, A settles C’s payment obligation in advance of the receipt of funds from C; that is, A extends intraday credit to C. A can potentially economize on collateral sought from C in respect of such a credit extension by carrying out monitoring.13 More specifically, A monitors to obtain a signal as to whether C has suffered a default shock in period 2. If monitoring reveals that a A

B

C C has an obligation to A: Probability  (1e) Payment in central bank money

A

B

C C has an obligation to B: Probability  e Payment in commercial bank money

Figure 9.1 Payment flows when C accesses the system via bank A.

144

J.P. Jackson and M.J. Manning

Period 0: The central bank sets its collateral policy with respect to intraday credit extensions to C.

Period 2: Nature determines the recipient of C ’s payment obligation and C ’s default shock is realised (but not revealed to agents). If settling for C, A monitors to obtain a signal as to whether C has suffered a default shock. Intraday credit is granted and C posts collateral if required to do so. Settlement occurs.

Time

Period 1: C decides whether to fulfil its payment obligation directly in the payment system; or indirectly via A.

Period 3: C ’s default shock is revealed to agents. Intraday credit repaid if no shock has occurred.

Figure 9.2 A time-line for actions.

default shock has occurred, collateral will be sought from C in respect of intraday credit granted; otherwise A will not require payments to be collateralized. Any monitoring costs incurred by A are passed on to C. If no default shock arises in period 3, all intraday credit is repaid. Otherwise, C defaults on the repayment of its intraday credit. Unless sufficient collateral has been posted, this will impose default costs upon agents in the system. These costs will be described below. The timeline of the model is shown in Figure 9.2. Agents’ actions and costs Characterization of bank C’s costs Bank C’s costs are characterized by: E[CC] = f (M,C,A), where M is A’s cost of monitoring C, which, to the extent incurred, will be passed on to C; C is C’s private opportunity cost of posting collateral either to A or the central bank; and A is A’s private opportunity cost of posting collateral to the central bank, which again will be passed onto C.14 We consider each in turn. As noted above, by incurring a monitoring cost, M, A can obtain a signal as to whether C has suffered a default shock and hence will be unable to repay an intraday credit extension. It is initially assumed that the signal obtained by monitoring is perfectly correlated with the shock; we later relax this assumption and allow for Type I and Type II errors. Private monitoring costs might be expected to be relatively low to the extent that the normal-course interconnection between financial institutions ensures a steady flow of information between them. Indeed, this implies that larger banks with diversified activities and a wide network of clients in non-payments-related businesses, will have access to more private information, and hence be better monitors. Furthermore, there are likely to be economies of scale in monitoring activity.

Central bank intraday collateral policy

145

Turning to collateral costs, it is worth noting that where the costs faced by a correspondent bank are lower than those faced by its customer banks, indirect participation may allow agents with high collateral costs to take advantage of lower costs enjoyed by other agents. Recall that in the model, any costs incurred by A in respect of settlements effected on behalf of C will be passed on in full; but the lower these costs are relative to C’s own direct collateral costs, the more efficient indirect participation will be relative to direct membership. Should C suffer a shock in period 3 and consequently fail to repay its intraday loan, additional private costs of default might be incurred. For simplicity such costs are normalized to zero. To further simplify the exposition, we also normalize to zero any fixed costs (technological and fees) associated with linking directly to the payment system or indirectly through A. When settling directly, C faces a collateral cost of C if the central bank requires full collateralization. Bank C’s costs under indirect participation depend on the relative costs of monitoring and collateralization, and the potential for internalization. We assume that A prices its correspondent services competitively, subject to a full cost–recovery constraint, and offers two alternative correspondent banking services: one involving monitoring; and another involving full collateralization. Bank C chooses the service that offers the lowest expected cost. We assume that C has no private information about its likelihood of experiencing a default shock, and that this is common knowledge. As the full collateralization service offered by A can never be cheaper than settling directly at the central bank (C faces cost, C, in both cases), and under the assumption that, where the costs of direct (D) and indirect participation (I) are equal, C will choose D, it is clear that C will never choose A’s full collateralization service. Hence, indirect participation will always be associated with monitoring. Where A does monitor C, and the signal obtained is perfectly correlated with the incidence of default, it will only be optimal for A to request collateral in default states (at a cost of C to C), but not otherwise. Bank C will, however, still have to compensate bank A for any additional collateral it may be required to post to the central bank, implying an extra cost above the direct cost of posting collateral to A of e(1 – )A. This assumes that  of the time A will simply use collateral posted by C to cover any requirement at the central bank. The expected cost to C is then: E[CC] = M + C + e(1 – )A

under (F,I)

E[CC] = M + C

under (Z,I)

Central bank’s costs We assume that the central bank makes decisions in order to minimize expected social costs. These costs are characterized by E[CCB] = f (M,C,A,S), where the

146

J.P. Jackson and M.J. Manning

first three cost parameters are private costs, as above, and S is the social cost of default by C when intraday credit is not fully collateralized. Monitoring and collateral costs are included here as these are deadweight costs to society arising from informational asymmetries and limited enforcement. A social cost S  1 arises when C cannot repay an uncollateralized intraday loan at the central bank and hence the central bank is forced to cover the resulting loss via taxation of unmodelled agents. The extent to which S exceeds unity reflects any distortion imposed by the tax. If, on the other hand, default occurs and collateral has been pledged, the defaulting agent’s creditor (either the central bank, or bank A under a tiered arrangement) may attach the defaulter’s collateral up to the amount of the pledge. For simplicity, we assume no market risk to the value of collateral posted.15

Agents’ decisions In this section, we apply the framework described above to establish equilibrium outcomes for central bank collateral policy and the degree of direct participation in payment systems. We consider three alternative cases: (i) perfect monitoring; (ii) imperfect monitoring; and (iii) imperfect monitoring with tiering spillovers. Case 1: perfect monitoring and agent-specific collateral costs In this case, we assume that monitoring reveals default states with certainty, and allow for banks A and C to face different opportunity costs of posting collateral. We solve the model by backwards induction, applying sub-game perfection as an equilibrium concept. Accordingly, we begin with C’s decision. Consistent with the earlier discussion, C faces the expected private costs detailed below: (Z,D): 0 (F,D): C (Z,I): min[M + C,C] (F,I): min[M + C + e(1 – )A,C] where, to establish expected costs, states of the world are weighted by the probabilities e and . Comparison of C’s expected costs immediately reveals that it will always be optimal for C to access the payment system directly if the central bank adopts a policy of zero collateralization (since min[M + C,C] > 0). In the event that the central bank chooses F, we see that C constitutes an upper bound for costs under indirect participation. Hence, in this case, C will certainly opt for indirect participation if the inequality in (1) holds.

Central bank intraday collateral policy M C > M + C + e(1 – )A ⇒ C >  + eA (1 – )

147 (1)

If this inequality does not hold, expected costs under (F,I) will equal C, leaving bank C indifferent. Again, we assume that, if indifferent, C will settle directly.16 It is clear from the inequality in (1) that C’s choice under full collateralization will depend on several parameters. In particular, the inequality in (1) is more likely to hold, and hence C is more likely to participate indirectly, the higher is C and the lower are , M, e and A. Inequality (1) illustrates that, with A < C, C can take advantage of A’s lower collateral costs by choosing to be an indirect member. With A < C = , this potential efficiency is no longer available and the inequality in (1) reduces to  > M/(1 – )(1 – e), which is less likely to hold. In period 0, the central bank chooses its actions, anticipating the choices C will make in response in period 1. The central bank’s expected (social) costs are given by: (Z,D): S (F,D): C (Z,I): min[M + C,C] (F,I): min[M + C + e(1 – )A,C] It is important to note that social and private costs are equivalent in all cases with the exception of when uncollateralized exposure is retained (and hence there is some probability that social default costs are suffered): i.e. the case with (Z,D). No externality exists under (Z,I) as A fully absorbs the shock of any default by C. This assumption will be relaxed later. Given that C will choose to participate directly if the central bank chooses Z, the central bank will compare expected social costs under (Z,D) with those under C’s optimal response to a policy of full collateralization. If the inequality in (1) holds, the relevant comparison is with expected costs under (F,I); if the inequality in (1) does not hold, the relevant comparison is with expected costs under (F,D). Depending on whether the inequality in (1) holds, the central bank will choose zero collateralization if either S < C or S < M + C + e(1 – )A (whichever is relevant, given parameter values); and full collateralization otherwise.17 Intuitively, then, ceteris paribus, zero collateralization and direct participation by C is more likely, the lower the probability that C defaults and the social costs associated with default; the higher the opportunity cost of posting collateral for either (or both) A and C; the lower the cost of monitoring; and the lower the probability of internalization.

148

J.P. Jackson and M.J. Manning

A graphical illustration It is instructive to illustrate agents’ choices graphically, so as to draw out their important determinants and identify potential sources of divergence of public and private interests. Figures 9.3 and 9.4 trace private and social costs with varying M, for given values of A, C, S,  and e. Figure 9.3 presents a case with a high degree of internalization (a 90 per cent probability that payments will be internalized). It is clear that, with zero collateralization, C will choose direct participation; and, with full collateralization, C will prefer indirect participation for all values of M up to the threshold X shown in the figure. Thereafter, direct participation will be chosen. Given these responses, the central bank will choose full collateralization: it is clear that the social cost of the combination (Z,D) is higher than that associated with C’s optimal choices under full collateralization, for all M. However, this is not the socially optimal outcome. It is clear from the figure that (Z,I) would maximize social welfare for all values of M shown. However, this first-best outcome is unachievable because C makes its choice after the central bank, and (Z,D) offers lower private costs for all M. Hence, under such a scenario, there is a wide interval of monitoring costs within which it seems that policy intervention might be justified to steer the market towards the socially optimal outcome of indirect participation when the central bank chooses not to require that credit extensions be collateralized. However, it should be recognized that this initial scenario has perfect monitoring and no tiering risks/spillovers, which, as we will show, will leave (Z,I) socially preferred for a narrower range of monitoring costs, if preferred at all. Also, we have assumed that A faces no risk of an exogenous default shock and hence the central bank’s credit extension to A is, in this scenario, riskless. 0.0020 (Z,D) Social

0.0018 0.0016

0.0012

X

0.0010

(F,I ) Social  Private

0.0008 0.0006 (Z,I ) Social  Private

Social/private cost

0.0014

(F,D) Social  Private

0.0004

(Z,D) Private

0.0002 0

0

0.0003

0.0006

0.0009

0.0012

0.0015

Monitoring cost (M)

Figure 9.3 High degree of internalisation (C = A = 0.0015; S = 1.1;  = 0.0015; e = 0.1).

Central bank intraday collateral policy

149

0.0020

X

0.0018

(F,D) Social  Private

0.0016

0.0012

(F,I ) Social  Private (Z,D) Social

Y

0.0010 0.0008 0.0006

(Z,I ) Social  Private

Social/private cost

0.0014 W

0.0004

(Z,D) Private

0.0002 0

0

0.0003

0.0006

0.0009

0.0012

0.0015

Monitoring cost (M)

Figure 9.4 Low default probability (C = A = 0.0015; S = 1.1; = 0.00075; e = 0.5).

Figure 9.4 presents a scenario with a low default probability. Here, the lowest expected private cost for C is again associated with the combination (Z,D). Under full collateralization, C would prefer indirect participation for values of monitoring cost up to threshold value X; above these values direct participation would be optimal. Given C’s responses, and the low default probability in this scenario, the central bank would choose full collateralization for values of monitoring cost up to threshold value W. Beyond this point, the central bank would favour zero collateralization. Interestingly, social and private preferences are aligned beyond Y in this case, reflecting C’s low default probability and hence the relatively low social costs associated with outcome (Z,D). Below Y, however, the first-best outcome (Z,I) is again unachievable, although the caveats noted above remain relevant in this regard. The profile of payment system participation in the UK The UK experience is consistent with the broad predictions of the model as stated above. In the United Kingdom the Bank of England requires full collateralization of intraday credit extensions.18 A sub-set of banks (the UK-owned banks) face very low opportunity costs of collateral due to the fact that assets held to meet prudential regulatory requirements can be used to back intraday liquidity needs. Furthermore, correspondent banking is highly concentrated, and becoming more so, with just three banks providing the bulk of these services. Thus, a high degree of internalization takes place.19 These correspondent banks are all UK-owned, and hence all benefit from, and offer, a low opportunity cost of collateral. Foreign-owned banks therefore have a strong incentive to participate indirectly, taking advantage of a high C – A and a

150

J.P. Jackson and M.J. Manning

low e (i.e. a high degree of internalization). The trend towards concentration in correspondent banking is thus largely self-fulfilling, particularly to the extent that economies of scale exist in monitoring. And, with the three large correspondent banks in the UK all major banks with diversified businesses, it is likely that they also have better access to their customer banks’ private information through other business lines than would smaller banks, and hence can offer a ‘cheaper’ correspondent service. Thus, despite the UK’s role as an international financial centre and a high foreign presence in sterling markets, just 15 per cent of daily value flowing through the large-value payment system is represented by the three foreign-owned direct participants.20 Case 2: imperfect monitoring and agent-specific collateral costs To assume perfect monitoring, as in case 1 is, perhaps, a bit strong. In this subsection, we relax this assumption to allow for Type I and Type II errors in monitoring. That is, with some probability, , bank A fails to ask for collateral and a default occurs (a Type I error); and with some probability, , bank A mistakenly identifies a state as a default state, and hence requests collateral unnecessarily (a Type II error). As bank A fails to obtain collateral in some default states, some uncollateralized exposure will be retained in the system under indirect participation. More specifically, when A monitors under full collateralization and indirect participation, C’s expected private costs are augmented by ( – )(C – eA) + . The first term captures the direct collateral cost incurred by C, adjusted for the correspondingly smaller pass-through of A’s collateral costs vis-à-vis the central bank. The second term captures costs imposed upon A, , in the event that C defaults and A is uncollateralized. Under the assumption that A is aware that it will retain uncollateralized exposure  of the time, these costs would be passed on to bank C. But, of course, with ( – )(C – eA) +  strictly positive when C > A, it is less likely that A’s correspondent service with monitoring will entail a lower cost to C than direct participation with full collateralization; that is, it is less likely that the inequality in (2) below will hold. C > M + ( +  – )C + e(1 – ( +  – ))A + 

(2)

Imperfect monitoring therefore has significant implications for both private and social costs under each of the policy/participation states involving indirect participation. As it has the effect of increasing costs, indirect participation is less likely than in case 1. More formally, bank C’s expected costs under the four possible strategy pairs become: (Z,D): 0 (F,D): C

Central bank intraday collateral policy

151

(Z,I): min[M + ( +  – )C + ,C] (F,I): min[M + ( +  – )C + e(1 – ( +  – ))A + ,C] It remains the case that, under zero collateralization, C will choose direct participation. With the imperfection in monitoring, however, it becomes less likely that indirect participation with monitoring will be a low-cost outcome and hence more likely that C’s costs under (F,I) will be equivalent to those under (F,D). Hence, given that, if indifferent, C will participate directly, (F,D) is more likely to be favoured. The central bank faces the following expected social costs: (Z,D): S (F,D): C (Z,I): min[M + ( +  – )C + ,C] (F,I): min[M + ( +  – )C + e(1 – ( +  – ))A + ,C] As before, it is known that C will choose to participate directly if Z is chosen and with (F,D) now more likely to be favoured by bank C in the event that the central bank chooses strategy F, the most relevant comparison may well be between expected social costs under (Z,D) and (F,D). Also, given that the costs associated with indirect participation and monitoring are higher in this scenario, the (unachievable) (Z,I) outcome is likely to be the socially preferred outcome for a much smaller interval of values for M (if, indeed, socially preferred at all). The foregoing has an interesting and important implication. In particular, given that A is more likely to resort to costly full collateralization when monitoring is imperfect and there is a risk of retaining uncollateralized exposure, the potential efficiency benefits associated with delegated monitoring are lost. Hence, it would appear that welfare could be improved if the banks accessing the system directly and providing correspondent banking services were ‘better monitors’: i.e. they had better skills or better information in this regard, which ensured that both and  (and particularly the latter) were low. To the extent that monitoring quality is improved when banks are large and diversified, and, hence, have better access to private information, such banks should be encouraged to participate directly and to provide correspondent banking services. Figure 9.5 illustrates that imperfect monitoring creates a smaller range of monitoring costs for which C will choose to be an indirect participant (the threshold value of M moves from Y to X) and shows that the effect of better monitoring would be to narrow the horizontal distance between the perfect and imperfect monitoring thresholds. Case 3, however, offers a qualification to this conclusion where tiering imposes a spillover.

152

J.P. Jackson and M.J. Manning 0.0020 (Z,D) Social

0.0018 0.0016

0.0012

X

0.0010

(F,I ) Social  Private

0.0008 0.0006 (Z,I ) Social  Private

Social/private cost

0.0014

(F,D) Social  Private

0.0004

(Z,D) Private

0.0002 0

0

0.0003

0.0006

0.0009

0.0012

0.0015

Monitoring cost (M)

Figure 9.5 The impact of imperfect monitoring (C = A = 0.0015; S = 1.1;  = 0.0015; e = 0.5; = 0.00075;  = 0.0005).

Case 3: imperfect monitoring and spillovers under indirect participation In the model, as presented in cases 1 and 2, indirect participation introduces no spillover risk to the payment system, or to the wider economy. Indirect participation is an equilibrium choice for C only when the central bank adopts a strategy of full collateralization; thus, even if C were to default with A uncollateralized, there would be no disruption to payments in the system and no spillovers. We do, however, allow for the possibility, under (F,I), that C’s default could impose costs on A, with an expected value of . While A seeks compensation from C for these expected losses (by passing through the expected cost), an actual loss of one unit would be suffered in those states of the world in which default occurred. This could cause liquidity, or even solvency, problems at A which might spill over into the system more widely. While we do not explicitly model the channels by which losses might be transmitted under such a scenario, we can attempt to capture possible spillovers by introducing an additional ‘tiering risk’ term, T, to expected social costs under indirect participation. By definition, T is an externality, and hence will not be internalized by C or A. It might be interpreted as reflecting the potential that A fails to fully internalize the risk that a failure of C will trigger A to suffer liquidity or solvency problems that could have knock-on effects on the financial system. In addition, it might capture the potential that A fails to internalize the risk that operational problems to itself would disrupt C’s payments. Expected social costs under (Z,I) and (F,I) thus become:

Central bank intraday collateral policy

153

(Z,I): If C > M + ( +  – )C + , then M + ( +  – )C + T (F,I): If the inequality in (2) holds: M + ( +  – )C + e(1 – ( +  – ))A + T otherwise: C Allowing for this tiering externality, a wedge is introduced between social and private costs under full collateralization and indirect participation (illustrated in Figure 9.6). Specifically, there exists a range of monitoring costs, between X and Y in the diagram, in which (F,I) is preferred to (F,D) for C, but (F,D) is preferred to (F,I) for the central bank. That is: ∑ + T > C > ∑ + 

(3)

where ∑ = M + ( +  – )C + e(1 – ( +  – ))A. The range of monitoring costs for which the inequality in (3) holds is increasing in T. However, we know from the inequality in (2) that, for C, (F,I) is only likely to be preferred to (F,D) for low values of . For larger values, the costs associated with indirect participation and monitoring would exceed those associated with full collateralization. Hence, it would never be optimal for indirect participation to be chosen by C, limiting the potential incidence of tiering externalities, and hence the potential impact of T. Nevertheless, the existence of this wedge indicates that there might be a role for additional policy intervention to mitigate tiering externalities. 0.0020 (F,I ) Social

0.0019 Y

(F,D) Social  Private

0.0018

0.0016 0.0015 0.0014 0.0013 (F,I ) Private

Social/private cost

0.0017

X

0.0012 0.0011 0.0010

0

0.0001

0.0002

0.0003

0.0004

0.0005

Monitoring cost (M)

Figure 9.6 The impact of tiering risk (C = A = 0.0015; S = 1.1;  = 0.0015; e = 0.5; = 0.00075;  = 0.0005; T = 1.4).

154

J.P. Jackson and M.J. Manning

Indeed, until now we have worked with a value of C’s payment obligation of unity. Generalizing this, by allowing C’s payment to take the value, P, we can capture the policy implications associated with payment system participants of different ‘sizes’. In particular, we find that the interval of monitoring costs over which social and private incentives under full collateralization are misaligned is increasing in the value of payments. That is, for a given error in monitoring, , and a given degree of spillover, T, higher payment values lead to higher potential uncollateralized exposures and hence potentially greater social losses. More formally, bank C’s and the central bank’s expected costs under full collateralization are given below. In these expressions, we assume that monitoring costs are invariant with respect to P.21 But we also recognize that total collateral costs and spillover costs will, ceteris paribus, be increasing in step with the value of payments. Hence, as P increases, the expected cost schedules shift upwards. Bank C: (F,D): PC (F,I): min{M + P[( +  – )C + e(1 – ( +  – ))A + ],PC} Central Bank: (F,D): PC (F,I): If M + P[( +  – )C + e(1 – ( +  – ))A + ] < PC, then M + P[( +  – )C + e(1 – ( +  – ))A + T]; otherwise, PC Figure 9.7, drawn in an analogous fashion to Figure 9.6, illustrates the implication of increasing payment values, or increasing size of payment system participants. Note the difference in the scales of the two figures, reflecting the fact that, with a ‘high P’, expected costs are significantly higher. The important observation from Figure 9.7 is that the interval X–Y covers a much wider range of monitoring costs and hence tiering spillovers may be a much more significant policy concern when ‘large’ payment system participants settle indirectly. Policy alternatives As a result, policymakers may wish to consider policy options that either reduce the size of the wedge, or encourage large payment system participants to settle directly. A number of policy options might be considered in this regard. First, banking supervisors might consider more stringent ex ante capital

Central bank intraday collateral policy

155

0.0090 (F,I ) Social

Y

(F,D) Social  Private

0.0085

X

0.0080

0.0070 0.0065

(F,I ) Private

Social/private cost

0.0075

0.0060 0.0055 0.0050 0

0.0003

0.0006

0.0009

0.0012

0.0015

Monitoring cost (M)

Figure 9.7 The impact of tiering risk with high payment value (C = A = 0.0015; S = 1.1;  = 0.0015; e = 0.5; = 0.00075;  = 0.0005; T = 1.4; P = 5).

and/or liquidity regulation to encourage correspondent banks to internalize the externalities associated with intraday credit extensions to indirect payment system participants. Second, steps might be taken to ensure the quality of monitoring carried out by correspondent banks, perhaps via improved accounting and disclosure standards for financial institutions. Third, direct participation might be encouraged by efforts to reduce either the quantum of liquidity required to effect payments in the system; or the opportunity cost of collateral faced by prospective direct members. The liquidity burden might be addressed by introducing a more liquidity-efficient payment system design; e.g. introducing queuing or netting algorithms, or allowing certain payments to be settled on a deferred net basis. And collateral costs might be lowered by broadening the eligible collateral list to include less liquid assets; or allowing greater fungibility of collateral across borders or across systems.22 Were such measures to be successful in encouraging direct participation of large and diversified banks that enjoyed superior access to private information and were such banks to then offer correspondent banking services, this might also ultimately achieve an improvement in monitoring quality.

Conclusions The model developed in this chapter can be used to examine the key factors influencing both an agent’s decision over whether to participate directly in a payment system and a central bank’s decision as to whether to require collateralization of intraday credit extensions to payment system participants.

156

J.P. Jackson and M.J. Manning

Consistent with the existing literature in this area, we show that a central bank will be more likely to require full collateralization of intraday credit when participants have high default probabilities, or face low collateral costs. However, a contribution of this chapter is to show that, for full collateralization to be a rational policy choice for the central bank, it is only necessary that a subset of agents have low collateral and monitoring costs; other agents can take advantage of these low costs if they become indirect participants. We also find that internalization is likely to make indirect participation more attractive under full collateralization. This suggests that economies of scale exist in correspondent banking and implies that significant concentration is likely to be observed in the provision of correspondent banking services (particularly where a subset of agents face particularly low collateral posting costs). And to the extent that economies of scale also exist in monitoring, one would expect banks to gravitate towards the larger, cheaper service-providers. Collateral pooling benefits, which we do not model here, are also likely to support such concentration. The model’s predictions are, prima facie, consistent with UK experience, with full collateralization, a high degree of indirect participation, and significant concentration in correspondent banking all key features of the UK landscape. Although the model does not capture legal and operational risks, and hence the full implications of internalization, our model can go some way towards offering some policy guidance as to whether such a profile of participation is desirable. Given that the agent’s and central bank’s decisions are taken sequentially, we show that, under certain circumstances, the first-best outcome might not be achievable. In particular, we find that zero collateralization and indirect participation might be optimal for a range of monitoring costs, but that, if the central bank chooses zero collateralization, the agent will always find it privately optimal to access directly. We do show, however, that zero collateralization and indirect participation is likely to be socially optimal for a smaller range of monitoring costs as monitoring becomes less perfect. When we allow for both imperfect monitoring and tiering spillovers, a wedge also emerges between private and social choices under full collateralization. In particular, we show that there will exist a range of monitoring costs in which the bank will prefer to participate indirectly under full collateralization, while the central bank would prefer direct participation. While we do not model the precise channels by which such spillovers arise, we can draw some broad policy conclusions in this regard. In particular, we show that the key determinants of the size of this wedge will be: the error in monitoring; the magnitude of any spillover, reflecting both the size of exposures arising through correspondent banking and the spillover per unit of exposure; and the cost of collateral if participating directly. This implies that policy to reduce the size of the wedge should be directed towards: (i) ensuring the capacity of correspondent banks to absorb either capital losses or liquidity shocks arising from the failure of a customer bank, perhaps via enhanced ex ante solvency and/or prudential regulation; (ii) encouraging correspondent banks

Central bank intraday collateral policy

157

to improve monitoring quality; and (iii) facilitating a low opportunity cost of posting collateral for all prospective payment system members. Our analysis could be improved by a more complete and sophisticated treatment of the interaction between payment system participants and the provision of payments services. For example, with just one agent making choices and a single payment made, we cannot address the implications of factors such as collateral pooling, or the intraday liquidity management game. And, with the direct members A and B playing only a passive role in this model, we cannot assess the risk they bring to the system, and hence cannot perform a complete welfare analysis. Finally, a more detailed analysis of tiering spillovers would be useful, in order to refine our policy conclusion that such spillovers should be addressed in regulatory design.

Notes 1 The views expressed in this chapter are those of the authors, and not necessarily those of the Bank of England. We would like to thank the following for helpful comments during the preparation of this work: Victoria Saporta, Steve Millard, Matthew Willison, Ana Lasaosa, Jochen Schanz and Will Roberds. 2 For example, in September 2005, participants of the CHAPS and CREST systems used an average of £65 billion of intraday liquidity at the Bank of England to facilitate settlement. This includes more than £50 billion generated by self-collateralizing repos in CREST. 3 See also BIS (2005) for a discussion of alternative hybrid system designs. 4 In some securities settlement systems, where the cash leg is settled gross (equally relevant to the issues considered in this chapter), the cost of generating cash liquidity is significantly reduced via the implementation of self- or auto-collateralization techniques. CREST and Euroclear France, for instance, apply such procedures, allowing the immediate pledge/repo of (eligible) securities to the central bank to generate liquidity to fund their own purchase. 5 Other possible central bank policies, such as restricting access or imposing quantity limits, are not considered in this chapter. 6 Our model cannot capture the potential effects of collateral pooling because we only consider the decision of a single agent. 7 Specifically we use the framework in arrangements 4 and 5 of Kahn and Roberds (2006), which deal with payments settling across central bank accounts. 8 We later generalize the value of the payment. 9 It is assumed that Bank C is eligible to participate directly in the system, but that the central bank’s preferences over direct versus indirect participation by Bank C, from the perspective of social cost, will be reflected in the collateral policy chosen. 10 We assume that only Bank C faces the possibility of an exogenous default shock in this model. A more complete framework might allow Bank A to suffer such a shock also. We note in our discussion the potential implications for our results of this simplification. 11 With only a single payment here, we rule out the possibility of strategic behaviour among settlement banks. In particular, we abstract from the possibility that banks delay outgoing payments until incoming payments have arrived so as to economize on collateral costs. This behaviour is well documented in the literature, e.g. Bech and Garratt (2003). 12 We assume that the central bank’s collateral policy is applied to all direct members. However as Banks A and B take no decisions and make no payments on their own

158

13 14

15 16 17 18 19 20

21 22

J.P. Jackson and M.J. Manning

behalf in this model, it is not necessary that the central bank consider the implications of its collateral policy on these agents’ behaviour. Consistent with empirical observation, we assume that the central bank either does not have the capacity, or finds it excessively costly, to monitor. In practice, these costs are often not passed on explicitly. Given that there is a clear economic rationale for full pass-through, it is likely that these costs are fully reflected in the price for a bundle of services provided by correspondent banks, which includes monitoring costs. Furthermore, explicit charging for intraday liquidity may become increasingly common in future, as payments become more time-critical intraday. Alternatively, it may be that the providers of these services simply face very low opportunity costs to posting collateral. It is worth noting that the linearity of all components of social cost implies that a central bank policy of partial collateralization will never be a dominant strategy. This might be justified if technological costs and the profit component of costs under indirect participation, which are normalized to zero here, exceeded the fixed costs of joining and accessing the payment system directly. It is worth noting that, under full collateralization, the central bank’s and Bank C’s expected costs are the same. Such an intraday credit policy is adopted by most G10 central banks, although some notable exceptions do exist, in particular the Federal Reserve, which charges an explicit fee on intraday overdrafts. It is thought that upwards of 20 per cent of sterling large-value payments are internalized across the accounts of correspondent banks. It is worth noting that in many comparable economies, large value payment arrangements are far less highly tiered. For example in the United States the Fedwire system has over 7,000 member banks, and the Japanese large-value payment system BOJ-NET has over 300 members, compared with 15 members of the UK CHAPS system and 14 members of the Canadian LVTS system. Several factors might help to explain the structural differences that can be observed between countries. In some jurisdictions authorities have made greater efforts to encourage wider membership of large value payment systems, either through imposing a specific regulatory requirement, applying moral suasion, or subsidizing the cost of such payment arrangements. In addition the impact of prudential liquidity requirements in encouraging concentration is not relevant in a number of jurisdictions. Finally, some countries have historically always had highly concentrated banking systems (this is true of the United Kingdom) while in others highly fragmented banking arrangements are observed. This assumption seems reasonable, although one could argue that a bank’s monitoring intensity might increase when the size of its potential exposures was greater. A more extreme policy option might be to simply compel certain payment system participants to settle directly. However, even if these banks were required to open settlement accounts with the central bank, it is not clear that they could be compelled to actually use them.

References Bank for International Settlements (BIS) (2005) New developments in large-value payment systems, Committee on Payment and Settlement Systems Publication No. 67. Bech, M. and Garratt, R. (2003) ‘The intraday liquidity management game’, Journal of Economic Theory, 109(2): 198–219. Chakravorti, S. (2000) ‘Analysis of systemic risk in multilateral net settlement systems’, Journal of International Financial Markets, Institutions and Money, 10: 9–30.

Central bank intraday collateral policy

159

Furfine, C.H. and Stehm, J. (1998) ‘Analyzing alternative intraday credit policies in realtime gross settlement systems’, Journal of Money Credit and Banking, 30 (4): 832–48. Harrison, S., Lasaosa, A. and Tudela, M. (2005) ‘Tiering in UK payment systems: credit risk implications’, Bank of England Financial Stability Review. Kahn, C.M. and Roberds, W. (2001) ‘Real-time gross settlement and the costs of immediacy’, Journal of Monetary Economics, 47: 299–319. Kahn, C.M. and Roberds, W. (2006) ‘Payments settlement: tiering in private and public systems’, unpublished thesis, University of Illinois. Manning, M.J. and Willison, M.D. (2006), ‘Modelling the cross-border use of collateral in payment systems’, Bank of England Working Paper No. 286. Mills, D.C. (2005), ‘Alternative central bank credit policies for liquidity provision in a model of payments’, Board of Governors of the Federal Reserve System Finance and Economics Discussion Series No. 2005–55. Willison, M. (2005), ‘Real-Time Gross Settlement and hybrid payment systems: a comparison’, Bank of England Working Paper No. 252.

10 Central banks’ interest calculating conventions Deviating from the intraday/overnight status quo George Speight, Matthew Willison, Morten Bech and Jing Yang1

Introduction Central banks lend central bank balances and accept deposits on terms designed to ensure that the overnight market interest rate is close to its ‘policy’ rate, which it sets to meet its monetary policy objectives. This rate anchors market interest rates for longer maturities. However, central banks also typically lend central bank balances and accept deposits at low or zero interest rates intraday. This anchors the intraday market interest rate at or close to zero, in so far as money is traded intraday at all. Dale and Rossi (1996) show that the central bank can set a low or zero intraday interest rate and ensure the overnight rate is close to its (higher) policy rate because all intraday lending is repaid by the end of the day; there is no spillover overnight. The existence of a distinction between intraday and overnight interest rates has its origins in the move from settling large-value payments on a deferred net settlement basis to settling them on a gross basis in real time. (A description of the diffusion of RTGS across the world’s large-value payment systems can be found in the chapter by Bech in this volume.) When banks exchanged central bank money at the end of each day there was simply no need for banks or the central bank to lend at maturities of less than one day because money was not being exchanged on a more frequent basis than once per day.2 But in a RealTime Gross Settlement (RTGS) payment system, and in equivalent ‘hybrid’ payment systems, it is possible to lend central bank money for periods strictly within a day. Indeed, the prospect of a genuine market in intraday money is one which generates great interest among market practitioners and policy makers alike. So far, central banks’ practice of providing intraday credit without charge or at a very low rate (and similarly not remunerating positive intraday balances) has been sufficient to hold back any such development.3 Broadly speaking, central banks have not charged interest on the basis of balances during the day because of a belief that if they did, banks would have a strong incentive to delay payments until late in the day. It would be individually

Banks’ interest calculating conventions

161

rational for banks to wait until late in the day in the hope that incoming payments would provide them with the liquidity they needed to make their payments, rather than having to borrow from the central bank. Of course, if all banks followed the same strategy none would actually gain from delaying payments. Payment delays are not desirable because they can imply the following risks and inefficiencies in the payment system: 1 2

3

the failure to receive payments in a timely manner may represent a liquidity risk to recipients; delay could also reduce the number of times liquidity is recycled in the payment system during the day, which may actually increase the amount of intraday credit required for payments to settle; delay of payments until late into the day implies greater exposure of the payment system to operational risk since if an operational problem hits later during the day there is less scope to overcome it in time for payments to settle.

This chapter questions this line of reasoning. It does this by analysing the effects of the central bank imposing its interest constraint more frequently. (We choose more rather than less frequently because it moves us towards the real-time paradigm, but the results are broadly generalizable to any frequency.) We find that banks would indeed delay payments, to the extent that they had no reason to send them more promptly. But it may be that in many cases, customers would opt to send payments in the morning rather than the afternoon, because money would now have value on a more frequent basis. To assume that banks would delay all payments until the end of the day is to take for granted that customers would continue in all cases to contract to make payments on an ‘end of day’ basis. However, this is not necessarily the case. Either way, payments would be compressed against deadlines to a greater extent than at present. But if customers opted to make some payments early, the payments would at least be spread across the day and would not be concentrated at the end of the day. Also, the change would likely affect the amount of intraday credit which banks extend to their customers. There may be customers who would be content for their banks systematically to delay their payments until later in the day. For these customers, settlement banks would extend less credit (at least on a duration-weighted basis) than if their payments were spread more evenly across the day: they would only incur intraday overdrafts towards the end of the day. And if the frequency with which settlement banks actively monitor customers’ accounts is anchored by how often they calculate interest, then a shift to calculating interest twice a day may mean that settlement banks’ charges to their customers become more closely related to the amount of credit the customers actually use. So if this bank-imposed ‘monitoring constraint’ contributes to credit being extended beyond the socially optimal level, the change in interest

162

G. Speight et al. Interest rate

Next possible policy change Policy rate

End of End half day of day

Time

Figure 10.1 Change in shape of yield curve.

calculating convention may generate financial stability benefits by easing this constraint. These effects could produce system-wide risk-reducing financial stability benefits if the private costs which settlement banks face in extending intraday credit are lower than the social costs. The chapter works through these effects in a simple scenario: where the central bank imposes its policy rate at the end of every half day rather than at the end of every day (Figure 10.1). Within each half day it provides credit at no charge and does not remunerate positive balances. The central bank now constrains the market rate of interest to trade at close to its policy rate between half-day periods.4 Our analysis in this chapter is very much in the spirit of a thought experiment, to begin exploring the issues. There are several important factors that we do not analyse. One is whether the possible financial stability benefits we identify could be achieved via other means. For example, as shown in Willison (2005), concentrating payments activity in smaller periods of the day in order to reduce the duration of exposures between settlement banks and customers could also be achieved by introducing liquidity saving features such as payment offsetting. Any wedge between private and social costs of intraday credit could be reduced by imposing capital requirements that bind more frequently than at the end of the day. It is also possible that the private sector may move to monitoring and charging of intraday credit on a more timely basis on its own accord as information technology costs decline. Another factor we do not take into account is the costs of shifting from one way of calculating interest to another (e.g. costs of upgrading IT systems). Of course, these costs would be mainly up front and would need to be contrasted with any recurring longer-term benefits. The chapter proceeds as follows. We first review the existing literature before summarizing the monetary stability implications of an increase in the frequency at which the central bank calculates interest. We then set out our model and work through the implications for financial stability of the change in how interest is calculated. Finally, we consider some further issues before offering some conclusions.

Banks’ interest calculating conventions

163

Literature review Angelini (1998), Bech and Garratt (2003) and Kobayakawa (1997) each develop game-theoretic models of an RTGS payment system which emphasize how banks’ decisions about when to make payments during the day depend on other banks’ decisions and that banks’ individual incentives can lead to an inefficient outcome when intraday credit is costly. They show that a bank may choose to delay payments when other banks, from which it receives payments, do not delay because the gain from reducing the amount of intraday credit it must obtain exceeds the cost of delaying. Since all banks have this incentive to delay, all banks delay their payments in equilibrium. But the equilibrium outcome is inefficient since intraday credit needs are unchanged and all banks incur delay costs. Banks would each be better off if none of them delayed. The central bank can reduce the likelihood of delay arising in equilibrium by reducing the cost of intraday credit; e.g. by setting a zero intraday interest rate. Other papers also explore the adverse effects of payment delays and the relationship these effects have with the cost of intraday credit using models where the timing of payments is treated as exogenous but where it is assumed that payment times are asynchronous. Freeman (1996, 2002) considers an economy where in any period agents need liquidity to make payments and before other agents from who they are due payments can make theirs. Agents can go to the market to obtain liquidity but if there is insufficient market liquidity they will be forced to delay some payments. Freeman (2002) shows that the liquidity constraints in the payment system can be alleviated if a central bank extends credit. Agents that need liquidity early borrow from the central bank and repay the central bank once they have received payments. Further, liquidity constraints in the payment system are completely eliminated when the central bank lends at a zero interest rate. Kahn and Roberds (2001), Martin (2004) and Zhou (2000) develop similar models and demonstrate that the central bank should lend intraday at a zero interest rate. In summary, most of the existing literature concludes that a central bank should always extend intraday credit at a zero interest rate to avoid liquidity constraints and delays occurring in RTGS payment systems.

Implications of our thought experiment for the implementation of monetary policy This section discusses the implications for implementation of monetary policy of the central bank calculating interest on settlement banks’ accounts at mid-day as well as at the end of the day. Broadly speaking, this would have relatively little effect on the central bank’s ability to implement monetary policy. Woodford (2003) shows that when central bank money is the ultimate settlement asset, central banks can control the price of their money at a particular maturity by standing willing to accept deposits at one rate and lend at another, thereby bounding the money market rate of interest for similar duration loans on

164

G. Speight et al.

the lower and upper sides, respectively. Central banks typically target the overnight interest rate. Dale and Rossi (1996) show that the implementation of monetary policy is not compromised by the central bank setting a lower intraday interest rate than the central bank’s overnight policy rate, provided that intraday credit has to be repaid in full by the end-of-day. By imposing this ‘quantity constraint’, the central bank ensures that overnight borrowers have to pay the overnight cost of central bank money. In our thought experiment, there is no reason why the central bank’s ability to implement monetary policy should be reduced. Banks face the same conditions for their holdings of central bank money at mid-day as at the end of the day under the current regime. It follows that banks will treat central bank money in the same way across the mid-day point as they do currently across the end-ofday point and the central bank would exercise the same control over market rates of interest as it does at present, but now starting from the maturity of half a day.5 The shortest maturity point on the yield curve would shift from one day to half a day. It is difficult to believe that the change would have any first-order macroeconomic implications, e.g. for the speed or effectiveness of the transmission of monetary policy. The inflation control mechanism would essentially be unaltered. However, to enforce a new regime with a shorter interest-calculating period, the central bank may need to implement operational changes. It would need to calculate interest on accounts and supply the necessary funds to meet the reserve maintenance requirement more regularly. It may also need to conduct open market operations (OMOs) more frequently, though this would depend on whether it had a reserve-averaging framework and, if so, on its length given that a central bank needs to observe the general principle that there should be at least one OMO round per reserve maintenance period.

Implications of our thought experiment for system-wide risk in the financial system The implications for system-wide risk or financial stability depend on the pattern of payments. The model sets out in a more formal way how the pattern of payments would be determined. Payments fall into two kinds: •

•

payments on behalf of customers where the customer does not express any preference for which time of day he wants his payment made; it seems reasonable to assume that end of day would remain the ‘default’ for these customers; payments which the bank has a particular reason for sending in a specific half of the day including customer payments, where the customer has requested the payment be sent in a particular half of the day, and some payments on the bank’s own account (e.g. for funding, trading or risk management purposes).

Banks’ interest calculating conventions

165

The risk implications are different for the different kind of payments. Overall, the key implications are for operational risk and credit risk in the payment system. There will be an increase in operational risk, regardless of the mix of the two kinds of payments. If a bank submits a payment to the payment system close to the deadline by which it should settle, and at that time there is an operational problem either with the bank or with the system itself, there is less scope to overcome the problem and allow the payment to be settled by the deadline. So as a general principle, as the time between a bank submitting a payment and the deadline for that payment to settle falls, operational risk increases. Payments of the first kind are systematically delayed until the second half of the day, reducing the average amount of time available for them to settle. Payments of the second type are submitted in the half of the day when they are expected to settle. Customers who had been expecting to receive a payment by a certain time but do not and who are not insured against this risk, clearly face a cost: either they will fail to meet their own commitments and face default or delay penalties, or they will have to resort to (possibly expensive) funding from alternative sources, or they may postpone consumption or investment. Sending customers whose payments fail may have to pay compensation, which may or may not cover the costs faced by receiving customers. Importantly, there may be knockon consequences for other agents, if the customer expecting to receive a payment needed it to fulfil an obligation to another customer, which is in turn delayed or fails. The social cost may be greater than the net private costs. A change in the way that the central bank calculates interest could also affect the amount of intraday credit which settlement banks provide to their customers and its price. Where banks systematically delay payments until late in the day, such payments will typically require less credit in order to settle. Holding other things constant, the duration of any extensions of credit occurring within the day will be shorter compared to when some payments are made earlier in the day. This will have implications for system-wide credit risk if the private cost to settlement banks of extending credit within the day does not fully reflect the social costs. There could be a wedge between private and social costs if banks fail to internalize the impact of a customer default on the wider financial system. If this externality exists, a fall in the duration of intraday credit may benefit financial stability. There could be further benefits for financial stability if the change in the central bank’s interest-charging convention affects how intensely banks monitor exposures to their customers. When the central bank only charges and remunerates at the end of the day, banks may not actively monitor customers’ actual balances through the day. Indeed, this has typically been the case: evidence presented in Harrison et al. (2005) suggests that, broadly speaking, banks have only tended to watch for limit breaches on a day to day basis. So any charge which they make for intraday overdrafts – including a charge to cover the credit

166

G. Speight et al.

risk they bear – may be only loosely linked to the amount of risk actually faced within the day, thereby contributing to any wedge between private and social costs. A central bank, by calculating interest at mid-day, and thus inducing settlement banks into doing the same, can ease this bank-imposed monitoring constraint and as a result reduce the private–social cost wedge. The remainder of this section develops these arguments more thoroughly, using a simple formal model to highlight the key issues. The model represents a single day in an RTGS payment system. There are two banks (bank 1 and bank 2) that are members of the RTGS system (they are ‘settlement’ banks), each with one customer (customer 1 and customer 2). Each customer has one payment with unit value to make to the other customer during the day. This is common knowledge. For simplicity we assume that settlement banks and customers both begin the day with zero balances. This implies that to make a payment the first settlement bank must run an overdraft with the central bank. Likewise, the first customer to make a payment must run an overdraft with their settlement bank. Time within the day is divided into two periods: morning and afternoon. When customers trade they write contracts with one another. A contract specifies the value of the payment (unity), and the period or periods within the day in which the payment should be made. The central bank requires extensions of intraday credit to be fully collateralized. We assume that a settlement bank faces a cost (C) when it posts collateral with the central bank. Given that time is modelled as discrete, we assume that a settlement bank has to obtain intraday credit from the central bank if it makes its payment in the period before it receives a payment and if it makes it in the same period as it receives a payment.6 Other costs include a delay cost (D), which a settlement bank incurs when it delays making its payment until the afternoon. This delay cost captures the greater risk that payments fail due to operational problems if they are delayed until the afternoon. A settlement bank also incurs a cost in extending credit to its customer from the morning to the afternoon (E) which it may not be able to control through other means and which is not fully reflected in the charges faced by customers. The final cost that settlement banks face is the interest incurred or received (R) on their accounts at the central bank at mid-day. Zero intraday interest rates The case where interest is calculated on account balances at the end of each day only – i.e. the current situation – serves as the benchmark case. Customers make their payment requests at the beginning of the day. The rationale for this is that they do not incur or receive interest according to their balance positions intraday, and that they wish to minimize delay costs. Each settlement bank then faces a choice between making its payment in the morning or the afternoon. Since each bank’s costs depend on the other bank’s decision as to when to make its payment, we model the situation as a simultaneous-move game. The banks’ costs are shown in the strategic-form game below.

Banks’ interest calculating conventions

167

Table 10.1 Pay-offs when central bank charges and remunerates at end of day only Bank 2 Bank 1

Morning Afternoon

Morning C,C D,C + E

Afternoon C + E,D C + D,C + D

Table 10.2 Equilibrium strategies when the central bank charges and remunerates at end of day only Bank 1’s decision

Bank 2’s decision

Providing that

Morning Afternoon Morning or afternoon Morning or afternoon

Morning Afternoon Afternoon or morning Morning or afternoon

C and E < D C and E > D E
We solve the game to find the Nash equilibrium. There are three possible equilibrium outcomes, shown in Table 10.2. Both banks make payments in the morning if the costs of doing so are lower regardless of the actions of the other bank when the other makes its payment. A bank prefers paying out in the morning when the other bank is doing the same if the possible saving in terms of collateral cost is less than the delay cost, i.e. C < D. A bank chooses to make its payment in the morning when the other bank makes its payment in the afternoon if the delay cost exceeds the cost associated with extending intraday credit to its customer; i.e. E < D. It follows that when C and E > D both banks choose to make payments in the afternoon since each prefers to delay irrespective of when the other is making its payment. When C < D < E it is possible that both banks make payments in the morning or that both make them in the afternoon. This is because when one bank makes its payment in the morning, the other prefers to do the same since the delay cost exceeds the cost of posting collateral. But when one bank makes its payment in the afternoon, the other prefers to delay because the delay cost is less than the cost of running an exposure to its customer in the morning. An equilibrium exists whereby the banks make payments in different periods – this occurs when E < D < C. Under these costs, a bank prefers to make its payment in the morning when the other makes its payment in the afternoon since the delay cost is greater than the cost associated with extending customer credit (E < D). But, a bank chooses to delay when the others choose the morning as the delay cost is less than the cost of posting collateral (D < C). Of course, there are other reasons why payments may be made in different periods of the day requiring settlement banks to extend intraday credit to their customers. One reason is that there could be different relative values of C, D and

168

G. Speight et al.

E for different payments. Another reason is that customers do not submit all of their payments at the start of the day. They also submit some in the afternoon. However, adding a stochastic process for customers to submit payment requests produces qualitatively similar results to the model used here. These results show that the two factors we are interested in from a financial stability perspective (extensions of intraday credit by settlement banks to customers, and payment delays) will depend on the particular values of the different costs. If delay costs are lower than costs related to collateral and extensions of intraday credit, all payments are delayed until the afternoon, which is detrimental to financial stability. When E < D < C one of the settlement banks will extend intraday credit to its customer from the morning until the afternoon. If settlement banks were to under-price intraday credit relative to the social costs, financial stability may be better served by having both payments made in the morning. Central bank charges and remunerates at mid-day and end of day The change in the way the central bank charges and remunerates to banks’ accounts will affect the pattern of payment activity over the day. If a bank’s customer does not specify when he would like his bank to make his payment, the bank has an incentive to send it late, earning interest on its account at the central bank by holding the customer’s balance across the middle of the day. If a customer requests the bank to make a payment in the morning, it would need to compensate its settlement bank accordingly – paying for an overdraft, or earning less interest in the case of a positive balance. Conversely, the recipient of this payment would insist that its own bank paid its interest. Competition would therefore establish a de facto shift to remunerating according to mid-day balances, with customers specifying in which half of the day they want their payment to be made. There would be demand from customers to make payments in the morning as well as in the afternoon for just the same reason as there is demand to make payments on Tuesday rather than on Wednesday. The money market would trade in half days, with interest calculated on a pro rata basis, i.e. 1/730 times the annualized rate per half day.7 But in practice, in many cases customers may continue to specify only that a payment be made by the end of the day, i.e. they may not specify which half of the day. By default, such contracts in effect specify the second half of the day, to the extent that settlement banks’ incentive in such cases is to delay.8 But it seems possible that for some types of customers and payments, ‘end of day’ could continue to be the default convention. This would be most likely for retail customers. As well as having no particular reason to send payments in the morning, these customers would probably face the highest costs to managing their balances on a more frequent basis. And it would tend to be more widespread in the short term, with behaviour possibly changing in the longer term. Below we analyse two extreme cases: first where customers are content to continue specifying ‘end of day’, and second where they specify one particular half of the day.

Banks’ interest calculating conventions

169

Customers do not specify which half of the day Suppose customers without a preference as to which half of the day their payments are settled in continue to make payment requests at the start of the day. They are concerned only that their payments are settled by the end of the day. Thus, the timings of payments during the day will still be determined by settlement banks. Settlement bank behaviour may differ from the case in which there is a zero intraday interest rate because they will have a greater incentive to delay making payments until the afternoon: if they delay, they earn interest on positive balances at mid-day (if they receive a payment in the morning) or avoid paying interest on an overdraft at mid-day (if they do not receive a payment in the morning), in addition to any collateral cost savings. The settlement banks’ costs for the different possible timings of payments are shown in the game below. The difference from the game in Table 10.1 is the interest cost R that is incurred when a settlement bank makes a payment in the morning but does not receive a payment in the morning. Conversely, when a settlement bank receives but does not send a payment in the morning, its costs are reduced by the interest it receives, R. There are the same three possible equilibrium outcomes as when the central bank sets a zero intraday interest rate. The equilibrium conditions are shown in Table 10.4. Comparing the third rows of Tables 10.2 and 10.4, we can see that the condition that needs to be met for both payments to occur in the afternoon is stronger when interest is charged at end of the day. This means that an equilibrium in which settling payments in the afternoon is a dominant strategy for both banks is more likely when interest is charged intraday and customers specify

Table 10.3 Pay-offs when central bank charges and remunerates at midday and at end of day Bank 2 Bank 1

Morning Afternoon

Morning C,C D – R,C + E + R

Afternoon C + E + R,D – R C + D,C + D

Table 10.4 Equilibrium strategies when the central bank charges and remunerates at midday and at end of day Bank 1’s decision

Bank 2’s decision

Providing that

Morning Afternoon Morning or afternoon Morning or afternoon

Morning Afternoon Afternoon or morning Morning or afternoon

C and E < D – R C and E > D – R E
170

G. Speight et al.

only that they want their payment to be made by the end of the day. Calculating interest at mid-day can give settlement banks greater incentives to delay for given values of C, D and E. For example, whereas a settlement bank prefers to make its payment in the morning when C and E < D when there is a zero intraday interest rate, the prospect of incurring or receiving interest at mid-day may induce it to delay (if C and E > D – R). That charging interest within the day increases payment system risk by increasing payment delays, is the result commonly found in the literature. However, the increased tendency for settlement banks to delay payments also means that equilibria in which one of the settlement banks extends intraday credit are less likely to occur. If there is a wedge between private and social costs of intraday credit, this represents a possible financial stability benefit. Customers specify which half of the day The other extreme case is that customers specify morning for half of their payments and afternoon for the other half. They and their banks face a cost if the payment is not made by the deadline, whether that was mid-day or the end of the day. In both cases, contracts are broken and money not received at the time expected.9 Operational risk is higher than in the benchmark case in which the central bank charges interest only at the end of the day and banks have all day available to them to settle the payments. Here, banks have only half a day to settle any payment. So there is a greater risk that operational problems will prevent some payments from being made by the deadline. But in the case where the central bank charges interest but customers do not express a preference, payments are systematically pressed up against the end-ofday deadline. Here they are spread across two deadlines. In some sense then the systemic risk is lower, since an operational incident which affected the functioning of the system for the second half of the day would only affect half of the day’s payments, not all of them. The worst-case scenario is not as bad. On the other hand, it may be easier to resolve operational problems in the afternoon because there is always the option to extend opening hours at the end of the day. It may be more problematic to extend the deadline for the morning session, since it would impinge on the afternoon. Whether this were the case would depend on the specific contingency arrangements in place for the system. But if it were the case, the delay cost for afternoon payments would be lower. There could be additional financial stability effects if, as we have supposed, any difference between the private and social costs of extending intraday credit is linked to how often interest is calculated on accounts. Settlement banks may not choose to actively monitor the evolution of their customers’ account balances through the day if they only calculate interest on these accounts at the end of the day. They may charge for the risk that exposures to their customers bring to them but only imperfectly because the charges will not be closely linked to the actual risks experienced during the day. A central bank’s decision to calcu-

Banks’ interest calculating conventions

171

late interest at mid-day as well as at the end of the day could help to ease this self-imposed monitoring constraint with wider benefits to the system. When the change is passed through by settlement banks and they calculate interest on customers’ balances at mid-day, they may be able to more closely align the charges they make for the risks from customer exposures to the actual exposures realized during the day. The result is that any wedge between private and social costs due to unmonitored interbank intraday exposures could be reduced by a central bank calculating interest more frequently.

Further issues Need the positive intraday rate be equal to the policy rate? The results in the previous sections depend on there being a positive intraday interest rate, but do not depend on that rate necessarily being equal to the central bank’s policy rate. (For simplicity, in our thought experiment we assume that it is.) It would be feasible for the central bank to lend at a positive ‘intraday’ rate within the day and lend over longer maturities at the policy rate as long as it ensured that all extensions of intraday credit were repaid by the end of the day (as it does now with a zero intraday rate) and that the intraday rate was less or equal to the policy rate. This situation is depicted in Figure 10.2. Of course, if the policy rate falls below the intraday rate, settlement banks and customers would have an incentive to substitute overnight credit for intraday credit. Comparison with Fedwire In the Fedwire payment system in the US, the Federal Reserve charges settlement banks a fee for intraday credit but does not remunerate them for running positive balances during the day. As Coleman (2002) states, fees are levied Interest rate

No change in policy rate possible until next scheduled decision

Slope reflects expectations and risk premium of future rate changes

CB policy rate Intraday rate

End of End half day of day

Next possible policy change

Time

Figure 10.2 An alternative change in the shape of the yield curve.

172

G. Speight et al.

according to banks’ overdrafts at the end of each minute. The fee can be interpreted as a positive intraday interest rate although it does not vary with changes in the Federal Reserve’s overnight, policy rate. Under these arrangements, in our model, a settlement bank does not have an incentive to delay its payment until the afternoon when it receives a payment in the morning because it does not receive any interest from having a positive balance at mid-day. As shown by Bech and Garratt (2003), this can lead to multiple equilibria in which settlement banks either make payments in the morning or in the afternoon (McAndrews and Rajan (2000) provide evidence that this happens in practice). If settlement banks can coordinate on the morning equilibrium, the delay costs could be lower than when the central bank both pays and charges interest on mid-day balances.

Conclusion Central banks typically set a zero interest rate intraday and impose their monetary policy rate overnight. This chapter undertakes a thought experiment to consider the possible effects of relaxing this intraday/overnight distinction. In our thought experiment, the central bank shifts from imposing its monetary policy rate only at the end of the day, with a zero intraday interest rate through the day, to imposing its monetary policy rate at mid-day and at the end of the day, with a zero rate in each half-day period. The general finding of the literature is that this would cause banks to delay payments systematically until the second half of the day, with an attendant increase in operational risk. We point out that this argument assumes that customers are content for banks to settle all their payments in the second half of the day. In effect, it takes for granted that customers instruct their banks to settle their payments by the end of the day. But this assumption may not be valid. Money would be traded by the half day rather than by the day, and in many cases customers could request a payment to be sent in the morning rather than the afternoon. Under competitive pressure, banks would be obliged to allow them to do this. Payments would still be compressed against deadlines to a greater extent than at present, but they would be spread across morning and afternoon. We also argue that the supply of intraday credit from settlement banks to their customers and the pricing of this credit could be affected. These effects could benefit financial stability if the private cost to settlement banks of providing intraday credit to their customers is less than the social cost. But the change in the central bank’s charging policy would not reduce its ability to implement monetary policy. It would have the same control over market rates of interest as it does at present, but now starting from a maturity of half a day.

Notes 1 The views expressed in this chapter are those of the authors, and do not necessarily reflect those of the Bank of England, the Federal Reserve Bank of New York or the Federal Reserve System.

Banks’ interest calculating conventions

173

2 In the case of Japan, where there were a series of net settlements through the day, the central bank did not extend intraday credit (see Hayashi, 2001). 3 The Swiss National Bank was an exception until 1999 since it provided no intraday credit to banks operating in its RTGS payment system (see Heller et al., 2000). 4 We assume throughout that the policy rate consistent with the central bank’s monetary policy objectives is strictly positive. Of course, if the policy rate were zero, there would be no discontinuity between intraperiod and interperiod interest rates. 5 We assume that the central bank’s new regime is passed through to all banks, whether they are direct settlement banks or customers of settlement banks. If this were not the case, we would have a more complicated situation in the money market. 6 This is a common assumption in the literature, e.g. Bech and Garratt (2003). 7 In sterling; conventions differ slightly across currencies. 8 An alternative contract type would leave open in which half of the day payment should occur, but with the sum paid adjusting to reflect interest foregone by the payee. If the interest were calculated at a pre-agreed fixed rate, then this contract has an option value for the payer, adding to its complexity. If the interest rate were tied to the market rate, it might still not accurately reflect the opportunity cost for the payee. So the standard contract would specify repayment within a particular half of the day. 9 The delay cost could be higher when there is a failure to make a payment in the morning. This is because, whereas there is always the option to extend opening hours at the end of the day, allowing delayed payments due in the afternoon to be completed by the end of the day, it is not possible to extend morning opening hours into the afternoon.

References Angelini, P. (1998) ‘An analysis of competitive externalities in gross settlement systems’, Journal of Banking and Finance, 22: 1–18. Bech, M.L. and Garratt, R. (2003) ‘The intraday liquidity management game’, Journal of Economic Theory, 109: 198–219. Coleman, S.P. (2002) ‘The evolution of the Federal Reserve’s intraday credit policies’, Federal Reserve Bulletin. Dale, S. and Rossi, M. (1996) ‘A market for intraday funds: does it have implications for monetary policy?’, Bank of England Working Paper No. 46. Freeman, S. (1996) ‘The payments system, liquidity and rediscounting’, American Economic Review, 86: 1126–38. Freeman, S. (2002) ‘Payments and output’, Review of Economic Dynamics, 5: 602–17. Harrison, S., Lasaosa, A. and Tudela, M. (2005) ‘Tiering in UK payment systems: credit risk implications’, Bank of England Financial Stability Review. Hayashi, F. (2001) ‘Identifying a liquidity effect in the Japanese interbank market’, International Economic Review, 42: 287–315. Heller, D., Nellen, T. and Sturm, A. (2000) ‘The Swiss interbank clearing system’, unpublished thesis, Swiss National Bank. Kahn, C.M. and Roberds, W. (2001) ‘Real-time gross settlement and the costs of immediacy’, Journal of Monetary Economics, 47: 299–319. Kobayakawa, S. (1997) ‘The comparative analysis of settlement systems’, Centre for Economic Policy Research Discussion Paper No. 1667. Martin, A (2004) ‘Optimal pricing of intraday liquidity’, Journal of Monetary Economics, 51: 401–24. McAndrews, J. and Rajan, S. (2000) ‘The timing and funding of Fedwire funds transfers’, Federal Reserve Bank of New York Economic Policy Review, 6 (2): 17–32.

174

G. Speight et al.

Willison, M. (2005) ‘Real-time gross settlement and hybrid payment systems: a comparison’, Bank of England Working Paper No. 252. Woodford, M. (2003) Interest and prices: foundations of a theory of monetary policy, Princeton, NJ: Princeton University Press. Zhou, R. (2000) ‘Understanding intraday credit in large-value payment systems’, Federal Reserve Bank of Chicago Economic Perspectives.

11 How should we regulate banks’ liquidity? Jean-Charles Rochet1

Introduction While the last 20 years have witnessed a remarkable trend towards harmonization of banks’ solvency regulations across countries, there is still considerable variation concerning liquidity requirements. Some of the existing requirements are based on stock measures (typically a minimum level of liquid assets in relation to the stock of liquid liabilities), while others are based on mismatch analysis (i.e. limiting the gaps between expected inflows and outflows of cash for short-term maturities). Several countries (including Australia, Germany, Singapore and the Netherlands) have recently reformed their systems by introducing new quantitative rules for banks’ liquidity regulation.2 Other countries, like the United Kingdom, are considering the implementation of such reforms. There are essentially three main reasons for this, which we now briefly discuss. The first reason for this renewed interest in liquidity regulation is the recent trend towards an increase in the concentration of the banking sector, as well as in the complexity and size of financial markets. The likely outcome of this trend is a small number of ‘large and complex banking organizations’ controlling a large number of interrelated markets. This system might be perfectly efficient during ‘normal times’ but it certainly leads to serious prudential concerns (of the ‘too big to be bailed out’ variety) should a crisis occur. A second reason for the increased attention of banking authorities, especially central banks, on the liquidity of banks is that these authorities want to encourage banks to use Real-Time Gross Settlement (RTGS) systems for large-value interbank payments, instead of Deferred Net Settlement (DNS) systems, which may be prone to systemic risk.3 These RTGS systems are highly liquidity intensive. For example, the daily turnover of a unit of liquidity on the US RTGS system, Fedwire, which is owned and operated by the Federal Reserve, is currently about 16, while that on CHIPS, the competing large-value payment system which has features that combine DNS and RTGS and which is owned and operated by private sector banks, is currently about 500: roughly speaking Fedwire requires 30 times more liquidity than CHIPS for a similar flow of payments.4 Finally, banking authorities are concerned by the fact that banks take huge positions on all kinds of derivative products, which are also very demanding in

176

J.-C. Rochet

terms of liquidity. As illustrated by several spectacular examples (e.g. the bail out organized by a consortium of German banks following the $1 billion losses made by Metalgesellschaft on mismatched derivatives contracts on oil in 1994 and the spectacular failure of the hedge fund, LTCM, which necessitated a coordinated bail out by commercial banks orchestrated by the Federal Reserve in 1998) inadequate liquidity management of derivatives positions can provoke disasters, especially if large banks adopt similar strategies and rely on similar market instruments to hedge their liquidity risks.5 This is also a source of prudential concern. Under the influence of the Basel Committee on Banking Supervision (BCBS) of the central banks of the G-10 countries, solvency regulations have received a lot of attention in the last 20 years, leading first to a harmonization across countries (Basel 1) then to an incredible degree of sophistication of solvency ratios (Basel 2). But are these ratios (whatever their complexity) sufficient to reduce the probability and extent of bank failures, especially in the face of exceptionally adverse conditions? Theoretical results and common sense suggest that liquidity requirements are a natural complement (or partial substitute?) to solvency requirements. In any case supervisors should consider a bank’s liquidity risk in conjunction with its capital adequacy: in the absence of any doubts on banks’ solvency, liquidity management would essentially reduce to a pure ‘plumbing’ problem.6 It is commonly accepted that central banks have to perform some kind of emergency liquidity assistance activity towards commercial banks. For several kinds of reasons (which will be developed later), interbank and financial markets may be insufficient providers of liquidity to banks in trouble. A liquidity requirement is a way to limit the need to use the lender of last resort (LLR) facility. A cost–benefit analysis of the LLR is thus needed to determine the appropriate extent of liquidity regulations. A priori the central bank is in a better position than commercial banks to provide liquidity assistance to banks in trouble, especially during systemic crises. However, given the lack of commitment power of public authorities, and the risk of forbearance under political pressure during crisis periods, there is value in limiting a priori the need for emergency liquidity assistance by the central bank. This could take the form of additional liquidity requirements, in order to cover exceptional liquidity needs under adverse circumstances. The plan of the rest of this chapter is the following: the next section examines the sources of liquidity risk for banks. We then list the main instruments that can be used by banks for managing their liquidity risks before briefly discussing the possible market failures in the provision of liquidity that may justify public intervention in the regulation and provision of liquidity to banks. We then explain why regulation of banks’ liquidity may be justified and discuss the way in which these regulations could be set.

How should we regulate banks’ liquidity?

177

Sources of liquidity risk for banks Like any other firm, a bank has to manage carefully its liquidity in order to be able to cover mismatches between future cash outflows and cash inflows. However the degree of uncertainty about these mismatches is clearly much higher in the banking sector. This has several sources, which we now examine in turn. Assets There is large uncertainty about the volume of new requests for loans (or renewal of old loans) that a bank will receive in the future. Of course the bank could refuse to grant these new loans but this would in general lead to the loss of profit opportunities. This would also be detrimental to the firm if it is creditrationed and more generally to the economy as a whole – banks are unique providers of liquidity to small- and medium-size enterprises, which constitute an important fraction of the private sector. This credit rationing would be especially costly if the firm is forced to liquidate, possibly resulting in additional losses for the bank itself. Liabilities There is also large uncertainty about the amount of withdrawals of deposits (including wholesale) or the renewal of rolled-over interbank loans. This is especially so when the bank is under suspicion of insolvency, when there is a temporary (aggregate) liquidity shortage or when the economy suffers from a macroeconomic shock. Off balance sheet operations Credit lines and other loan commitments are crucial to borrowers and involve a large liquidity risk to the banks. Similarly, banks are major participants in derivative markets (swaps, futures, options), which may turn out to be a highly liquidity-intensive activity. Payment systems For large-value interbank payments, central banks favour the use of RTGS over DNS systems, because they are less prone to systemic risk. However, RGTS systems are highly liquidity-intensive and can only function properly if banks hold sufficient collateral to back credit lines, either from the central bank or from other participants.7 The failure of a large participant in a large-value payment system (LVPS) could provoke a big disruption to the financial system. Even a liquidity shortage or a ‘gridlock’ due to a temporary stop in the payment activity of a large bank could have dramatic consequences. This creates a ‘too

178

J.-C. Rochet

big to fail’ issue since it is likely that the central bank would be forced to intervene in such a situation. To avoid or simply to mitigate such problems, ex ante regulation of the liquidity of large participants in RTGS systems seems warranted. Securities settlement A related issue concerns the industrial organization of the financial sector. If banks are allowed to merge with central security depositories or CSDs (forming what is known as International Central Security Depositories or ICSDs), additional liquidity requirements may be needed since CSDs can be viewed as essential infrastructures enabling security trading.

Instruments of liquidity management for banks In addition to their cash reserves, banks can rely on their other assets as sources of liquidity as we describe in this section. Government securities These can be used as collateral for borrowing liquidity. However, these securities are also used as collateral for LVPSs. This raises the question of cross pledging of collateral. Such cross pledging is in general warranted, since it allows diversification between different sources of risk for economizing on collateral. However, it requires sufficient independence between payment risks and other forms of liquidity risk, as well as coordination between the central bank (which is often in charge of monitoring the LVPSs) and the prudential supervisory authority, which in many countries is different from the central bank. Marketable securities Such securities (equities, interbank loans) can be sold easily in normal circumstances but can become illiquid in adverse circumstances (this is related to the notion of ‘fire sales premium’) which may provoke insolvency in extreme cases. This may justify LLR interventions (see Rochet and Vives, 2004). Securitizable loans Such assets, in principle, can also be a source of liquidity but securitization operations are costly and have to be planned in advance. They cannot provide liquidity in emergency situations.

How should we regulate banks’ liquidity?

179

LLR interventions and credit lines Of course the central bank can provide emergency liquidity assistance but this facility is often improperly used to bail out insolvent banks.8 Goodfriend and Lacker (1999) have argued that banks could instead grant each other credit lines. Why is this seldom seen in practice? Finally, note that liquidity needs can be strongly reduced by the use of appropriate risk management methods (see Froot and Stein, 1998).

Market failures in the provision of liquidity This section briefly discusses the possible market failures that may justify public intervention in the regulation and provision of liquidity. Opaqueness of banks’ assets and moral hazard Banks have two fundamental characteristics: they play a crucial role in the financing of small and medium firms that do not have direct access to financial markets and they rely principally on external sources (deposits) for financing these loans. The fact that banks have to screen and monitor their borrowers creates an opaqueness of banks’ assets: as shown convincingly by Morgan (2002), these bank assets are difficult to evaluate by external analysts. This opaqueness generates possibilities of moral hazard, in the form of insufficient effort by banks on screening their borrowers, or on monitoring their activities after the loan has been granted. Modern corporate finance theory (e.g. Tirole, 2005) has shown that in such a situation, liquidity needs (due, for example, to cost over-runs in the borrowers’ projects or to deposit withdrawals in the banks themselves) are covered insufficiently by financial markets. Rochet (2004) shows that several institutional arrangements are possible to solve this market failure. For example, private contractual arrangements such as pools of liquidity accompanied with interbank committed credit lines can be used to mitigate this inefficiency. This can be a substitute for emergency liquidity assistance by the central bank, at least in the absence of aggregate shocks (see below). Coordination failures Opaqueness of banks’ assets also creates an externality between lenders on the interbank markets, payment system participants, or uninsured depositors. The decision to renew a short-term interbank loan, a debit cap on participants of a LVPS or a wholesale deposit depends not only on fundamental uncertainty (the quality of the bank’s assets) but also on strategic uncertainty (what other lenders or depositors will do). Freixas et al. (2000) study the consequence of such strategic uncertainty on the risk of contagion in an interbank LVPS. In such a context, liquidity requirements can be a way to limit systemic risk. Allen and Gale (2000) also show how contagion can emerge when interbank markets are

180

J.-C. Rochet

incomplete. Using the methodology of global games (already used by Morris and Shin (1998) for studying currency crises) Rochet and Vives (2004) show that a combination of liquidity requirements, solvency requirements and LLR interventions may prevent the occurrence of co-ordination failures on interbank markets.9 Such co-ordination failures arise when some (large and uninsured) depositors decide to withdraw, not because they think the bank is likely to be insolvent, but because they anticipate others will withdraw. The rationale behind liquidity requirements is that they reduce the impact of strategic uncertainty on the final situation of the bank, since they allow the bank to withstand larger withdrawals. The same is true for solvency requirements and LLR intervention. The difficulty (yet to be solved) is to determine the appropriate combination of these three instruments that minimizes the total costs of prevention of such coordination failures. Macroeconomic shocks and systemic issues Some form of government intervention may be needed in case of macroeconomic shocks such as recessions, devaluations, stock market crashes and the like. The same is true for disruptions in the payment system. Anticipating this kind of intervention, banks may rationally decide to take an excessive (from the point of view of social welfare) exposure to such risks, knowing that they are likely to be bailed out if the risks materialize. Rochet (2004) studies this question and shows that ex ante regulation of banks’ liquidity may be a way to mitigate this behaviour. We develop his analysis in the next section, where we discuss a possible rationale for the regulation of banks’ liquidity.

Why regulate banks’ liquidity? After having established that banks need liquid reserves, in particular because access to financial and interbank markets may not always cover their short-term financing needs, it remains to understand why regulation is needed, i.e. why the managers and shareholders of these banks do not choose by themselves the appropriate level of liquid reserves for their bank. In fact, like solvency regulations, liquidity regulations can be justified by two forms of externalities: the first is associated with the protection of small depositors, who are likely to be hurt by the failure of their bank, but are not in a position to monitor or influence the decisions of its managers. This explains why, in the vast majority of countries around the world, small depositors are insured and banks are regulated and supervised by supervisors or central banks, in charge of protecting the interests of depositors, or minimizing the liability of the deposit insurance fund. The second justification for banking regulations has to do with the protection of financial stability, i.e. the guarantee that the payment and financial systems are able to channel funds appropriately between economic agents, even if the country is hit by a large shock, like a recession, an asset price crash, a devaluation or a terrorist attack.

How should we regulate banks’ liquidity?

181

This dichotomy between two types of rationales is clear for solvency regulations, which are probably the most important of banking regulations. Indeed, these solvency regulations are usually justified by two reasons: as a complement to deposit insurance schemes in the protection of small depositors (banks’ capital is useful to limit both the probability and impact of banks failures and also to limit the incentives of under capitalized banks to take too much risk) and as a protection against systemic risk, i.e. any risk that might endanger the stability of the banking and financial system (including payment and securities settlement systems). Thus, there is a micro-prudential aspect and a macroprudential aspect to ex ante regulation. Similarly, liquidity regulations can be justified by micro- and macroprudential reasons: they are a complement to the LLR facility, since they limit the need for emergency liquidity assistance when an individual bank is in trouble. Also they are useful during banking crises or in case of macroeconomic shocks, since they limit the need for a generalized bailout. This is especially so because of the commitment problem of governments who typically feel inclined to intervene ex post during a banking crisis. To limit this tendency, liquidity requirements should be conditioned on the bank’s exposure to macro shocks (Rochet, 2004). In practice this means that uniform liquidity requirements (like the Stock Liquidity Requirement imposed on banks operating in the United Kingdom) could be replaced by more flexible systems, where the liquidity requirement may be more or less stringent according to the bank’s solvency and/or to simple measures of the bank’s exposure to several types of macroeconomic shocks, deduced, for example, from Value-at-Risk calculations under different scenarios.10 An important issue concerns the need for public (as opposed to private) regulation, i.e. whether banks could regulate themselves, like participants in a clearing house. Holmström and Tirole (1998) show that the private solution can be sufficient if there are no aggregate shocks. However a purely private solution is likely to be relatively complex to implement. It would consist of requiring banks to form pools of liquidity (like in the case of the German Liko-bank) and to sign multilateral credit line commitments, specifying clearly the conditions under which an illiquid bank would be allowed to draw on its credit line. By contrast, emergency liquidity assistance by the central bank is probably simpler to organize, but may be prone to forbearance under political pressure. In any case, due to the possibility of macro-shocks, some form of government intervention is needed. The difficulty is then to avoid excessive intervention, such as ex post bailouts of insolvent banks. We discuss this question in the next section. As already noted, liquidity regulation of large participants in the payment system is also warranted in order to limit the risk of needing massive liquidity injections by the central bank in case of a disruption in the payment system. Two policy questions arise: •

Is it necessary to impose an additional liquidity requirement (on top of the SLR, that is aimed at covering potential liquidity problems over a short period, say a week) to cover also intraday liquidity needs?

182 •

J.-C. Rochet If the answer to the first question is yes, how to set this additional liquidity requirement, taking into account that banks have the possibility to ‘bypass’ the RTGS system by either entering into bilateral netting agreements with other banks or using competing DNS systems, which could be more prone to systemic risk?

Finally, it should be noted that systemic risk in payment systems and interbank markets could be eliminated altogether if the central bank decided to insure interbank transactions and payments finality against credit risk. This system was implicitly in place in many countries during most of the last century. Thus the only logical explanation for the recent movement towards RTGS systems and limitation of LLR interventions is that central banks and prudential supervisory authorities want to promote peer monitoring by banks. The same reason may explain the surprising reliance of commercial banks on short-term finance. However, Rochet and Tirole (1996) show that the effective implementation of peer monitoring among banks may be difficult, due to commitment problems by governments. Liquidity requirements may be a useful way to mitigate these commitment problems.

How to regulate banks’ liquidity? As we have seen, there are two essential motivations for regulating banks’ liquidity, one being micro-prudential (i.e. limiting the externality associated with individual bank failures) and the other being macro-prudential (i.e. limiting excessive exposures to macroeconomic shocks by banks, under the expectation of a generalized bailout by the government). A simple liquidity ratio such as the SLR seems to be appropriate to cover the first objective, with the possible qualification that undercapitalized banks could be subject to more stringent requirements. This would be in the spirit of the ‘Prompt Corrective Action’ methodology imposed by the FDIC Improvement Act (1991) to US supervisors, i.e. the idea of some progressiveness in the restrictions imposed on problem banks, forcing supervisors to act before it is too late. However, the macro-prudential objective of liquidity regulation seems more difficult to attain, given, in particular, the difficulty in forecasting precisely the liquidity needs of banks during a crisis. One particular component of these liquidity needs is, of course, related to the intraday needs of the banks for channelling their large-value payments on the RTGS systems, but it has to be stressed that other liquidity needs, equally important to meet during a crisis, may materialize only after two to five days (for example, refinancing on the interbank markets). This implies that the crucial distinction is not in terms of time horizon (intraday versus two to five days) but rather between individual shocks, for which there is no reason to extend emergency liquidity assistance to banks that are insolvent (and therefore simple, uniform, liquidity ratios should be enough), and macroeconomic shocks, for which a massive liquidity injection by the central bank (and maybe a partial recapitalization of some of the banks by the finance ministry) may be warranted.

How should we regulate banks’ liquidity?

183

Thus there seems to be a need for a second type of liquidity requirement, based on some indices of exposure to macroeconomic shocks by individual banks, and intended to limit the need for an ex post liquidity injection by the central bank. These indices should be designed ex ante (and adjusted regularly) by the prudential supervisory authority, possibly after using the internal risk model of each bank and different sorts of stress tests. One difficulty would be, of course, to avoid regulatory arbitrage, i.e. ‘window dressing’ or manipulations of accounting information by the banks, in order to minimize their liquidity requirements, without effectively decreasing their exposure to macroeconomic shocks. In the context of LVPS, it would mean, for example, requiring cooperation and information sharing between the RTGS and any privately run competitor, and computing collateral requirements on an aggregate basis. However, additional liquidity requirements aimed at mitigating macroeconomic shocks could constitute a ‘waste’ of liquidity, given that they would be used only under exceptional circumstances. A superior solution may be, in this case, for the central bank to commit to provide conditional credit lines under the strict control of an independent prudential regulator or other independent authority. The characteristics of these credit lines (maximum amount, commitment fee, conditions under which they can be used) would be specified ex ante by the independent authority. The associated loans could be made senior to all other liabilities, thus limiting the risk of recourse to taxpayers’ money.

Conclusion Like solvency regulations, liquidity regulations for banks can be justified by two different motives: one is to limit the risk and the impact of individual bank failures, the other is to limit the need for massive liquidity injections by the central bank in case of a macroeconomic shock. In normal times, the pool of marketable securities that can provide liquidity to the banks is substantial. Therefore a simple, uniform liquidity ratio like the SLR may be all that is needed, with the possible qualification that the prudential supervisor could require additional liquidity for undercapitalized banks, in the spirit of the ‘Prompt Corrective Action’ implemented in the United States. As for macro-prudential purposes – that is, anticipating what can occur in the case of large macroeconomic shocks – it is probably necessary to go further, and either require additional liquidity, or secure a credit line by the central bank, both based on the exposure of each individual bank to such macroeconomic shocks and specified by the prudential supervisor or another independent authority. The definition of appropriate indices of such exposure to macroeconomic shocks (possibly using stress tests and worst case scenarios) is an important empirical challenge. Similarly, some form of cost–benefit analysis of LLR interventions would be useful in order to evaluate the exact costs of liquidity provision by the central bank, and the social costs of excessive liquidity.

184

J.-C. Rochet

Notes 1 I acknowledge useful comments by Victoria Saporta. I am solely responsible for any remaining mistakes. 2 See Bank of England (2003). 3 This is criticized in the chapter by Selgin reprinted in this volume which disputes the suggestion that DNS systems are intrinsically subject to systemic risk, at least in the absence of government intervention. 4 However the composition of payments in the two systems is different: the average payment on Fedwire is much bigger than on CHIPS. (I thank Victoria Saporta for providing these figures.) 5 See http://bodurtha.georgetown.edu/enron/derivatives_events.htm for a list of other derivatives ‘disasters’. 6 However, Allen and Gale (2004) show that liquidity requirements for banks may be needed when the financial markets for transferring aggregate risks among agents are incomplete. 7 See the chapters by Bech, McAndrews and Lester et al. in this volume for more detail on the evolution of different large-value interbank payment systems and on the systemic risk and liquidity properties of RTGS versus DNS systems. For a contrarian view on the systemic risk benefits of the evolution of RTGS systems, see the article by Selgin reprinted in this volume. 8 For example, Hoffman and Santomero (1998) show that in the United States the discount window (the lender of last resort facility) was often used improperly to rescue banks that subsequently failed. 9 This methodology is extremely fruitful. For example, Morris and Shin (2004a) used it to model debt pricing, Morris and Shin (2004b) to model liquidity crises on asset markets, Haldane et al. (2005) for analysing sovereign debt restructuring, Goldstein (2005) for modelling twin crises, and Goldstein and Pauzner (2005) for modelling bank runs. 10 As of January 2006, the SLR requires that UK-owned retail banks hold liquid assets to cover worst-case net wholesale outflows in sterling over the next five days plus 5 per cent of sterling retail deposits. Up to half of the net wholesale outflow can be met from holdings of sterling certificates of deposit, but the remainder must be assets eligible at the Bank of England.

References Allen, F. and Gale, D. (2000) ‘Financial contagion’, Journal of Political Economy, 108(1): 1–33. Allen, F. and Gale, D. (2004) ‘Financial intermediaries and markets’, Econometrica, 72: 1023–61. Bank of England (2003) ‘Strengthening financial infrastructure’, Bank of England Financial Stability Review, 15: 80–100. Freixas, X., Parigi, B. and Rochet J.C. (2000) ‘Systemic risk, interbank relations and liquidity provision by the central bank’, Journal of Money Credit and Banking, 32(2): 611–38. Froot, K. and Stein, J. (1998) ‘A new approach to capital budgeting for financial institutions’, Journal of Financial Economics, 47: 55–82. Goldstein, I. (2005) ‘Strategic complementarities and the twin crises’, Economic Journal, 115: 368–90. Goldstein, I. and Pauzner, A. (2005) ‘Demand-deposit contracts and the probability of bank runs’, Journal of Finance, 60(3): 1293–327.

How should we regulate banks’ liquidity?

185

Goodfriend, M. and Lacker, J. (1999) ‘Limited commitment and central bank lending’, Federal Reserve of Richmond Working Paper 99–2. Haldane, A., Penalver, A., Saporta, V. and Shin H.S. (2005) ‘Analytics of sovereign debt restructuring’, Journal of International Economics, 65(2): 315–33. Hoffman, P. and Santomero, A. (1998) ‘Problem bank resolution: evaluating the options’, The Wharton Financial Institutions Centre Working Paper 98–05-B. Holmström, B. and Tirole, J. (1998) ‘Private and public supply of liquidity’, Journal of Political Economy, 106(1): 1–40. Morgan, D. (2002) ‘Rating banks: risk and uncertainty in an opaque industry’, American Economic Review, 92(4): 874–88. Morris, S. and Shin, H.S. (1998) ‘Unique equilibrium in a model of self-fulfilling currency attacks’, American Economic Review, 88(3): 587–97. Morris, S. and Shin, H.S. (2004a) ‘Coordination risk and the price of debt’, European Economic Review, 48(1): 133–53. Morris, S. and Shin, H.S. (2004b) ‘Liquidity black holes’, Review of Finance, 8(1): 1–18. Rochet, J.C. (2004) ‘Macroeconomic shocks and banking supervision’, Journal of Financial Stability, 1(1): 93–110. Rochet, J.C. and Tirole, J. (1996) ‘Interbank lending and systemic risk’, Journal of Money, Credit and Banking, 28: 733–61. Rochet, J.C. and Vives, X. (2004) ‘Coordination failures and the lender of last resort: was Bagehot right after all?’, Journal of the European Economic Association, 6(2): 1116–47. Tirole, J. (2005) Corporate Finance, Princeton, NJ: Princeton University Press.

Part IV

Policy perspectives on the future of payments

12 The diffusion of real-time gross settlement Morten L. Bech

Introduction Payments are an integral part of a modern market economy as most transactions involve the use of cash, checks or electronic transfers. Commercial banks and the central bank provide in a symbiotic relationship the infrastructure through which payments flow. An efficient payment system is a prerequisite for a wellfunctioning economy. Historically, central banks have played an active role in the payment system. They continue to do so both as a provider of payment services and as an overseer of private sector systems. The extent to which the central bank is involved in the payment system varies from country to country. Most central banks at least provide the medium to settle small payments, i.e. cash. In addition, central banks tend to support an interbank payment system that settles large, time-critical wholesale payments. Many central banks, such as the Federal Reserve, also process retail payments including checks and Automated Clearing House (ACH) transfers. However, central banks tend not to issue debit cards or credit cards. Historically, interbank payments have been settled via netting (end-of-day) systems. The volume of interbank payments increased dramatically throughout the 1980s, 1990s and early 2000s as a result of rapid financial innovation and the integration and globalization of financial markets. As the volume of transactions increased, central banks became worried about the risks inherent in netting systems.1 Hence, in the last couple of decades many countries have modified the settlement procedure employed by their interbank payment system with a view to reduce both settlement and systemic risks. Most central banks opted for the implementation of Real-Time Gross Settlement (RTGS) systems. By 1985, three central banks had implemented an RTGS system; a decade later, the number had increased to 16. Since the mid-1990s, the rate of adoption of RTGS systems has increased significantly. Until the late 1990s, RTGS was a phenomenon utilized predominately by industrialized countries, but both transitional and developing countries have begun investing heavily in improving their financial infrastructures and payments systems. At the end of 2005, 91 central banks had adopted RTGS systems.

190

M.L. Bech

In this chapter, we take an in-depth look at the current state of affairs in the area of interbank payment systems. Based on a survey of central bank web sites, we map out the diffusion process of the RTGS technology among the central banks of the world. We compare the diffusion process to that of other technologies and discuss the key drivers of this evolution. We briefly discuss the emergence of two alternative technologies and speculate on how the future of central bank operated interbank payments systems might look in terms of settlement methodology. At the apex of the financial system are a number of critical financial markets that provide the means for agents to allocate capital and manage their risk exposures.2 Instrumental to the smooth functioning of these markets are a set of wholesale payment systems and financial infrastructures that facilitate clearing and settlement. Most of these infrastructures and systems use central bank money for final settlement of obligations. Moreover, the central bank uses the interbank payment system to implement monetary policy and it serves as the platform for the interbank money market. A poorly designed system can either create or magnify the impact of shocks to the financial system, potentially resulting in contagion within and across economies.3 Furthermore, an efficient and resilient payment system ensures that the central bank can conduct monetary policy effortlessly and without regard to other concerns. Efficiency increases the responsiveness of the monetary system to impulses from the central bank and decreases transaction costs for agents of the economy. Resiliency ensures that the central bank can act swiftly and in a timely manner to monetary shocks. Hence, a sound interbank payment system is a precondition for the successful conduct of monetary policy as well as financial stability. In the United States, there are two principal systems that settle interbank payments: the Federal Reserve’s Fedwire Funds Transfer System® (Fedwire) and the Clearing House Interbank Payments System (CHIPS) – a private sector enterprise. Today more than 9,500 participants use Fedwire to initiate funds transfers. Participants use Fedwire to handle large-value, time-critical payments, such as payments for the settlement of interbank purchases and sales of federal funds; the purchase, sale and financing of securities transactions; the disbursement or repayment of loans; and the settlement of real estate transactions. Several ancillary payment and securities settlement systems use Fedwire to both prefund their respective settlement processes and square final positions over the course of the business day.4 In Fedwire participants initiate funds transfers that are immediate, final and irrevocable when processed. Fedwire is a RTGS system. In fact, Fedwire was the world’s first RTGS system. Its origins go back to 1918 when the Federal Reserve inaugurated a network of wire communications among the individual Reserve Banks. The new system of wire-initiated book entries allowed funds to be transferred on behalf of the member banks and significantly reduced the need for physical shipment of gold and currency. In the early 1970s, the Fedwire system migrated to a fully computerized platform, and settlement in real time was achieved.

191

600

90

500

75

400

60

300

45

200

30

100

0

Fedwire  CHIPS (RHS) CHIPS Fedwire 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005

Multiple of GDP

$ trillion

The diffusion of real-time gross settlement

15

0

Year

Figure 12.1 Value of transfers originated on Fedwire.

During the first year of operation the Federal Reserve Bank of New York processed around 100 wires per day which increased to about 600 per day, ten years later. Today, an average of over 525,000 transfers is originated every day over Fedwire. The value of transfers originated has seen tremendous growth as well. As shown in Figure 12.1, the annual turnover increased from just over $100 trillion in 1985 to over $520 trillion in 2005. As a multiple of GDP, the value of transfers originated over Fedwire and CHIPS went from 45 to 70 over the same period. The interbank payment systems in other countries traditionally settled payments using end-of-day net settlement.5 In a deferred net settlement (DNS) system payment orders are accumulated throughout the day. Settlement of the net amount takes place typically once, at the end of the day. By reducing the number and overall value of payments, netting substantially reduces the amount of money needed to settle a given set of obligations. However, a well-established drawback of (unprotected) DNS systems is the higher risk involved. Finality of settlement is only achieved at the settlement period and thus there is no certainty that the payments will be settled until that point in time. If one participant fails to meet its payment obligation when due, all processed payment orders could be unwound with the consequent risk of other participants failing to be able to meet their obligations in turn. As interbank payment systems around the world saw growth in the value of payments settled similar to that of Fedwire, steps were taken to reduce the amount of settlement risk by changing the settlement procedure to RTGS.

192

M.L. Bech

The diffusion of real time gross settlement In the 1980s, a number of Western European countries began implementing RTGS systems. Denmark started the trend in 1981, and the Netherlands and Sweden followed suit in 1985 and 1986, respectively. By 1988, RTGS systems operated in five of the six major currencies (all but sterling), as SIC was implemented in Switzerland (1987), EIL-ZV in Germany (1987) and BoJ-NET in Japan (1988). As BoJ-NET provided both DNS and RTGS, the RTGS mode was seldom used in practice due to higher liquidity costs.6 The last country to implement an RTGS system in the 1980s was Italy in 1989. During the early 1990s, RTGS adoption continued at a rate of roughly one country per year with Finland in 1991, Czechoslovakia and Turkey in 1992, Poland in 1993 and South Korea in 1994. In 1992, the Treaty of Maastricht created the foundation for Economic and Monetary Union (EMU) in Europe. A year later, the central banks within the European Union agreed that each member state should have an RTGS system. Furthermore, in 1995 the central banks decided to interlink the national RTGS systems through the TARGET system in order to facilitate a single monetary policy. These decisions led to a flurry of new systems and upgrades to exiting ones. TARGET went live on January 4, 1999.

Figure 12.2 Adoption of RTGS in Europe – 1995.

The diffusion of real-time gross settlement

193

In order to promote the use of the euro, the EU countries that did not join the EMU were allowed to participate in TARGET. Denmark, Sweden and the United Kingdom implemented separate euro RTGS systems alongside RTGS systems for their domestic currencies. Greece eventually joined the euro on January 1, 2001, when its new RTGS system, Hermes, went live. The Eurosystem is currently developing the next generation of TARGET, TARGET2. TARGET2 will dispose of the national RTGS systems and run on a single technical platform. All euro-system central banks will participate in TARGET2. Denmark will participate, but Sweden and the United Kingdom will not. TARGET2 is planned to go live in the second half of 2007. As the European Central Bank (ECB) made RTGS a prerequisite for membership of the EMU, the prospective members in the rest of Europe began to implement RTGS as well. RTGS was implemented in Slovenia in 1998, Hungary in 1999, Latvia in 2000, Bulgaria, Estonia and Malta in 2002, Slovakia in 2003, Lithuania in 2004 and Romania in 2005. The continent-wide adoption also encouraged RTGS implementation in countries outside of the sphere of the European Union and accession countries. Norway implemented RTGS in 1997, Belarus in 1998, Iceland in 2000, Azerbaijan and Georgia in 2005 and Ukraine in 2005. As hostilities ended in the Balkans in the late 1990s, governments started efforts to rebuild their respective economies and establishing sound and

Figure 12.3 Adoption of RTGS in Europe – 2005.

194

M.L. Bech

efficient financial systems was considered a priority. RTGS systems were – with support from the European Union and the World Bank – implemented in Croatia in 1999, Bosnia and Herzegovina in 2001, Macedonia in 2001, Serbia in 2003, Albania in 2004 and Montenegro in 2005. With ongoing projects in Russia, Cyprus and Moldova, the diffusion of RTGS in Europe is nearly complete. Outside Europe the rate of adoption of RTGS has been equally impressive since the mid-1990s. Australia and New Zealand implemented in 1998 and they remain the only countries in Oceania that have gone live with RTGS. In Asia, the rate of RTGS implementation has been fairly steady. On average, about one

Figure 12.4 Adoption of RTGS in Asia – 2005.

The diffusion of real-time gross settlement

195

Asian country has adopted RTGS per year. Besides the early adopters Japan and Korea, the following countries have implemented RTGS: Thailand in 1995, Hong Kong and Kazakhstan in 1996, Singapore in 1998, Malaysia in 1999, Indonesia in 2000, Taiwan, China and the Philippines in 2002, Sri Lanka and India in 2003. In 2001, the Bank of Japan reconfigured its interbank payment system making RTGS the only mode of settlement for funds and abolishing DNS.7 In addition, the State Bank of Pakistan is now in the process of introducing a RTGS system. Saudi Arabia was the first country in the Middle East to implement RTGS in 1997. Additional Middle Eastern countries have since adopted RTGS as well: Qatar in 2000, United Arab Emirates in 2001, Jordan in 2002, Kuwait in 2004 and Oman in 2005. Israel expects its new system to be operational in the first quarter of 2007. The adoption of RTGS is well under way in Asia. In Africa, the South African Reserve Bank (SARB) spearheaded adoption in 1998. Through the South African Development Community (SADC) the SARB has participated in developing and strengthening the financial infrastructure in the rest of southern Africa.8 Among the goals for the SADC is the implementation of RTGS systems. RTGS has been implemented so far in the following SADC member states: Mauritius in 2000; Namibia, Malawi and Zimbabwe in 2002; Botswana in 2003 and Tanzania and Zambia in 2004. The Central Bank of West African States (BCEAO) launched an RTGS system in 2004 for the members of the West African Economic and Monetary Union (WAEMU).9 The Bank of Central African States (BEAC), which shares the Franc CFA as currency with the BCEAO, is in the process of implementing a similar RTGS system. The BEAC is the common central bank for the Central Africa Economic and Monetary Community (CEMAC).10 In addition to these regional efforts, four countries have implemented RTGS systems. These countries are Ghana in 2002, and Uganda, Kenya and Nigeria in 2005. Moreover, the central banks of Algeria, Egypt and Libya have all started RTGS projects. Africa has seen tremendous growth in the adoption of RTGS in the last five years and will continue to do so in the near term. Many of the projects in Africa have received support from the World Bank. In the Western Hemisphere, Canada is an interesting case. It is the only Group of Ten (G10) country that has decided not to implement a RTGS system. Instead, Canada opted for a system that employs multilateral netting by novation.11 The Canadian Large Value Transfer System (LVTS) is considered to be equivalent to RTGS in terms of finality as the Bank of Canada provides an explicit guarantee of settlement. Mexico began a substantial modernization of its payment systems in 1994. At that time, the Bank of Mexico operated both an electronic interbank settlement system and a manual check clearinghouse. A principal objective of the reforms was to replace large-value checks with electronic transfers through the Bank’s existing RTGS system (SIAC), This was accomplished in stages culminating with the launch of a new RTGS system (SPEI) in 2005.

196

M.L. Bech

Figure 12.5 Adoption of RTGS in Africa – 2005.

In South America, Uruguay was the first country to adopt RTGS in 1995. The trend of implementation has since been about one country every two years. Argentina’s payment system was substantially reformed in 1997, when the central bank established a new framework for private clearinghouses to modernize the traditional paper-driven systems. At the same time the Argentine central bank also implemented a RTGS system. Colombia followed in 1998, Peru in 2000, Brazil in 2002, and Bolivia and Chile in 2004. In other words, seven of 13 countries have adopted RTGS in South America. RTGS implementation in Central America and the Caribbean has only started recently. The Netherlands Antilles and Cuba began using RTGS in 2001 and 2002, respectively. The following countries have also implemented RTGS:

1995

2005

Figure 12.6 Adoption of RTGS in South America.

Figure 12.7 Adoption of RTGS in Central America – 2005.

198

M.L. Bech

Barbados and Costa Rica in 2002, Trinidad and Tobago in 2004 and Guatemala in 2005. The Inter-American Development Bank is assisting the efforts to implement RTGS systems in the region.

Technology diffusion Technology is a factor in the production of payment services just like labor and capital. However, it is intangible and difficult to measure. Theories of technological change emphasize two aspects of technology. First, technology is nonrival in the sense that the marginal costs for an additional agent to use the technology are negligible. Second, the return to technological investments is partly private and partly public.12 The private return to individual agents facilitates innovation but innovation also benefits other agents through knowledge spillovers. According to Rogers (1995), diffusion is the process by which a technology is communicated through certain channels over time among members of a social system. Thinking of the community of central banks as a social system suggests that the diffusion of RTGS described above can be compared to the diffusion of other technologies. A stylized fact from empirical studies of technology diffusion is that the rate of adoption of new technology follows a predictable intertemporal pattern.13 At first the rate of adoption is slow but at some point it takes off if the technology eventually is successful. The rapid adoption continues until a sizeable share of the members of the social system have adopted the technology, at which point the rate of adoption levels off and eventually begins to fall. This pattern of adoption implies that the share of adopters in a social system follows a sigmoidal or S-shaped curve as a function of time as illustrated by Figure 12.8. The Morgan Stanley Central Bank Directory 2005 lists 174 central banks and monetary authorities. In the 1980s the adoption rate was one central bank every other year and this increased to one per year in the early 1990s. In the years following 1995, the annual RTGS adoption rate has not dipped below five new central banks per year. The rate peaked between 2001 and 2002, when a total of 21 central banks implemented new RTGS systems. The number of adopters and the cumulative frequency distribution is shown in Figure 12.9. The diffusion of RTGS among central banks appears (so far) to be consistent with the stylized fact. Members of a social system have different capacities to adopt a new technology. The capacity can, in an international context such as this, reflect different stages of economic development or underlying cultural attitudes toward technological change. Rogers (1995) classifies agents’ capacity to adopt a technology into five categories under the assumption that the rate of adoption follows a bell curve. The first 2.5 percent of adopters are labeled innovators. The following 13.5 percent are labeled opinion leaders or early adapters. The early majority is the next 34 percent of adopters up to the median while the late majority is the 34 percent above the median. The remaining 16 percent of adopters are called laggards or late adopters.14 Rogers (1995) argues that opinion leaders tend to

Diffusion starts to take off

Innovators (2.5%)

Opinion leaders (13.5%)

Early majority (34%)

Late majority (34%)

Laggards (16%)

Figure 12.8 S-curve and adopter groups. 175 100 150 80

60

100

75 40 50 20 25

0 1970

1980

1990

2000 Year

Figure 12.9 Adoption of RTGS in central banking.

2010

0 2020

Number of central banks

Adoption (%)

125

200

M.L. Bech

hold the key to technology diffusion as a whole. Opinion leaders are agents who provide advice and information about an innovation to members of the social system. These agents tend to support the norms of the social structure and serve as a model for others. Opinion leaders are at the center of the communication network of the social system and can reach a large number of members with information. Applying Roger’s classification implies that central banks that adopted RTGS prior to 1987 would be considered innovators. Central banks that adopted RTGS after 1986 but before 1998 are opinion leaders. Central banks that adopted RTGS from 1998 through 2004 belong to the early majority and the most recent adopters are in the late majority. Among the innovators and opinion leaders, we find the members of the Committee on Payment and Settlement Systems (CPSS), with the exception of Canada.15 CPSS serves as a forum for the central banks of the G10 countries to monitor and analyze developments in domestic payment, settlement and clearing systems as well as in cross-border and multicurrency settlement schemes. The CPSS has been influential in promoting RTGS by providing advice and distributing information on payment system design. For example, in 1997, the CPSS published a report on RTGS that laid out general features as well as specifics of the systems in operation in the CPSS countries.16 In addition, the CPSS has been instrumental in defining the norms of the central bank community in area of payments. The CPSS has published best practices and guidelines such as the Lamfalussy report on interbank netting schemes, and the set of core principles for systemically important payment systems.17 Many central banks belonging to the early and late majorities explicitly cite the recommendations put forward by the CPSS as a reason for implementing RTGS. Moreover, the CPSS recommendations are part of the toolkit of the Financial Sector Assessment Program (FSAP) jointly established by the International Monetary Fund (IMF) and the World Bank in 1999. The FSAP specifically looks at countries’ financial sectors, assessing strengths and vulnerabilities in order to reduce the potential for crisis. One view in the literature on international technology diffusion is that there is a common pool of knowledge to which all countries have access, so technology diffusion is constrained only by the receiving country’s ability to understand and make use of the new technology. However, the empirical literature on technology diffusion across countries suggests that it has a spatial dimension. In other words, geography matters and diffusion tends to be localized or clustered around innovative centers.18 One reason is that typically only a broad outline of a given technology is codified and readily available. Thus successful implementation requires ‘know how’ or skills that are tacit. These non-codified aspects require face to face interaction in order to be transferred.19 On the other hand, the same literature also finds that physical or cultural distances tend to matter less as consequence of improved communication technology, lower transportation costs and multicultural interactions. It is outside the scope of this article to provide statistical evidence for localization and the importance of bilateral linkages. However, casual inspection of

The diffusion of real-time gross settlement

201

the maps shown above does suggest that the adoption was clustered at least in Europe and southern Africa. We alluded to the importance of the EU and the SADC for these phenomena above. A potential example of the importance of bilateral ties is the fact that central banks of Australia, Hong Kong, New Zealand, Singapore and South Africa quickly followed the decision of the Bank of England to implement RTGS. All of these central banks have close historical ties to the Bank of England and extensive staff exchange programs. An important channel of international technology diffusion is trade. Trade gives access to foreign goods or, implicitly, technologies. Instead of developing an RTGS system on its own a country may simply decide to import it from abroad. While central banks often provide technical expertise to one another they tend not to sell systems to each other. However, as the RTGS technology matured and the adoption took off, private companies entered the market place and began to offer off-the-shelf or standardized software solutions. Currently, there are at least four providers that have built RTGS systems in more than one country. They are LogicaCMG plc of the United Kingdom, CMA Small System AB of Sweden, the joint venture of Perago Ltd of South Africa and SIA SpA of Italy, and Montran Corporation of the United States. The possibility of sharing development costs across customers, and competition among providers, have presumably lowered the cost of implementing RTGS and hence made it feasible for more countries to adopt. Countries that have implemented solutions from these four providers are listed in Table 12.1. A total Table 12.1 Imported RTGS systems LogicaCMG

CMA

Perago/SIA

Montran

Luxembourg ECB Ireland Azerbaijan Hungary Turkey Latvia Bosnia and Herzegovina India United Arab Emirates United Kingdom France Monaco Andorra Saudi Arabia New Zealand Slovenia Croatia

Montenegro Oman BCEAO Macedonia Moldova Pakistan Libya Algeria BEAC

South Africa Namibia Malawi Zimbabwe Zambia Uganda Sweden

Romania Georgia Mauritius Kuwait Bulgaria Barbados Netherlands Antilles Ghana Albania Bahamas Tanzania Angola Guatemala

Source: Company websites and newswires. Note Italics imply contract has been signed.

202

M.L. Bech

of 37 central banks have implemented solutions from one of these providers and within the early and late majority groups it is 31 out of 63 or close to half.

The future of real time gross settlement A new technology is adopted over time by a social system and technology itself evolves over time. Technological innovations tend to occur in bursts and, typically, when the limitations of the previous technology become binding. A new technology might emerge before the previous one has been fully adopted by the social system. Undoubtedly, the large-value payment system will continue to evolve. Globalization and financial integration that have contributed to recent developments are likely to be important drivers for years to come. The current set up of heterogeneous settlement structures across different currencies is likely to come under pressure. Flexible and interoperable solutions will be in demand as legacy systems are replaced. In terms of RTGS there is countervailing evidence as to whether it will continue to be the settlement technology of choice for central banks. At least two alternative technologies are vying to be the way of the future. One is that of hybrid systems, which have already been implemented in several countries around the world. The other is the e-settlement vision laid out by Leinonen in the next chapter of this book. Hybrid systems employ advanced settlement algorithms that combine components of both net and real time gross settlement.20 The hybrid system technologies emerged because, the elimination of settlement risk in an RTGS system results in an increased need for intraday liquidity to smooth the nonsynchronized payment flows. While central banks provide intraday liquidity, it is costly: either in the form of explicit fees or implicitly as the opportunity cost of pledged collateral. Hence, banks manage their liquidity closely throughout the day. In order to suppress the demand for intraday liquidity, several systems have developed different types of liquidity savings features. Such features include advanced queue management such as different priority levels and the possibility of reordering payments in the queue, urgent and non-urgent payment streams, bilateral limits to manage credit exposures and reciprocity, bilateral offsetting of payments and full or partial multilateral netting.21 Prominent examples include RTGS plus in Germany, LVTS in Canada, CHIPS in the USA, Paris Net Settlement (PNS) in France and BI-REL in Italy. An overview of the different features available in the systems is provided in Table 12.2. In the short run, more payment systems are likely to adopt liquidity savings features similar to the ones mentioned previously. In fact, many of the RTGS systems mentioned above already encompass gridlock resolution mechanisms.22 In the world of e-settlement, the current centralized structure built around accounts at the central bank will be replaced by a decentralized network-based environment. The fundamental idea is that the settlement asset (a secure digital stamp) will be attached to payment messages. Messages will travel over a dedic-

The diffusion of real-time gross settlement

203

Table 12.2 Liquidity savings features RTGS+

Liquidity savings features Queue management Multiple payment streams Bilateral limits Bilateral offsetting Multilateral netting

Priority level Reordering

Full Partial

x x x x x x x

LVTS

BI-REL

PNS

x x x x x x

x

CHIPS x x

x x x

x x

Source: Bank for International Settlements (2005).

ated decentralized internet-like network. Payments would be akin to emails, and banks would operate payments and accounts servers just as companies manage email and mailbox servers. Under e-settlement, payments will settle gross and in real time much like cash changing hands. The e-settlement technology will have wide-ranging consequences for the structure of the interbank payment system as we know it today. Presumably, the distinction between commercial bank and central bank money would become mute. Moreover, the clearing role of the central bank would be greatly reduced, if not cease to exist, as a consequence of the decentralized structure. However, the central bank would still provide the settlement asset (and potentially credit) by generating the e-settlement stamps and distributing them to banks. The big question is whether the new distributed IT technologies will make the net benefits of decentralization greater than those of the current centralized computing structure. If that is the case then real-time gross settlement is here to stay, albeit under a new technological umbrella.

Conclusion Our analysis shows that RTGS has been very successful over the last three decades and has become the de facto standard in terms of settlement methodology among the central banks of the world. Moreover, we find that the diffusion process of technology within central banking – at least at the macro level – shares many similarities with other observed diffusion processes of technology. So far the adoption process of RTGS is consistent with the standard S-curve prediction. Moreover, some anecdotal evidence was found of clustering in the diffusion process. We identified in CPSS an opinion leader that has performed many of the roles suggested by theory in terms of disseminating information about the new technology and setting standards for the social system. We described hybrid systems as the emergence of a competing technology that potentially will replace RTGS. However, we also noted that on the horizon is a new technology that while implying substantive changes in the operation of the interbank payment system actually would turn the tide back in favor of RTGS.

204

M.L. Bech

Notes 1 See, for example, Bank for International Settlements (1990). 2 These markets include overnight interbank money, foreign exchange, commercial paper, government and agency securities, corporate debt, equity securities and derivatives. 3 The events of September 11, 2001 underscored the importance of a resilient interbank payment system. See McAndrews and Potter (2002) and Lacker (2004). 4 These systems include the Depository Trust & Clearing Corporation (DTCC), the Clearing House Interbank Payments System (CHIPS) and Continuous Linked Settlement (CLS). 5 CHIPS used to be an end-of-day net settlement system but in 2001 the system was changed to a prefunded net settlement system with multiple settlement cycles. 6 See Bank of Japan (2001). 7 In November 2005 the Bank of Japan released a consultation document on the proposal for the next generation of BoJ-NET. 8 The member states of the SADC are: Angola, Botswana, Democratic Republic of Congo, Lesotho, Malawi, Mauritius, Mozambique, Namibia, Swaziland, Seychelles, South Africa, Tanzania, Zambia and Zimbabwe. 9 WAEMU members are: Benin, Burkina, Côte d’Ivoire, Guinea Bissau, Mali, Niger, Senegal and Togo. 10 CEMAC members are: Cameroon, Central African Republic, Congo, Gabon, Equatorial Guinea and Chad. 11 ‘Novation’ is defined in Bank for International Settlements (2003) as: Satisfaction and discharge of existing contractual obligations by means of their replacement by new obligations (whose effect, for example, is to replace gross with net payment obligations). 12 Keller (2004). 13 See, for example, Griliches (1957) and Mansfield (1968). 14 The classification reflects the fact that for a normal distribution 68 percent of the observations are within one standard deviation from the mean (or median) and 95 percent are within two standard deviations. 15 The members are the central banks of France, Germany, Belgium, Italy, Japan, the Netherlands, Sweden, the United Kingdom, the United States, Canada and Switzerland. In 1997 the Hong Kong Monetary Authority and the Monetary Authority of Singapore joined the Committee. 16 In 2005, the CPSS published a new report on recent developments in large-value payments that focused on trends since the 1997 RTGS report. 17 Bank for International Settlements (1990, 2001). 18 Keller (2004) provides a review of the evidence of geographic effects on international technology diffusion. 19 Keller (2004). 20 The term was coined by McAndrews and Trundle (2001). 21 New settlement algorithms continue to be developed. See, for example, Johnson et al. (2005) for an introduction to the receipt reactive settlement methodology. 22 Bech and Soramäki (2002) discuss gridlock resolution mechanisms.

References Bank for International Settlements (1990) Report of the Committee on Interbank Netting Schemes of the central banks of the Group of Ten countries (Lamfalussy Report), Committee on Payment and Settlement Systems Publication No. 4. Bank for International Settlements (1997) Real-time gross settlement, Committee on Payment and Settlement Systems Publication No. 22.

The diffusion of real-time gross settlement

205

Bank for International Settlements (2001) Core principles for systemically important payment systems, Committee on Payment and Settlement Systems Publication No. 43. Bank for International Settlements (2003a) A glossary of terms used in payments and settlement systems, Committee on Payment and Settlement Systems. Online, available at: www.bis.org/publ/cpss00b.pdf (accessed November 1, 2006). Bank for International Settlements (2003b) The role of central bank money in payment systems, Committee on Payment and Settlement Systems Publication No. 55. Bank for International Settlements (2005) New developments in large-value payment systems, Committee on Payment and Settlement Systems Publication No. 67. Bank of Japan (2001) ‘Real-time gross settlement (RTGS) in Japan: an evaluation of the first six months’, Bank of Japan Quarterly Bulletin, 9(4). Bech, M. and Soramäki, K. (2002) ‘Liquidity, gridlocks and bank failures in large value payment systems’, in R. Pringle and M. Robinson (eds) E-Money and Payment Systems Review, London: Central Banking Publications. Emmons, W.R. (1997) ‘Recent developments in wholesale payment systems’, Federal Reserve Bank of St. Louis Review, 79(6): 23–44. Gilbert, A.M., Hunt, D. and Winch, K.C. (1997) ‘Creating an integrated payment system: the evolution of Fedwire’, Federal Reserve Bank of New York Economic Policy Review, 3(2): 1–7. Green, E.J. (2001) ‘Clearing and settling financial transactions, circa 2000’, in Santomero, A.M., Viotti, S. and Vredin, V. (eds) Challenges for Central Banking, Dordrecht: Kluwer. Green, E.J. and Todd, R.M. (2001) ‘Thoughts on the Fed’s role in the payment systems’, Federal Reserve Bank of Minnesota Quarterly Review, 25(1): 12–37. Keller, W. (2004) ‘International technology diffusion’, Journal of Economic Literature, 42(3): 752–82. Lacker J. (2004) ‘Payment system disruptions and the Federal Reserve following September 11, 2001’, Journal of Monetary Economics, 51: 935–65. McAndrews, J. and Potter, S. (2002) ‘Liquidity effects of the events of September 11, 2001’, Federal Reserve Bank of New York Economic Policy Review, 8(2): 59–79. McAndrews, J. and Trundle, J. (2001) ‘New payment system designs: causes and consequences’, Bank of England Financial Stability Review, 11: 127–36. Rogers, E. (1995) Diffusion of Innovations, New York: The Free Press.

13 E-settlement Soon a reality? Harry Leinonen1

Introduction Current payment clearing conventions, although now using electronic automation, have evolved out of paper-based physical processing and transportation. In order to further increase efficiency and improve services, payment systems need to be re-engineered. Although the technology necessary for improved payment systems has been available for many years, development in general and at the international level, in particular, has been very slow. The banking sector does not have the proper incentives for rapid migration to more efficient processing conventions. The current payment industry and service structures contain a massive barrier against change. The network externalities, monopolistic processing nodes, almost fixed demand of a complementary good and cooperation requirement among competitors fortify current practices. The rules of open competition do not function properly in the payments industry. The first objective of this chapter is to present the technical possibilities available to improve interbank payment infrastructures, network-based payment systems and decentralised settlement functions (e-settlement). The second objective is to analyse the barriers for change and how these could be circumvented in order to speed up development. History has shown that even the toughest barriers and walls do tumble down eventually. It is in the interest of the general public that the possibilities of modern technology are applied to the payment industry at an early stage and to ensure that old legacy investments and conventions do not delay development.

Development trends Using modern technology could improve customer service in payment systems in several ways. There are clearly visible trends, which can be found in other industries, and which have been introduced partially in advanced payment systems or employed by non-banking competitors. Payment processing could become faster and ultimately real-time delivery could be the norm. Already there are some countries in which most payments

E-settlement

207

are processed almost in real-time;2 PayPal3 is already providing this service worldwide. There are no technical barriers to real-time payment processing, but in most countries payment processing times are still long, typically several days.4 Processing costs should fall drastically for electronic payments. The costs of telecommunication, IT processing and data storage are all expected to continue to decrease. Payments are, from the IT-processing point of view, quite simple transactions. The data volume per transaction and the processing complexity are low compared with other types of documents and processing on the internet. The most dramatic change should be seen in interconnectivity. All companies and most private customers are likely to have PCs and servers connected to the internet and other networks. Everyone could be addressable and reachable over the network. The next generation of mobile phones will have broadband interfaces and processing capacities, which are sufficient for all kinds of payment processing. Efficient electronic straight-through-processing interfaces between banks and their customers could automate the processing of payments, not only for corporate customers but also for private customers. E-invoices could arrive as emails and could be processed automatically and dispatched to banks for payment over the internet. Both payments received and paid out will automatically be updated in the accounting systems. Completely automated payments will become the norm. The electronic payment services should be both reliable and secure. There have been some security and reliability problems with the internet and some of the first electronic payment implementations. However, the situation has improved over recent years and there are already solutions and standards in place to provide secure quality service to customers. This modern payment scenario can already be found in some national payment systems, especially in the Nordic and Baltic countries, where more than 50 per cent of customers use network-based e-payment solutions and over 90 per cent of bank payments are introduced via paperless means.5 The pressure on the banking sector to improve customer service in payments will increase in the coming years especially with e/m-payments developments. It is in the interest of the authorities to further development in order to ensure efficient payment services for the general public.

Network-based payment systems In this section, we discuss some important features that an efficient future payment system must possess. In particular, we argue that such a system needs to be based around a network on which everyone can be uniquely identified, that allows for straight-through processing, and on which settlement can be decentralised. It is this final feature that defines e-settlement and that we will discuss in more detail below.

208

H. Leinonen

Connecting networks Today everybody is connected via networks. Using the internet and e-mail we can send different kinds of messages, virtually in real-time, to most of the bank employees in the world. However, we lack these possibilities for payment messages. By extending the decentralised network concept of the internet to payments using secure and dedicated TCP/IP networks like SWIFTnet, we can build a new end-to-end and Straight Through Processing (STP) payment process without any intervening clearing and sorting centres. The network sorts the payments by transporting them to the given network address in the same way that an e-mail is routed to the given network server as described in Figure 13.1. In the email system banks have e-mail/mailbox servers. In a network-based payment system, banks will need payment/account servers, otherwise the concept is very similar.6 International account number (IBAN) In order to route payment messages efficiently, in network-based solutions as well as in any traditional system, we need a universal account number convention that clearly indicates the address of any account in the system. This could be compared to the address of mailboxes on the internet or international GSMtelephone numbers for routing SMS-messages internationally, as described in Figure 13.2a. An international account number (IBAN)7 seems to have become the preferred option.8 If all banks implement IBANs properly, the customer will only need to state the correct IBAN and the payment will be routed to the right account. This will require the systems to have search tables from which the right bank identifier code (BIC), bank name, network address, etc. can be found based on the IBAN. These kinds of modules and search tables are emerging.9 Without

Payment

Bank 1 Payment/ account server

TCP/IP Internet network

Bank 2 Payment/ account server

Payment

Payment Payment

Bank 3 Payment/ account server

Bank n Payment/ account server

Figure 13.1 Direct interbank communication in a network-based infrastructure.

E-settlement

209

(a) Payments can be made by transferring funds by addressing directly the receiving account in a common account number space.

(b) The common account number space is divided into sub-spaces belonging to service providers which are connected via interoperable system bridges. Figure 13.2 The common account number space.

a universal account number-space, efficient cross-border STP cannot be reached. Interoperable bridges are needed to route payments between the accounts of different service providers as described in Figure 13.2b. Immediate credit transfers The interbank payment transfer and settlement should be seen as one essential part of the whole payment (credit transfer) circle as described in Figure 13.3. The payment process starts and ends at the customers. In Figure 13.3, (1) the payer receives a bill or other instruction from the beneficiary concerning a payment to be made. (2) The payer then sends the instruction to his bank for processing and routing to the beneficiary’s bank. (3) This interbank leg includes e-settlement, so that the beneficiary’s bank receives both the payment message and the final settlement. (4) The beneficiary’s bank can then inform the beneficiary as to the incoming/final payment.10 This credit-push/credit-transfer type of payment is the most convenient and efficient in the network real-time world. It has fewer processing and transportation legs than electronic credit/debit card payments, direct debits or electronic

210

H. Leinonen Payer’s bank

Including e-settlement

Beneficiary’s bank

3 Bank-to-bank

4 Customerto-bank

Bank-tocustomer

E-banking

Statement receipts

2 Customer-to-customer

1

Beneficiary

Payer E-billing

Figure 13.3 E-settlement is part of the credit transfer circle, which provides efficient electronic communications between participants in a payment.

cheques. In credit transfers, the payer’s bank identifies its customer, checks the payment instruction and debits the payer’s account; the beneficiary’s bank checks the settlement and credits the beneficiary’s account. In the future realtime world, payments will be processed within seconds in the same way as e-mail and SMS-messages are now processed. The simple credit transfer/ credit-push structure will become the dominant payment method in the future. Cheques and complex direct debit schemes will be crowded out. Because card payments have a dominant position today, they will probably convert to creditpush technology over a period of years. The emerging mobile payment systems that survive will probably be based on the credit-push convention. Common interoperable standards There is a need for interoperable and common standards which support this automated billing and payment process. Modern IT technology is able to contain and transport all the relevant payment and invoicing information at very low cost. However, the content has to be specified clearly and the formats for formatted fields have to be agreed upon. In addition to the sender’s and receiver’s account address, amount and due date, there is, in particular, a need for a formatted reference number for automated reconciliation. Between service providers an audit trail code is required in order to track payments during and after the processing. There is also a need for standards between

E-settlement

211

service providers as well as between service providers and their customers, each supporting each other. From the customers’ point of view, common standards are the very basis for efficient interfaces and open markets. With common and open standards the automated general ledgers, receivables, pay roll systems, etc. can directly communicate with payment systems. Currently there are a large number of existing old domestic standards and proposed new standards, which are not interoperable. These have often been designed for some specific purposes and are often based on old paper-based concepts and IT limitations existing, when they were implemented.11 The new concepts all have in common the fact that they favour the XML-type (Extensible Markup Language)12 of design. Most of them build on legacy systems in which the redesigning benefits are not properly exploited. All the currently-used payment processes (as well as future ones) could be designed based on one basic message structure derived for an e-invoice. The e-invoice would have all the necessary billing and payment details in standardised formats. (The current credit transfer, direct debit and card payment messages contain a subset of data due to previous IT storage and communication limits). This standardised payment record could be used in the following payment initialisation processes: •

• •

•

payer input via e-banking: i.e., a normal credit transfer where the payer may have received the billing information electronically from the payee, for example, via the internet; payee initialisation via an e-billing process, which presents the bill for explicit payer acceptance, for example, via an e-banking interface; payee initialisation via a direct debit process in which the payee’s service provider sends the payment information to the payer’s service provider for automated booking under the condition that the payer has pre-authorised the payee, identified via IBAN, to debit the payer account within a given amount and time limit; payee initialisation via card-based authorisation, in which the billing information is accompanied by encrypted authorisation information provided by the authorisation modules in the merchant’s/payee’s payment terminal and payee’s card.

The payment data necessary for the different processing conventions are identical. The only difference can be found in the initialisation and acceptance processes. The payment and billing processes can be simplified and structured to a clear message dialogue containing the essential information for efficient enduser processing. The billing data would accompany the payments and vice versa through the complete end-to-end process in order to enable a combined automatic payment, invoicing and accounting process both by the payer and the payee.

212

H. Leinonen

Clearing in decentralised infrastructures The network environment is decentralised and telecommunication builds bridges between the different independent but interoperable entities in the network. As regards payments, these bridges must transfer payment messages as well as interbank settlements. This is the main difference as compared with other messaging systems. In a decentralised network-based environment the settlement method should also be decentralised in order to be efficient. In this environment, the interbank settlement method will need to entail immediate finality between all the different participating service-providing institutions (mainly banks). The interbank settlement process itself will have to be a well-integrated part of the payment process, with end-to-end control from sending to receiving bank. In order to support rapid payment transfers, the settlement method must also support real-time processing. An efficient settlement system also supports continuous reconciliation for immediate error detection. It is also important that most of the security and control features be built into the system, to enable immediate reaction.

The e-settlement proposal The new e-settlement model introduces new automated and electronic possibilities for interbank settlements in a network-based decentralised payment infrastructure.13 E-settlement is a proposed new settlement method for the next generation of payment systems, which can nonetheless be employed already in current payment systems. It has the potential to generate large benefits for the wider economy. In this section we first discuss how e-settlement works as well as its needs in terms of an interbank network, central bank liquidity and security features. We then discuss the benefits it affords the economy. The general layout The e-settlement solution should be seen as part of the future payment infrastructure that supports an increase in e-commerce via the internet, real-time security and money market deals and transfers, mobile payments (currently GSM-based but soon UMTS-based) and cross-border payments. The payment world (for all kinds of payment) will undergo change, as will all other messaging systems, from slower-paced batch processing to immediate real-time service, integrated directly with user systems in a global network community. E-settlement provides a solution that can be integrated into current systems, using part of the existing infrastructure, and hence facilitates a gradual change from current structures to new e-based structures. The fundamental idea of e-settlement is attachment of a digital e-settlement stamp to the current payment messages, as shown in Figure 13.4 The e-settlement stamp is added to the payment message and serves to trans-

E-settlement

213

Beneficiary’s bank address Payment envelope Payment details ... Central bank e-settlement stamp End of payment

Figure 13.4 The digital e-settlement stamp is part of the payment message.

Payment  Interbank network

Bank 1

Bank 2

Digital CB cover

Figure 13.5 The digital encrypted stamp with central bank cover will follow the payment message through the network.

fer central bank money from payer’s bank to beneficiary’s bank. Final settlement is part of the payment message, in the form of electronic central bank money for interbank settlement purposes. The electronic stamp will accompany the payment on its route through the interbank payment network to the receiving bank as described in Figure 13.5. The electronic stamp can be seen as a modern version of a central bank draft. It is the cover in central bank money of the payment(s) it accompanies. The stamp is protected by very strong and modern cryptographic technology (e.g. Public Key Infrastructure: PKI14). These stamps are produced and decoded by e-settlement modules situated close to banks’ payment systems, as shown in Figure 13.6. The e-settlement modules are tamper-resistant devices provided by central banks to each bank. These are closely integrated with banks’ payment systems, e.g. directly integrated with the SWIFTnet access platform (CBT). This makes settlement transfers a highly automated part of payment processing. Integrating the new settlement process will be quite straightforward, given that it will be done on the access platform (e.g. SWIFT CBT) level. In traditional RTGS systems, banks’ settlement accounts are located in the centralised RTGS system. In the e-settlement system, each bank’s settlement account is distributed to the bank’s own processing site in a central bank-controlled e-settlement module. This module should be regarded as a completely automated central bank branch, serving one customer with one account. Each bank has access to its own

214

H. Leinonen

Sending bank

Receiving bank Network account platform

Bank payment system E-settlement module

Payment network

Network account platform

Bank payment E-settlement system module

Customer account

Customer account

STP, network based, end-to-end, all through real-time process

Figure 13.6 E-settlement stamps are produced by e-settlement modules, which are closely integrated with banks’ payment systems.

account, as before, but is much more closely integrated in a more automated and efficient way. The distribution of central bank money in electronic format to banks’ payment platforms is the essential feature of the e-settlement approach. The distributing e-settlement modules need to be highly secure and to meet at least the same security standards as do traditional RTGS systems. The system should also be generally open and independent, to support the various payment networks used by banks. E-settlement is especially suitable for large volumes as there are no centralised bottlenecks and can therefore be used both for largevalue and low-value payments. Interbank network A dedicated interbank network (Figure 13.7) is needed to link together all participating banks and the central bank, for the purpose of processing payments. In a network-based payment system, the most essential element is the interbank communication network. All banks can address each other directly and send payments to each other without a centralised processing and routing site. This is the essential new paradigm introduced by internet communications (TCP/IP-networks). All participants can operate independently; they need only enough networking capacity to meet their own needs. System administration is needed only for administration purposes, but, for example, not for payment processing. The new SWIFTnet network, introduced by SWIFT, is one that can support direct communications between all participants. There are also national dedicated payment networks with the same capability, e.g. the interbank network Pankkiverkko2 in Finland.

E-settlement

Bank 1

Bank 2

Bank 3

215

Bank 4

Interbank network

Central bank

System administration site

Figure 13.7 A dedicated interbank network connects all banks and the central bank with each other for payment processing.

Immediate liquidity For settlement purposes, banks need liquidity. Liquidity is transferred by the central bank to the system (e-settlement modules) at the start of the day. It can be increased during the day by the central bank via liquidity transfers or payments to the banks. At the end of the day, liquidity is transferred back to the central bank. The liquidity in the settlement modules is thus composed of positive balances of central bank money, originally in the traditional form of reserve deposits, intraday credits, etc., but transferred from the centralised system in the morning to distributed e-settlement modules to be employed during the day in the e-settlement system. In the evening the liquidity will be transferred back to the centralised accounts for overnight bookings. When, at some point in the future, interbank payment systems provide a 24-hour/seven-days-a-week service, as has been predicted, the e-settlement modules could start to run continuously by only reporting the balance at the turn of accounting day. In a true real-time environment, there is generally little scope for the various types of liquidity saving features, based on delaying or queuing of payments. Customers are waiting for direct confirmation of their payments. A bank that is often obliged to inform its customers that payments are queued (that is, waiting for liquidity) will lose customers. In the real-time environment, customers expect direct delivery in the same way as in e-mailing. Still, the e-settlement module could contain basic queuing facilities for situations in which the available liquidity is not sufficient or customers are willing

216

H. Leinonen

to accept delays. These would be decentralised queues, designed for different levels of complexity. Bilateral or multilateral netting could be accomplished in the distributed e-settlement system through bilateral or multilateral netting requests to check whether there are transactions also queued at the other end. Different types of netting and advanced liquidity saving features would complicate the system. It is advisable to keep the basic system very simple. High security and availability The system’s security features must be carefully designed. The settlement balance and all security keys need to be in tamper-resistant environments and all the encryption algorithms must be highly reliable. There should be no possibility of system intrusion, and any type of ‘hacking’ should be immediately detectable. The system will be closed, with settlement money circulating among a limited number of trustworthy users. The system will include automated reconciliation at end-of-day, from time to time during the day and in connection with each transaction. In a network-based environment, all parts – centralised and distributed – must be well secured. A digital/electronic version of the four eyes principle has to be implemented. Two different controlling programs and two different encrypted and tamperresistant storage devices protect the money balances. The separate controlling functions are monitoring each other so that there are no individual IT developers or development teams that have all the necessary programming codes in their possession. In case of intrusion attempts, all critical information will be destroyed at that site and there will be a security alert to the system administrator(s). High availability must also be ensured in the distributed system. In a distributed system, a malfunction will generally affect only one participant at a time and only those payments to and from that particular participant. In order for the participant to re-establish normal operations quickly, there should be back-ups and mirrored devices for all critical components. Redundant information in the e-stamps gives the possibility to parse the information of completely destroyed IT-sites. It also supports a direct switch of functionalities and services to working parts of the network. This makes distributed systems more robust than traditional centralised systems. E-settlement benefits The main benefit of e-settlement is that it enables redesign of the whole payment system process in an efficient way, using new network possibilities. It thereby creates the next generation of payment systems infrastructures and makes the settlement process more efficient. Payment systems will change considerably in the near future due to modern technology and it would be an advantage to modernise the settlement conventions at the same time. The decentralised network-based model facilitates direct real-time communication. General standards will result in a STP process. The best example is the

E-settlement

217

current e-mail system and its standards. These are applied worldwide and give a really low cost, efficient and rapid communication system. The e-mail costs are so low that they are considered part of general overheads; nobody bothers to report separately on them. The same type of infrastructure with additional security elements and the settlement function could be developed for the payment sector. The cost-advantage of the e-settlement system is in the low processing costs of adding the e-settlement stamp that enables instant final settlement in central bank money. The extra processing cost of adding the e-settlement stamp will be practically nil. It will be an integral part of the payment process itself. Banks need only invest in low cost equipment. The very low transaction costs of e-settlement will also enable banks to transfer payment flows from centralised processing centre systems to more efficient decentralised network-based communications. The bottlenecks created by centralised resources will disappear and even the dependence on critical centralised resources will be dramatically reduced. E-settlement could offer a solution for integrating the euro-area payment systems, and a multi-currency version could serve an even larger area. In order to achieve large-scale benefits via the e-settlement model, the number of participating banks and the payment flows must be sufficiently large. The e-settlement approach will also reduce settlement risk, because all settlements are undertaken in central bank money with immediate finality. The reconciliation and control functions in the system will also reduce the possibility of errors or at least speed up the process of finding them. In general decentralised systems are more robust than centralised systems.

Delaying and promoting development factors in payment systems Competition in the market should normally result in optimal utilisation of new technology. However, a considerable implementation lag is often found in the payment industry. When one compares payment developments in different countries, they go in different directions and at a different pace, despite the same technology being available to everyone. There currently seem to be much stronger forces for maintaining the current payment conventions than for enhancing them. This section examines why this might be the case. In particular, it argues that this is the result of market failures in the payments industry. Negative economic provider interest A payment is a complementary product, that is, payments are completely dependent on the underlying economic transactions. Any improvement in payment service will affect the total number of payments very little or not at all. It is a zero-sum game, where the volumes are just shifted from one payment instrument to another. Payment system costs are mostly fixed.

218

H. Leinonen

This results in an overall market situation, where the current service providers have very little interest to change current payment services, because the economic benefits are larger, when the current services can be used for a longer time. The margins are high in a system with a high proportion of sunk costs. Any investment in new technology will ‘cannibalise’ old investments and services and the service provider has to maintain several parallel technologies for the changeover period, which tends to be quite long in payment services. Investments in new technology have their own risks and these risks can be decreased by waiting for times of more certainty. The general structure of the payment systems, as will be explained in the next sections, are such that an individual service provider experiences difficulties in changing the technology employed due to interbank dependencies. It is seldom beneficial to any service provider to start the change from a competitive point of view, as the other providers need to be incorporated. The result would be increased investment costs for everyone and sinking margins, as any new technology will not only have lower total costs, but probably also lower customer tariffs. A typical example would be moving to more rapid interbank transfers, which would result in costs of change and decreasing float-income from the banks’ perspective. Network externalities Network externalities are strong in payment services. This is especially true for electronic services. Bringing new instruments and standards to the market is difficult as the payment infrastructure in itself is huge and complex. The chip card EMV15 undertaking by the credit card industry is a good example. All of the hundreds of millions of payment cards and millions of payment terminals throughout the entire world need to be equipped with chip card technology in order to shift from the prevailing magnetic strip technology to the improved chip card technology. The new technology will not work if there are not enough users both among payers and payees that are willing to use it and service providers that are willing to implement the technology in the first place. Every new technological design has to fight for its existence and overcome the initial chicken-and-egg situation. It is also in the interest of service providers of the old technology to increase these initial barriers as much as possible as will be described in the next section. The old payment instruments also have an advantage over the new, because in investment comparisons past sunk costs are often written off. The costs of current instruments are, in alternative investment calculations, based on variable costs only and sometimes even on marginal costs only, while the costs for new instruments are based on average costs including investments and based on smaller initial volumes, because of the probable long implementation period. Network externalities can be abused by powerful service providers in dominant positions. If a payment instrument is popular among payers it puts pressures on payees to accept it. For example, the popularity of credit card use by tourists forces tourist service providers to accept credit cards and thereby gives credit

E-settlement

219

card service providers a chance to debit high merchant fees. The larger the card base, the more powerful negotiation position the card service provider will have. Structural barriers Payment service provision is often restricted via regulation and reserved for credit institutions out of stability concerns. However, at the same time, this raises the barrier to new entrants. Payment services are provided to customers via a layered structure. The customers are served by the service providers having direct customer relationships. There are then specialised infrastructure providers serving this first layer of service providers. The layered structure can be deeper with large banks providing correspondent banking service for smaller banks. In most countries the layered structure has resulted in monopolistic ACH structures for interbank payment processing resulting in one route and convention for interbank payment transfers, which is decided upon by the ACH. As the larger players often have a larger ownership and more user power in the ACH, the decisions tend to promote the larger players. The service area of an ACH can also vary and the more customer-related services the ACH performs the less room there is for genuine competition in the market. The new entrants often face high member fees or other membership requirements, which are difficult to be met in the start up phase.16 Network externalities and layered structures require competitors to cooperate on new standards and service conventions.17 New products or services will seldom be successful on the market if the large players do not agree upon common standards and processing conventions. Merely by delaying cooperation the current status quo can be maintained. New payment conventions will generally shift the current accepted balance of benefits and costs for example; more rapid processing will decrease the float income more for some system participants than for others. A ‘don’t rock the boat’ policy is often the result of this accepted balance among competitors. Psychological pricing structures The pricing of payment services is, to a large extent, non-transparent and often contains elements of cross-subsidisation and hidden factors. Traditionally in many countries making payments has been a non-priced service and it is only during recent years that different kinds of visible payment tariffs have emerged. Cash, especially is often viewed by the payers as a free payment method, because the costs are embedded in the merchants’ product prices. This is partly due to the legal tender status of cash and past dominant position when it was the most efficient payment method on the market. Central banks and banks have often subsidised cash processing, which results in lower costs for banks and merchants. Banks often cross-subsidise payment services out of current account interest margins. Compared with other accounts the interest rate on payment accounts is

220

H. Leinonen

lower and often even zero. In addition to this, banks use non-transparent value dates for paying interest on customers’ accounts. Cross-subsidisation and non-transparent pricing of old payment instruments, results in a difficult psychological pricing barrier for more efficient new payment methods. The improved cost efficiency is not visible to the end-user unless negative prices are introduced. For end-users who make the final payment instrument choice there is no premium in selecting a low cost service if all service forms have the same – mostly zero – external prices. Customers then make the choice based on convenience and out of habit, which generally favours traditional payment instruments. Sometimes this zero-pricing convention has been required by authority regulation and the extensive cheque volumes of the United States and France is mainly a result of their zero-price regulations. Cross-subsidisation implies that there is not enough competition among service providers, i.e. banks in a regulated market. There would not be a high enough margin in deposit and credit services to support payments in a competitive market. Having to cross-subsidise free payment services would make it difficult for an individual service provider to start pricing payment services. In a market where cheques are free it is generally difficult for a single service provider to start pricing cheques to its customers as this might result in customers moving their deposits to another bank. The payments industry has also introduced different kinds of common interchange fees, which are levied on interbank transactions but generally also on intrabank transactions. Common interchange fees make it difficult for individual service providers to undercut the common price level. There have been several studies18 supporting the use of interchange fees and splitting up the card payments market in particular among payers and payees/merchants. This has been especially in the interest of international card payment schemes. However, payment systems are just a transportation mechanism. The market functions better when direct transparent pricing is applied and when the end-user can select the most efficient transportation method. It is quite clear when one makes a comparison with other transportation systems with transparent pricing, that no support can be found for non-transparent pricing or complex sender/receiver pricing schemes to ensure efficient development of transportation systems. The merchant fees result in a form of cross-subsidising within the merchants, i.e. the price of goods includes an average payment fee and the payer cannot select the more efficient based on transparent prices. The MIF part of merchant tariffs is mostly non-negotiable and there is not a clear service-price relationship. The card transactions fees and especially merchant fees can thereby be higher than those that would transpire in a transparent and competitive market. There is also a psychological barrier for end-users to move from what seems to be a free service to an explicitly priced service. If merchants were to add a payment service fee on each purchase, customers would probably react negatively, stating that this is an extra add-on. They would not be able to observe any simultaneous decrease in the merchant prices as these could only be observed

E-settlement

221

during a short change-over period. Customers seem to find it easier to accept a higher hidden cost than to be reminded with every purchase about the payment service fee and thereby the need to make an explicit choice.

Authority actions Authorities face the problem of selecting the correct solution to the market failure situation. There are different tools available relating to research and studies, moral suasion, recommendations, subsidies, initial investments, operational involvement and regulations.19 Studies, research, moral suasion and recommendations are softer measures in the push for the developments desired. Via these, authorities act in the role of catalyst, trying to help the market solve market failures. By way of subsidies, investments and operational involvement authorities can change the economic situation of the market. These can be helpful in introducing new payment instruments and methods, when the chicken-and-egg-situation restrains private developments. However, returning to normal market practices and full cost recovery should be considered as soon as the initial barriers are overcome, in order to avoid preventing further development via artificial support of a given payment method. Regulatory measures are often effective but also risky, since they can result in developments in the wrong direction. Regulating for development is more difficult than for bringing order to an existing situation. Regulations need also to be reviewed over time in order to keep up with developments. Regulations can be used in three distinctly different ways: to correct market incentive structures, to impose limitations and to enforce technical standards. As far as efficiency developments are concerned, the most important path seems to be correcting market incentives. Without proper incentives private service providers will not have any interest in developing systems. Payment system developments would probably be speedier if there were clearer limitations on barriers to entry as well as rules for monopolistic activities. Implementing detailed technical standards has proven to be a lengthy process in all private systems. In some cases it may be faster to use a regulatory approach, but it will require deep technical knowledge in the regulatory process. Market failure: the cause for authority actions Authority actions can be necessary to overcome market failures. There can be market failures from stability and efficiency points of view and the problem in both cases is to assess the size of the problem and to find a satisfactory remedy. Currently the payment system concerns are mostly on the efficiency side. There are studies20 showing large gains to the electrification of payment systems. However, these seem to be too intangible for many service users. One problem seems to be that the gains consist of a large number of small benefits and, thus, they are not given the attention they deserve. A gain of 10 cents per transaction,

222

H. Leinonen

for example, will add up at total economy level to considerable benefits, but for a payer with 10–20 transactions per month the extra cost may seem affordable. The gains are often viewed separately for the different stakeholders and for different improvement proposals without an overall picture and without balancing the gains and costs over all stakeholders in the process. The developments are also dependent on each other and on advance payment services, e.g. e-billing will take off only when e-payments/giros are working and e-commerce payments need rapid e-payment services with different payment/delivery guarantee models. Market failure, due to the negative development incentives, builds up over the years. Generally, technological developments are continuous and when one business sector does not exploit the available possibilities the gap between the provided and the potential service level will gradually increase (see Figure 13.8). In a market failure situation the systems operate far from the optimal level. The wider the gap, the greater the reason for authority involvement and the larger the potential public benefits of authority actions. The technological developments have been tremendous during the last decade. The current inefficiency gap in payment systems varies from country to country, but in most cases customers experience it concretely as • • • • •

low processing speed; high processing costs; insufficient data content; poor system integration and therefore low straight-through-processing level; inefficient and non-standardised customer interfaces.

In terms of both the processing speed and processing costs PayPal and e-mail services in general show the real-time benchmark that is attainable. Using internet-technology transactions can be processed immediately and at very low costs. There are many more e-mails sent than payments and nobody sees this as a cost issue although most e-mails have large enclosures (more than 10 megabytes), which could contain the data of more than 1,000 payments. Costs

Current level Efficiency and stability gap Possible level Time

Figure 13.8 The increasing market failure gap.

E-settlement

223

The current data content of payments has been squeezed down due to old communication and data storage restrictions, which have disappeared completely. Payments could include a lot of beneficial data for the customers. The most important data would be invoicing data connected to the payment. This would reduce reconciling needs considerably and customers would get an integrated e-archive for payments and invoices. Just establishing a good e-invoicing service will save about C10–30 per invoice and in the European Union there are more than 60 billion invoices processed every year. The potential for savings are therefore considerable. In order to integrate systems and promote straight-through-processing all transactions and accounts need clear addressing systems. These should preferably operate on an international level. Good benchmarks can be found in other industries. For airway cargo the airway bill number (AWB) identifies all delivery notes, invoices, etc., referring to a given cargo. All airfields and flights have clear international identifiers. Even the normal surface parcel mail uses international codes and each parcel can be traced online. Payments would need an international bank account scheme, a payment/transaction identification code as well as payer’s and payee’s references in order to provide good integration coding. As customers’ payment processing is computerized and most private customers are internet-based, there is a need for common international e-banking and e-payment standards. E-mail is used for sending all kinds of information today. Very few people remember the old days, with different incompatible e-mail systems. Common payment standards would make payment systems interoperable and make it possible for software houses to develop integrated software solutions for customers. This would automate the processing and reduce the costs both for banks and customers. The basic question for the authorities is, therefore, has the market failure gap become so large that it is necessary for authority actions in order to reduce it for the benefit of the whole economy and, if so, what would be the most suitable actions. Possible actions The development of payment systems can be made more rapid by reducing the impact of the negative factors described in the previous section. However, doing this will generally require some external push, because the current market is in a market failure trap. The push could come from the big payment service users, i.e. big multinational companies as their internal costs would decrease considerably with more efficient payment methods. However, even for these big companies it seems to be true that the user community for payments is so scattered and fragmented that a strong European or international push is difficult to muster. The main focus in these companies is their own production and development facilities and not that of the banking or other sectors, where they will not be able to generate a competitive edge.

224

H. Leinonen

The push could also come from non-bank competitors like telcos, merchant chains or special service providers like PayPal. The main business focuses of these companies are in other areas. Although there have been several trials in almost all countries, the network barriers (chicken-and-egg situation) have been at least until now too high to cross. The trials have generally been with specialised payment instruments for given environments and usages lacking the possibilities for establishing general purpose payment schemes attracting large volumes. This leaves the authority push as the most viable solution. The authority push can be delivered using three general methods (separately or in combination) • • •

opening up payment systems to competition; direct regulation; operational involvement.

It is important that this push is delivered only on a temporary basis in order to get the industry over the current development barrier. After this has been achieved the authority involvement, for example, through direct regulation, could be withdrawn. The probability of the new and more efficient payment processes having a retarding effect on the old ones is very small. Instead there is a risk that old regulation could hinder future necessary developments. There are examples of this from the past especially in cheque processing. Opening up payment systems to competition Payment systems seem to contain a built-in tendency towards having only limited competition. It works at four separate levels • • • •

customers are tied to service providers; banking regulations limit service provision possibilities; interbank systems contain access limitations; and interbank systems tend to be monopolies or oligopolies

This results in underinvestment, high centralised costs, bundling of services and cross-subsidising. Barriers to innovation favour large established providers and enable them to establish barriers for new and small providers (e.g. a burdensome rule system and/or participant criteria). In order to decrease the limitations of competition special actions are needed to promote competition and decrease the possible abuse of market powers. Customers get tied to service providers via fixed account number and addressing conventions. When the customer changes service providers, the need to communicate the new account numbers to all payment partners requires extra effort and cost. Service providers try to increase the reluctance to change provider by cross-subsidised pricing. The marginal tariffs are low and it is difficult for a customer to change provider for an individual banking service. With

E-settlement

225

complex and non-transparent pricing conventions, it is difficult for customers to make effective price comparisons. Electronic interfaces require customer training and special hardware and software features. It means additional efforts for customers when changing providers and keeping parallel providers for an interim period. Customer ties to service providers can be reduced by requiring transferable account number conventions, by stipulating transparent and easy-to-compare tariff structures and by demanding open and general minimum technical standards. Banking regulations generally limit deposit taking to banks. As most payments are made out of different kinds of deposit accounts with or without overdraft facilities, payment service provision will, at least de facto, be limited to banks. Banking regulations have been issued in order to secure customer deposits. This protection of deposits can be maintained and competition increased if funds may be transferred easily from bank accounts to special payment accounts maintained by specialised payment service providers. This is, in fact, the way in which many innovative payment schemes work, but it also brings added complexity to the payment system. A strong enough risk of competition, however, may bring sufficient development incentives to banks as service providers that more efficient payment methods will be implemented earlier. Providing payment services requires access to interbank clearing and settlement systems, because the customers want to reach payment receivers that also have accounts with other service providers. The access criteria to interbank clearing and settlement can show a large variety of entrance barriers including high entrance fees, high fixed member fees, special club-type extra criteria, geographical requirements and complicated technical standards. Interbank settlement services are often provided by central banks and access to central bank money accounts may include additional access criteria. Forcing openness and fair access criteria onto these services increases the opportunities for competition. The interchange fees and surcharge prohibitions result in extensive crosssubsidising and build barriers for new entrants and solutions. Abolishing interchange fees within the infrastructures would open up the competition considerably and would make it possible for the users to compare the true costs of different payment alternatives. In most cases interbank payment infrastructures operate under almost monopolistic conditions. There is generally only one ACH per country although sometimes the load can be divided among specialised ACHs, e.g. one for giros and direct debits and another for interbank card payments. There are, de facto, only two international interbank card processing networks, which are very much closed clubs. Interbank settlement is a part of the clearing process, which is often done in central bank money on their RTGS systems. The access rules to central banks’ RTGS system can therefore become a barrier to competition.

226

H. Leinonen

Direct regulation Regulation can be used in different ways. Opening systems to competition will generally require regulatory measures, when there are high barriers to competition in place. Direct and technical regulation can be used to solve different kinds of technical and standardisation issues e.g. establishing clear account number practices and security requirements. It is often difficult to get standards implemented by all participants within a given timeframe and in a harmonised manner. The regulatory approach can solve these kinds of issues. As a comparison, road traffic is a typical area of communication in which a large number of regulations can be found. In a rapidly developing area like payment systems the regulatory measures have to be very flexible and updated when necessary due to different developments. Regulations in the area of standards and competition issues would probably be efficient in speeding up development. Operational involvement In a pier to pier type of network the operational involvement of authorities will be rather small. The network services will be provided by the normal telecommunication providers. In order to ensure stability, authorities would still need to provide and supervise licensees for infrastructural participants. However, there is also an opportunity for central banks to become the high-level infrastructure administrator by maintaining the master register for infrastructure participants as part of the interbank settlement services. Central banks could still have a significant role in the settlement process by providing central bank money based e-settlement solutions. In order to promote efficiency, central bank services also need to move into the internet age and support decentralised processing. Inefficient central bank service could otherwise become a barrier for general service development. Overall efficiency can only be reached in payment processing when all infrastructure parties are employing technology efficiently.

Conclusion When will e-settlement become a reality? New technology always eventually pushes obsolete, inefficient alternatives out of the market. There are three basic alternative scenarios for the development of network and e-settlement based development. First, the current market players may start to use modern technology. This has mostly been the case for example in mobile telephone services. The second alternative is that new entrants grab the market. This happened when mechanical calculators were replaced by electronic calculators. The third alternative is strict authority regulation. This has been common in the past regarding payment system developments both as regards cash and cheques, but less during recent years. Based on current developments in the European market a mix of all three scenarios seems to be the most probable.

E-settlement

227

As far as the timing of development is concerned, history shows a larger number of completely wrong prognoses. Technical progress has become faster, especially when large economic benefits are at stake. E-payments, e-invoicing and e-commerce imply large savings for companies and consumers. The pressure will grow over time and non-bank service providers will have a growing interest in the market, if banks do not supply efficient services. Authorities have tried for several years to speed up developments and their efforts have lately increased. Both the Eurosystem and European Commission have become active in Europe. The time-to-market for new products is generally only six months in other industries. It is longer in the payment industry, but it is clearly decreasing. The technical development required for implementing a network-based structure including e-settlement is estimated to be about two to three years in an efficiently-run project based on utilising current network services such as SWIFTnet. This could be rather soon in the payment system development context, in which the initial decision-making process seems to be the main obstacle. Sending payments can be made as easy as sending e-mails and at the same level of cost. It is time to move from the current legacy systems to the new dominant technology in payment processing.

Notes 1 The views expressed are those of the author and do not necessarily reflect those of the Bank of Finland. 2 For examples from Finland and some of the new EU member countries, which have built new banking systems based on real-time processing, see ECB Blue Book and www.pankkiyhdistys.fi. 3 See www.paypal.com. 4 The ECB Blue Book gives details on European systems. For example in Belgium, Finland and the Netherlands real-time or near real-time interbank express transfers already exist. In the Netherlands and Norway normal payments are also cleared the same day. In most other countries clearing and settling real-time payments takes at least one day but often more. According to EU Commission studies on cross-border payments, the average delivery time was 3.3 days in 2001. 5 The self-service automation level reached in Nordic and Baltic countries was, in 2002, already over 80 per cent and in the most efficient countries about 95 per cent. Customers prefer to send payments over the internet using e-banking instead of using branch services (see ECB Blue Book statistics). The growth has been very strong since the end of 1990s when the internet became generally popular. 6 Modern technology has made it possible to administrate and handle large network structures. SWIFTnet is a good example of this. The message-based network connecting airline companies, travel agencies and airports for bookings, ticketing and checkin is another good example of an interoperable distributed network with numerous participants. 7 Information on IBAN can be found on the web-page of the European Committee for Banking Standards (ECBS): www.ecbs.org. 8 However, there are also competing options, e.g. the international card number used by Visa, Mastercard, etc., the e-mail addresses used by PayPal and phone numbers used by mobile payment systems. Competing account numbers make the structures more complex.

228

H. Leinonen

9 For IBAN-information, see the ECBS web-page (www.ecbs.org) and the Thomson Financial Publishing web-page (www.tfp.com/payment.shtml). 10 See Leinonen (2000) for details. 11 See, for example, www.swift.com, www.fba.fi, www.ecbs.org, www.twiststandards. com, www.rosettanet.org, www.x12.org and www.ifxforum.org. 12 See info on XML www.w3schools.com/xml/default.asp and www.w3schools.com/ schema/default.asp. 13 See Leinonen et al. (2002). 14 See Schneier (1996). 15 Visa and MasterCard have developed the chip card standards for credit cards, which have caught the interest of other credit card companies. See www.mastercard.com and www.international.visa.com. 16 Kemppainen (2003) and McAndrews (1995, 1997). 17 McAndrews (1995, 1997) and Rochet and Tirole (2002). 18 Wright (2004), Schmalensee (2002) and Rochet and Tirole (2003). 19 Bank for International Settlements (2001, 2003). 20 Humphrey et al. (2001) and Guiborg and Segendorf (2002).

References Bank for International Settlements (2001) Core principles for systemically important payment systems, Committee on Payment and Settlement Systems Publication No. 43. Bank for International Settlements (2003) Policy issues for central banks in retail payments, Committee on Payment and Settlement Systems Publication No. 52. European Central Bank (1999) Improving cross-border retail payment services – The Eurosystem’s view, Frankfurt am Main: European Central Bank. Online, available at: www.ecb.int/pub/pdf/other/retailpsen.pdf (accessed 22 January 2007). European Central Bank (2000) Improving cross-border retail payment services – Progress report, Frankfurt am Main: European Central Bank. Online, available at: www.ecb.int/pub/pdf/other/retailps2000reporten.pdf (accessed 22 January 2007). European Central Bank (2001) Towards an integrated infrastructure for credit transfers in euro, Frankfurt am Main: European Central Bank. Online, available at: www.ecb.int/pub/pdf/other/credtransfeuroecofinen.pdf (accessed 22 January 2007). European Central Bank (2003) Towards a Single Euro Payments Area – Second progress report, Frankfurt am Main: European Central Bank. Online, available at: www. ecb.int/pub/pdf/other/singleeuropaymentsarea200306en.pdf (accessed 22 January 2007). European Central Bank (2004) Towards a Single Euro Payments Area – Third progress report, Frankfurt am Main: European Central Bank. Online, available at: www.ecb.int/pub/pdf/other/singleeuropaymentsarea200412en.pdf (accessed 22 January 2007). Guibourg, G. and Segendorf, B. (2002) ‘Do prices reflect costs? A study of the price and cost structure of retail payment services in the Swedish banking sector 2002’, Sveriges Riksbank Working Paper No. 172. Humphrey, D., Mosche, K. and Vale, B. (2001) ‘Realizing the gains from electronic payments: costs, pricing and payment choice’, Journal of Money, Credit and Banking, 33: 216–34. Kemppainen, K. (2003) ‘Competition and regulation in European retail payment systems’, Bank of Finland Discussion Paper No. 16/2003.

E-settlement

229

Leinonen, H. (2000) ‘Re-engineering payment systems for the e-world’, Bank of Finland Discussion Paper No. 17/2000. Leinonen, H., Lumiala, V.-M. and Sarlin, R. ‘Settlement in modern network-based payment infrastructures: description and prototype of the e-settlement model’, Bank of Finland Discussion Paper No. 23/2002. McAndrews, J. (1995) ‘Antitrust issues in payment systems: bottlenecks, access and essential facilities’, Federal Reserve Bank of Philadelphia Working Paper No. 96–19. McAndrews, J. (1997) ‘Network issues and payment systems’, Federal Reserve Bank of Philadelphia Business Review. McCreevy, C. ‘The wind has changed’, speech at EUROFI-Banking and Finance in Europe Annual Conference 2005. Rochet, J.-C. and Tirole, J. (2002) ‘Competition among competitors: the economics of payment card associations’, Rand Journal of Economics, 33: 549–70. Rochet, J.-C. and Tirole, J. (2003) ‘Platform competition in two-sided markets,’ Journal of the European Economic Association, 1: 990–1029. Schmalensee, R. (2002) ‘Payment systems and interchange fees’, Journal of Industrial Economics, 50: 103–22. Schneier, B. (1996) Applied Cryptography Second Edition: Protocols, Algorithms, and Source Code in C, Mississauga, Ontario: John Wiley & Sons. Tumpel-Gugerell, G. (2004) ‘Time to act: clear objectives and a convincing roadmap for the Single Euro Payments Area’, speech at the Coordination Committee of the European Payments Council. Wright, J. (2004) ‘The determinant of optimal interchange fees in payment systems’, Journal of Industrial Economics, 52: 1–26.

14 Real-time liquidity management in a globally-connected market Richard Pattinson

El

ec tr T Al onic rad i t O er n co tion ve at m a r t ive mu l e he t n xc co rad ica han un ing tio ge te s ns s r t ys ne ra te tw ST di m o P ng s r k Pr s ov id er s

This chapter discusses risks in global financial system infrastructures. It argues that the creation of the Continuous Linked Settlement (CLS) bank service, and other initiatives to eliminate settlement risk, have resulted in an increasingly globally-connected world. This means that liquidity/operational events in systems can rapidly spread to other systems around the world causing major problems. As a result, there is a need for a global liquidity and communications bridge; this issue should be actively pursued by all market participants, public and private sectors and proposals agreed as soon as possible. The plumbing system in the major global markets is shown in Figure 14.1.1 It ks or s tw ge ne s an s em ch ion st ng ex at sy di al ic g tra n n on u di r iti m tra nte u ad om e Tr c c tiv co ni r na the s tro lte er er ec A Ov id ov Pr P ST

CDiCG

El

London Clearing House

EuroCCP

Monte Titoli OeKB CREST

Clearnet

SITEME Interbolsa Clearstream

CHAPS BI REL ARTIS SPGT RTGSPLUS LIPS-Gross

Euroclear CCPs

BOGS/ CSD HERMES

ELLIPS TBF/ TOP

SIS

APK

BoF-RTGS DEBES

CSDs

SIC NSCC

DTC

VP

SLBE

TARGET

SCLV

CLS Bank

FedWire

E/ K RIX RTGS

CSD Links

OM-VPC

LVTS SWIFT-RTGS

CDS

MEPS

CLS

2nd wave CLS currency

VPS RTGS

RITS

CDP

BoJ-Net

s er id ov ks or Pr s tw P ge ne s ST an s em ch ion st ng ex at sy di al nic ng tra on u di r iti m tra nte u ad om e Tr c c tiv e co ni r na th tro lte er ec A Ov

El

Austraclear

CCAS

JSCC

ST P Pr ov tr T id er Al onic rad s te c itio O rn o n ve at m a r t ive mu l e he t n xc h r i c a co d a an un ing tio ge te s ns s r t ys ne ra te tw di m o ng s r k s

HKSCC

ec

JasDec

El

Chess

Figure 14.1 Links between world market infrastructure (illustrative only).

Real-time liquidity management 231 is a complex web of linked real-time settlement systems crossing national boundaries and many time zones. The payments and settlements world is no longer a set of domestic silos standing alone, by and large, immunised from external system contagion. The world is now hard-wired; it is connected in a live environment bringing together payment and securities settlement systems from Sydney to Seattle. This transformation from domestic silo to global connectivity has virtually happened by stealth and is a product of the last three years since the launch of the CLS bank service. This is not an implied criticism of CLS, rather recognition that CLS was created by a large number of leading foreign exchange (FX) trading banks with active encouragement from the central bank/regulatory community to eliminate the settlement risk inherent in foreign exchange transactions, known by the term ‘Herstadt Risk’. CLS bank service is the first global infrastructure to be simultaneously connected to the systems of fifteen currencies with a potential for more currencies to be added to the service over the coming years. Domestic large-value payment systems are commonly ‘connected’ to domestic central securities depositories (CSDs) to facilitate real-time delivery versus payment (DvP) in the securities market, again as a methodology for reducing settlement risk. In turn a number of CSDs are connected to central counterparties all of which are illustrated in the concentric rings in Figure 14.1. Theoretically it is possible for an operational incident in a central counterparty infrastructure on one side of the world to have an impact on a large-value payment system on the other side of the world in a matter of minutes through this connected/network environment. In the rather comfortable legacy world of domestic silos any operational incident would be handled by a limited number of people at the centre of those systems; infrastructure provider, settlement bank, central bank, etc. Participants in the impacted system would be relatively relaxed and assume that the operational difficulty would be fixed. Business would continue pretty much as normal. In the new world of increasingly connected cross-border systems, with a growing number of time-specific obligations to meet, the story is very different. It is difficult to know where the problem originated, who is responsible for fixing it, who should be contacted about it and so on. A sustained lack of information can easily lead to unilateral action being taken by market participants in order to protect their liquidity. Is this just an operational issue or is there a real credit issue involved? An information vacuum is a dangerous thing and unilateral action by a large number of participants will lead to sub-optimal management of the event. Payment system liquidity has to be available in the right place, at the right time and in the right currency. Historically the value date of a particular transaction was the key fact; for a growing number of payments the specific time on a particular value date is the key fact. The payments and settlements world is moving from a value date specific environment to a time sensitive one. Globally-connected markets supported by globally-connected infrastructure means that operational failures are much more visible and much more immediate. Depending on where the failure occurs there will be significant reputational, regulatory and financial risks.

232

R. Pattinson

These risks can be illustrated by the analogy of a balloon. Let us assume that we have a risk balloon which in the payments and settlements world contains three risks; settlement (credit) risk, liquidity risk and operational risk. Industry has been very successful at addressing settlement risk especially in the FX and securities markets with the introduction of real-time payment versus payment for eligible FX transactions via CLS and real-time DvP in some securities markets. In effect the end of the risk balloon containing settlement risk has been squeezed very hard but unfortunately the balloon has not been deflated, rather the air representing the two remaining risks, liquidity and operational, has been pushed up the other end. The result is that the balloon containing the two remaining risks is much more susceptible to exploding. We have, unfortunately, addressed the risks sequentially rather than in parallel. The challenge now is not to go backwards but to put in place arrangements that can manage the remaining risks, delivering a safer global payments and settlements platform. Liquidity and especially intraday liquidity risk can be supported by a global liquidity bridge. Much work has been done on this subject by the New York Payments Risk Committee (PRC): www.ny.frb.org/prc/. The PRC has proposed a number of solutions by which intraday liquidity could be obtained to support global payment system liquidity demands. These proposals include solutions that require central bank support and others that are primarily private sector solutions. The underlying theme of all the proposals is to bring forward arrangements that support the liquidity demands of global high-value payment systems. Global payment systems support global trade but there is a real danger that without such liquidity bridges liquidity could get trapped in individual systems in times of stress. The PRC continues to liaise with the Committee on Payment and Settlement Systems, and work with a number of infrastructure providers to see if a permanent solution can be put in place to support global payment systems. Clear and concise communication is a key to understanding and reacting rationally to a stressed situation. It is evident that the globally-connected world requires rather more than the legacy environment of domestic silos. Knowing what is the problem, where is the problem, who is managing the problem and an expectation of when the problem will be resolved goes a very long way to calming nerves in the connected world described above. It is unlikely that market participants will be as relaxed as they were historically if they do not have decent information to hand from a trusted source. As described previously there will be a tendency to take unilateral action if there is an incident in the payments world but with little or no information on time to resolution. Therefore, the creation of a global communications bridge to facilitate information exchanges during times of stress will be enormously beneficial to all market participants.

Note 1 For simplicity the many hundreds of users, SWIFT and custodial links are not shown.

15 Will central banking survive electronic money? Stefan W. Schmitz

Introduction and motivation The emergence of new technology has led to a renewed interest in the potential economic impact of the parallel use of multiple units of account.1 The diffusion of the internet could increase the costs of enforcement of national legislation. Electronic money could be issued in foreign jurisdictions, where national legal restrictions on the issue of banknotes do not apply and cannot be enforced, and could be a close substitute for banknotes and coins.2 The diffusion of internet usage and advances in encryption technology reduce the costs of issuance and distribution of electronic money relative to the issuance and distribution of physical banknotes and coins. The transaction costs associated with the parallel use of multiple units of account and the exchange of real assets decrease. Relative prices of goods and financial assets in different units of account, and the different units of account themselves, could be calculated (almost) instantaneously at low marginal costs due to continuous trading of units of account, goods and financial assets, the real-time availability of price information and the low costs of computer power to conduct the necessary calculations. Furthermore, the units of account, the goods and the financial assets can be exchanged at low marginal costs due to online markets (including securities settlement systems and central securities depositories), where trading and execution take place instantaneously and continuously.3 This chapter discusses the impact of the diffusion of electronic money on the future of central banking, i.e. on the efficacy of monetary policy. The analysis is based on the method of institutional analysis. Placing the diffusion of electronic money in the broader framework of institutional change in payment systems poses questions that are relevant for the analysis of the future of central banking: e.g. what are the main drivers of institutional change in the payments system and what are the instruments at the discretion of central banks to cope with it? What is the likely institutional structure of electronic money schemes and how does it impact on the essential elements of effective monetary policy? And what potential instruments of monetary policy implementation are available in a world without central bank money? The chapter will be structured along the following lines. I first develop the

234

S.W. Schmitz

conceptual framework and apply it to historical and recent innovations in the payments system before investigating the relationship between payment systems, central banks and monetary policy. I then apply the proposed conceptual approach to derive the likely institutional structure of e-money schemes, its implications for the demand for central bank money and for the role of central bank money as the generally accepted medium of exchange and uniform unit of account. Finally, I propose an approach to monetary policy in a world without central bank money (even though the previous sections show that this is highly unlikely) before concluding.

Conceptual approach to institutional change in the payments system The conceptual approach is based on the systematisation of the central institutional characteristics of the payments system – defined as the economy wide web of payment systems and instruments – and on the analysis of the dynamics of their change over time. The evolution of the payments system is subject to ongoing institutional change, e.g. the emergence of coinage, transferable deposits and banknotes, fiat money and credit card systems. The diffusion of electronic money schemes is a further instance of institutional change. The method of institutional analysis is the appropriate concept to investigate the likely consequences of the diffusion of e-money. The evolution of the retail payment system is path dependent. The existence of a generally accepted medium of exchange and a uniform unit of account are information networks that exhibit network effects.4 In the current state of payment systems (in advanced economies) a generally accepted medium of exchange prevails in the respective market, where it also entails the function of the uniform unit of account. Given positive transaction costs, a generally accepted medium of exchange will further reduce transaction costs relative to an economy without a generally accepted medium of exchange. An analysis of the effects of the diffusion of e-money schemes has to derive the necessary and sufficient conditions for a transition from one generally accepted medium of exchange and the associated unit of account to another and the effects of the diffusion of new technologies on the evolution of payment systems with respect to these conditions. In other words, will the diffusion of e-money lead to a sufficient reduction in the marginal costs of adopting a potentially emerging new generally accepted medium of exchange in a decentralised manner, i.e. individually by each agent?5 If so, how would the payments system operate in the phase of transition from one generally accepted medium of exchange to another? Is the parallel use of multiple units of account efficient and sustainable? An appropriate methodology to address the individual decisions at the margin – i.e. the individual choice of medium of exchange and unit of account in a given institutional arrangement – is based on New Institutional Economics, i.e. methodological individualism, transaction and information costs and an explicit analysis of the process of transition between equilibria.6

Will central banking survive electronic money?

235

Before turning to the question of the likely institutional structure of e-money schemes, I first address the following question: What are the relevant forces shaping institutional change in wholesale as well as retail and small-value interbank payment systems?7

Main drivers of institutional change in the payments system The main interdependent drivers of institutional change in the payments system are broadly categorised in two groups: policy initiatives and changing demand by banks (e.g. minimising opportunity costs of holding reserves), as well as by their customers (i.e. increasing demand for cross-border payment services due to globalisation).8 Central banks’ objective to control the monetary system – in order to ensure the effective implementation of monetary policy, the maintenance of financial stability, the smooth operation of the payment system and the collection of seignorage – is in general thought to require commercial banks to hold some reserve of central bank money. The commercial banks’ objective of profit maximisation requires them to economise on such reserves. The design of the payments system involves a trade-off between settlement risk and liquidity costs. An analysis of recent developments suggests that central banks and commercial banks as well as final customers have diverging preferences with respect to the optimal risk/cost combination in payment systems, due to a divergence between the social and private costs of disruptions of the payments system.9 New technologies have an impact on institutional change by changing the incentive and cost structure underlying particular institutional arrangements in payment systems and, hence, by enabling the development of new products, new markets and new governance structures furthering the politico-economic interests of those who promote them.10 I conceptualise technology as a production technology that transforms inputs (i.e. labour, capital) into outputs (i.e. payment services such as clearing and settlement). Rather than reducing the term technology to hardware and software (i.e. computers, telecommunication infrastructure), this conceptualisation also encompasses organisational structures, rules and procedures in the production of payment services.11 While general purpose technologies such as information and telecommunication technologies can be assumed to be exogenous to the politicoeconomic tensions that drive institutional change in the payments systems, this does not hold true for more specific payment technologies.12 The latter are endogenous to the process of institutional change: they are developed by R&D efforts of payment system participants and they depend on complementary innovations to become productive. First, these can be necessary at the level of the individual payment service provider: the development and adoption of new payment technologies necessitates adaptations at this level in areas such as organisational structures, internal governance mechanisms and risk management models as well as skills. These complex and costly endeavours are the result of conscious decisions made under uncertainty. Second, they also involve private complementary investments at the market level such as private institutions

236

S.W. Schmitz

monitoring credit histories of users of payment instruments (credit registers). Complementary innovations are also necessary at the payments system level and involve political decisions: these are institutional innovations such as the governance structure of the payments system (i.e. regulation and oversight of new payment institutions and technologies) and the general legal framework (e.g. privacy protection and liability issues in electronic payment systems). The development and adoption of specific payments technologies and the complementary institutions at the firm, the market and at the payments system level are endogenous to the politico-economic tensions that drive institutional change in the payments system.

Central institutional characteristics of the payments system The central institutional characteristics of the payments system encompass the medium of final settlement in the payments system and its relation to the generally accepted medium of exchange in the economy as well as the characteristics of the interbank clearing and settlement institution.13 The latter include conditions of access to its accounts, conditions of access to its credit facilities, and the nature of its clearing and settlement process (i.e. real-time gross settlement (RTGS) with or without intraday credit, deferred net settlement (DNS), hybrid systems). In addition, the surrounding institutional environment, in which the interbank payment system operates, is of importance: the state of development of the interbank money market and the sophistication of participants’ treasury management. But also some features of monetary policy implementation have repercussions on the institutional characteristics of this system. The reserve maintenance system is of particular relevance in this respect.14 These characteristics can be interrelated in important ways. The relationship between the generally accepted medium of exchange and the medium of final settlement as well as the relationship between the clearing and settlement institution and the issuer of the generally accepted medium of exchange can influence credit and liquidity risk of the interbank payment system. If the medium of final settlement is not the generally accepted medium of exchange, potential demand for exchanging the medium of final settlement into the generally accepted medium of exchange imposes a liquidity risk on the participants of this system, as the generally accepted medium of exchange is by definition the most liquid asset in the relevant market. If the clearing and settlement institution is not the issuer of the generally accepted medium of exchange, its opportunity costs of holding sufficient reserves are positive and it can, in principle, go bankrupt, which exposes participants to credit and liquidity risk. Institutional characteristics influence the operational characteristics of the payments market, such as its efficiency, stability and reliability, the concentration of payment flows, the nature and intensity of competition among payment systems, the structure and level of costs of access to the interbank payment system and to intraday credit and the degree of tiering in the payments system.

Will central banking survive electronic money?

237

Longer-term developments in payments systems In this section, I use the institutional approach to analyse the development of payment systems. The early evolution of non-cash means of payment in medieval Europe can be explained well within the proposed conceptual framework: payment innovations were the response to changing trading patterns; they were driven by the interaction between the desire of merchants to reduce the opportunity and transaction costs associated with the generally accepted medium of exchange and the motives of public authorities; they required complementary legal and institutional innovations; they reduced the demand for the generally accepted medium of exchange and increased (i.e. credit) risk in the payments system. Sales credit reduced the demand for the generally accepted medium of exchange by bilateral offsetting of debt due to mutual credit exposure and by the assignment of third-party debt (book credit) as a means of payment.15 Periodic settlement in coin served as a monitoring and disciplining device to address credit risk. The observability of the quality of debt at acceptable costs was a prerequisite for low transaction costs, so that its applicability was confined to trade mainly among well acquainted trading partners. With a changing trading environment, two important payment innovations emerged: the bill of exchange and the transfer of deposits at deposit banks. As interregional trading volume grew, cambium contracts – a legal innovation – evolved in trade between Italy and the fairs of Champagne.16 The contract was drawn by a notary and both parties had to be present in person. Bilateral trust was a precondition for such a transaction. Trading patterns changed, interregional trade grew and took place with strangers rather than within small communities of well-acquainted merchants. The organisation of trade changed, too, and the individual merchant was replaced by trading companies, which operated through branches and agents in commercial centres. The merchants themselves did not have to travel anymore; two trading partners would not necessarily be present at the same location at the same time to visit the notary to draw a cambium contract. Kohn (1999) argues that the trading companies evolved into merchant banks, which replaced the notary and the presence of both parties, by means of internal communication: bills of exchange (lettere di cambio). It helped to separate sales transactions from credit provision in (wholesale) trade. The transaction costs of the assignments of third-party debt were substantially reduced by legal provisions: the transferability and negotiability of debt.17 The demand for the generally accepted medium of exchange was reduced, as claims to the merchant bank served as means of payment. The merchant bank issuing bills of exchange acted as a trusted third party, so that trust between merchants was no longer a precondition for the acceptance of bills of exchange.18 Similarly, the origin of deposit transfers relates to the high transaction costs of payment in coins, which required substantial expertise to assess the quality and weight of the coins and to count them. Moneychangers specialised in this

238

S.W. Schmitz

service and charged fees. Each transfer of coins would require their assaying and weighing by the moneychanger. To reduce these costs, merchants deposited the coins at the moneychanger once they were assayed and weighed. They transferred the deposits as means of payment and avoided the costs of re-assaying. In some areas deposit banks engaged in international transactions and deposit transfers served as an alternative to the bill of exchange. Which of the alternatives prevailed at the commercial centres of medieval Europe was mainly subject to local and interregional trading patterns and politico-economic pressures (e.g. the prohibition of deposit banking in Antwerp), but not technological innovations. Abstract units of account (‘imaginary money’ or ‘political money’) existed throughout Europe from about 800 to 1800. They were defined in terms of the generally accepted medium of exchange (gold or silver), although not actually coined.19 Coins served as means of payment, each defining a monetary unit. Prices were quoted in the abstract unit of account and had to be recalculated in terms of the monetary units of the coins in circulation. How was the abstract unit of account linked to the generally accepted medium of exchange? The exchange rate of each coin in terms of the abstract unit of account was set by the public authorities and it was illegal to exchange coins at deviating rates.20 Abstract units of account were politico-economic instruments to tackle the following problem: many different local and foreign coins circulated in the various political areas in Europe and no uniform unit of account existed. The introduction of an abstract unit of account reduced the number of exchange rates to the number of coins in circulation rather than (half) the set of cross-rates. It was hard to keep a bimetallic system in equilibrium, as relative prices of precious metals – or the gold and silver content of the coins – frequently changed.21 With an abstract unit of account, official exchange rates could easily be adjusted by crying coins up or down, respectively, in terms of the abstract unit of account.22 In the long term evolution of payment systems, one of the most important instances of institutional change was the foundation of central banks. From the foundation of their precursors to that of the Bank of England in 1694, the Federal Reserve in 1914 and the European Central Bank (ECB) in 1998, they constituted institutional innovations explained by politico-economic considerations (e.g. public finance motives, smooth functioning and efficient payments system, seigniorage, financial stability or political and economic unification) rather than technological innovations.23 Similarly the establishment of common currencies from the circulation of federal reserves notes in 1914 to the circulation of euro notes in 2002 were based on widespread technologies, but represented institutional innovations based on politico-economic reasoning. Humphrey et al. (1996) describe the evolution of payment systems in Europe, Japan and the US in the nineteenth and twentieth centuries. The authors explain the dominance of credit transfers in Europe by banking concentration, nationwide networks and cooperation among banks as a response to the development of postal giro services across Europe. In the case of Japan, the authors consider the lower crime rate as the major reason for the larger reliance on cash at the

Will central banking survive electronic money?

239

point of sale compared with the United States. The evolution of the Japanese payments system was largely driven by policy initiatives and by strategic banking sector cooperation.24 The check system dominated in the United States. The credit transfer system failed to pick-up, despite banking services having been offered relatively early to the general public. The major reasons were regulatory constraints (i.e. branching restrictions), the resulting low concentration, and the involvement of the Federal Reserve in cheque clearing and the subsidisation of the cheque system. Lacker et al. (1999) argue that the Federal Reserve’s entry into cheque clearing was a consequence of a legal privilege – the sole right of mail presentment at par – which in turn was motivated by the desire to attract new members to the Federal Reserve system. In interregional business-tobusiness payments, bank drafts dominated until the late nineteenth century. Prescott and Weinberg (2003) argue that their replacement by cheque was due to demand by merchants, institutional innovations (i.e. credit reporting services), which enabled merchants to evaluate the quality of cheques offered by previously unknown counterparties, and technological advances (i.e. development of the telegraph), which reduced the costs of gathering information concerning credit quality from credit registers.

Major recent developments in payments systems According to Bank for International Settlements (2005), liberalisation, globalisation and consolidation have enormously increased the volumes handled in wholesale (large-value) payment systems and have thus increased awareness of potential threats to systemic stability. Fry (1999) reports that unprotected DNS systems dominated the large-value payment market internationally until the 1980s. Bank for International Settlements (1990) highlighted the associated stability risks and suggested ‘Core Principles’ for cross-border DNS systems to contain such risks: in particular, systems should be able to settle even in the case of failure of the largest net debtor. McAndrews and Trundle (2001) argue that the remaining risks and the associated costs evident even in protected DNS systems led to the adoption of RTGS in all EU and G10 countries in the 1990s. In addition, Bank for International Settlements (2001) also encourages all payment and settlement systems that meet the criteria of systemic importance to settle in central bank money. Federal Reserve System (2006) highlights that while these developments have reduced systemic risk they have also increased the reliance on central bank money and raised the intraday demand for central bank money significantly. Most central banks provide subsidised intraday credit against collateral for two reasons. First, intraday liquidity can be interpreted as a public good.25 Second, central banks want to attenuate the increase of banks’ opportunity costs of intraday liquidity associated with the move to RTGS and to contain their opposition against the move (e.g. routing payments through private systems). The higher costs of liquidity in RTGS also gave rise to hybrids (e.g. continuous net settlement and queue-augmented RTGS).26 The wholesale money market is the only financial market in the EU which is effectively integrated.27

240

S.W. Schmitz

The establishment of the European large-value payment system – ‘TransEuropean Automated Real-time Gross settlement Express Transfer’ (TARGET) – in 1999 laid the foundations for this integration and, thereby, for the ECB to implement monetary policy effectively across the euro area. The levels of tiering and concentration in the value of payments differ widely between 29 large-value payment systems in the G10.28 In conjunction with differences in minimum reserve requirements, this may contribute to explaining the large differences in banks’ reserves at the respective central bank between the euro area (4.7 per cent of narrow money), the United Kingdom (0.2 per cent of narrow money), and the United States (1.5 per cent of narrow money) and the share of banks’ deposits at other banks (2.0 per cent of narrow money in the United States, 21.6 per cent in the euro area and 57.6 per cent in the United Kingdom).29 The demand for international large-value payments and for liquidity bridges across systems and across borders increases. As a response to regulatory concerns about systemic risk arising from foreign-exchange (FX) settlement risk, the financial industry developed a real-time settlement system in 2002 that ensures same-day final and irrevocable settlement in foreign exchange transactions (payment-versus-payment settlement). Continuous Linked Settlement (CLS) was designed to reduce transaction costs and settlement risk in foreign exchange markets.30 The system allows for multilateral netting as payment instructions must be submitted before the settlement cycle starts so that member banks have to fund only the net positions in each currency; they can also overdraw their accounts at CLS in some currencies as long as their overall position across currencies remains positive (taking into account FX market volatility haircuts). Both features substantially reduce funding requirements for member banks and their opportunity costs of holding reserves in central bank money. While the transactions are settled across the books of CLS Bank International, clearing takes place across the books of the relevant central banks, where CLS Bank International holds accounts (via remote access). More recently large international banks highlighted the problem of liquidity islands, i.e. banks have to hold increasing amounts of liquidity in various systems in many countries, but find it costly to transfer liquidity across systems and across borders. In order to decrease the opportunity costs of liquidity of large international banks at the international level and to make use of scale and scope effects in global funding risk management, they called for new services to be offered by central banks, such as the development of new intraday liquidity services, the liberalisation of remote access to central bank accounts and to intraday credit (including the acceptance of foreign assets as collateral) as well as the establishment of multicurrency facilities.31 Bank for International Settlements (2003a) summarised recent trends in small-value payment systems in the G10 countries and in Australia: •

A shift from cash and paper-based instruments to non-cash electronic payment methods (electronification).

Will central banking survive electronic money? •

•

•

241

The evolution of product innovation in the context of new payment methods (e-money, m-payments) and in the area of access products (ATMs offer additional services, such as reloading prepaid mobile phone cards; internet banking). New market entrants (e.g. mobile phone companies, telecommunication operators, net-based scratch card companies) are often particularly innovative, and are more active in the area of new payment instruments, despite the fact that banks remain the main players in the payment system. New market players and new products are usually regulated as banks or e-money institutions in the European Union and to some extent in the United States, where large differences prevail across states. In the European Union, policy initiatives constituted the major drivers of change in cross-border retail payments.32

If technology were the main driver of institutional change in the payments system and the G10 countries had access to similar payments technology, the institutional structure of their payments systems would be similar. However, the institutional and organisational structures of the economy-wide payments system differ across economies. Indeed, the chapter by Bech in this volume shows how the diffusion of payments technology itself differs widely across G10 countries. Despite large differences in institutional structure across them, all countries have in common that central bank money serves as the generally accepted medium of exchange and the unit of account and all economically relevant payment systems are eventually linked to central bank money via the banking system.

Institutional change in the payments system and monetary policy The impact of the institutional characteristics of the payments system on monetary policy can be categorised along four dimensions. First, institutional characteristics of the payments system affect the level of demand for central bank money as well as its structure, predictability, velocity and its sensitivity with respect to central banks’ instruments (i.e. the interest elasticity of demand for central bank money). Second, the operational efficiency of the payment system is a precondition for the emergence of deep and liquid interbank markets. Both are prerequisites for the effective implementation of monetary policy, as a large and unstable float can lead to higher and more volatile reserves both at the level of individual banks and at the aggregate level. Third, the payment system should not be a source of unforeseen and unpredictable shocks to the quantity and costs of liquidity with ensuing direct and indirect ramifications for monetary policy. Fourth, if central bank money is the medium of final settlement in largevalue payment systems and participants have access to its accounts, the design of the settlement mechanism and the liquidity management strategies of the participants affect the demand for intraday credit.33 A higher demand for

242

S.W. Schmitz

intraday credit can increase the risks for monetary policy, due to the increasing risk of spillover of intraday credit to the overnight money market. In principle, central banks implement monetary policy by manipulating the short-term interest rate, i.e. the overnight interest rate in the interbank market.34 Despite the small size of their repurchasing operations on interbank markets relative to total turnover, their impact is sufficient to steer the market. This is mainly due to their ability to issue the generally accepted medium of exchange at zero marginal cost. But central banks have additional instruments at their discretion that increase their grip on the money market by imposing a structural liquidity deficit. They can influence demand for their own liabilities by minimum reserve requirements and by legal restrictions concerning the issuance of banknotes as well as by (in some countries) ‘moral suasion.’ The main instruments of monetary policy implementation are open market operations, minimum reserve requirements and standing (lending and deposit) facilities. Today central banks also routinely employ announcements of levels of their main operating target in monetary policy implementation. These instruments can be adapted to cope with institutional change in the payment system. But they also have an impact on the institutional characteristics of payment systems, and can, therefore, be employed by central banks to proactively shape institutional change in payment systems.35

Institutional change in the payments system and payment system policy In addition to their choice of monetary policy instruments central banks can influence institutional change in payment systems by their own payment system policy and by their influence on the legal framework governing the payments system. Their payment system policy consists of three main categories. First, central banks encourage systemically important payment systems to settle in central bank money in order to reduce systemic, credit and liquidity risk as well as to ensure service continuity (settlement policy).36 Second, central banks’ access policies to central bank money (in the form of central bank accounts) are the core instrument of their payment system policy with respect to payment system participants. Third, Bank for International Settlements (2003b) reports that, in general, access to central bank accounts also implies access to intraday credit at the central bank and the underlying considerations are very similar. One of the main drivers of institutional change in the payments system is the legal framework governing it, which falls into the competence of legislatures. Nevertheless, central banks exert a high level of influence in drafting rules (soft law) at the international level and in shaping national legislation by consulting governments and legislature.37 Furthermore, legal frameworks in the European Union and United States transfer substantial regulatory discretion concerning the regulation and oversight of payment systems to central banks (e.g. reporting requirements, ECB Oversight Standards, Regulation E). In the end, however, the role of central banks in the legislative process and their regulatory discretion are conditional on the public’s support for the current monetary constitution.

Will central banking survive electronic money?

243

To summarise, central banks have a large range of instruments at their discretion to react to but also to influence institutional change in the economy-wide payment system. They are heavily involved in the legal and political process shaping the broad legislative framework concerning payment instruments and they transfer substantial regulatory power within this framework. In addition, central banks can adapt the instruments of monetary policy implementation and their own payment systems policies to cope with institutional change in the payments system. In recent history central banks have demonstrated their determination and their political ability to maintain control of the monetary system in the face of institutional change in the payments system and to actively shape it.

The likely institutional structure of e-money schemes The following section focuses on the impact of the evolution of electronic money on the incentives and costs concerning core characteristics of the institutional structure of payment systems, the choice of the generally accepted medium of exchange and the unit of account.38 The potential separation of the generally accepted medium of exchange and parallel units of account has gained increased attention in the literature on e-money due to the emergence of new technology.39 In an overview of the literature Schmitz (2002b) shows that the parallel use of multiple units of account is not desirable and that their competitive supply in the case of fiat-type currencies is not feasible.40 The transaction costs of the co-ordination of individual plans are reduced and the transparency of markets is increased by the existence of a uniform unit of account (in the respective market). I demonstrate that users and issuers face strong strategic incentives not to opt for an alternative, generally accepted, medium of exchange and unit of account in electronic money schemes under current inflation rates. On the one hand, this result is due to network effects, sunk costs, information costs and switching costs which are characteristic for retail payment systems and the choice of the unit of account. On the other hand, the argument rests on the findings regarding the underlying mechanism of price formation. In the case of a price matching strategy, nominal prices for all goods in the generally accepted medium of exchange are multiplied by the exchange rate of the e-money unit. The existence and sufficient liquidity of markets for all goods and for the e-money units relative to the dominant unit of account (in order to establish relative prices of e-money units of account in the dominant unit of account) are necessary preconditions for e-money schemes – denominated in alternative units of account – to be able to quote goods prices in the alternative unit of account. Even if both preconditions are met, trading on markets denominated in alternative units of account involves higher prices due to a spread in exchange between the dominant unit of account and the alternative ones. In the case of a price discovery strategy, nominal prices for all goods in terms of the e-money units of account are determined by market exchange. Initially, each market denominated in the e-money unit of account would be less liquid relative to the one denominated in the dominating unit of account. Thus,

244

S.W. Schmitz

the intensity of competition and the information content of prices would be lower, and the spread between bid and ask prices higher. Under both strategies – price matching and price discovery – the real prices for all goods in the e-money units of account are higher than in the generally accepted medium of exchange and the dominant unit of account. The institutional analysis of e-money and monetary policy analyses the choice of unit of account in an environment of a dominant unit of account. At moderate levels of inflation, participants in the payment system have no incentive to switch from the dominant unit of account to one or more emerging alternative(s) in the relevant market. Consequently, the most likely institutional structure of emerging e-money schemes includes denomination in the dominant unit of account and redeemability, which is argued to be a necessary but not a sufficient precondition for the sustainable exchange of e-monies for central bank money at par. The role of national currencies as units of account will not be diminished by the diffusion of e-money at the current moderate levels of inflation. As central banks hold on to the monopoly of the supply of the generally accepted medium of exchange at zero marginal costs, they retain control of its supply and its purchasing power, in principle. The balance sheet of central banks will shrink relative to a world without electronic money, which is mainly a positive sign as institutional change in the payments system (e.g. electrification of retail payments systems, tiering in wholesale payment systems) increases efficiency of the economy-wide payments system. This implies that monetary policy could become more rather than less effective.41 However, the diffusion of e-money can also lead to an increase in the relative interest elasticity of the real demand for money vis-à-vis the interest elasticity of the demand for goods, which would have the opposite effect on the effectiveness of monetary policy.42 The net effect is ambiguous. In order to preserve a structural liquidity deficit in the generally accepted medium of exchange in a situation of decreasing aggregate demand for central bank money, central banks would have to reduce aggregate supply correspondingly. Given a structural liquidity deficit, the current institutional structure of monetary policy implementation can likewise be applied to a world with a very low but positive demand for central bank money. Central banks have coped well with institutional change in the payment system in the past (e.g. diffusion of credit and debit cards, elimination of reserve requirements in Australia, Canada, New Zealand, Sweden and the UK).43

Fundamentals of monetary policy in a world without money However spectacular recent innovations in payment systems are depicted, a world without central bank money is not in sight. This notwithstanding, it is important for policy makers as well as researchers to investigate the potential implications of such an evolution, even if it is deemed unlikely at the moment. This section provides a conceptualisation of monetary policy in a world without central bank money based on a generally accepted medium of exchange that also serves as a medium of final settlement.44

Will central banking survive electronic money?

245

In a world with central bank money, the generally accepted medium of exchange also functions as the medium of final settlement in the interbank payment system. Schmitz (2002b) argues that for efficiency reasons a single generally accepted medium of exchange and a unified unit of account in the relevant market also prevails in a world without central bank money (e.g. based on a commodity standard). All means of payment are claims to the medium of final settlement. In order to reduce the spread between bid and ask interest rates in the interbank market by reducing credit, liquidity and market risk, the respective generally accepted medium of exchange will also serve as the medium of final settlement in the interbank market. It is the only medium that is not a direct or indirect claim on future resources and that ensures settlement finality in the interbank payment system.45 Monetary policy in a world without central bank money becomes feasible through a combination of minimum reserve requirements in the medium of final settlement and interest paid or charged on these. These instruments are available to central banks, because they are public authorities with certain regulatory competencies transferred to them by the respective legislature. These competencies are independent of the loss of central banks’ monopoly to issue the generally accepted medium of exchange at zero marginal cost and can (and already do) entail the competence to impose obligations on third parties such as minimum reserve requirements in the medium of final settlement, as well as to specify an interest rate paid or charged on these for the purpose of monetary policy implementation. The opportunity costs of holding additional reserves in the medium of final settlement are determined by the marginal costs of obtaining it on the market minus the (positive or negative) remuneration of minimum reserve requirements at the margin. Irrespective of the loss of the monopoly provision of the medium of final settlement central banks can manipulate the opportunity costs of holding reserves at the margin. Rather than assuming the money market rate to be the main policy target, central banks can treat the market rate of the medium of final settlement as exogenous and steer liquidity conditions (i.e. the opportunity costs of holding reserves at the margin) by manipulating the interest rate paid or charged on minimum reserves held by market participants directly. Comparable with the implicit taxation of financial intermediation by imposing minimum reserve requirements in a world with central bank money, remuneration paid or fees charged on minimum reserve requirements in a world without central bank money correspond to a subsidy and tax, respectively, on the liabilities (which are subject to minimum reserve requirements) of money market participants (mainly banks). An increase (or decrease) of the interest charged on minimum reserves shifts the stock of reserves held on average over the maintenance period and, hence, the aggregate demand for the medium of final settlement downwards (or upwards) at a given market rate. The supply schedule of the aggregate stock of the medium of final settlement is unaffected by changes in the opportunity costs of holding reserves, as it is determined by marginal costs of supply of the

246

S.W. Schmitz

medium of final settlement (e.g. marginal costs of production in the case of a commodity standard). The equilibrium price in the market for the medium of final settlement decreases (increases). Under the precondition that the supply of the medium of final settlement is not infinitely inelastic, the equilibrium price decreases (or increases) less than the interest rate on minimum reserves; thus, the opportunity costs of the stock of minimum reserves increases (or decreases) at the margin. This tightens (or eases) liquidity conditions for market participants. In addition to the aggregate stock of the medium of final settlement, banks supply end-of-day excess reserves on the overnight market. How will the supply of excess reserves influence the marginal costs of aggregate supply? The demand for and the supply of excess reserves are unplanned residuals of the payments processed during the opening hours of the interbank payment system. After the realisation of end-of-day balances banks lend excess reserves, which are not remunerated, or borrow to cover deficiencies in the overnight market. As interest is neither paid nor charged on excess reserves in the proposed regime, their supply and demand are independent of the opportunity costs of holding the stock of minimum reserves. If the time it takes to adjust the aggregate stock of the medium of final settlement is below the maintenance period, arbitrage opportunities ensure that market participants have no incentive to borrow from each other at costs above the marginal costs of the medium of final settlement. Analogously to the determination of the opportunity costs of holding reserves in a world with central bank money, the opportunity costs of holding the stock of aggregate minimum reserves are determined by the marginal costs of supplying these reserves and not by the interest rate on the flow of the medium of final settlement due to demand for and supply of excess reserves or the interest charged on minimum reserves. The analysis of the current legal framework governing the operations of the Bank of England, the ECB and the Fed reveals that potential politico-economic objections to the transfer of the necessary competencies to implement the proposal are misplaced. In principle, these central banks already possess the necessary regulatory authority and discretion to impose reserve requirements and to charge and pay interest thereon and only minor adaptation to the current framework would be necessary.46 In principle, central banks can manipulate the opportunity costs of holding reserves at the margin but with less accuracy. This is due to the fact that additional instruments to absorb liquidity shocks and to stabilise money market rates (i.e. standing facilities, intraday credit) are not available to central banks, which do not have the monopoly to supply the generally accepted medium of exchange at zero marginal cost. Central banks lose control of the supply of the medium of final settlement, such that supply shocks add to the uncertainty central banks face in monetary policy implementation in a world without central bank money.

Will central banking survive electronic money?

247

Conclusions The analysis of recent and long-run institutional change in payment systems suggests that its main drivers are politico-economic factors and the demand of commercial banks and final customers rather than technological innovations. The institutional structure of payment systems displays large differences between countries, despite similar technology being available. Central banks have a large number of policy instruments at their discretion to cope with institutional change and have proved to do so effectively in the past. The politicoeconomic analysis of institutional change in payments systems, nonetheless, is still in its infancy. A research programme of theoretical and empirical analysis is called for to investigate the preliminary results concerning the dynamics of change in payments systems presented here. Such a programme should transcend the boundaries of economics. It ought to include financial and political history as well as political science to provide additional historical and contemporary case studies; it ought to be transdisciplinary in nature. The diffusion of electronic money is a further instance of institutional change in payment systems. Consequently, the appropriate conceptual framework to investigate the impact of its diffusion on the efficacy of monetary policy and the future of central banking is the method of institutional analysis. The essential element of effective monetary policy is the monopoly provision of the generally accepted medium of exchange – with its incidental functions as uniform unit of account and medium of final settlement – at zero marginal cost. The investigation of the future of central banking, therefore, builds on an analysis of the impact of the diffusion of electronic money on the institutional characteristics of the payments system. I demonstrate that users and issuers face strong disincentives to switch from an established generally accepted medium of exchange and uniform unit of account in the respective market to an alternative generally accepted medium of exchange or an economy without a generally accepted medium of exchange. The most likely institutional structure of electronic money schemes includes the denomination of electronic money and redeemability in the dominant generally accepted medium of exchange and unit of account. Thus, central banks will retain their monopoly to issue the generally accepted medium of exchange with its incidental functions as the uniform unit of account and the medium of final settlement at zero marginal cost. Monetary policy will remain effective, in principle. Although the reduction of the demand for central bank money to zero is highly unlikely, this chapter provides a sketch of potential instruments of monetary policy implementation in a world without central bank money. It builds on the regulatory competencies of central banks, which do not depend on its monopoly in providing the generally accepted medium of exchange at zero marginal cost. In principle, monetary policy would be feasible with a combination of minimum reserve requirements in the generally accepted medium of exchange and interest paid, or charged, on these. It rests on the result that – unless the economy resembles a Walrasian economy with zero transaction costs – the

248

S.W. Schmitz

existence of a generally accepted medium of exchange will increase the efficiency of the payments system. The question whether central banks will survive e-money is one of political economy rather than one of technology diffusion, that is, it is one of the ability to present their case to the respective legislature relative to the respective capabilities of banks and their final customers. In recent history central banks have demonstrated their determination and their political ability to maintain control of the monetary system in the face of institutional change. To sum up, central banking will survive electronic money due to the expected institutional structure of electronic money schemes and the large number of instruments at the discretion of central banks to cope with institutional change in the payments system.

Notes 1 See inter alia Browne and Cronin (1996), Centi and Bougi (2003), Cohen (2002), Costa Storti and De Grauwe (2003), Crede (1995), England (1996), Freedman (2000), Friedman (1999, 2000), Goodhart (2000), Henckel et al. (1999), King (1999), Kobrin (1997), Krozner (2003) and Palley (2002). Schmitz (2006a) takes a critical view of the theoretical coherence of these models, arguing that they collapse to either a Walrasian economy, or variants of a commodity standard, or the current monetary system based on central bank money. 2 Electronic money is defined in the E-money Directive (2000/46/EC) Article 1 (3)(b) as monetary value as represented by a claim on the issuer which is: (i) stored on an electronic device; (ii) issued on receipt of funds of an amount not less in value than the monetary value issued; (iii) accepted as a means of payment by undertakings other than the issuer. 3 According to the advocates of the parallel use of multiple units of account the various issuers of electronic money units of account would compete in three areas: (i) the management of the electronic payments system (i.e. marketing the system to a large number of attractive trading partners, non-pecuniary benefits of participation, reliability and security of the technological and organisational infrastructure, liability for costs in cases of unauthorised payments, loss or fraud), (ii) the financial performance of the reserves backing the systems (i.e. rate of return, volatility etc.), and (iii) the choice of regulatory regime, if such choice is indeed legally possible. The relative performance of reserve assets would then translate into the relative purchasing powers of alternative units of account. 4 See Menger (1909), Krüger (1999), Schmitz (2002b) and Selgin and White (2002). 5 Another potential direction of research would address the following question: will institutional change in the payments system reduce the marginal costs of a socially concerted adoption of a new generally accepted medium of exchange and a new unit of account? This question is, however, beyond the scope of this chapter. 6 Schmitz (2002a) argues that current neoclassical models of money (i.e. search models, OLG models and spatial separation models) based on comparative static analysis are inappropriate to analyse institutional change in the payments system as their institutional structure is exogenously given and static, i.e. institutional change cannot be conceptualised and the transition dynamics between equilibria cannot be analysed. 7 This section builds on Schmitz and Wood (2006). 8 Some examples of policy initiatives are the CPSS Core Principles, the Single Euro Payment Area, the EU New Legal Framework, revisions to Federal Reserve policy on

Will central banking survive electronic money?

9 10 11 12

13

14 15

16

17

18 19

20 21

22

249

payment system risk, the Uniform Money Services Act and amendments to Money Transmitter Laws in many US states. In addition to policy initiatives directly addressing payment systems, privacy, consumer protection and anti-money-laundering laws, to name but a few, also affect payment systems and can influence institutional change in payment systems. Fry (1999) page 82. McAndrews and Trundle (2001). See Rip and Kemp (1998) for a discussion of sociological, philosophical and economic concepts and theories of technological change. See, for example, European Commission (2003), which argues that Regulation 2560/2001/EC ‘has provided an incentive for the payment industry to modernize their EU-wide payment infrastructure’. The adoption of new technology is itself driven by politico-economic determinants. Menger (1909) defines the generally accepted medium of exchange as the most liquid good in the economy, the good with the highest marketability and, thus, the lowest spread. Settlement finality refers to an unconditional and irrevocable payment (EU Final Settlement Directive 98/26/EC). That is, the averaging of minimum reserve requirements, the averaging period, its relation to the interval of central banks’ refinancing operations and the potential employment of minimum reserves for settlement purposes. White (2006) argues that credit cards are conceptually similar to sales credit: the growth of multi-outlet retailers (not a technological innovation!) led to the formalisation of standing credit authorisations. Banks adopted the model in the 1950s in the United States, acted as trusted third parties and helped to separate the sales transaction from the credit transaction. The cambium contract involves a credit transaction, a remittance service, and an exchange transaction. It is similar to a bill of exchange, except for higher transaction costs due to the requirement that both parties to the contract must be present in person at the notary. If third-party debt is only assignable but not transferable, only the original creditor had the right to sue the debtor, but not its last holder. If it is assignable and transferable, the current holder has the right to sue only the initial creditor but not the assigners. If it is negotiable as well, the current holder can sue all assigners if he does not receive payment by the initial creditor. The probability of default, and credit risk, are effectively reduced with each assignment. Trust was still an important ingredient in this institutional arrangement. Merchants were exposed to credit risk vis-à-vis the merchant bank. Its reputation was essential for the conduct of business and performed the function of a disciplining device. The most prominent example (the pound or livre or lira) emerged through coinage though. Under Charlemagne, 240 pennies were cut from one pound of silver; a shilling consisted of 12 pence, and a pound of 20 shillings. The standard remained in use even long after the pennies ceased to exist. Einaudi (1953) page 245. Nevertheless, it had to be close to the going market rate of the coins in circulation in terms of the generally accepted medium of exchange. Otherwise, arbitrage would have driven the undervalued coins out of circulation. Without an abstract unit of account two adjustment mechanisms prevailed: coins could be reminted, with an adjusted gold or silver content to reflect changes in market prices in the bullion market (or changes of the specie content), or adjustments of the nominal purchasing power of each circulating coin could re-establish the appropriate relative prices of the coins. In both cases adjustment and transaction costs are very high. Although abstract units of account reduce transaction costs in a setting of many circulating coins with variable exchange rates, a uniform unit of account based on the generally accepted medium of exchange would further reduce transaction costs. See Schmitz (2002b).

250

S.W. Schmitz

23 Kohn (1999) and Fratianni and Spinelli (2006) highlight the roles of the public finance motive in the evolution of early public banks in medieval Europe; Kindleberger (1984) stresses the public good considerations of well-functioning and efficient payment systems. Goodhart (1988), pages 105–84, provides accounts of public finance motives for the foundations of various central banks in Europe, including the Banque de France 1800 (government finance), Oesterreichische Nationalbank 1816 (re-organising the currency system of the Austro-Hungarian Empire after inflationary war financing) and the Reichsbank 1875 (unification and organisation of the note issue and the payment system in the newly established German Empire). 24 Policy initiatives include the government-pushed development of a modern banking system after 1868; the National Centralized Domestic Exchange Settlement System (NCDE) operated by the Bank of Japan in 1943 as a small-value payment system; the BOJ-NET in 1988 as a large-value payment system. An example of banking sector cooperation was the replacement of the paper-based NCDE by the electronic ZENGIN system in 1973. 25 Banks applying for intraday credit – rather than postponing payments – increase aggregate intraday liquidity in the system, which then circulates in the system for the rest of the day and reduces the liquidity costs of all other participants (a multilateral non-attributable positive externality). 26 Bank for International Settlements (2005). 27 European Commission (2003). 28 Bank for International Settlements (2003b). 29 Data is for 2004 (except share of banks’ deposits at other banks of narrow money for the euro area which is for 2002). Sources: European Central Bank (2006), Bank for International Settlements (2006) and own calculations. 30 CLS has about 70 shareholders of which 26 provide settlement services in 15 currencies in cooperation with the respective central banks. More than 700 banks, funds and corporations worldwide are indirect participants in the systems (March 2006). CLS Bank International is incorporated under US law and regulated by the Federal Reserve Bank of New York. 31 Payments Risk Committee (2003). The banks also evaluated a number of private sector solutions, but concluded that it was not clear that there would be a business case for any of these and called for the socialisation of the related costs via services provided by central banks. Some central banks have already followed the banks’ demands, since they partly overlap with central banks’ growing concerns about the increasing reliance of banks on intraday credit (e.g. the ECB Correspondent Central Banking Model; the liquidity bridge between EURO 1 and TARGET; the acceptance of some foreign assets as collateral by the central banks of Denmark, England, Norway, Sweden, Switzerland and the United States). 32 That is, the introduction of the euro banknotes, Regulation 2560/2001/EC, the Single Euro Payments Area (SEPA) initiative, and the New Legal Framework. 33 Central banks operate large-value payment systems and/or encourage systemically important private clearing and settlement systems to settle in central bank money, too. Some of them require prefunding in central bank money by participants (e.g. CHIPS and CLS), which also raises demand for intraday liquidity and intraday credit at central banks. 34 Schmitz (2006b) presents a more detailed conceptualisation of the instruments of monetary policy implementation relevant for the analysis. 35 Descriptions of the monetary policy instruments of the ECB and the Fed can be found in European Central Bank (2006a) and Federal Reserve System (2002, 2004). 36 Bank for International Settlements (2001), in its 6th Core Principle, states that ‘Assets used for settlement should preferably be a claim on the central bank; where other assets are used, they should carry little or no credit risk and little or no liquidity risk.’ 37 For example, CPSS Core Principles, Angell Report, Lamfalussy Report, Recommen-

Will central banking survive electronic money?

251

dations for Central Counterparties, Recommendations for Securities Settlement Systems. Article 105(4) of the EU Treaty establishing the European Union stipulates that the ECB shall be consulted on any proposed Community Act in its fields of competence; and by national authorities regarding any draft legislative provision in its fields of competence. 38 This section builds on Schmitz (2002b). 39 Inter alia, King (1999), page 26, illustrates the focus on technology in reasoning about institutional change in the payment system: The key to such developments [final settlement by the transfer of real wealth] is the ability of computers to communicate in real time to permit instantaneous verification of the credit worthiness of counterparties, thereby enabling private sector real-time gross settlement to occur with finality. Any securities for which electronic markets exist could be used as part of the settlement process. 40 The argument does not provide a rationale for legal barriers against potential currency competition, though. Berentsen (2006) shows that the private provision of fiat-type currency would be feasible under very restrictive assumptions. It is still considered suboptimal in his model. 41 Selgin and White (2002) argue that monetary policy becomes even more effective as the elimination of currency would reduce the variability of the money multiplier and, thus, increase the predictability of the relationship between central bank money and nominal spending. Furthermore, the ratio of central bank money to broad money is reduced so that each unit change becomes more effective at the margin. 42 Goodhart (1989), page 271. 43 See Freedman (2000), Bindseil and Würtz (2006), White (2006), Sellon and Weiner (1997) and Woodford (2002). 44 Schmitz (2006b) provides a more detailed presentation and derivation of the ideas included in this section. 45 Irrespective of the fact that in extended netting systems private clearing and settlement institutions allow for the extension of settlement and the exchange of debt instruments (often highly liquid government bonds) as collateral in net payment systems to economise on central bank reserves, final settlement takes place in the generally accepted medium of exchange, eventually. 46 Schmitz (2006b).

References Bank for International Settlements (1990) Report of the Committee on Interbank Netting Schemes of the Central Banks of the Group of Ten Countries (Lamfalussy Report), Committee on Payment and Settlement Systems Publication No. 4. Bank for International Settlements (2001) Core principles for systemically important payment systems, Committee on Payment and Settlement Systems Publication No. 43. Bank for International Settlements (2003a) Policy issues for central banks in retail payments, Committee on Payment and Settlement Systems Publication No. 52. Bank for International Settlements (2003b) The role of central bank money in payment systems, Committee on Payment and Settlement Systems Publication No. 55. Bank for International Settlements (2005) New developments in large-value payment systems, Committee on Payment and Settlement Systems Publication No. 67. Bank for International Settlements (2006) Statistics on payment and settlement systems in selected countries – figures for 2004, Committee on Payment and Settlement Systems Publication No. 74.

252

S.W. Schmitz

Berentsen, A. (2006) ‘On the private provision of fiat currency’, European Economic Review, 50: 1683–98. Bindseil, U. and Würtz, F. (2006) ‘Payment systems from the monetary policy implementation perspective’, in S.W. Schmitz and G.E. Wood (eds) Institutional Change in the Payments System and Monetary Policy, London: Routledge. Browne, F.X. and Cronin, D. (1996) ‘Payment technologies, financial innovation, and laissez-faire banking: a further discussion of the issues’, in J.A. Dorn (ed.) The Future of Money in the Information Age, Washington, DC: Cato Institute. Centi, J.P. and Bougi, G. (2003) ‘The possible economic consequences of electronic money’, in J. Birner and P. Garrouste (eds) Austrian Perspectives on the New Economy, London: Routledge. Cohen, B.J. (2002) ‘Monetary instability: are national currencies becoming obsolete?’, in J. Busumtwi-Sam, M. Griffin Cohen, L. Dobuzinskis and S. McBride (eds) Turbulence and New Directions in Global Political Economy, London: Palgrave Macmillan. Costa Storti, C. and De Grauwe, P. (2003) ‘Monetary policy in a cashless society’, in M. Balling, F. Lierman and A. Mullineux (eds) Technology and Finance, Challenges for Financial Markets, Business Strategies and Policy Makers, London: Routledge, 241–60. Crede, A. (1995) ‘Electronic commerce and the banking industry: the requirement and opportunities for new payment systems using the internet’, Journal of Computer Mediated Communication, 1(3). Online, available at: http://jcmc.indiana.edu/vol1/issue3/ crede.html (accessed 1 November 2006). Einaudi, L. (1953) ‘The theory of imaginary money from Charlemagne to the French Revolution’, in F.C. Lane and J.C. Riemersma (eds), Enterprise and Secular Change: Reading in Economic History, London: Allen & Unwin. England, C. (1996) ‘The future of currency competition’, in J.A. Dorn (ed.) The Future of Money in the Information Age, Washington, DC: Cato Institute. European Central Bank (2006a) The Implementation of Monetary Policy in the Euro Area, Frankfurt am Main: European Central Bank. Online, available at: www.ecb.int/pub/pdf/other/gendoc2006en.pdf (accessed 1 November 2006). European Central Bank (2006b) Payment and Securities Settlement Systems in the European Union and in the Acceding Countries: Addendum Incorporating 2004 Data, Frankfurt am Main: European Central Bank. Online, available at: www.ecb.int/ pub/pdf/other/bluebook2006addenden.pdf (accessed 1 November 2006). European Commission (2003) Communication from the Commission to the Council and the European Parliament Concerning a New Legal Framework for Payments in the Internal Market (Consultative Document), COM (2003) 718. Federal Reserve System (2002) Alternative Instruments for Open Markets and Discount Window Operations, Washington, DC: Federal Reserve System. Federal Reserve System (2004) The 2004 Federal Reserve System Payments Study, Washington, DC: Federal Reserve System. Federal Reserve System (2006) Consultation Paper on Intraday Liquidity Management and Payment System Risk Policy, Washington, DC: Federal Reserve System. Fratianni, M. and Spinelli, F. (2006) ‘Did Genoa and Venice kick a financial revolution in the Quattrocento?’, Oesterreichische Nationalbank Working Paper No. 212. Freedman, C. (2000) ‘Monetary policy implementation: past, present, and future: will the advent of electronic money lead to the demise of central banking?’, International Finance, 3: 211–27. Friedman, B. (1999) ‘The future of monetary policy: the central bank as an army with only a signaling corps?’, International Finance, 2: 321–38.

Will central banking survive electronic money?

253

Friedman, B. (2000) ‘Decoupling at the margin: the threat to monetary policy from the electronic revolution in banking’, International Finance, 3: 261–72. Fry, M. (1999) ‘Risk, costs and liquidity in alternative payment systems’, Bank of England Quarterly Bulletin, 39(1): 78–86. Goodhart, C.A.E. (1988) The Evolution of Central Banks, Cambridge, MA: MIT Press. Goodhart, C.A.E. (1989) Money, Information, and Uncertainty, London: Macmillan. Goodhart, C.A.E. (2000) ‘Can central banking survive the IT revolution’, International Finance, 3: 189–209. Henckel, T., Ise, A. and Kovanen, A. (1999) ‘Central banking without central bank money’, IMF Working Paper No. 99/92. Humphrey, D.B., Sato, S., Tsurumi, M. and Vesala, J.M. (1996), ‘The evolution of payments in Europe, Japan, and the United States’, IMF Policy Research Working Paper No. 1,676. Kindleberger, C.P. (1984) A Financial History of Western Europe, London: Allen & Unwin. King, M. (1999) ‘Challenges for monetary policy: old and new’, Bank of England Quarterly Bulletin, 39(4): 397–415. Kobrin, S.J. (1997) ‘Electronic cash and the end of national markets’, Foreign Policy, 107: 65–77. Kohn, M. (1999) ‘The Origins of Western Economic Success: Commerce, Finance, and Government in Pre-Industrial Europe’, unpublished thesis, Dartmouth University. Online, available at: www.dartmouth.edu/~mkohn/ (accessed 1 November 2006). Krozner, R.S. (2003) ‘Currency competition in the digital age’, in D. Altig and B. Smith (eds) Evolution and Procedures in Central Banking, New York: Cambridge University Press. Krüger, M. (1999) ‘Towards a moneyless world?’, University of Durham Department of Economics & Finance Working Paper No. 99–16. Lacker, J.M., Walker, J.D. and Weinberg, J.A. (1999) ‘The Fed’s entry into check clearing reconsidered’, Federal Reserve Bank of Richmond Economic Quarterly, 85: 1–31. McAndrews, J. and Trundle, J. (2001) ‘New payment system designs: causes and consequences’, Bank of England Financial Stability Review, 11: 127–36. Menger, C. (1909) ‘Geld’, Handwörterbuch der Staatswissenschaften, Jena; trans. (2002) ‘Money’, in M. Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Payments Systems: From Barter to Electronic Money, Cheltenham: Edward Elgar. Palley, T.I. (2002) ‘The e-money revolution: challenges and implications for monetary policy’, Journal of Post Keynesian Economics, 24: 217–33. Payments Risk Committee (2003) Managing Payment Liquidity in Global Markets: Risk Issues and Solutions, New York: Payments Risk Committee. Online, available at: www.newyorkfed.org/prc/manage.pdf (accessed 1 November 2006). Prescott, E.S. and Weinberg, J.A. (2003) ‘Incentives, communication, and payment instruments’, Journal of Monetary Economics, 50: 433–54. Rip, A. and Kemp, R. (1998) ‘Technological change’, in S. Rayner and E.L. Malone (eds) Human Choice and Climate Change, Columbus, OH: Batelle Press. Schmitz, S.W. (2002a) ‘Carl Menger’s “Money” and the current neoclassical models of money’, in M. Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Payments Systems: From Barter to Electronic Money, Cheltenham: Edward Elgar. Schmitz, S.W. (2002b) ‘The institutional character of electronic money schemes’, in M. Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Payments Systems: From Barter to Electronic Money, Cheltenham: Edward Elgar. Schmitz, S.W. (2006a) ‘Models of worlds without money: a critical assessment of the

254

S.W. Schmitz

literature’, in S.W. Schmitz and G.E. Wood (eds) Institutional Change in the Payments System and Monetary Policy, London: Routledge. Schmitz, S.W. (2006b) ‘Monetary policy in a world without central bank money’, in S.W. Schmitz and G.E. Wood (eds) Institutional Change in the Payments System and Monetary Policy, London: Routledge. Schmitz, S.W. and Wood, G.E. (2006) ‘Institutional change in payment systems and monetary policy: an introduction’, in S.W. Schmitz and G.E. Wood (eds) Institutional Change in the Payments System and Monetary Policy, London: Routledge. Selgin, G.A. and White, L.H. (2002) ‘Mengerian perspectives on the future of money’, in M. Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Payments Systems: From Barter to Electronic Money, Cheltenham: Edward Elgar. Sellon, G.H. and Weiner, S.E. (1997) ‘Monetary policy without reserve requirements: case studies and options for the United States’, Federal Reserve Bank of Kansas City Economic Review, Second Quarter: 6–30. White, L.H. (2006) ‘Payment system innovations in the United States since 1945 and their implications for monetary policy’, in S.W. Schmitz and G.E. Wood (eds) Institutional Change in the Payments System and Monetary Policy, London: Routledge. Woodford, M. (2002) ‘Financial markets efficiency and the effectiveness of monetary policy’, Federal Reserve Bank of New York Economic Policy Review, 8(1): 85–94.

16 Payment systems and central banks Where are we now and where will e-payments take us? Charles Freedman1

This chapter focuses on two major issues. The first relates to the trade-off between risk containment, on the one hand, and cost or efficiency, on the other hand, in the design of large-value clearing and settlement systems for payments, securities, and foreign exchange. In the course of the discussion of this issue, we touch upon the role of central banks in the oversight or regulation of such systems and of retail payment systems. The second issue involves the implications for central banks and monetary policy of a world of electronic money.

Risk and efficiency in payment systems Over the past 25 years, there has been enormous progress in both industrialized countries and emerging economies in the design and construction of electronic clearing and settlement systems for large-value transactions in payments, securities, and foreign exchange. The motivation behind these developments has been twofold, with both elements related to the very rapid growth in the volume of financial transactions over the period. The first involved efficiency considerations: the necessity of developing electronic systems in order to avoid being overwhelmed by the volume of paper that would have been generated by the growth of transactions. In fact, it is clear that paper-based systems could not have supported the rapid growth that has been seen in financial transactions in a number of areas, such as securities markets. The second was the necessity from a public policy point of view of risk-proofing the emerging electronic systems in order to eliminate or at least minimize the systemic risk that would have resulted from the concentration of clearing and settlements in systems with potentially very large interbank exposures. The risk-proofing demanded by the authorities when building these systems or authorizing them (in cases such as Canada where the central bank did not build or own them) ended up being costly in a number of dimensions. One major cost arose from the requirement in almost all systems for the collateralization of the exposure of the system to the risk of failure of the largest participant or of all participants, depending on the system. A second, more recent cost resulted from the pressure for improved business

256

C. Freedman

continuity planning in clearing and settlement systems, especially in the lead-up to Y2K and following the September 11, 2001 terrorist attacks on the United States. Both of these costs can be attributed to the need to cope with the externalities for the financial system and the economy of a financial or operational failure in a clearing and settlement system. In the course of constructing or overseeing the design of clearing and settlement systems, central banks faced a challenge related to the trade-off between cost and risk mitigation. As William McDonough once commented at a meeting of the Committee on Payment and Settlement Systems (CPSS), there is no point in building a system that has been made so costly by the requirements of risk-proofing that no private financial institution is prepared to use it. In order to induce financial institutions to participate in such systems, it was therefore important to try to find ways of keeping down costs, particularly the cost of the collateral needed to minimize the exposure of the system to a participant default and to generate the necessary liquidity to settle the system in case of such a default. This was done in a number of ways. In the United Kingdom, liquid assets that the banking supervisor required to be held for prudential purposes were allowed to serve double duty as collateral underlying the RTGS system. In the United States, where for historical reasons collateral was not required for Fedwire transactions, even after the Fed acted to reduce its enormous intra-day exposures it did not impose collateral requirements except on financial institutions about which there were questions of creditworthiness. In Canada, rather than build an RTGS system, the financial institutions chose to build a Large Value Transfer System (LVTS) based on netting, but it was done in such a way as to have the same attributes of intra-day finality as an RTGS system, with payments through the system being unconditional and irrevocable. While the design of the LVTS introduced an element of survivors-pay in case of a participant failure (instead of the system being completely based on defaulter-pays), the rationale provided by participants for taking on this extra risk was that it would significantly reduce the collateral cost of operating the system.2 Other ways in which central banks cooperated with financial institutions to reduce collateral costs included: (i) providing interest-free intra-day liquidity (with the sole exception to my knowledge being the Swiss system prior to 1999); (ii) allowing the use of certain types of assets as collateral for which the banks had no alternative use and hence were virtually costless (such as mortgagebacked securities); (iii) being prepared in certain circumstances (such as Y2K and situations of extreme liquidity pressure) to accept assets as collateral that otherwise would not have been acceptable; (iv) in some cases accepting assets denominated in foreign currency that could therefore be used in more than one system; and (v) constructing the securities and payment systems to permit the transfer of collateral across the systems in such a way that the doubling of collateral requirements for the settlement of securities transactions in a type two system would be avoided. In addition, although there has not yet been much progress in this area, consideration has been given to the possibility of centraliz-

Payment systems and central banks

257

ing the holding of collateral so as to avoid the necessity of holding many different pools of collateral in different locations and in different currencies. The trade-off between risk-proofing and cost has shown up in the recent development of hybrid systems, where the offsetting of transactions is used to reduce the amount of collateral required from participants. In the case of securities clearing and settlement systems, the use of the purchased securities as collateral (or self-collateralization) has served to lessen the cost pressures of collateral requirements. And, in the case of CLSB, the system was cleverly constructed in such a way as to minimize the size of the cash positions in the various currencies needed to underpin the extremely high value of transactions without introducing risk into the system. Central banks, both separately and through joint efforts such as the preparation and distribution of CPSS reports, have played an extremely important role in stimulating and overseeing these developments. This was a natural consequence of the relationship between the structure of payment systems and the way that monetary policy is implemented. Equally important, once such clearing and settlement systems were in place and played a central role in the financial system, their risk-proofing had to be good enough that the failure of any participant could not, through exposures within the system, bring down the system or lead to multiple failures. Otherwise, the authorities might one day have been faced with the options of either watching the system fail, or preventing the failure of the system by bailing out the failing participant or the system itself. The option of allowing the failure of a system that was central to the financial sector and the economy is not acceptable, nor is the moral hazard associated with having to bail out participating institutions to avoid the failure of the system: hence, the necessity of risk-proofing the system. Another way of making this point is to consider the externalities for the surviving participants and for the financial system as a whole of the failure of a major participant in a key clearing and settlement system. The choice of the point on the trade-off curve between cost and risk containment in the case of systems dealing with large-value transactions emphasized the safety considerations and resulted in a very high degree of risk-proofing (albeit with attempts made to reduce the cost of collateral, as discussed earlier). In the case of retail systems with small-value payments, the authorities have typically not insisted on the same very high degree of risk-proofing, normally allowing the suppliers and users of such systems to choose the point on the trade-off curve themselves. The reason is that such systems do not carry systemic risk, since the values flowing through them are much smaller than those in large-value systems and the failure of a participant would be much less likely to bring down the system or other participants. At the same time, the volumes flowing through such systems are much larger, putting a premium on efficiency and cost considerations. More recently, the concept of system-wide risk has been added to that of systemic risk in discussing such systems. That is, even though a participant failure might not lead to collapse of a retail payment system or the failure of other

258

C. Freedman

participants, it might result in the loss of confidence in the system. For example, the inability of those members of the public who bank with the failed institution to access funds or make payments through the retail payment system might lead to the unwillingness of other transactors to use the system and might have a significant impact on the operation of the system, and thereby have deleterious effects on the economy. This still leaves open the question as to whether the responsibility for the oversight of retail payment systems should be assigned to the central bank or to some other authority (perhaps acting together with the central bank). On the one hand, given the importance of efficiency in such high volume, low value systems and given that there are other bodies responsible for competition and efficiency in most countries, it can be argued that the central bank should not be involved in overseeing such systems. On the other hand, some have argued that given the expertise of central banks in payments and other financial clearing and settlement systems, and given the linkages between most retail payment systems and the RTGS system or the central bank balance sheet, a case could be made for central banks taking some responsibility for the oversight of retail payment systems, in addition to their responsibility for large-value wholesale payment systems. In countries where central banks were given this responsibility, they would either have to develop expertise in competition issues or work jointly with the body responsible for competition issues. In deciding where the responsibility for overseeing such systems should lie, one should not forget that the operators and participants in such systems have strong incentives to protect them because, in case of serious problems in a system, their investment would be at risk and they would be subject to reputational risk.

Electronic money and central banks Any discussion of electronic money needs to begin with some practical points. In spite of the technical brilliance of some of the initiatives in the e-money sphere, such as Mondex in the case of stored value cards, and a number of network products of the digital cash type, in practice such products have not made much of an impression on the payments landscape. This is not a question of technical capability, but rather the lack of an economic space for such developments to break into the payments area, given the wide variety of payment instruments already available, including credit cards, debit cards, pre-authorized debits, automated credit transactions, and the venerable banknote.3 Indeed, in Canada, Mondex has ceased its limited trial operations, at least for the moment. The business case may simply not be there at this point of time for yet another payment instrument, given the costs of developing and promoting the spread of such a system. In analyzing the requirements for the effective implementation of monetary policy, it is important to understand both the structure of the payment system and the legal arrangements underlying the implementation of monetary policy. In Canada, the elimination of reserve requirements in the early 1990s led to

Payment systems and central banks

259

appreciable changes in the details of the way policy was implemented. And the introduction in the late 1990s of the electronic large-value payment system to replace the use of cheques for large-value transactions required even more significant changes in the mechanism for policy implementation. But neither development hampered the ability of the Bank of Canada to influence the overnight interest rate on which its policy actions are focused. One could generalize from this experience to suggest that central banks should have relatively little difficulty in adjusting their policy tools in response to changes in payment systems and/or the spread of electronic money, at least up to the theoretical point where their balance sheet totally disappeared. Bank for International Settlements (1996) came to the conclusion that while the spread of electronic money might require some changes in the techniques of policy implementation, it would not fundamentally alter the ability of central banks to conduct monetary policy. To quote a key paragraph from that research: The ability of central banks to achieve the desired level of very short-term interest rates should not be hampered in a significant way by the spread of e-money. However, techniques to adjust the supply of reserves or settlement balances might have to be modified if the amount of bank notes on the central bank balance sheet declined very substantially. In particular, central banks might have to issue other kinds of liabilities, such as central bank bills, to give them the resources to purchase securities for use in openmarket transactions. The chapter by Schmitz in this volume uses the traditional model in which the overnight interest rate is determined by the intersection of the demand curve and the supply curve for reserves or base money. While appropriate for some systems, such as those in place in the United States and in the Euro area, it is not particularly relevant for the corridor-type system that is now in place in countries such as Canada, Australia, and New Zealand that have eliminated reserve requirements. In the Canadian system, for example, as part of its standing liquidity facilities, the central bank stands ready to make whatever loans are needed by banks short of funds as a result of payment system outcomes at an interest rate of 25 basis points above the target policy interest rate, and to accept all deposits that banks wish to hold with it at an interest rate of 25 basis points below the target policy interest rate. Effectively, the infinitely elastic supply of loans and the infinitely elastic demand for deposits by the Bank of Canada at the prescribed interest rates force the overnight interest rate to fall within the corridor or band. There are (at least) two important aspects of this type of system. First, the supply and demand for base money (or settlement balances) becomes a secondary feature of the system, with the setting of the corridor interest rates the primary mechanism of central bank influence. Second, in some versions of the system, the total supply of settlement balances can be set equal to zero without, in any way, lessening the ability of the central bank to control the overnight

260

C. Freedman

interest rate. Indeed, in the case of the Bank of Canada mechanism, the overall supply of settlement balances was for some period of time actually set equal to zero. More recently, for reasons of operational efficiency, the banks have asked the Bank of Canada to supply a small amount ($50 million) of settlement balances at the end of each day. The question of whether a central bank can continue to influence the very short-term interest rate in a case where its balance sheet shrinks to zero because its liabilities are replaced by some form of e-money, as the means of final settlement in payment systems, has been the subject of some discussion and disagreement.4 I would argue that even in such unlikely circumstances the central bank could continue to influence the very short-term interest rate, although the mechanisms it could use might have an artificial element about them. One possibility would involve the central bank insisting on settlement of its own transactions on its own balance sheet and refusing to settle through alternative mechanisms. This type of arrangement would be strengthened if the central bank continued to act as banker for the government. A second possibility would involve a situation in which the central bank continued to establish a corridor for the overnight interest rate by providing standby facilities in which it was prepared to accept overnight deposits and to extend overnight loans at prescribed rates of interest. If the overnight rate of interest in the market were tending to decline below the central bank’s deposit rate, market participants would choose to hold overnight deposits with the central bank. Conversely, if the overnight rate were tending to increase above the top of the band, borrowing institutions would turn to the central bank for funds. The central bank would fund these loans by issuing its own liabilities (either marketable paper or deposits on its books). More fundamentally, as Woodford (2000) argues, the unit of account in a purely fiat system is defined in terms of the liabilities of the central bank. Because of this, the central bank can clearly define the nominal yield on overnight deposits in settlement accounts as it chooses. The special feature of central banks in this view is that they are entities, the liabilities of which happen to be used to define the unit of account in a wide range of contracts that people exchange with one another. Woodford goes on to argue that the question about such an arrangement is how much central banks’ monetary policies would matter. And that would depend on how many people still choose to contract in terms of the currencies, the values of which central banks continued to determine. That in turn would depend on the cost of transacting in different types of units of account. While all this is theoretically fascinating, it is very unlikely to become a practical issue in any foreseeable future. As long as central banks provide stable value for their currencies, it is likely that the public will continue to transact in those currencies, and alternative units of account with equally low transaction costs are not likely to develop.

Payment systems and central banks

261

Notes 1 I would like to thank Clyde Goodlet of the Bank of Canada for helpful comments on an earlier draft of these remarks. 2 It is worth noting that some central banks were critical of the LVTS arrangements since the LVTS differed in structure from an RTGS. This may have reflected a lesser concern with collateral costs than was the case in Canada. 3 Among the challenges to the spread of electronic money were the high initial fixed costs that would have been needed to establish the networks and the fact that the pricing of competing instruments might have made it more difficult to charge for e-money services (such as reloading cards, home readers, etc.). 4 The debate about the influence on interest rates of a central bank in a world in which e-money replaces central bank liabilities can be found in Friedman (1999) and the symposium in the July 2000 issue of International Finance.

References Bank for International Settlements (1996) Implications for central banks of the development of electronic money, Basel: Bank for International Settlements. Friedman, B.M. (1999) ‘The future of monetary policy: the central bank as an army with only a signal corps?’, International Finance, 2: 321–38. Woodford, M. (2000) ‘Monetary policy in a world without money’, International Finance, 3: 229–60.

Index

abstract units of account 238 account-based systems 3, 8–9 Admanti, A. and Pfleiderer, P. 103 agent: actions and cost 144–6; decisions 146–55; economic 23 agent-specific collateral costs 146–52 aggregate shocks 84 airway bill number (AWB) 223 Allen, F. and Gale, D. 179–80 Allen, H. et al. 5 Angelini, P. 163 anonymity 33 Arrow–Debreu model 76 asset 177; collateral 46, 53, 256; illiquid 46; liquid 5; settlement 16; ultimate settlement 10, 17, 19, 20, 34, 163 authority actions 221–6 automated clearing houses (ACHs) 37–8, 71, 130, 189, 219, 225; G10 models for intervention in 38 Bagehot, W. 20 bank: Boston 17; buyer’s 90; central 6, 7, 15–39; central see central bank; commercial 142; correspondent 7–8; Federal Reserve 21; foreign-owned 149; German 176; insolvent 179; Japanese 123; liquidity holdings 8; liquidity regulation 175–83; multilateral net position 126; opaqueness of assets 179; public regulation 181; pursuit of profit 18–19; second-tier 8, 141; seller’s 90; settlement 161, 162, 165, 167, 170 Bank of Canada 20, 54–5, 123, 260 Bank of Central African States (BEAC) 195 Bank of England xvii, 17, 19, 25, 149, 238 Bank for International Settlements 101, 103, 112, 239, 240, 242

Bank of Japan 87, 123, 195 Bank of Mexico 195 Bank of New York (BoNY) 25 bankers’ associations 124 banking: school 75; supervision 20; theory 68 bankruptcy law 141 Barings episode (1890) 19 barter economy 2, 16, 22 Basel Committee on Banking Supervision (BCBS) 176 Bech, M.L. 7, 138, 160–72, 189–203; and Garratt, R. 6, 50, 138, 163 behavioral economics 127 Belgium 32 Bellman equations 91–2 Biasis, B. et al. 112 bills of exchange 237 Bolt, W. and Humphrey, D. 26 Boston banks 17 Buiter, W.H. 33, 34 call market 102 Calomiris, C.W.; and Kahn, C.M. 18 Canada 32, 195, 258; Bank of 20, 54–5, 123, 260; banks 18, 54–5; Department of Finance 55; Large Value Transfer System (LVTS) 55, 195, 256; Mondex 258 Canadian Payments Association 123 card-based authorization 211 cash 33; substitute 8 cashless economy 85 Cavalcanti, R. 85; and Wallace, N. 48, 58, 75–85 Central Africa Economic and Monetary Community (CEMAC) 195 central bank 6, 7, 15–39, 45–55; collateral policy 142–55; commercially-oriented

Index 263 20; comparative advantage of 46, 51; costs 146–55; definition 46; end of day charges and remuneration 168–71; equilibrium strategies (end of day) 167; equilibrium strategies (mid-day) 169; foundation of 238; interest rate policy 160–72; intraday collateral policy 138–57; liabilities 19; mid-day charges and remuneration 168–71; modern 22; monetary policy 163–4; monopoly 47; pay-offs (end of day) 167; pay-offs (mid-day) 169; and payment systems 255–60; proto 18, 20–1; role in payment economics 68–72; role in payment systems 45–55; social costs 147–8, 151 Central Bank of West African States (BCEAO) 195 central banking and e-money 233–48 Central Securities Depository (CSD) as a central bank 35–6 central security depositories (CSDs) 178 Chakravorti, S. 138 Clearing House Interbank Payment System (CHIPS) 120–3, 127–31, 175, 190; participants 128 clearinghouse activities 69 cognitive psychology 127 Coleman, S.P. 171 collateral 142–6, 256–7; agent-specific costs 146–52; assets 46, 53, 256; costs 145; pooling 140 collateralization 140, 141, 145; full 139, 149, 153; zero 146, 147, 151 commercial bank 142 Committee on Payment and Settlement Systems (CPSS) 200, 256 commodity money 2 competitive market and the neoclassical model 49–50 competitive private systems 34 confidence, loss of 26 consumption, first period 58–9 contestable markets theory 50 continuous auction market 102 Continuous Linked Settlement (CLS) system 9, 34, 230–2, 240 controlling stake 27 Core Principles for Systemically Important Payment Systems (CPSS) 48 correspondent bank 7–8; hierarchy 51 cost: collateral 145; efficiency 255–8; monitoring 153, 154; private 147–8, 153, 165; private monitoring 144; social 147–8, 151, 153

credit: intraday 72, 122, 124–5, 138–9, 167, 172; lines 70, 179, 183; risk 88, 165; sales 237; short-term 46, 55; trade 59; transfer circle 210; transfers 209–10 cross-subsidization 219–20 crossing market 102 currency school 75 custodian sector 35 Dale, S. and Rossi, M. 160, 164 data storage restrictions 223 daylight overdrafts 119–20, 122, 124 dealer markets, inventory effects in 105 debit caps: bilateral 131; multilateral 131 default probability, low 149 default shock 143, 148 deferred net settlement (DNS) 6, 7, 87–98, 101–2, 121–32, 175, 192, 239; equilibrium 94; existence of DNS and RTGS equilibria 97; government failure 132; intraday payments 124; unwind 127–8 deposit transfers 237–8 deregulation 50 Diamond–Dybvig theorem 65 digital e-settlement stamp 213 direct debit 211 direct participation 147, 149 discount-window loan 123 dominant unit of account 244 double coincidence of wants 16, 22 Dowd, K. 34 Drehmann, M. et al. 33 e-banking 211, 223 e-billing 211, 222, 223 e-money 33, 233–48, 243–4, 258; institutional structure of schemes 243–4 e-payment 23, 255–60; standards 223 e-settlement 9, 202–3, 206–27; benefits 216–17; cost-advantage 217; credit transfer circle 210; digital encrypted stamp 212–14, 213; general layout 212–14; high security and availability 216; modules 213–15, 214 economic agents 23 Economic and Monetary Union (EMU) 192–3 economics of payments 2, 5; see also payment economics economy: barter 16; industrialized 45; of scope 49, 51, 52–4 efficiency, cost 255–8 electronic limit order books 112

264

Index

emergency liquidity assistance 176, 179, 181, 182 equilibrium analysis 92–5 equilibrium price 246 Euroclear 32 Europe, Western 192–3 European Central Bank (ECB) 193, 238 European Commission 227 European Union (EU) 192–3 Eurosystem 227 Faster Payments Service 31 federal funds: average rate 107–8, 108; trades 105–6, 106; transactions 107–11 Federal Reserve 20, 47, 52, 121, 122, 171, 238–9; Bank of Chicago 48; Bank of New York 191; banks 21; involvement in US payment systems 49; service provision 71 Fedwire 50–1, 61, 104, 120–1, 125, 171–2, 190, 256; intraday credits 122; value of transfers originated on 191 final settlement 23, 236, 246, 247 finance theory, modern corporate 179 financial asset, transfer 36 financial instability 22 Financial Sector Assessment Program (FSAP) 200 financial stability 20 fiscal theory 47, 48 Folkerts-Landau, D. 122, 131 foreign-owned banks 149 free banking market 18 Freedman, C. 255–60 Freeman, S. 46, 58, 163; model 46, 54 Freixas, X.: 179; and Parigi, B. 103 Friedman, B.M. 34 Fry, M.J. 87, 239 Fujiki, H. 67 Furfine, C.H. and Stehm, J. 139 Gale, D. and Allen, F. 179–80 game-theoretic model 6, 76, 163 Garman, M. 105 Garratt, R. and Bech, M. 6, 50, 138, 163 Germany, banks 176 globalization 202 globally-connected market 130–2 good funds 119 Goodfriend, M. 70; and Lacker, J. 179 Goodhart, C.A.E. 69 government: failure 129–30; securities 178 Green, E.J. 45–72, 47; and Todd, R. 17, 47, 49, 70

Greenfield, R.L. and Yeager, L.B. 35 Haldane, A.G. et al. 1–10 hard money 75 Harrison, S. et al. 139, 141, 165 He, P. et al. 4, 16, 88 Holmström, B. and Tirole, J. 181 Humphrey, D.B. 123, 128, 238; and Bolt, W. 26 illiquid asset 46 illiquid collateral 46, 52, 53, 55 immediate credit transfers 209–10 immediate liquidity 215–16 imperfect monitoring 78, 84, 150–5; and agent-specific collateral costs 150–2; impact 152; and spillovers under indirect participation 152–5 imperfections, market-based 129 indirect participation 149; spillovers under 152–5 industrial organization 57, 68, 100 industrialized economies 45 inside money 4 insolvent banks 179 institutional change: in monetary policy 241–2; in payments system 241–3; in payments system policy 242–3 Inter-American Development Bank 198 inter-bank payments 17–18, 87, 90–1; hierarchy 15 interbank network 214–15, 215 interbank payment systems 189–203 interest, negative economic provider 217–18 interest rate: alternative change in shape of yield curve 171; central bank policy 160–72; change in shape of yield curve 162; overnight market 160; positive intraday 171–2; zero intraday 163, 166–8, 172 intermediaries, payment 58–9, 62 internalization of payments 140; high degree of 148 international account number (IBAN) 208–9; common account number space 209 International Central Security Depositories (ICSDs) 178 International Monetary Fund (IMF) 200 intervention in payment systems: G10 models 37; models of 26–31; operation 27; oversight or regulation of system 27; ownership of system 27; public sector

Index 265 25–31; ranking models of 31; stylized models of 28 intraday collateral policy 138–57 intraday credit 72, 122, 124–5, 138–9, 167, 172; exposures 139; extensions 138, 149, 155 intraday liquidity 7, 239, 256 intraday overdrafts 165–6 inventory models 105–6 invisible goods 98 Jackson, J.P. and Manning, M.J. 138–57 Japan: Bank of 87, 123; banks 123 Japanese payment system 239 Kahn, C.M. 103; and Calomiris, C.W. 18; and Roberds, W. 3, 8, 103, 122, 138, 139, 142, 163 Kaufman, G. 129 key policy issues 6 King, M.A. 34, 48 Kiyotaki, N. and Moore, J. 35; and Wright, R. 16, 62 Kobayakawa, S. 163 Kocherlakota, N.R. 35; and Wallace, N. 84 Kohn, M. 237 Lacker, J. 5, 239; and Goodfriend, M. 179 Lagunoff, R. and Schreft, S.L. 64 Lamfalussy provisions 131 Lamfalussy report 200 large-value payment system (LVPS) 38, 64, 101, 104, 123, 141, 160, 177–8, 240 least-cost avoider principle 126 legal tender currency 33 Leinonen, H. 35, 206–27 lender of last resort (LLR) 176, 178, 181; interventions 179 Lester, B. et al. 87–98 liabilities 177 limit payment order system 102, 112, 113 liquid assets 5 liquidity: assistance 176; constraints 163; emergency assistance 176, 179, 181, 182; ex post 8; immediate 215–16; instruments of management 178–9; intraday 7, 239; macro-prudential motivation for regulation 181–2; market failures in provision of 179–80; microprudential motivation for regulation 181–2; policy 5; real-time management 230–2; regulation 175–83; risk 9, 161; risk sources 177–8; savings features 203; sinks 26; transfers 215

loan, discount window 123 loss of confidence 26 McAndrews, J. 100–13; and Rajan, S. 6, 50, 61, 103; and Trundle, J. 7, 239 McDonough, W. 256 macroeconomic shocks 8, 180–3 macroeconomic variables 4 macroeconomy 3, 4–5 Manning, M.J. and Jackson, J.P. 138–57; and Willison, M.D. 139 market: call 102; competitive 49–50; contestable 50; continuous auction 102; crossing 102; dealer 105; failure 25–6, 32, 71, 221–3; failure gap 222; free banking 18; globally-connected 130–2; microstructure 63–4, 110–13; private 127–8; security 101–4; sterling 150 market microstructure 63–4, 110–13; models 105–6 market-based imperfections 129 marketable securities 178 Martin, A. 163 Mengle, D.L. 122, 126 Mexico, Bank of 195 microstructure of financial markets 6, 110–13 Millard, S.P. 6, 7, 9, 87–98; and Saporta, V. 15–39 Mills, D.C. 84 mirage externalities 123–6 mobile phone airtime 8 model: Arrow–Debreu 76; game-theoretic 6, 76; good economic 57; moral-hazard 76; overlapping-generations 58; random-matching 62, 89; searchtheoretic 4, 7; settlement-friction 58, 60, 62, 64, 65 Mondex 258 monetarism 75 monetary authorities 130 monetary economics 57, 100 monetary policy 163–4; fundamentals 244–6; institutional change in 241–2 monetary stability 20, 22 monetary system 68 money: definition 22; electronic 33; microstructure of 100–13 money holdings 80–3 monitored person 79 monitoring: imperfect 78, 84; perfect 146–50 monitoring costs 153, 154 Moore, J. and Kiyotaki, N. 35

266

Index

moral hazard 70–1, 179; model 76 Morgan, D. 179 Morgan Stanley Central Bank Directory 198 Nash equilibrium 79, 167 natural pyramiding 16–18 negative economic provider interest 217–18 netting systems 6 network 208; externalities 26, 49, 218–19; interbank 214–15, 215 network-based payment systems 207–12; direct interbank communication 208 New Institutional Economics 234 New York 17–18; Bank of (BoNY) 25; banks 18; Clearing House system 20; Payments Risk Committee (PRC) 232; Stock Exchange 102 New Zealand, Reserve Bank of 123 Norman, B. et al. 7 null hypothesis 107 off balance sheet operations 177 open market operations (OMOs) 164 operational risk 33–4, 165, 170, 172 optimal settlement rules 87–98 Ostroy, J. 76 overlapping-generations model 58 overnight market interest rate 160 Pagano, M. 103; and Roell, A. 104 Pakistan, State Bank of 195 Parigi, B. and Freixas, X. 103 participant, second-tier 140 participation: direct 147, 149; indirect 149; in UK 149–50 Patinkin, D. 76 Pattinson, R. 230–2 payment: in bank 23; common interoperable standards 210–11; economics 57–8, 68–72, 100; electronic (e-payment) 23, 223, 255–60; flows 142–3, 143; formulation of better basic models 57–63; good economic model 57; industrial organization of 3, 5; intermediation 62; internalization 140, 148; primitive system 2; processing costs 222; processing speed 222; pyramid 16, 20, 35; regular 23; research 57–66; retail purchase 23; risk 64–6; size 24; substitute systems 24; types 24; wholesale 34, 119–32 PayPal 222

perfect monitoring 146–50; and agentspecific collateral costs 146–50 Pfleiderer, P. and Admanti, A. 103 planner 83–4 positive intraday interest rate 171–2 Prescott, E.S. and Weinberg, J.A. 239 pricing structures, psychological 219–20 private agents 124 private costs 147–8, 153 private information shocks 84 private market 127–8 private money 80 private monitoring costs 144 processed payment orders 6 production technology 235 Prompt Corrective Action 182, 183 proto central bank 18, 20–1 provisional payments 124 prudential regulator 32 psychological pricing structures 219–20 public sector intervention 25–31; see also intervention in payment systems pyramiding 16–18 Rajan, S. and McAndrews, J. 6, 50, 61, 103 random-matching model 62, 89 rational choice theory 127 real-time gross settlement (RTGS) 6, 7, 64–5, 87–98, 101–2, 138–57, 160, 166, 175, 189–203, 239; adoption in Africa 195, 196; adoption in Asia 194; adoption in Central America 196, 197; adoption in central banking 199; adoption in Europe (1995) 192; adoption in Europe (2005) 193; adoption in South America 196, 197; daylight overdrafts (RTGS-DO) 119–20, 122, 124; diffusion of 192–203; equilibria 94–5; existence of DNS and RTGS equilibria 97; future of 202–3; good funds (RTGS-GF) 119, 127; imported systems 201; s-curve and adopter groups 199; systems providers 201; tiering in 138–57 regression results 109–10 regular payments 23 Reichsbank 20 Reserve Bank of New Zealand 123 retail purchase payments 23 risk: containment 255–8; credit 88, 165; liquidity 9, 161; operational 165, 170, 172; payment 64–6; spillover 141, 152; system-wide 164–71; systemic 6, 7, 64, 126–8, 182, 239, 255; tiering 152–5

Index 267 risk-efficiency trade-off 7, 255–8 risk-mitigating actions 30 Roberds, W. 122; and Kahn, C.M. 3, 8, 103, 122, 138, 139, 142, 163 Rochet, J.-C. 175–83, 179; and Tirole, J.T. 122 Roell, A. and Pagano, M. 104 Rogers, E. 198 Rosenblat, T.S. 8 Rossi, M. and Dale, S. 160, 164 sales credit 237 Samuelson’s classic model 59 Saporta, V.: 1–10; and Millard, S. 15–39 Schanz, J. 5 Schmitz, S.W. 233–48, 243, 245, 259 Schoenmaker, D. 88, 127 Schreft, S.L. and Lagunoff, R. 64 Scott, H.S. 126 search-theoretic model 4, 7 second-tier banks 8, 141 second-tier participants 140 secured net settlement 127, 130–1 securities: government 178; marketable 178; settlement 178 securities settlement systems, G10 models for intervention in embedded payment systems of 39 securitizable loans 178 security markets 101–4 seignorage tax 47 Selgin, G.A. 88, 119–32; and White, L.H. 35 service providers 224 settlement: banks 161, 162, 165, 167, 170; deferred net (DNS) 6, 7, 87–98, 101–2, 121–32, 175, 192, 239; electronic (esettlement) 9, 202–3, 206–27; final 236, 246, 247; methodology 203; optimal frequency of 128; process 90; real-time gross (RTGS) 6, 7, 64–5, 87–98, 101–2, 138–57, 160, 166, 175, 189–203, 239; secured net 127, 130–1; securities 178 settlement asset 16; ultimate 10, 17, 19, 20, 34 settlement institution 17, 18–20; privatelyowned 20 settlement rules; optimal 87–98 settlement-friction model 58, 60, 62, 64, 65 Shi, S. 77 shock: default 143, 148; macroeconomic 8, 180–3 short-term credit 46, 55

single coincidence meetings 90 size of payments 24 Skeie, D. 104 Smith, B. 48 Smith, V. 18 social costs 147–8, 151, 153 social security systems 61 social welfare 96–8, 148 South African Development Community (SADC) 195 South African Reserve Bank (SARB) 195 specialization environment 77 Speight, G. et al. 160–72 spillover: risk 141, 152; systemic 141–2; under indirect participation 152–5 State Bank of Pakistan 195 Stehm, J. and Furfine, C.H. 139 sterling markets 150 Stock Liquidity Requirement (SLR) 8, 140 store-of-value systems 3, 8–9 stored value 58 substitute systems 24 Suffolk Bank system 17, 18 Swiss National Bank 20 system-wide risk 164–71 systemic risk 6, 7, 64, 126–8, 182, 239, 255; externalities 25, 36 systemic spillovers 141–2 technology: diffusion 198–202; innovations 202; production 235 telecommunications industry 53 Thornton, H. 19 tiering 138–57; impact of risk 153; impact of risk with high payment value 155; risk 152–5 time-line for actions 144 Tirole, J.T. 185; and Holmström, B. 181; and Rochet, J.-C. 122 Todd, R.M. and Green, E.J. 17, 47, 49, 70 trade: credit 59; federal funds 105–6, 106 trade meetings, second period 59 Trans-European Automated Real-time Gross settlement Express Transfer (TARGET) 193, 240 transfer: deposit 237–8; electronic 1; immediate credit 209–10; liquidity 215 Trejos, A. and Wright, R. 77 Trundle, J. and McAndrews, J. 7 ultimate settlement asset 10, 17, 19, 20, 34, 163 uncollateralized exposure 147, 150, 151, 154

268

Index

Uniform Commercial Code 130 United Kingdom (UK): Bank of England xvii, 17, 19, 25, 149, 238; Banking Act (1844) 19; government accounts 17; Office of Fair Trading 31; payment system participation 149–50; Peel’s Act (1844) 75 United States of America (USA): banking system 49; Census Bureau 63; Federal Reserve Act (1913) 21; inter-bank payments 17–18 unwind 127–8; zero risk 129 valuable monies 78 valuable private money 79, 80 value, stored 58 value-at-risk calculations 181 Van den Bergh, P. and Veale, J.M. 127 Vayanos, D. 103 Veale, J.M. and Van den Bergh, P. 127 Wallace, N. 35, 48, 58, 77; and Cavalcanti, R. 48, 58, 75–85; and Kocherlakota, N. 84 weakly implementable allocations 78–80 Weinberg, J.A. and Prescott, E.S. 239 welfare 98; social 96–8, 148

West African Economic and Monetary Union (WAEMU) 195 Western Europe 192–3 White, L.H. and Selgin, G.A. 35 wholesale payment 34, 119–32; private contracts versus government policy 132 Willison, M.D. 87–98, 112, 139, 160–72; and Manning, M.J. 139 Winston, C. 50 Woodford, M. 9, 34, 48, 163, 260; theory 47 World Bank 195, 200 world market infrastructure, links between 230 Wright, R. and Kiyotaki, N. 16, 62; and Trejos, A. 77 Yang, J. et al. 160–72 Yeager, L.B. and Greenfield, R.L. 35 zero collateralization 146, 147, 151 zero intraday interest rate 163, 166–8, 172 zero marginal costs 244, 247 zero unwind risk 129 zero-pricing convention 220 Zhou, R. 56, 65, 87, 163