Вы находитесь на странице: 1из 287

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 through-
out the world. This volume draws on wide-ranging contributions from promi-
nent 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

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 8 Organisational Change and

Lynn Bicker Retail Finance
An ethnographic perspective
2 Bank Deregulation and Richard Harper, Dave Randall
Monetary Order and Mark Rouncefield
George Selgin
9 The History of the Bundesbank
3 Money in Islam Lessons for the European Central
A study in Islamic political Bank
economy Jakob de Haan
Masudul Alam Choudhury
10 The Euro
4 The Future of European A challenge and opportunity for
Financial Centres financial markets
Kirsten Bindemann Published on behalf of Société
Universitaire Européenne de
5 Payment Systems in Global Recherches Financières (SUERF)
Perspective Edited by Michael Artis,
Maxwell J. Fry, Isaak Kilato, Axel Weber and
Sandra Roger, Elizabeth Hennessy
Krzysztof Senderowicz,
David Sheppard, Francisco Solis 11 Central Banking in Eastern
and John Trundle Europe
Edited by Nigel Healey and
6 What is Money? Barry Harrison
John Smithin
12 Money, Credit and Prices
7 Finance Stability
A characteristics approach Paul Dalziel
Edited by David Blake
13 Monetary Policy, Capital Flows 20 Central Banking Systems
and Exchange Rates Compared
Essays in memory of Maxwell Fry The ECB, the pre-Euro
Edited by William Allen and Bundesbank and the Federal
David Dickinson Reserve System
Emmanuel Apel
14 Adapting to Financial
Globalisation 21 A History of Monetary Unions
Published on behalf of Société John Chown
Universitaire Européenne de
Recherches Financières (SUERF) 22 Dollarization
Edited by Morten Balling, Lessons from Europe and the
Eduard H. Hochreiter and Americas
Elizabeth Hennessy Edited by Louis-Philippe Rochon
and Mario Seccareccia
15 Monetary Macroeconomics
A new approach 23 Islamic Economics and Finance:
Alvaro Cencini A Glossary, 2nd Edition
Muhammad Akram Khan
16 Monetary Stability in Europe
Stefan Collignon 24 Financial Market Risk
Measurement and analysis
17 Technology and Finance Cornelis A. Los
Challenges for financial markets,
business strategies and policy 25 Financial Geography
makers A banker’s view
Published on behalf of Société Risto Laulajainen
Universitaire Européenne de
Recherches Financières (SUERF) 26 Money Doctors
Edited by Morten Balling, The experience of international
Frank Lierman, and financial advising 1850–2000
Andrew Mullineux Edited by Marc Flandreau

18 Monetary Unions 27 Exchange Rate Dynamics

Theory, history, public choice A new open economy
Edited by Forrest H. Capie and macroeconomics perspective
Geoffrey E. Wood Edited by Jean-Oliver Hairault
and Thepthida Sopraseuth
19 HRM and Occupational Health
and Safety 28 Fixing Financial Crises in the
Carol Boyd 21st Century
Edited by Andrew G. Haldane
29 Monetary Policy and 37 The Structure of Financial
Unemployment Regulation
The U.S., Euro-area and Japan Edited by David G. Mayes and
Edited by Willi Semmler Geoffrey E. Wood

30 Exchange Rates, Capital Flows 38 Monetary Policy in Central

and Policy Europe
Edited by Peter Sinclair, Miroslav Beblavý
Rebecca Driver and
Christoph Thoenissen 39 Money and Payments in Theory
and Practice
31 Great Architects of Sergio Rossi
International Finance
The Bretton Woods era 40 Open Market Operations and
Anthony M. Endres Financial Markets
Edited by David G. Mayes and
32 The Means to Prosperity Jan Toporowski
Fiscal policy reconsidered
Edited by Per Gunnar Berglund 41 Banking in Central and Eastern
and Matias Vernengo Europe 1980–2006
A comprehensive analysis of
33 Competition and Profitability in banking sector transformation in
European Financial Services the former Soviet Union,
Strategic, systemic and policy Czechoslovakia, East Germany,
issues Yugoslavia, Belarus, Bulgaria,
Edited by Morten Balling, Croatia, the Czech Republic,
Frank Lierman and Hungary, Kazakhstan, Poland,
Andy Mullineux Romania, the Russian Federation,
Serbia and Montenegro, Slovakia,
34 Tax Systems and Tax Reforms Ukraine and Uzbekistan
in South and East Asia Stephan Barisitz
Edited by Luigi Bernardi,
Angela Fraschini and 42 Debt, Risk and Liquidity in
Parthasarathi Shome Futures Markets
Edited by Barry A. Goss
35 Institutional Change in the
Payments System and Monetary 43 The Future of Payment Systems
Policy Edited by Andrew G. Haldane,
Edited by Stefan W. Schmitz and Stephen Millard and
Geoffrey E. Wood Victoria Saporta

36 The Lender of Last Resort

Edited by F.H. Capie and
G.E. Wood
The Future of Payment

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)

List of figures x
List of tables xii
Notes on contributors xiii
Foreword xv
Acknowledgements xvi

General introduction: the future of payment systems 1


Payment systems and public policy 13

1 Central banks and payment systems: past, present

and future 15

2 The role of a central bank in payment systems 45


3 Some challenges for research in payments 57


4 Payment economics and the role of central banks 68


New approaches to modelling payments 73

5 New models of old (?) payment questions 75

viii Contents
6 Optimal settlement rules for payment systems 87

7 The microstructure of money 100


Current payment policy issues 117

8 Wholesale payments: questioning the market-failure

hypothesis 119

9 Central bank intraday collateral policy and

implications for tiering in RTGS payment systems 138

10 Central banks’ interest calculating conventions:

deviating from the intraday/overnight status quo 160

11 How should we regulate banks’ liquidity? 175


Policy perspectives on the future of payments 187

12 The diffusion of real-time gross settlement 189


13 E-settlement: soon a reality? 206


14 Real-time liquidity management in a

globally-connected market 230

15 Will central banking survive electronic money? 233

Contents ix
16 Payment systems and central banks: where are we
now and where will e-payments take us? 255

Index 262

1.1 Stylised models of intervention in payment systems 28

6.1 DNS equilibrium 94
6.2 RTGS equilibrium ( = 0) 94
6.3 RTGS equilibrium ( = 0.005) 95
6.4 RTGS equilibrium ( = 0.05) 95
6.5 Existence of DNS and RTGS equilibria 97
6.6 Welfare 98
7.1 Intraday pattern of activity: total value and volume of federal
funds traded 106
7.2 Fifth and 95th percentiles of the average federal funds rate
minus the target rate 108
9.1 Payment flows when C accesses the system via bank A 143
9.2 A time-line for actions 144
9.3 High degree of internalisation 148
9.4 Low default probability 149
9.5 The impact of imperfect monitoring 152
9.6 The impact of tiering risk 153
9.7 The impact of tiering risk with high payment value 155
10.1 Change in shape of yield curve 162
10.2 An alternative change in the shape of the yield curve 171
12.1 Value of transfers originated on Fedwire 191
12.2 Adoption of RTGS in Europe – 1995 192
12.3 Adoption of RTGS in Europe – 2005 193
12.4 Adoption of RTGS in Asia – 2005 194
12.5 Adoption of RTGS in Africa – 2005 196
12.6 Adoption of RTGS in South America 197
12.7 Adoption of RTGS in Central America – 2005 197
12.8 S-curve and adopter groups 199
12.9 Adoption of RTGS in central banking 199
13.1 Direct interbank communication in a network-based
infrastructure 208
13.2 The common account number space 209
Figures xi
13.3 E-settlement is part of the credit transfer circle, which provides
efficient electronic communications between participants in a
payment 210
13.4 The digital e-settlement stamp is part of the payment message 213
13.5 The digital encrypted stamp with central bank cover will
follow the payment message through the network 213
13.6 E-settlement stamps are produced by e-settlement modules,
which are closely integrated with banks’ payment systems 214
13.7 A dedicated interbank network connects all banks and the
central bank with each other for payment processing 215
13.8 The increasing market failure gap 222
14.1 Links between world market infrastructure 230

1.1 Ranking models of intervention 31

1.A1 G10 models for intervention in large-value payment systems 37
1.A2 G10 models for intervention in ACHs 38
1.A3 G10 models for intervention in the embedded payment systems
of securities settlement systems 39
7.1 Regression results 109–10
10.1 Pay-offs when central bank charges and remunerates at end of
day only 167
10.2 Equilibrium strategies when the central bank charges and
remunerates at end of day only 167
10.3 Pay-offs when central bank charges and remunerates at midday
and at end of day 169
10.4 Equilibrium strategies when the central bank charges and
remunerates at midday and at end of day 169
12.1 Imported RTGS systems 201
12.2 Liquidity savings features 203

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 Eco-
nomics 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
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 Commit-
tee; 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
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
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, Rout-
ledge (with Geoffrey E. Wood) and Carl Menger and the Evolution of
Payment Systems: From Barter to Electronic Money, Edward Elgar (with M.
George Speight. Senior Manager, Systemic Risk Reduction Division, Bank of
Neil Wallace. Professor of Economics, The Pennsylvania State University.
Matthew Willison. Economist, Systemic Risk Assessment Division, Bank of
Jing Yang. Senior Economist, International Finance Division, Bank of England.

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 formalisa-
tion of their financial and monetary stability objectives.
The role of central banks in payment systems has, however, changed signific-
antly 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
With these questions in mind, the Bank of England hosted a two-day confer-
ence 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 suc-
cessful 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

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 invalu-
able at various stages of the project. Of course, all remaining errors and omis-
sions 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, ‘Whole-
sale 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 ques-
tion 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 con-
siders 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? His-
torically, 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 require-
ments 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 com-
pletely 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 defi-
nitions 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 poten-
tially 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 institu-
tions. 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 well-
functioning payment systems, measured in terms of output, employment
and inflation?
• The industrial organization of payments: What is a suitable industrial struc-
ture 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 micro-
economics 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
But the payment system in He et al. (2005) is risk free. Two more imperfec-
tions 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 pay-
ments 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 implica-
tions 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 cost-
effective 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 fea-
tures are likely to imply a tendency towards concentration and a lack of
These potential externalities are discussed extensively and rigorously in the
chapters by Millard and Saporta, Green and Lacker. But the scale of these prob-
lems, and hence the appropriate solution to them, remain vexed questions. The
chapters by Green and Lacker argue that the importance of many of these exter-
nalities 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 sys-
temic 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 particip-
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 conta-
gious 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 pay-
ments 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’ equilib-
rium where all banks may delay payments and no banks save liquidity by so-
doing. 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 liter-
ature 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 pay-
ments (Part III)? As Millard and Saporta discuss, historically, central banks’
relationship with payment systems has extended beyond providing their liabili-
ties 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 apprecia-
tion of the ways in which different risks within a system – credit, liquidity, oper-
ational, 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 set-
tlement 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 sledgeham-
mer approach to cracking a nut. Moreover, although RTGS systems all but elim-
inate 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 frame-
work 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 set-
tling 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
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 reg-
ulate 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 limit-
ing 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 require-
ment’ 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 anony-
mous 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 whole-
sale 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 reg-
ulators have fully taken into account the systemic risk implications of this
increased interconnection between systems is, however, a moot point. In particu-
lar, liquidity risk – and the accompanying need for effective cross-border liquid-
ity management – may have increased inadvertently as a result of this increased
interconnection between infrastructures.
However wholesale payment systems develop, there are likely to be implica-
tions 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 corres-
ponds 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 ulti-
mate 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.

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 lan-
guage built up around them. With luck, readers might even feel spurred to tackle
some of these fascinating and important issues for themselves.

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
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.

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 set-
tlement 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’, unpub-
lished 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
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 con-
sequences’, 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
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
Past, present and future
Stephen Millard and Victoria Saporta1

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 influen-
cing 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 inter-
ested 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 liabili-
ties 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 safe-
keeping. The banks would then issue their customers with receipts. These
receipts represented a form of debt and, eventually, this debt became ‘transfer-
able’ 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 settle-
ment 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 man-
aging 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 ‘cus-
todian’ 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 Govern-
ment would step in to cover any resulting loss. For example, in 1890 the Chan-
cellor 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’ bal-
ances – 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 natu-
rally take this role at the top of the hierarchy? Does the market, in each cur-
rency, 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 settle-
ment 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 unit-
bank system – that is, a system consisting of a large number of small independ-
ent 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 inter-
est. 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 evid-
ence that suggests that when the provider of the ultimate settlement asset is also a
commercial bank, conflicts of interest ensue, especially during periods of finan-
cial 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 develop-
ment 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 com-
modity 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 fea-
tures (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 govern-
ment 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 responsibil-
ities 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 func-
tions imply for their modern interest, and active involvement, in payment systems?
Historically, privately-owned settlement institutions that supplied the settle-
ment 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’. Histor-
ical 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 commercially-
oriented 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 set-
tlement 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 estab-
lish 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 incen-
tives 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 specif-
ically 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 (includ-
ing 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 commer-
cial 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’ origin-
ally 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 pay-
ments to be made with ‘bank money’ – banknotes, bills of exchange and pay-
ments ‘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. Even-
tually, 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 ulti-
mate 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, securi-
ties purchases), payments that are made electronically or by paper, regular pay-
ments 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 con-
fidence 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 expo-
sures 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 pay-
ments 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 pay-
ments is what creates a demand on the part of the banks for the ultimate settle-
ment 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-eli-
gible 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 prob-
lems elsewhere in the financial system more generally – that is, financial insta-
bility – 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 Sep-
tember 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 func-
tions 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 ade-
quate 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 regula-
tion, 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

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 fail-
ures 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 finan-
cial 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 fail-
ures 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 indi-
vidual 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 indi-
vidual 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 charac-
terised 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 co-
operatives 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 prob-
lems; 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 pro-
vided 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 owner-
ship 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, imple-
mentation and operation of all, or a sub-set, of the elements such as software, hard-
ware, 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 main-
tains 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 com-
pliance with a set of minimum standards and design principles (e.g. the Core
principles for systemically important payment systems set by the Bank of Inter-
national 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 formal-
ity 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 com-
plete 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


Model 4 Model 5


Model 6 Model 3


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 inter-
vention 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 pro-
viding 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 identi-
fied 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 cir-
cumstances, oversight offers incremental value in achieving central bank object-
ives. 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 opera-
tion of the monetary and financial system are free from systemic risk externali-
ties. 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 underpro-
vision of a public good in the absence of negative externalities, a priori it
appears less appropriate when private sector solutions are plagued by negative
In addition, a central bank might not be indifferent between models with and
without operation. First, and most critically, outsourced operation of the infra-
structure raises the question of control over access to central bank money settle-
ment. 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 out-
sourced, the central bank can only control such exposure at arms’ length. Differ-
ent 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 con-
siderations 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 perform-
ance. 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 recov-
ery: oversight essentially constitutes taxation of the members. Finally, as noted
above, a model with mutual ownership might encourage the continued engage-
ment 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 over-
sight might expose the central bank to conflicts of interest, either apparent or
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
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 appropri-
ately 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 regu-
lator 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 address-
ing 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 benefi-
cial 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 opera-
tion, however, there may be a lower efficiency cost to public ownership. Taking
all public sector objectives together, an oversight-only model might be pre-
ferred, 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 Degree of effectiveness

in mitigating systemic in achieving additional
risk externalities public sector objectives

Model 1: owner/operator High/medium Low

Model 2: owner High/medium Medium
Model 3: operator Medium/low Medium/low
Model 4: overseer High/medium High
Model 5: overseer/operator High Medium/low
Model 6: laissez-faire 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 statutory-
based 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 introduc-
ing 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 large-
value 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 con-
tribution 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 pro-
tection 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 ten-
tatively 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 declin-
ing, 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, equiva-
lently, 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 pro-
vided 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. Green-
field 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 ensur-
ing 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. govern-
ment debt.26 Wallace (1988) makes the same point by noting that the acceptabil-
ity 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 Deposit-
ory (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, govern-
ments 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 broad-
brush 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, includ-
ing 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 regula-
tion 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
Tables 1.A1 to 1.A3 explore existing models for intervention in selected coun-
tries’ 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) Model 1 Model 2 Model 3 Model 4 Model 5

Australia HVCS ✓
Belgium ELLIPS ✓
Canada LVTS ✓
France TBF ✓
Germany RTGSplus ✓
Italy BI-REL ✓
Japan BOJ-NET ✓
Netherlands TOP ✓
Norway NICS ✓
New Zealand SCP ✓
Sweden RIX ✓
Switzerland SIC ✓
US Fedwire ✓

*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) Model 1 Model 2 Model 3 Model 4 Model 5

Australia BECS ✓
Belgium CEC ✓
Canada ACSS ✓
France SIT ✓
Germany RPS ✓
Italy BI-COMP* ✓
Japan Zengin ✓
Netherlands Interpay ✓
Norway NICS Retail ✓
New Zealand ISL ✓
Sweden BGC ✓
Switzerland DTA ✓

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

Country Domestic SSS(s) Model 1 Model 2 Model 3 Model 4 Model 5 Notes

Australia Austraclear ✓ Interfaced model

Belgium NBB Clearing ✓
CIK ✓ Interfaced model
Canada CDSX ✓ Settles through LVPS
Sweden VPC ✓ Settles through LVPS
France Euroclear France ✓ Integrated model
Germany Clearstream Banking AG ✓ Interfaced model
Italy Monte Titoli ✓ Interfaced model
Japan BOJ-NET JGB Services ✓ Settles through LVPS
JASDEC ✓ Settles through LVPS
Netherlands Necigef ✓ Interfaced model
Norway VPO ✓ Interfaced model
New Zealand Austraclear NZ ✓
Switzerland SWX ✓ Settles through LVPS
UK CREST ✓ 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
US Fedwire Securities ✓
DTC ✓ Cash transfers onto DTC accounts take
place through Fedwire
40 S. Millard and V. Saporta
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
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 sup-
plies 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,
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 mutually-
owned 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 ‘superimpos-
ing’ 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 infra-
structure 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 pay-
ments 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,
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
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 announce-
ment 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/
21 This is the current arrangement in the United Kingdom. A publicly available memo-
randum 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 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.
22 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 pre-
vented 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 estab-
lished (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.
23 Kahn and Roberds (2005) analyse this issue in the context of a model of a simple
economy with credit arrangements.
24 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.
25 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.
26 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 essen-
tially vanishes and fiscal and monetary policy would be, literally, the same thing.

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 import-
ant 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 eccen-
tricities, 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
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 macro-
economic 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 Eco-
nomics, 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 eco-
nomics’, 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.
Lacker, J. (2004) ‘Payment system disruptions and the federal reserve, following Septem-
ber 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’, Man-
chester 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,
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
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

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 industri-
alised 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 finan-
cial 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 dif-
ferent 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 insti-
tutional arrangement for meshing these various contributions functions essen-
tially 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 con-
ditions 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 con-
straint 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 overlapping-
generations 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

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 ser-
vices? However, a central bank is the natural candidate to play the macroeco-
nomic 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
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 eco-
nomics, 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 seignor-
age 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 mon-
opolist 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 discus-
sion 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 prin-
ciples that was submitted to the CPSS by the Federal Reserve Bank of Chicago
(2000). This contrast evidences a more fundamental distinction. The policymak-
ers’ 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
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 argu-
ments, 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 com-
mittee, 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 thou-
sands 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 occa-
sion 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 them-
selves 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 econo-
mists that it must have a monopoly position. In particular, there is no consensus
among economists that private clearinghouses should be prevented or discour-
aged 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 technolo-
gies 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 information-
transmission infrastructure. On the demand side, there are ‘network externali-
ties’ – 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 con-
testable markets, according to which fairly nonintrusive regulation supports effi-
cient 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, pre-
dicts 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 inef-
ficient 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 pay-
ments 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 trans-
actors, 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 mod-
elled 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 out-
comes. The gist of my preceding discussion of market structure is that this pre-
sumption 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 infla-
tion 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 busi-
ness 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 trans-
actions. 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 prin-
ciple 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 account-
management 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 effi-
ciently 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 dis-
count 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 opera-
tor 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 disquali-
fied 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 col-
lateral 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 dif-
ficult 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 protec-
tion regulation. Central banks seem to have no comparative advantage in regula-
tion 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 exam-
ines 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 regula-
tions 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 reg-
ulating others. There is a potential for conflict of interest. This potential is illus-
trated 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 reab-
sorb 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.

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 scepti-
cal 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 collat-
eral) 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 regu-
lates 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 insti-
tutions are able to provide services complementary to these, the social con-
sequence of broader central bank involvement in the payment system is likely to
be a marginal gain, at best.

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 sys-
temically 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
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 Mone-
tary 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 doc-
trine versus the quantity theory reconsidered’, Journal of Political Economy, 96, pages
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
Edward J. Green

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

1 Formulating better basic models.

2 Making market-microstructure data publicly available.
3 Providing sound advice about payment systems risk.
4 Understanding the relationship between payments and other business

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 opportun-
ities to enlarge the feasible gains;
• centralised, command-and-control institutions or all-encompassing contin-
gent contracts that would provide prior resolution of all decisions are infea-
sible; 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 feas-
ible 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 imple-
ment. The equilibria of various institutional frameworks can then be studied, and
recommendations about optimal institutions can be made in a way that is forth-
right 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 intermedi-
aries’ 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 interme-
diaries. Each of these models is now examined, to see how they fit the prin-
ciples 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. For-
tunately, 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 second-
period 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 contempor-
aries 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 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.
2 The trade credit has to be nominal – that is, settled in fiat money rather than
in a commodity.
3 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 intermedi-
ary 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 iden-
tical 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 inter-
mediation – provide further, accurate details about this basic fact that were men-
tioned 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, poten-
tially 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 eco-
nomic efficiency is to facilitate the operation of the payment system by regulat-
ing 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 settlement-
friction 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 coexis-
tence 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 settlement-
friction 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 con-
tinuum 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 assump-
tion 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 ver-
sions of the model – adapted in ways such as specifying 30 or 40 periods (inter-
preted 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 opera-
tions 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 co-
ordinate their settlements directly, and that therefore make it socially valuable
for Fedwire to grant intraday credit to the banks that use it. To analyse convinc-
ingly 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 intermedi-
aries 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 multi-
lateral 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’
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 interme-
diary 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 accept-
able 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

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 circum-
stances, ranging from ‘triple witching days’ in markets for cash-settled deriva-
tives 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 main-
taining 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 safe-
guards. 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 man-
agement 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 payment-
economics 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 counter-
party 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 counter-
parties 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 collat-
eral. 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. Think-
ing about collateral in terms of the Diamond–Dybvig theorem (along lines sug-
gested 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 liq-
uidity 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 partic-
ipant with very low transaction costs might be able to profit from doing some-
thing 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 daylight-
trading 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 under-
standing the benefits of net settlement and on the transaction costs of payment
float. Yet, for moderate-size commercial payments as well as for household pay-
ments, 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 pur-
poses 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 under-
standing what drives the choices that transactors and their intermediaries make
among payment options.

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 Mon-
etary Economics, 38, pages 101–115.
Freeman, Scott and Paula Hernandez-Verme (2004) ‘Default and fragility in the pay-
ments 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 eco-
nomics, 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 implica-
tions. 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 effect-
ively 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 emerg-
ing 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 set-
tlement 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 consis-
tent with efficiency – that is, with Pareto optimality. Central bank intervention
sometimes alters the distribution of net benefits among payment system particip-
ants. 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 trans-
fer of central bank account balances.
The second core idea is that many, if not most, of the private sector institu-
tions 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 wide-
spread 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, opera-
tional 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 institu-
tions 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 insur-
ance 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 opera-
tions, 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 docu-
ment that claim, except to note the extent to which access to central bank settle-
ment 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 monopo-
lized interbank settlement. In this, I find Green and Todd (2001) persuasive
when they argue that the rationale for more extensive central bank service provi-
sion depends on the extent to which there are economies of scope between addi-
tional 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 ques-
tion 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 external-
ity, 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 pro-
vides 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 provid-
ing 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 con-
ditions 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 collateraliza-
tion 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.

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
Lacker, J.M. (2004) ‘Payment System Disruptions and the Federal Reserve After Septem-
ber 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
Ricardo Cavalcanti and Neil Wallace1

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 seem-
ingly 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 pro-
duction 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 essen-
tially 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 bor-
rowing 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, there-
fore, 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 moni-
tored 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-the-
bearer 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 con-
sequences 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 consis-
tent 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 valu-
able 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 weak-
ening 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 specialization-
type 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. Through-
out, 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 moni-
tored. 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
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 con-
sidered.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
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
Given a suggested allocation, the following game is played. Consider a pair-
wise 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 indi-
vidual defections. For example, we are ruling out permanent autarky for the
entire economy as a response to an individual defection. Notice that our defec-
tion 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
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 imple-
mentable 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 representative-
agent 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 distri-
bution 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 con-
sumers is an arbitrary part of the allocation and will be discussed below.) More-
over, 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 distribu-
tion 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 dis-
counted 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

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 start-
ing 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 par-
ticipation 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)), partici-
pation constraint (equation (4)) implies participation constraints (equation (5)).
Next, we describe necessary conditions for duplicating the above consump-
tion and production pattern without private money. In order to duplicate the
pattern, each monitored person must begin a period with outside money. Other-
wise, 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 allo-
cation, 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 para-
meters 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
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 meet-
ings – 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 alloca-
tion with the same positive output in all meetings except those in which the
unmonitored producer has money is not implementable. Given such an alloca-
tion, 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 immedi-
ately 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

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 mis-
leading. 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 unmon-
itored 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
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 unmoni-
tored producer than does an unmonitored consumer in any single-coincidence
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 recogniz-
ability. 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? Presum-
ably, 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.

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 intro-
duced 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.

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 manage-
ment 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, Minneapo-
lis, MN: University of Minnesota Press.
Hammond, P. (1987) ‘Markets as constraints: multilateral incentive compatibility in con-
tinuum 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 record-
keeping 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 dis-
sertation, Pennsylvania State University.
Ostroy, J. (1973) ‘The informational efficiency of monetary exchange’, American Eco-
nomic Review, 63: 597–610.
Patinkin, D. (1951) ‘The invalidity of classical monetary theory’, Econometrica, 19:
Shi, S. (1995) ‘Money and prices: a model of search and bargaining’, Journal of Eco-
nomic 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 Proce-
dures 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

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 opera-
tional 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 large-
value 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 func-
tioning 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 partici-
pant 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 set-
tlement (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 pay-
ments. 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 provid-
ing 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 settle-
ment on CHIPS – at the time a DNS system – to those of an otherwise equival-
ent 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 equi-
libria 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 interest-
bearing 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 impli-
cation 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 settle-
ment rules. This assumption partly derives from the first – sellers cannot use
prices to induce buyers to switch settlement rules – but goes further by prohibit-
ing 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 ques-
tion 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 dis-
cussing equilibria within it. We then compare welfare over regions of the para-
meter 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.

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 mod-
elled 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.

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 sub-
periods. 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 cer-
tainty. However, the timing of the settlement process is crucial if there is a pos-
sibility 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 con-
sequently 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 trans-
ferred 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 inter-
bank 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
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
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
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 + ]

– (5)
r((1 + r) – r) + xP( ,)

[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 + ]

– (6)
r((1 + r) – r) + xP( ,)

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 increas-
ing 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 0.1 x 0.2

Figure 6.1 DNS equilibrium.




0 0.1 x 0.2

Figure 6.2 RTGS equilibrium ( = 0).

Optimal settlement rules for payment systems 95




0 0.1 x 0.2

Figure 6.3 RTGS equilibrium ( = 0.005).




0 0.1 x 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 equilib-

rium 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 reduc-
tions in  would expand the range of parameter values for which there exist both
RTGS and DNS equilibria.
96 B. Lester et al.
In this section we will analyse the level of social welfare under the two settle-
ment 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 com-
patibility 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 – )
(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 ration-
ality and incentive compatibility conditions hold. The values of 1 and 2 are

rx(1 – )[(1 – M1)u + M1c]

1 = – a
(1 – M1)xu – [(1 – M1)x + r((1 + r) – r]c
2 = –a.

It is straightforward to show that

(1 – M1)x
(<)2 if u (>)c.
r + (1 – M1)x
Optimal settlement rules for payment systems 97

RTGS Autarchy


DNS and

Min {1, 2} 

Figure 6.5 Existence of DNS and RTGS equilibria.

When c is relatively low, the relevant constraint on  is the banking con-

straint; the cost of settlement must be sufficiently small so that banks can con-
tinue to pay interest on deposits. Alternatively, when c is relatively high, the
relevant constraint are the seller’s incentive constraints because  must be suffi-
ciently 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 equilib-
rium 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.


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 equilib-
rium, 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 interme-
diate 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 coordi-
nate on a welfare-superior settlement rule and any role for public policy in pre-
venting 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
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 ().

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
Bank of England (2006) Payment systems oversight report 2005, London: Bank of
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
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

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 insti-
tutional arrangements for the exchange of financial instruments. This study has
examined the efficiency of different market structures. Various measures of effi-
ciency 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 altern-
ative 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 inven-
tory 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 queue-
augmented system, the importance of links to ancillary payment systems, dis-
count 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 exam-
ining 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 con-
trast, 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 charac-
terized. 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 compar-
ison 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 deliv-
ered outside of the payment system. For example, the title to a building or secur-
ity 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 accumu-
lated 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 dis-
count 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 equi-
libria can occur, however, call markets can assist in selecting a particular clus-
tering 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 dis-
cussed 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) con-
sider 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 fragmenta-
tion 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 pay-
ments 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 borrow-
ing 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 price-
dependent. 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 con-
straints 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. Has-
brouck and Sofianos (1993) and Madhavan and Smidt (1993) show mean rever-
sion 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 14
1,600 Value 12

Volume (thousands)
US$ (billions)

1,000 8

800 6

0 0
12:30 2:30 4:30 6:30 8:30 10:30 12:30 2:30 4:30 6:30
a.m. a.m. a.m. a.m. a.m. a.m. p.m. p.m. p.m. p.m.

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).
Values shown are the average values in a calendar minute calculated over the days of the sample

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 pur-
chases 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 Mendel-
son (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 posi-
tion 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 supple-
ment 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 bid-
ask 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 bid-
ask 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
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

5th percentile
95th percentile

Average midpoint target (%)





12:30 2:30 4:30 6:30 8:30 10:30 12:30 2:30 4:30 6:30
a.m. a.m. a.m. a.m. a.m. a.m. p.m. p.m. p.m. p.m.

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).
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
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 devia-
tions in its transactions balance, both at 4.00 and 4.30 p.m.

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 Midpoint – Deviation Midpoint – Deviation

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

Transaction –0.456 –0.075

balances (0.242)* (0.146)
Deviation from –0.426 –0.068
mean transaction (0.248)* (0.146)

January 0.613 0.633 –1.179 –1.179

(0.73) (0.731) (0.509)** (0.509)**
February 0.407 0.427 –0.38 –0.379
(0.731) (0.731) (0.518) (0.518)
March –0.116 –0.092 –1.463 –1.461
(0.729) (0.729) (0.507)*** (0.507)***
April 0.844 0.846 –0.396 –0.394
(0.739) (0.739) (0.513) (0.512)
May 1.232 1.234 –0.826 –0.825
(0.739)* (0.739)* (0.508) (0.508)
June 0 0 0.398 0.4
(0) (0) (0.521) (0.521)
July 0.137 0.13 0.435 0.435
(0.749) (0.749) (0.513) (0.513)
August 0.598 0.587 –0.286 –0.286
(0.78) (0.78) (0.521) (0.521)
September 0.787 0.777 0.046 0.047
(0.775) (0.775) (0.524) (0.524)
October 0.784 0.779 –0.754 –0.755
(0.729) (0.729) (0.504) (0.504)
November 1.87 1.874 0 0
(0.759)** (0.759)** (0) (0)
December –0.817 –0.817 –1.031 –1.031
(0.741) (0.741) (0.513)** (0.513)**
Pre-holiday –3.57 –3.559 –3.712 –3.713
(0.79)*** (0.79)*** (0.589)*** (0.589)***
Post-holiday 4.939 4.939 3.794 3.793
(0.822)*** (0.822)*** (0.579)*** (0.579)***
Good Friday 0.134 0.081 5.672 5.671
(2.967) (2.967) (2.639)** (2.639)**
1st Thursday 0.435 0.437 –0.295 –0.295
(0.667) (0.667) (0.481) (0.481)
1st Friday –4.364 –4.366 –4.361 –4.361
(0.68)*** (0.68)*** (0.471)*** (0.471)***
1st Monday 0.466 0.473 –1.021 –1.021
(0.685) (0.685) (0.472)** (0.472)**
1st Tuesday –3.87 –3.869 –3.578 –3.577
(0.667)*** (0.667)*** (0.469)*** (0.469)***
1st Wednesday –2.617 –2.613 –3.683 –3.682
(0.677)*** (0.677)*** (0.476)*** (0.476)***
110 J. McAndrews
Table 7.1 continued

Dependent Midpoint – Deviation Midpoint – Deviation

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

2nd Thursday –1.556 –1.555 –1.381 –1.381

(0.691)** (0.691)** (0.468)*** (0.468)***
2nd Friday –2.22 –2.217 –1.934 –1.933
(0.669)*** (0.67)*** (0.488)*** (0.488)***
2nd Monday 0 0 0 0
(0) (0) (0) (0)
2nd Tuesday –2.153 –2.157 –2.927 –2.927
(0.686)*** (0.686)*** (0.479)*** (0.479)***
2nd Wednesday –0.468 –0.466 –2.101 –2.101
(0.675) (0.675) (0.475)*** (0.475)***
First of month 3.987 3.986 6.458 6.459
(0.806)*** (0.806)*** (0.526)*** (0.526)***
Last of month 9.207 9.214 8.382 8.384
(0.818)*** (0.819)*** (0.581)*** (0.581)***
End-of-quarter –3.398 –3.401 –0.276 –0.275
(1.617)** (1.617)** (1.082) (1.082)
Fed funds target –12.078 –12.101 –2.746 –2.746
increase (2.978)*** (2.979)*** (1.879) (1.879)
Fed funds target 11.512 11.533 16.586 16.588
decrease (1.75)*** (1.75)*** (1.868)*** (1.868)***
Constant 1.525 –145.181 3.108 –139.205
(0.73)** (0.722)*** (0.507)*** (0.506)***
Observations 1640 1640 2648 2648
Adjusted R2 0.21 0.21 0.23 0.23

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 doub-
ling 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), Hamil-
ton (1996), Lee (2003) and Demilralp et al. (2004), most of the reserve mainte-
nance 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 mid-
point 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 estima-
tion. 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 effi-
cient 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 transac-
tions) 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

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, depend-
ing 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 over-
drafts, may require continuous trading in money markets to assist banks in main-
taining 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 techno-
logy 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 com-
munication and computation that make these technologies feasible and economi-
cal 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 develop-
ment 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 Inter-
national shows. These links have assisted banks in managing their positions in dif-
ferent 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.

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 mid-
point 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 effi-
cient 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 develop-
ments 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.

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):
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-versus-
payment systems in which both a security and account balances are exchanged
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

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 inven-
tory’, 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, Commit-
tee 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 micro-
foundations, 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:
Hasbrouck, J. and Sofianos, G. (1993) ‘The trades of market makers: an empirical analy-
sis of NYSE specialists’, Journal of Finance, 48: 1565–93.
Ho, T. and Stoll, H. (1981) ‘Optimal dealer pricing under transactions and return uncer-
tainty’, 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 inven-
tory 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 con-
sequences’, Bank of England Financial Stability Review.
Pagano, M. (1989). ‘Trading volume and asset liquidity’, Quarterly Journal of Eco-
nomics, 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 Eco-
nomic Studies, 66(2): 219–54.
Willison, M. (2005) ‘Real-time gross settlement and hybrid payment systems: a compari-
son’, Bank of England Working Paper No. 252.
Part III

Current payment policy

8 Wholesale payments
Questioning the market-failure
George Selgin1

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

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 Settle-
ment (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 Settle-
ment (RTGS-GF) systems require their members to possess clearing balances or
‘good funds’ sufficient to cover all payments, while less strict versions (RTGS-
DO) 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, includ-
ing the Fed, have actively promoted the change both by recommending or man-
dating 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 ini-
tiated in 1994, the frequently made claim that unregulated DNS systems suffered
from an analogous but independent externality problem overlooks crucial differ-
ences 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 them-
selves, 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

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 pay-
ments. 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 intra-
day credit, including the risk of settlement failure, were therefore borne approxi-
mately by the Federal Reserve System (which might find itself holding
Fedwire-generated claims on a failed institution) and ultimately by the general
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). Nonethe-
less, 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 spon-
sored 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 offi-
cials’ 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 mon-
etary 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 intra-
day 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
The tendency of regulators and economists to treat traditional or ‘unsecured’
DNS systems as close cousins of Fedwire, with its reliance upon free or under-
priced intraday credits or ‘daylight overdrafts,’ is evident throughout the liter-
ature 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 over-
drafts 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 particip-
ants to credit risks as they extend large volumes of payments-related intra-
day 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, contrac-
tual agreements would generally call for a payments ‘unwind.’ An unwind com-
mences 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 can-
celled. 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 wholesale-
payments network.6 Economists and policy-makers therefore had to rely on sim-
ulations to assess the likely consequences of a DNS-system unwind. According
to Humphrey’s (1986) influential simulation, the failure of a major CHIPS par-
ticipant, 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 recom-
mendations, 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 Pay-
ments 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 collateral-
ized 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 sug-
gests 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 Fedwire-
coloured glasses.
Putting those glasses aside allows us to get an undistorted picture of tradi-
tional, voluntary contractual arrangements of commercial banks and private
clearinghouses, where by ‘voluntary’, I mean free from pressure, edicts, or (sub-
sidized) competition from government agencies. As noted earlier, the over-
whelming historical tendency has been for financial firms to establish
clearinghouses that rely on unsecured DNS to handle both small- and large-
value 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 clear-
inghouse, ‘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 settle-
ment 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
• 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 tradi-
tional 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 benefi-
ciaries 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 over-
look 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 credit-
worthiness as well as receiving banks’ low estimate of the probability of a settle-
ment 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 inter-
vene 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 beneficia-
ries’ 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 suffi-
cient 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. Econo-
mists’ 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 activ-
ities, they may also have been exposed to losses resulting from other particip-
ants’ 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 arrange-

ment 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 settle-
ment. 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 expecta-
tions 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 collat-
eral 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
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 partici-
pating 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 pos-
sibility 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 pre-
diction concerning the direction of the bias, and therefore supplies little guid-
ance 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 aug-
menting 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 prob-
able than ever, if banks could not bring themselves to contemplate the possibil-
ity 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 real-
time 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 bene-
ficiaries’ 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 institu-
tions (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 industri-
alized 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 con-
taining 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 argu-
ments 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 pre-
sumably 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 uncer-
tain 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 pay-
ments 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 pay-
ments. The CHIPS exception, however, allowed its members to revoke pay-
ments 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 multi-
lateral 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 pay-
ments less efficient. Nor does the fact that many DNS systems, including
CHIPS, who have adopted CAPS voluntarily, necessarily contradict this conclu-
sion. 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, surviv-
ing participants are required to share responsibility for its net settlement obliga-
tions. 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 pre-
viously, 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 fail-
ures, 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 private-
market 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 cus-
tomers; 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 spon-
sored 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 guar-
antees 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.

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 sugges-
2 In January 2001, CHIPS switched from deferred net settlement to near-continuous
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 valu-
able for its excellent critique of models (e.g. Freeman, 1999) that claim to demon-
strate 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 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!
5 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
6 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.
7 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.
8 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.
9 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
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 con-
nection 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
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 complica-
tions. 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 dis-
cussed 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 releas-
ing 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 particip-
ants 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 par-
ticipates in a DNS network indirectly, via a correspondent, because it is not reason-
able 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 gener-
ally 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 partici-
pant, 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 fre-
quent 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 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.
18 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 presum-
ably go far in further reducing the (already low) probability of a DNS settlement
19 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 occur-
ring following a single participant’s failure was 96 percent. This was again assuming,
following Humphrey, zero recovery of provisional (intraday) credit to account
20 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.
21 For a more general discussion of the incentive advantages of net settlement see
Roberds (1999) and Kahn et al. (2003).
22 In personal correspondence Roberds acknowledges the inherent risk-control advan-
tages 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.’
23 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 pre-
miums] were too great.’
24 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).
25 From 1857 to 1907, when interbank lending markets were poorly developed,
members of CHIPS’ predecessor, the New York Clearing House Association, oper-
ated an effective coinsurance scheme by agreeing to accept fully collateralized clear-
inghouse ‘loan certificates’ in lieu of gold for settlements during financial
emergencies (Timberlake, 1984).
26 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).
27 Dale (1998: 229) points out that the very involvement of central banks ‘in the settle-
ment 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 con-
tained by means of settlement arrangements in which central bank involvement ‘is
least evident.’
28 In CHIPS this particular externality problem was partly avoided by linking each par-
ticipant’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
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
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
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 Eco-
nomics, 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:
Humphrey, D.B. (1986) ‘Payments finality and the risk of settlement failure,’ in A. Saun-
ders and L.J. White (eds) Technology and the regulation of financial markets, Lexing-
ton, 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):
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: effi-
ciency, 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,’ unpub-
lished 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

In this chapter we present a model of a Real-Time Gross Settlement (RTGS)
payment system with tiered membership where settlement is facilitated by intra-
day credit extensions from the central bank. RTGS systems process and settle
payment instructions individually in real time, ensuring intraday finality. Fur-
thermore, 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 settle-
ment has taken place. The model allows us to examine the key factors that influ-
ence 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 collater-
alization 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 exten-
sive 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 signific-
ant, 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 sub-
stantial 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 pre-
funding 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 expo-
sure 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 provi-
sion 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 sug-
gests 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
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 phenom-
enon, known as ‘tiering’, is a commonly observed feature in many RTGS
systems internationally. CHAPS Sterling, with only 13 out of around 350 com-
mercial 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 col-
lateral 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 associ-
ated 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 diversifi-
cation, 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
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 intra-
day 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 set-
tlement 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 obliga-
tion 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 indi-
rect 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 cost-
efficiency 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 maxi-
mized 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 frame-
work. We then apply this framework to analyse agents’ decisions under altern-
ative 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 set-
tlement members of the payment system, with A also a potential provider of cor-
respondent 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 corre-
spondent banking services provided by direct payment system participant, A. Its
set of potential actions is then {D,I}, where D = direct participation; and I = indi-
rect 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 oblig-
ation 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 pre-
vents 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 col-
lateral 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 econo-
mize on collateral sought from C in respect of such a credit extension by carry-
ing 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


C has an obligation to A: C has an obligation to B:
Probability  (1e) Probability  e

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


Period 1: C decides Period 3: C ’s default

whether to fulfil its shock is revealed to
payment obligation directly agents. Intraday credit
in the payment system; repaid if no shock has
or indirectly via A. occurred.

Figure 9.2 A time-line for actions.

default shock has occurred, collateral will be sought from C in respect of intra-
day 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 mon-
itoring 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
Should C suffer a shock in period 3 and consequently fail to repay its intra-
day 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 ser-
vices: 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 participa-
tion 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 col-
lateral 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 result-
ing 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

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 proba-
bilities 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 147
C > M + C + e(1 – )A ⇒ C >  + eA (1)
(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 com-
parison 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 (which-
ever 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 collater-
alization, 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.
(Z,D) Social
Social/private cost

(F,D) Social  Private

X 0.0012
(F,I ) Social  Private
(Z,I ) Social  Private
(Z,D) Private 0.0002
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

(F,D) Social  Private

Social/private cost
(F,I ) Social  Private 0.0012

Y (Z,D) Social 0.0010

(Z,I ) Social  Private 0.0004
(Z,D) Private 0.0002
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 moni-
toring 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 collater-
alization 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 correspon-
dent 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 sub-
section, we relax this assumption to allow for Type I and Type II errors in moni-
toring. 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 participa-
tion. 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 par-
ticipation. 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

(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 partici-
pation. 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 pre-
ferred 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 moni-
toring 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 encour-
aged 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

(Z,D) Social

Social/private cost
(F,D) Social  Private
X 0.0012

(F,I ) Social  Private 0.0010

(Z,I ) Social  Private
(Z,D) Private 0.0002
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

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 partici-
pation is an equilibrium choice for C only when the central bank adopts a strat-
egy 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
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 liquid-
ity 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 pre-
ferred 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 increas-
ing 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 associ-
ated with full collateralization. Hence, it would never be optimal for indirect
participation to be chosen by C, limiting the potential incidence of tiering exter-
nalities, 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.

(F,I ) Social 0.0019
(F,D) Y
Social  Private
Social/private cost

(F,I ) Private
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 poten-
tial uncollateralized exposures and hence potentially greater social losses.
More formally, bank C’s and the central bank’s expected costs under full col-
lateralization 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

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 implica-
tion 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

(F,I ) Social Y
(F,D) Social  Private 0.0085

X 0.0080

Social/private cost


(F,I ) Private


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 stand-
ards 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 opportun-
ity 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 pay-
ments 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.

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 collateral-
ization 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
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 partici-
pation 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 moni-
toring 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) ensur-
ing the capacity of correspondent banks to absorb either capital losses or liquid-
ity 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 treat-
ment 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 col-
lateral 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.

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 Willi-
son, 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 facili-
tate 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 rel-
evant to the issues considered in this chapter), the cost of generating cash liquidity is
significantly reduced via the implementation of self- or auto-collateralization tech-
niques. CREST and Euroclear France, for instance, apply such procedures, allowing
the immediate pledge/repo of (eligible) securities to the central bank to generate liq-
uidity 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 sim-
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 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.
13 Consistent with empirical observation, we assume that the central bank either does
not have the capacity, or finds it excessively costly, to monitor.
14 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.
15 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.
16 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.
17 It is worth noting that, under full collateralization, the central bank’s and Bank C’s
expected costs are the same.
18 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.
19 It is thought that upwards of 20 per cent of sterling large-value payments are internal-
ized across the accounts of correspondent banks.
20 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 liquid-
ity requirements in encouraging concentration is not relevant in a number of juris-
dictions. Finally, some countries have historically always had highly concentrated
banking systems (this is true of the United Kingdom) while in others highly frag-
mented banking arrangements are observed.
21 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.
22 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.

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 real-
time 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 immedi-
acy’, 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 com-
parison’, 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

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 Real-
Time 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 bal-
ances 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 pay-
ments would provide them with the liquidity they needed to make their pay-
ments, rather than having to borrow from the central bank. Of course, if all
banks followed the same strategy none would actually gain from delaying pay-
Payment delays are not desirable because they can imply the following risks
and inefficiencies in the payment system:

1 the failure to receive payments in a timely manner may represent a liquidity

risk to recipients;
2 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;
3 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

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 para-
digm, 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, cus-
tomers 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, pay-
ments 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 dura-
tion-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 calcu-
lating interest twice a day may mean that settlement banks’ charges to their cus-
tomers 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 Time

half day of day

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 experi-
ment, 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 offset-
ting. 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 inter-
est is calculated. Finally, we consider some further issues before offering some
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 rela-
tionship 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 con-
straints 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 intra-
day 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 settle-
ment 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 con-
straint’, 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-of-
day 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 macroeco-
nomic implications, e.g. for the speed or effectiveness of the transmission of
monetary policy. The inflation control mechanism would essentially be
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 pay-
ments 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
• 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 pay-
ments on the bank’s own account (e.g. for funding, trading or risk manage-
ment 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
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
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. Pay-
ments 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 knock-
on 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 set-
tlement 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
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 remuner-
ates at the end of the day, banks may not actively monitor customers’ actual bal-
ances 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 set-
tlement banks into doing the same, can ease this bank-imposed monitoring con-
straint 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 cus-
tomers 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 speci-
fies 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 collateral-
ized. 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 set-
tlement 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 intra-
day, 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

Morning Afternoon
Bank 1 Morning C,C C + E,D
Afternoon D,C + E 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 Morning C and E < D

Afternoon Afternoon C and E > D
Morning or afternoon Afternoon or morning E<D<C
Morning or afternoon Morning or afternoon C<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 cus-
tomers, 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 intra-
day 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 cus-
tomer 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 cus-
tomer 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 bal-
ances, 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 pay-
ments 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 annual-
ized 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 pay-
ments 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 pay-
ments 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 settle-
ment 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 con-
dition 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 equilib-
rium 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

Bank 2

Morning Afternoon
Bank 1 Morning C,C C + E + R,D – R
Afternoon D – R,C + E + 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 Morning C and E < D – R

Afternoon Afternoon C and E > D – R
Morning or afternoon Afternoon or morning E<D–R<C
Morning or afternoon Morning or afternoon C<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 intra-
day 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 com-
monly 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 pay-
ments 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
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-of-
day deadline. Here they are spread across two deadlines. In some sense then the
systemic risk is lower, since an operational incident which affected the function-
ing 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 bal-
ances 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 cus-
tomers’ 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 settle-
ment 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
Slope reflects
expectations and
No change in policy risk premium of
rate possible until next future rate changes
scheduled decision

CB policy rate
Intraday rate

End of End Next possible Time

half day of day policy change

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 inter-
preted 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.

Central banks typically set a zero interest rate intraday and impose their mone-
tary policy rate overnight. This chapter undertakes a thought experiment to con-
sider 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 pay-
ments 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 pay-
ments 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 provid-
ing 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.

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 stan-
dard 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

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 Eco-
nomic 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’, Inter-
national 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 immedi-
acy’, 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 compari-
son’, 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’
Jean-Charles Rochet1

While the last 20 years have witnessed a remarkable trend towards harmon-
ization 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 rela-
tion to the stock of liquid liabilities), while others are based on mismatch analy-
sis (i.e. limiting the gaps between expected inflows and outflows of cash for
short-term maturities). Several countries (including Australia, Germany, Singa-
pore 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 inter-
bank 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 co-
ordinated 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 pru-
dential 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 coun-
tries (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 exception-
ally adverse conditions? Theoretical results and common sense suggest that
liquidity requirements are a natural complement (or partial substitute?) to sol-
vency 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) facil-
ity. A cost–benefit analysis of the LLR is thus needed to determine the appropri-
ate 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, espe-
cially 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 cir-
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

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 credit-
rationed 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.

There is also large uncertainty about the amount of withdrawals of deposits
(including wholesale) or the renewal of rolled-over interbank loans. This is espe-
cially so when the bank is under suspicion of insolvency, when there is a tempo-
rary (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 deriv-
ative 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 inter-
vene 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

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), addi-
tional liquidity requirements may be needed since CSDs can be viewed as essen-
tial 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 securi-
ties 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

Marketable securities
Such securities (equities, interbank loans) can be sold easily in normal circum-
stances 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 appro-
priate 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 stra-
tegic 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 co-
ordination failures.

Macroeconomic shocks and systemic issues

Some form of government intervention may be needed in case of macroeco-
nomic 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 ques-
tion and shows that ex ante regulation of banks’ liquidity may be a way to miti-
gate 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 devalu-
ation or a terrorist attack.
How should we regulate banks’ liquidity? 181
This dichotomy between two types of rationales is clear for solvency regula-
tions, 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 macro-
prudential aspect to ex ante regulation.
Similarly, liquidity regulations can be justified by micro- and macro-
prudential 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 inter-
vene ex post during a banking crisis. To limit this tendency, liquidity require-
ments 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 meas-
ures 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) regu-
lation, i.e. whether banks could regulate themselves, like participants in a clear-
ing 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 organ-
ize, 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 inter-
bank 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. Liq-
uidity 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 quali-
fication that undercapitalized banks could be subject to more stringent require-
ments. 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 liq-
uidity needs is, of course, related to the intraday needs of the banks for chan-
nelling 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 co-
operation 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 macroeco-
nomic 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 author-
ity. 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.

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 fail-
ures, 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 author-
ity. 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 inter-
ventions would be useful in order to evaluate the exact costs of liquidity provi-
sion by the central bank, and the social costs of excessive liquidity.
184 J.-C. Rochet
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
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 sys-
temic 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 dis-
count 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 eli-
gible at the Bank of England.

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:
Bank of England (2003) ‘Strengthening financial infrastructure’, Bank of England Finan-
cial 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):
Froot, K. and Stein, J. (1998) ‘A new approach to capital budgeting for financial institu-
tions’, 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 cur-
rency 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 Finan-
cial 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):
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
Morten L. Bech

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 well-
functioning economy. Historically, central banks have played an active role in
the payment system. They continue to do so both as a provider of payment ser-
vices 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
Hence, in the last couple of decades many countries have modified the settle-
ment 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 transi-
tional 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 technolo-
gies and discuss the key drivers of this evolution. We briefly discuss the emer-
gence of two alternative technologies and speculate on how the future of central
bank operated interbank payments systems might look in terms of settlement
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 expo-
sures.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 precon-
dition for the successful conduct of monetary policy as well as financial stability.
In the United States, there are two principal systems that settle interbank pay-
ments: 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 disburse-
ment 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.
The diffusion of real-time gross settlement 191

600 90

500 75

400 60

Multiple of GDP
$ trillion

300 45

200 30

100 Fedwire  CHIPS (RHS) 15

0 0
1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005

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 pay-
ments 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 pay-
ments 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 imple-
mented 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 imple-
ment 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 Eurosys-
tem is currently developing the next generation of TARGET, TARGET2.
TARGET2 will dispose of the national RTGS systems and run on a single tech-
nical 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 mem-
bership 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 imple-
ment 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 opera-
tional in the first quarter of 2007. The adoption of RTGS is well under way in
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 implementa-
tion 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 con-
tinue 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 nova-
tion.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 elec-
tronic 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 modern-
ize 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 imple-
ment 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 techno-
logical change emphasize two aspects of technology. First, technology is non-
rival 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 facili-
tates 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 com-
pared 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 adop-
tion 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
Members of a social system have different capacities to adopt a new techno-
logy. The capacity can, in an international context such as this, reflect different
stages of economic development or underlying cultural attitudes toward techno-
logical 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 Opinion Early Late Laggards

(2.5%) leaders majority majority (16%)
(13.5%) (34%) (34%)

Figure 12.8 S-curve and adopter groups.



Number of central banks
Adoption (%)

60 100




0 0
1970 1980 1990 2000 2010 2020

Figure 12.9 Adoption of RTGS in central banking.

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
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 multi-
currency 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 explic-
itly cite the recommendations put forward by the CPSS as a reason for imple-
menting 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 vulner-
abilities 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 techno-
logy 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 con-
sequence of improved communication technology, lower transportation costs
and multicultural interactions.
It is outside the scope of this article to provide statistical evidence for local-
ization 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 competi-
tion 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 Montenegro South Africa Romania

ECB Oman Namibia Georgia
Ireland BCEAO Malawi Mauritius
Azerbaijan Macedonia Zimbabwe Kuwait
Hungary Moldova Zambia Bulgaria
Turkey Pakistan Uganda Barbados
Latvia Libya Sweden Netherlands
Bosnia and Herzegovina Algeria Antilles
India BEAC Ghana
United Arab Emirates Albania
United Kingdom Bahamas
France Tanzania
Monaco Angola
Andorra Guatemala
Saudi Arabia
New Zealand

Source: Company websites and newswires.

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, typ-
ically, 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 con-
tinue 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 com-
ponents 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 non-
synchronized 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
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

Liquidity savings features RTGS+ LVTS BI-REL PNS CHIPS

Queue management Priority level x x x

Reordering x x x
Multiple payment streams x x
Bilateral limits x x
Bilateral offsetting x x x x
Multilateral netting Full x x x x
Partial 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.

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 methodol-
ogy 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 pre-
diction. Moreover, some anecdotal evidence was found of clustering in the dif-
fusion 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
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 Settle-
ment (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 pro-
posal 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, Equato-
rial 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
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 Switzer-
land. 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 pay-
ments 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.

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), Com-
mittee 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 set-
tlement 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 San-
tomero, A.M., Viotti, S. and Vredin, V. (eds) Challenges for Central Banking, Dor-
drecht: 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 Septem-
ber 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 con-
sequences’, 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

Current payment clearing conventions, although now using electronic automa-
tion, 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 competi-
tors 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 objec-
tive 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 bar-
riers 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 conven-
tions 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
Processing costs should fall drastically for electronic payments. The costs
of telecommunication, IT processing and data storage are all expected to con-
tinue to decrease. Payments are, from the IT-processing point of view, quite
simple transactions. The data volume per transaction and the processing com-
plexity are low compared with other types of documents and processing on the
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 inter-
faces and processing capacities, which are sufficient for all kinds of payment
Efficient electronic straight-through-processing interfaces between banks and
their customers could automate the processing of payments, not only for corpor-
ate 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
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 ser-
vices 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 decen-
tralised. 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 mes-
sages. By extending the decentralised network concept of the internet to pay-
ments 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 pay-
ments 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 e-
mail system banks have e-mail/mailbox servers. In a network-based payment
system, banks will need payment/account servers, otherwise the concept is very

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 conven-
tion 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 GSM-
telephone 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

Bank 2
Bank 1 Payment server
account TCP/IP
server Internet
network Payment

Bank n
Bank 3 Payment/
Payment/ account
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 dif-
ferent 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 bene-
ficiary concerning a payment to be made. (2) The payer then sends the instruc-
tion 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 transporta-
tion legs than electronic credit/debit card payments, direct debits or electronic
210 H. Leinonen

Payer’s bank Including e-settlement Beneficiary’s bank


Customer- Bank-to-
to-bank customer Statement
E-banking receipts


1 Beneficiary

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 real-
time 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 credit-
push 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 standard-
ised formats. (The current credit transfer, direct debit and card payment mes-
sages 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 pro-
vided by the authorisation modules in the merchant’s/payee’s payment ter-
minal and payee’s card.

The payment data necessary for the different processing conventions are identi-
cal. 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 end-
user processing. The billing data would accompany the payments and vice versa
through the complete end-to-end process in order to enable a combined auto-
matic payment, invoicing and accounting process both by the payer and the
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 mes-
saging 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 con-
tinuous 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 possi-
bilities 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 infra-
structure 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
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.

Bank 1 Bank 2

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 Network
Bank account account Bank
payment platform Payment platform payment
system E-settlement network
E-settlement system
module module
Customer Customer
account 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 net-
works used by banks. E-settlement is especially suitable for large volumes as
there are no centralised bottlenecks and can therefore be used both for large-
value and low-value payments.

Interbank network
A dedicated interbank network (Figure 13.7) is needed to link together all partic-
ipating banks and the central bank, for the purpose of processing payments.
In a network-based payment system, the most essential element is the inter-
bank 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 pro-
cessing. 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 215

Bank 1 Bank 2 Bank 3 Bank 4



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 pay-
ments 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 posit-
ive 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 con-
tinuously 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 situ-
ations 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 compli-
cate 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 distrib-
uted – must be well secured. A digital/electronic version of the four eyes prin-
ciple has to be implemented.
Two different controlling programs and two different encrypted and tamper-
resistant 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 pos-
session. 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 distrib-
uted 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 mod-
ernise the settlement conventions at the same time.
The decentralised network-based model facilitates direct real-time communi-
cation. 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 secur-
ity elements and the settlement function could be developed for the payment
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
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 partici-
pating banks and the payment flows must be sufficiently large.
The e-settlement approach will also reduce settlement risk, because all settle-
ments 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 decen-
tralised systems are more robust than centralised systems.

Delaying and promoting development factors in payment

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 coun-
tries, 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 indi-
vidual 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 dif-
ficult 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 domin-
ant 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 cus-
tomers 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 provid-
ing 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 gener-
ally shift the current accepted balance of benefits and costs for example; more
rapid processing will decrease the float income more for some system particip-
ants 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 competit-
ive market. Having to cross-subsidise free payment services would make it diffi-
cult 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 cus-
tomers moving their deposits to another bank.
The payments industry has also introduced different kinds of common inter-
change 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 pay-
ments 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 transporta-
tion 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
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 nega-
tively, 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, opera-
tional 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 introduc-
ing 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 dif-
ficult 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 devel-
opments 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 dif-
ferent 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 pay-
ments need rapid e-payment services with different payment/delivery guarantee
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 inter-
net-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.


Current level

Efficiency and stability gap

Possible level


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 com-
pletely. 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 integ-
rated 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 prefer-
ably operate on an international level. Good benchmarks can be found in other
industries. For airway cargo the airway bill number (AWB) identifies all deliv-
ery notes, invoices, etc., referring to a given cargo. All airfields and flights have
clear international identifiers. Even the normal surface parcel mail uses inter-
national 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
As customers’ payment processing is computerized and most private cus-
tomers 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 multina-
tional 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 special-
ised payment instruments for given environments and usages lacking the possi-
bilities for establishing general purpose payment schemes attracting large
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 diffi-
cult 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 train-
ing 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
Banking regulations generally limit deposit taking to banks. As most pay-
ments are made out of different kinds of deposit accounts with or without over-
draft 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
Providing payment services requires access to interbank clearing and settle-
ment 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, geo-
graphical requirements and complicated technical standards. Interbank settle-
ment 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
The interchange fees and surcharge prohibitions result in extensive cross-
subsidising and build barriers for new entrants and solutions. Abolishing inter-
change 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 monop-
olistic conditions. There is generally only one ACH per country although some-
times 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 competi-
tion 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 imple-
mented by all participants within a given timeframe and in a harmonised
manner. The regulatory approach can solve these kinds of issues. As a compari-
son, 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 telecom-
munication 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 other-
wise become a barrier for general service development. Overall efficiency can
only be reached in payment processing when all infrastructure parties are
employing technology efficiently.

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 techno-
logy. 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 altern-
ative 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 pres-
sure 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 struc-
ture 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 SWIFT-
net. 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.

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
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 connect-
ing airline companies, travel agencies and airports for bookings, ticketing and check-
in is another good example of an interoperable distributed network with numerous
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
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/
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).

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 pay-
ments, 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
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
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 pay-
ments: costs, pricing and payment choice’, Journal of Money, Credit and Banking, 33:
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 Euro-
pean 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

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

ng s r k

ec A Ov
di m o
ra te tw

tro lte er
te s ns s

Tr c c tiv co
un ing tio ge
r t ys ne

ad om e
ni r na the
co rad ica han

iti m tra nte

on u di r
he t n xc

al ic g tra
r t ive mu l e


ex at sy di
ve at m a

ch ion st ng
O er n co tion

an s em
Al onic rad


ge ne s
tr T

s tw










EuroCCP Monte Titoli

Clearstream CHAPS BI REL
Euroclear LIPS-Gross BOGS/ CSD

CLS VPS 2nd wave
CLS currency





ng s r k
ec A Ov


di m o
ra te tw
tro lte er


te s ns s
Tr c c tiv e co

un ing tio ge


r t ys ne
ad om e
ni r na th

Austraclear CCAS

co d a an
iti m tra nte
on u di r

he t n xc
al nic ng tra

r t ive mu l e
ex at sy di

ve at m a

ch ion st ng

O rn o n
te c itio

an s em

Al onic rad

ge ne s

tr T
s tw



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 transac-
tions, known by the term ‘Herstadt Risk’.
CLS bank service is the first global infrastructure to be simultaneously con-
nected 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 inci-
dent 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 diffi-
cult 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 transac-
tion 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 repu-
tational, 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 rep-
resenting 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 Pay-
ments 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 Settle-
ment 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 ration-
ally 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 communica-
tions bridge to facilitate information exchanges during times of stress will be enor-
mously beneficial to all market participants.

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 phys-
ical 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 differ-
ent 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 poten-
tial 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 concep-
tual 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

The conceptual approach is based on the systematisation of the central institu-
tional 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 dif-
fusion of e-money schemes has to derive the necessary and sufficient conditions
for a transition from one generally accepted medium of exchange and the associ-
ated unit of account to another and the effects of the diffusion of new technolo-
gies 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 trans-
ition 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 inter-
bank 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 mainte-
nance 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 com-
mercial 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 incen-
tive 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 inter-
ests 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 set-
tlement). 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 politico-
economic 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 inno-
vations 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 gover-
nance 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 devel-
opment and adoption of specific payments technologies and the complementary
institutions at the firm, the market and at the payments system level are endoge-
nous to the politico-economic tensions that drive institutional change in the pay-
ments 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 gener-
ally accepted medium of exchange in the economy as well as the characteristics
of the interbank clearing and settlement institution.13 The latter include con-
ditions 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 relation-
ship 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 concentra-
tion 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 frame-
work: 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
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 pre-
requisite 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, interre-
gional 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 (whole-
sale) 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 mer-
chant 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
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 trans-
ferred 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 alternat-
ives 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
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 polit-
ical 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 considera-
tions (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 circula-
tion of euro notes in 2002 were based on widespread technologies, but represen-
ted 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, nation-
wide 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 regu-
latory constraints (i.e. branching restrictions), the resulting low concentration,
and the involvement of the Federal Reserve in cheque clearing and the subsidis-
ation 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-to-
business 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 previ-
ously 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, globalisa-
tion 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 – ‘Trans-
European 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 dif-
ferences 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
The demand for international large-value payments and for liquidity bridges
across systems and across borders increases. As a response to regulatory con-
cerns 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 transac-
tions (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 over-
draw 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 inter-
national 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 multic-
urrency 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
• New market entrants (e.g. mobile phone companies, telecommunication
operators, net-based scratch card companies) are often particularly inno-
vative, 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 rele-
vant payment systems are eventually linked to central bank money via the
banking system.

Institutional change in the payments system and monetary

The impact of the institutional characteristics of the payments system on mone-
tary 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 large-
value 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 rela-
tive 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 discre-
tion 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 pay-
ments system. In recent history central banks have demonstrated their determi-
nation 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 institu-
tional 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 liter-
ature 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 characteris-
tic 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. Ini-
tially, 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 pur-
chasing power, in principle. The balance sheet of central banks will shrink rela-
tive to a world without electronic money, which is mainly a positive sign as
institutional change in the payments system (e.g. electrification of retail pay-
ments 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 corre-
spondingly. 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 rel-
evant 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 avail-
able to central banks, because they are public authorities with certain regulatory
competencies transferred to them by the respective legislature. These competen-
cies are independent of the loss of central banks’ monopoly to issue the gener-
ally 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
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 pro-
posal are misplaced. In principle, these central banks already possess the neces-
sary regulatory authority and discretion to impose reserve requirements and to
charge and pay interest thereon and only minor adaptation to the current frame-
work 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 addi-
tional 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
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 insti-
tutional change and have proved to do so effectively in the past. The politico-
economic 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 tran-
scend 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 elec-
tronic 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 mone-
tary policy implementation in a world without central bank money. It builds on
the regulatory competencies of central banks, which do not depend on its mon-
opoly 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 effi-
ciency 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 cap-
abilities 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

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 Wal-
rasian 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, reliabil-
ity 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
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? 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 address-
ing 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.
9 Fry (1999) page 82.
10 McAndrews and Trundle (2001).
11 See Rip and Kemp (1998) for a discussion of sociological, philosophical and eco-
nomic concepts and theories of technological change.
12 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.
13 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).
14 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.
15 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 formalisa-
tion 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 transac-
tion from the credit transaction.
16 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.
17 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 transfer-
able, the current holder has the right to sue only the initial creditor but not the assign-
ers. 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.
18 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.
19 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.
20 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.
21 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.
22 Although abstract units of account reduce transaction costs in a setting of many circu-
lating 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; Kindle-
berger (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 curren-
cies 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 propos