Академический Документы
Профессиональный Документы
Культура Документы
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
policy.
List of figures x
List of tables xii
Notes on contributors xiii
Foreword xv
Acknowledgements xvi
PART I
Payment systems and public policy 13
PART II
New approaches to modelling payments 73
PART III
Current payment policy issues 117
PART IV
Policy perspectives on the future of payments 187
Index 262
Figures
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
England.
Jeffrey Lacker. President of the Federal Reserve Bank of Richmond.
Harry Leinonen. Adviser to the Board of the Bank of Finland and the Finnish
representative in the Eurosystem Payment and Settlement System 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
England.
James J. McAndrews. Vice President and Head of the Money and Payment
Studies Function in the Research and Statistics Group of the Federal Reserve
Bank of New York. He received a PhD in economics from the University of
Iowa.
xiv Contributors
Richard Pattinson. Head of Regulatory and Industry Issues, Global Payments,
Barclays Bank. Holds a number of external positions in the industry primarily
concerned with payments and settlements and the management of payment
system liquidity including: Chairman, CHAPS Clearing Company Limited
(UK); Deputy Chairman, SWIFT (UK) Limited (UK); Director, Voca
Limited (UK); Director, EBA Association (France); Director, CLS Group
Holding AG (Switzerland); Director, CLS Bank International (USA);
Member Bank of England Money Market Liaison Group; Member UK
Market Advisory Committee.
Jean-Charles Rochet. Professor of Economics and Mathematics at Toulouse
School of Economics (Toulouse University) and Research Director at Institut
D’Economie Industrielle, Toulouse, France.
George Selgin. Professor of Economics at the University of Georgia’s Terry
College of Business. His latest book, Good Money: Private Enterprise and
the Beginnings of Modern Coinage, is forthcoming from the University of
Michigan Press.
Stefan W. Smitz. Currently at Oesterreichische Nationalbank. Co-editor of
Institutional Change in the Payments System and Monetary Policy, Rout-
ledge (with Geoffrey E. Wood) and Carl Menger and the Evolution of
Payment Systems: From Barter to Electronic Money, Edward Elgar (with M.
Latzer).
George Speight. Senior Manager, Systemic Risk Reduction Division, Bank of
England.
Neil Wallace. Professor of Economics, The Pennsylvania State University.
Matthew Willison. Economist, Systemic Risk Assessment Division, Bank of
England.
Jing Yang. Senior Economist, International Finance Division, Bank of England.
Foreword
The Bank of England has two core purposes – monetary stability and financial
stability. Payment systems are the mechanism by which money is transferred to
enact both real transactions (such as buying bread) and financial transactions
(such as buying bonds). So robust payment systems are integral to both of the
Bank’s core purposes. The same is true in central banks around the world.
Indeed, in many central banks their payment system role predated the 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
today.
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.
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.
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.
Conclusion
We have not solved the conundrum with which we began this introduction. We
hope, however, that this volume is some small step towards beginning to close
the gap between the perceptions of practitioners, academics and the general
public on the role and importance of payment systems. In particular, we hope
this volume helps raise the interest of those who had previously thought
payment systems obscure and demystified, at least to a degree, the technical lan-
guage built up around them. With luck, readers might even feel spurred to tackle
some of these fascinating and important issues for themselves.
Notes
1 Lacker (2006) makes a similar point, arguing that the interest rate should be related to
the rate at which agents discount the utility they obtain from consumption if incentives
– in this case, banks’ payment incentives during the day – are not to be adversely
affected.
2 Definitions all taken from Bank for International Settlements (2003).
3 In his chapter in this volume, Rochet also mentions two other factors that have led to
an increase in liquidity risk: increases in banking sector concentration, as well as in the
complexity and size of financial markets, and the increased use of derivative products,
which generate a large demand for liquidity.
References
Allen, H., Christodoulou, G. and Millard, S.P. (2006) ‘Financial infrastructure and
corporate governance’, Bank of England Working Paper No. 316.
Bank for International Settlements (2003) A glossary of terms used in payments and 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
Pennsylvania.
McAndrews, J. and Rajan, S. (2000) ‘The timing and funding of Fedwire funds transfers’,
Federal Reserve Bank of New York Economic Policy Review.
McAndrews, J. and Trundle, J. (2001) ‘New payment system designs: causes and 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
England.
Norman, B., Shaw, R. and Speight, G.E. (2006) ‘The history of interbank settlement
arrangements: exploring central banks’ role in the payment system’, unpublished
thesis, Bank of England.
Rosenblat, T.S. (1999) ‘What makes the money go round?’, PLD thesis, Massachusetts
Institute of Technology.
Schanz, J. (2006) ‘Innovation and ownership structure in payment systems’, unpublished
thesis, Bank of England.
Woodford, M. (2004) Interest and Prices: Foundation of a Theory of Monetary Policy,
Princeton, NJ: Princeton University Press.
Part I
Introduction
Central banking and payment systems – mechanisms that enable the transfer of
monetary value – are inextricably linked. In the past, institutions that developed
into modern central banks stood at the top of the inter-bank payments hierarchy,
providing the ultimate settlement asset exchanged by commercial banks when
settling payments with each other. At present, modern central banks devote a
considerable proportion of their resources to operating, overseeing and 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.
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
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.
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
OWNS
Model 4 Model 5
OVERSEES
Model 6 Model 3
OPERATES
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.
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.
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
systems
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:
Australia HVCS ✓
Belgium ELLIPS ✓
Canada LVTS ✓
France TBF ✓
PNS ✓
Germany RTGSplus ✓
Italy BI-REL ✓
Japan BOJ-NET ✓
Netherlands TOP ✓
Norway NICS ✓
New Zealand SCP ✓
Sweden RIX ✓
Switzerland SIC ✓
UK CHAPS* ✓
US Fedwire ✓
CHIPS ✓
ECB TARGET ✓
EURO1 ✓
Note
*It should be noted that, although operated and overseen by the Bank of England, CHAPS does not
map perfectly to model 5 as oversight is conducted without formal powers of enforcement.
38 S. Millard and V. Saporta
Table 1.A2 G10 models for intervention in ACHs
Australia BECS ✓
CECS ✓
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 ✓
LSV ✓
UK BACS ✓
US EPN ✓
ACH ✓
Note
*Operation of one of the two sub-systems settling across BI-COMP is carried out by SIA on behalf
of the Bank of Italy
In the case of large-value systems, the majority of central banks own and
operate (model 1). The most notable exceptions here are the United Kingdom,
Australia, New Zealand and Belgium, where the infrastructure for the 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
References
Bagehot, W. (1873) Lombard Street: A description of the money market, London: Henry
S. King.
Bank for International Settlements (BIS) (2001) Core principals for systemically 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
Press.
Goodhart, C.A.E. (2004) ‘Financial supervision from an historical perspective’, Paper
given to Bank of Finland Conference on ‘The structure of financial regulation’.
Green, E.J. and Todd, R. (2001) ‘Thoughts on the Fed’s role in the payment system’,
Federal Reserve of Minneapolis Quarterly Review, Winter, pages 12–27.
Greenfield, R.L. and Yeager, L.B. (1993) ‘A laissez-faire approach to monetary stability’,
in White, L (ed.), Free banking. Volume 3: Modern theory and policy, Aldershot: Elgar.
Haldane, A., Hall, S., Saporta, V. and Tanaka, M. (2004) ‘Financial stability and 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.
99–03.
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,
mimeo.
Rajan, R. and Zingales, L. (2001) ‘The firm as a dedicated hierarchy: a theory of the
origins and growth of firms’, Quarterly Journal of Economics, Vol. 116, pages
805–51.
44 S. Millard and V. Saporta
Selgin, G.A. and White, L.H. (1994) ‘How would the invisible hand handle money?’,
Journal of Economic Literature, Vol. 32, No. 4, pages 1718–49.
Selgin, G.A. and White, L.H. (2005) ‘The future of fiat money: a Mengerian perspective’,
University of Georgia, mimeo.
Smith, V. (1936) The rationale for central banking, London: P&S King Ltd.
Thornton, H. (1802) An enquiry into the nature and effects of paper credit of Great
Britain, London: Hatchard.
Timberlake Jr, R.H. (1984) ‘The central banking role of clearinghouse associations’,
Journal of Money, Credit and Banking, Vol. 16, No. 1, pages 1–15.
Trivoli, G. (1979) The Suffolk Bank: a study of a free-enterprise clearing system,
London: Adam Smith Institute.
Wallace, N. (1988) ‘A suggestion for oversimplifying the theory of money’, Economic
Journal, Vol. 390, pages 25–36.
Woodford, M. (2004) Interest and prices: foundation of a theory of monetary policy,
Princeton, NJ: Princeton University Press.
2 The role of a central bank in
payment systems
Edward J. Green
Introduction
The central banks of all industrialised countries specialise to some extent in
what roles they play in their respective payment systems. Each defers to other
entities in its respective economy, both private and public, to assume roles that it
elects not to fill. Nevertheless, there is considerable variation across 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.
• 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
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.
Conclusion
A central bank does several things that are of immense value to the payment
system, and it is better for the central bank to focus resources on doing those
crucial things excellently than to dissipate resources by taking on additional
tasks. Such a policy of deliberate focus is a difficult one for any organisation to
adopt and maintain. Indeed, many central bankers probably believe that it would
be difficult in practice to adopt such a policy. There are too many demands from
the banking industry, the legislature, and various other stakeholders to embark
on additional tasks, and those demands cannot be ignored. To rebut those 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.
References
Bech, M.L. and Garrett, R. (2003) ‘The intraday liquidity management game’, Journal of
Economic Theory, 109, pages 198–209.
Cavalcanti, R. and Wallace, N. (1999) ‘Inside and outside money as alternative media of
exchange’, Journal of Money, Credit and Banking, 31, pages 443–57.
Committee on Payment and Settlement Systems (CPSS) (2001) ‘Core principles for 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
2006).
Freeman, S. (1996a) ‘The payments system, liquidity, and rediscounting’, American
Economic Review, 86, pages 1126–38.
Freeman, S. (1996b) ‘Clearinghouse banks and banknote over-issue’, Journal of 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
3–19.
Wallace, N. (2005) ‘From private banking to central banking: ingredients of a welfare
analysis’, International Economic Review, 46, pages 619–31.
Weinberg, J. (1997) ‘The organization of private payment networks’, Federal Reserve
Bank of Richmond Economic Quarterly, 83, pages 25–44.
Winston, C. (1993) ‘Economic deregulation: days of reckoning for microeconomists’,
Journal of Economic Literature, 31, pages 1263–89.
Woodford, M. (2003) ‘Interest and prices: foundations of a theory of monetary policy’,
Princeton, NJ: Princeton University Press.
Zhou, R. (2000) ‘Understanding intraday credit in large-value payment systems’, Federal
Reserve Bank of Chicago Economic Perspectives, 24, pages 29–44.
3 Some challenges for research in
payments
Edward J. Green
In this chapter I discuss four directions in payment research that provide particu-
lar challenges in both pure and applied economics, chosen from among the
many important topics in this active field. They are:
• 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:
References
Cavalcanti, Ricardo and Neil Wallace (1999) ‘Inside and outside money as alternative
media of exchange’, Journal of Money, Credit, and Banking, 31, pages 443–457.
Cavalcanti, Ricardo, Andres Erosa and Ted Temzilides (1999) ‘Private money and
reserve management in a random-matching model’, Journal of Political Economy, 107,
pages 929–945.
Freeman, Scott (1996a) ‘The payments system, liquidity, and rediscounting’, American
Economic Review, 86, pages 1126–1138.
Freeman, Scott (1996b) ‘Clearinghouse banks and banknote over-issue’, Journal of 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
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.
References
Goodfriend, M. and Lacker, J.M. (1999) ‘Limited Commitment and Central Bank
Lending’, Federal Reserve Bank of Richmond Economic Quarterly, 85, 4, pages 1–27.
Goodhart, C.A.E. (1988) The Evolution of Central Banks, Cambridge: The MIT Press.
Green, E.J. and Todd, R.M. (2001) ‘Thoughts on the Fed’s Role in the Payments
System’, Federal Reserve Bank of Minneapolis Quarterly Review, 25, pages 12–27.
Kocherlakota, N.R. (1998) ‘Money is Memory’, Journal of Economic Theory, 81, pages
232–251.
Lacker, J.M. (2004) ‘Payment System Disruptions and the Federal Reserve After 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
questions
Ricardo Cavalcanti and Neil Wallace1
Introduction
In both the United States and the United Kingdom, a monopoly on ‘currency’
issue grew out of a system in which there were many issuers of banknotes. In the
United Kingdom, that monopoly was created in 1844, and was accompanied by
a 100 per cent specie marginal reserve requirement against banknote issue. The
1844 law, Peel’s Act, was a victory for the currency school, whose members
advocated some version of hard money, or what much later came to be called
monetarism. The 1844 law was opposed by members of the banking school:
those who advocated some versions of laissez-faire in intermediation. Among
the questions alluded to in the debates were: Was the private note-issuing system
accomplishing anything? If it was, then would it be desirable to have the Bank
of England manage its monopoly so as to emulate what the private note system
was accomplishing? In this chapter, we revisit those questions and do so for at
least three reasons. First, one test of progress in monetary theory is its ability to
provide new insights about old questions that have never been satisfactorily
resolved. Second, those old questions have modern analogues: should central
banks operate lending facilities and, if so, how? Should stored value, and other
modern analogues of private note-issue, be regulated and, if so, how? Third, the
modelling ideas that throw light on those questions have implications for 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.
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.
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
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:
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
spending.
Although the above comparison is suggestive, it is not decisive even about
this simple setting with the special {0,1} money holdings. The example does not
establish that private money is necessary for an optimum, even among stationary
allocations. Even with y = q*, the allocation described above does not maximize
welfare. In that allocation, a monitored producer does not produce for an
unmonitored consumer who has no money. But some production in such 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}.
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.
Notes
1 We are indebted to Stacey Schreft of the Federal Bank of Kansas City and to John
Moore of the University of Edinburgh for helpful comments on an earlier draft.
2 Distinctions among the money issued by subsets of monitored people are discussed in
Wallace (2003).
3 In a sense, excluding the unmonitored people from meeting the planner and others after
pairwise meetings is without loss of generality. Because unmonitored people can hide
money, the planner can at best give non-negative transfers to them that are weakly
86 R. Cavalcanti and N. Wallace
increasing in their money holdings. And even that can be regarded as problematic.
How does the planner prevent the same unmonitored person from showing up many
times at a date for a transfer?
4 There is a literature on matching models that avoids heterogeneity of money holdings.
One route is the so-called large family model (Shi, 1997). Another is the device 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.
References
Cavalcanti, R. and Wallace, N. (1999a) ‘A model of private banknote issue’, Review of
Economic Dynamics, 2: 104–36.
Cavalcanti, R. and Wallace, N. (1999b) ‘Inside and outside money as alternative media of
exchange’, Journal of Money Credit and Banking, 31(2): 443–57.
Cavalcanti, R., Erosa, A., and Temzelides, T. (1999) ‘Private money and reserve 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:
134–51.
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
Introduction
Payments are transfers of value between agents. For all payments that are not made
in cash, finalization of payment occurs separately to the exchange of goods and will
involve a payment system: a specification for when and how the actual funds are
delivered consisting of a settlement asset, credit arrangements, infrastructure and
rules. Indeed, Zhou (2000) defines a payment system as a ‘contractual and 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.
Banks
We build on this simple random matching model of money by allowing agents
either to hold money in the form of cash or to deposit it in a bank. Banks are 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.
Settlement
Whether an agent chooses to hold his money in the form of cash or a bank
deposit it has implications for when money is transferred between buyers and
sellers. To capture this we assume that each period is divided up into two 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)
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:
We derive conditions for an equilibrium in which trade occurs with only a DNS
payment system and one in which trade occurs with only an RTGS payment
system. We then compare the conditions under which each type of equilibrium
exists.
Such equilibria will exist if the following conditions hold:
Individual rationality: the value of being a seller or a buyer is at least as good
as leaving the market and living in autarky: V0
0 and V1
0.
Incentive compatibility: a seller has an incentive to produce his good in
exchange for one unit of money and a buyer has an incentive to trade his unit of
money for one unit of the good: V1 – V0
c + and u
V1 – V0.
Banking constraint: agents choose to deposit money in a bank: ( ) 0.
The probability that trade occurs equals one in any single coincidence
meeting when we restrict attention to pure-strategy equilibria. Thus, the only
difference between holding money as cash and depositing it in the bank is the
banking fee. Buyers have incentives to deposit money in banks as long as they
receive a return from doing so; that is, the banking fee is negative. Hence,
( ) 0 is both necessary and sufficient for agents to choose to deposit money.
The buyer’s individual rationality constraint holds if the seller’s individual
Optimal settlement rules for payment systems 93
rationality constraint and the buyer’s incentive compatibility constraints are both
satisfied. Thus, we only need to derive the conditions under which V0
0,
V1 – V0
c + , u
V1 – V0 and ( ) 0 hold.
Now, it is not feasible for all agents without money to borrow from a bank
because then there would be no sellers. So, in equilibrium, the loan market
clears at a rate at which sellers are indifferent between remaining a seller and
borrowing. As stated in He et al. (2005), they will be indifferent when the cost
of borrowing, , is equal to the difference between the value of being a buyer
and the value of being a seller, V1 – V0. Using this result, equations (1) and (4)
then imply the following expressions for the values of being a buyer and of
being a seller, respectively:
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
0.03
0.02
0.01
0
0 0.1 x 0.2
0.03
0.02
0.01
0
0 0.1 x 0.2
0.03
0.02
0.01
0
0 0.1 x 0.2
0.03
0.02
0.01
0
0 0.1 x 0.2
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
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
(1 – M1)x
1
(<)2 if u (>)c.
r + (1 – M1)x
Optimal settlement rules for payment systems 97
RTGS Autarchy
2
DNS and
DNS
RTGS
2
(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.
References
Bank for International Settlements (2005) New developments in large-value payment
systems, Committee on Payment and Settlement Systems Publication No. 67.
Bank of England (2005) Payment systems oversight report 2004, London: Bank of
England.
Bank of England (2006) Payment systems oversight report 2005, London: Bank of
England.
Fry, M.J., Kilato, I., Roger, S., Senderowicz, K., Sheppard, D., Solis, F. and Trundle, J.
(1999) Payment systems in global perspective, London: Routledge.
He, P., Huang, L. and Wright, R. (2005) ‘Money and banking in search equilibrium’,
International Economic Review, 46: 637–70.
Lester, B. (2005) ‘A model of interbank settlement’, unpublished thesis, University of
Pennsylvania.
Schoenmaker, D. (1995) ‘A comparison of alternative interbank settlement systems’,
London School of Economics Financial Markets Group Special Paper, No. 204.
Zhou, R (2000), ‘Understanding intraday credit in large-value payment systems’, Federal
Reserve Bank of Chicago Economic Perspectives, 24(3): 29–44.
7 The microstructure of money
James McAndrews1
Introduction
In his chapter in this volume, Ed Green suggested that ‘payment economics
comprises the topics that pertain to both monetary economics and industrial
organization’: loosely paraphrasing, the study of the industrial organization of
money. In this chapter, I approach the study of payment systems as the study
of the market microstructure of money. In doing so, I will also use the lens of
market microstructure to address issues relating to trends in payment systems.
The study of the microstructure of financial markets has focused on the 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.
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
Volume
1,600 Value 12
1,400
10
Volume (thousands)
1,200
US$ (billions)
1,000 8
800 6
600
4
400
2
200
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.
Time
Figure 7.1 Intraday pattern of activity: total value and volume of federal funds traded (2
October 2001–31 September 2004) (source: Federal Reserve Bank of New
York: author’s calculations).
Note
Values shown are the average values in a calendar minute calculated over the days of the sample
period.
activity averaged over the sample period, from October 2001 through September
2004. The levels of the lines in the chart are per-minute averages for each
minute of the Fedwire operating day, averaged across the sample period.9
There is a great deal of clustering in federal funds trading activity, as there
is in payments activity generally in Fedwire. Our test of inventory effects
focuses on banks that are active on both sides of the market for Fedwire during
the period of heaviest trade in federal funds. Recognizing that large banks are
active on both sides of the market, we treat the banks as dealers. There may
well be economies of scope between making and receiving large amounts of
payments and buying and selling federal funds. It may be that these economies
of liquidity generation can explain the concentration of banks actively buying
and selling federal funds. We treat our sample of banks as dealers, and test
whether their balances have an effect on the transaction prices of their 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
effects.
The null hypothesis is that there are no effects on the prices at which the bank
transacts. The interpretation of this hypothesis is that a finding of no inventory
effects is consistent with an efficient market: banks can substitute expected
funds for current balances and make loans even with low current balances late in
the day. This suggests that banks have strong expectations for their ability to
borrow later in the day. As shown in Figure 7.2, which plots the 5th and 95th
percentile of the excess of the average federal funds rate (calculated as the
average federal funds rate in the particular minute across the days of the sample
period) over the target federal funds rate, it is clear that the very sparse trading
in the early morning hours leads to greater variability in observed trades during
those periods.
The regression also includes a number of calendar and event effects. In
particular each regression we report has dummies for the month, for days that
precede or follow a holiday, for Good Friday (a day of low activity), for the days
of the reserve maintenance period, for the first and last banking days of the
month, and for the end of the quarter. Finally, we include dummies for days of
108 J. McAndrews
4.00
5th percentile
95th percentile
3.00
1.00
0.00
1.00
2.00
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.
Time
Figure 7.2 Fifth and 95th percentile of the average federal funds rate minus the target
rate (2 October 2001–30 September 2004) (source: Federal Reserve Bank of
New York: author’s calculations).
Note
Values shown are the 5th and 95th percentile of the distribution of the average federal funds rate
minus the target rate for each minute calculated over the days of the sample period.
federal funds target rate increases and, separately, one for decreases in the target
rate.10
I perform two other tests. First, I conduct the same test described above for
the 4.30 to 4.45 time period. As the 4.00 time was chosen somewhat arbitrarily
among those times in which federal funds trading is active, I wanted to test
whether any findings were robust to changes in the time considered. Second, I
also regress the deviations in the midpoint minus target rate on the bank’s devia-
tions in its transactions balance, both at 4.00 and 4.30 p.m.
Results
The results of these tests are shown in Table 7.1 (all coefficients and standard
errors in Table 7.1 have been multiplied by 10,000 to reduce the number of zeros
in the Table). First, we find weak results in favor of inventory effects, with the
coefficient on the transactions balance variables always negative and significant
at the 10 percent level for the 4.00 p.m. estimations. The estimated coefficient is
negative but smaller in absolute value, and statistically insignificant, for the
The microstructure of money 109
Table 7.1 Regression results
Notes
Standard errors in parentheses: * significant at 10%; ** significant at 5%; *** significant at 1%.
All coefficients and standard errors reported at e + 4 level.
4.30 p.m. regressions. The estimated coefficients are not particularly significant
in an economic sense for the 4.00 p.m. estimate. Column 1 suggests that a 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.
Summary
The microstructure approach to the study of payment systems views payment
systems as analogous to security settlement systems. Payment system designs
have analogues in the designs of securities settlement systems. Similarities in the
economics of payment systems and security settlement systems suggest that
further study of payment systems, borrowing from the financial market
microstructure literature, would be fruitful.
The tests for inventory, calendar, and event effects in the US federal funds
market suggested that there are only slight effects of these variables on the 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.
Notes
1 I wish to thank Adam Ashcraft, Bruno Biais, and Xavier Freixas, and the participants
at the Future of Payments Conference at the Bank of England for helpful comments. I
thank Kurt Johnson for excellent research assistance. The views expressed in this
chapter are those of the author and do not necessarily reflect the views of the Federal
Reserve Bank of New York or the Federal Reserve System.
2 This rough characterization does not fully specify a DNS as it does not describe the
rules of the DNS in the event of a default. See Bank for International Settlements
(1989, 1990) for more complete descriptions of DNS systems.
3 If there are not sufficient balances on account, a payment order may be queued,
pending the arrival of additional balances, or rejected. See Bank for International
Settlements (1997) for a full discussion of various alternatives and implementations
of RTGS systems.
114 J. McAndrews
4 From the RTGS plus website of the Deutsche Bundesbank (surveyed April 20, 2005):
www.rtgsplus.de/en/leistungsumfang/limitsteuerung/inhalt_e.htm.
5 See McAndrews and Trundle (2001) and Bank for International Settlements (2005)
for more detailed descriptions of the range of alternatives for these payment systems.
6 Again, I confine myself here to ‘pure’ payment systems, and ignore delivery-versus-
payment systems in which both a security and account balances are exchanged
simultaneously.
7 See Regulation D, 12 CFR 204, Board of Governors of the Federal Reserve System.
8 The method for identifying federal funds purchases from the Fedwire transactions
journal was pioneered by Furfine (1999), and applied by Demilralp et al. (2004), as
well. The method I use is very similar to Demilralp et al. (2004) ‘N-to-N’ method.
We search for payments from a bank to its counterparty on day t that is in round
values of $100,000, and a matching return payment on day t + 1 that is slightly larger
than the first payments, and within a reasonable range given federal funds rates
reported by brokers for that day. Specific details are available from the author.
9 On May 22, 2004, the Fedwire operating day lengthened from its previous hours of
12.30 a.m. to 6.30 p.m. to 9.30 p.m. the previous day to 6.30 p.m. We do not include
the hours of 9.30 p.m. to 12.30 a.m. in this chart.
10 We note that open market operations are usually conducted in the late morning, well
before the 4.00 p.m. time of our tests.
11 Demilralp et al. (2004) have a good discussion of who is eligible to participate in
federal funds trading.
12 See www.e-mid.it/index.php for a description of the screen-based money market
platform.
References
Admanti, A. and Pfleiderer, P. (1988) ‘A theory of intraday patterns: volume and price
variability’, Review of Financial Studies, 1: 3–40.
Amihud, Y. and Mendelson, H. (1980) ‘Dealership market: market-making with 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:
80–97.
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
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)
Introduction
When a bank, acting on its own behalf or that of a client, needs to transfer a
large sum, say $10 million, to another bank, the transfer will almost certainly be
made through a ‘wholesale’ payments system. Wholesale payments systems
receive payments messages from sending banks, relay them to receiving banks,
and oversee the final settlement of interbank accounts.
Two basic kinds of wholesale payments systems exist. A Deferred Net 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
second-best.
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:
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.
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
risk.’17
What about the potentially catastrophic consequences of a DNS unwind? Are
there not sufficient grounds for preferring RTGS to DNS even if the former
arrangement is much more costly and does not completely eliminate systemic
risk? One reply to this question asks, ‘What catastrophic consequences?’ To
repeat: there has never been a settlement failure and consequent unwind on
CHIPS or any other major DNS wholesale-payments system; and the chances of
a DNS-system participant failing suddenly enough to precipitate a settlement
crisis (as can happen if other participants have no inkling of troubles at a 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.
Notes
1 The author thanks Bill Bergman, Sandra Haasis, David Humphrey, George Kaufman,
David Mustard, Harold Nitsch, Will Roberds, Art Snow, Larry Wall, Ron Warren,
Lawrence H. White, several anonymous referees, and participants in the University of
Georgia Economics Department workshop for their helpful comments and sugges-
tions.
2 In January 2001, CHIPS switched from deferred net settlement to near-continuous
settlement.
3 Since 1986 the Fed has also imposed limits, known as ‘net debit caps,’ on the
maximum overdraft individual banks could obtain from it without being subject to
special administrative actions. These caps appear, however, to have had only a very
limited effect on the overall value of overdrafts (Hancock and Wilcox, 1996).
4 According to Zhou (1999), allowing for deductibles Fedwire’s ‘imputed average
annual rate [was] only around 11 basis points’ after 1995. Zhou’s study is also 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
participant.
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
members.
10 Even today most US banks provide their customers with ‘deposit account agreement
and disclosure’ statements that include language like the following (from the Seaway
National Bank of Chicago):
All non-cash items (for example, checks) deposited to your Account are posted
subject to our receipt of final payment by the payor bank. If final payment is not
received . . . you authorize us to charge any of your Accounts, without prior notice
and at any time, for the amount of the returned item, our returned item fee, any
interest paid on the item, and any other fee we pay or incur . . .. With respect to wire
transfers . . . you agree to enter into and comply with our wire transfer agreement. . . .
Credit given by us to you with respect to [a] wholesale (wire) funds transfer entry is
provisional until we receive final settlement for such entry through a Federal
Reserve Bank. If we do not receive final settlement, you are hereby notified and
agree that we are entitled to a refund of the amount credited to your account in 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.
[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
failure.
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
holders.
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
References
Angelini, P., Maresca, G., and Russo, D. (1996) ‘Systemic risk in the netting system,’
Journal of Banking and Finance, 20(5): 853–868.
Board of Governors of the Federal Reserve System (1988) Controlling risk in the
payments system, Washington, DC: Board of Governors of the Federal Reserve
System.
Cannon, J.G. (1910) Clearing houses, Washington, DC: Government Printing Office.
Coase, R.H. (1988) The firm, the market and the law, Chicago, IL: University of Chicago
Press.
Dale, R. (1998) ‘Risk management and public policy in payment, clearing and settlement
systems,’ International Finance, 1(2): 229–259.
Dowd, K. (1994) ‘Competitive banking, bankers’ clubs, and bank regulation,’ Journal of
Money Credit, and Banking, 26(2): 289–308.
Folkerts-Landau, D. (1997) ‘Wholesale payments and financial discipline, efficiency, and
liquidity,’ International Monetary Fund Working Paper No. 154.
Freeman, S. (1999) ‘Rediscounting under aggregate risk,’ Journal of Monetary 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:
870–908.
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):
30–39.
Kahn, C.M., McAndrews, J. and Roberds, W. (2003) ‘Settlement risks under gross and
net settlement,’ Journal of Money Credit and Banking, 35: 591–608.
Kaufman, G. (1999) ‘Do lender of last resort operations require bank regulation?,’
unpublished thesis, Loyola University Chicago.
Mengle, D.L. (1990) ‘Legal and regulatory reform in electronic payments: an evaluation
of payment finality rules,’ in D.B. Humphrey (ed.) The US payment system: efficiency,
risk and the role of the Federal Reserve, Boston, MA: Kluwer.
Mengle, D.L., Humphrey, D.B., and Summers, B.J. (1987) ‘Intraday credit: risk, value,
and pricing,’ Federal Reserve Bank of Richmond Economic Review, 73(1): 3–14.
Richards, H.W. (1995) ‘Daylight overdraft fees and the Federal Reserve’s payment
system risk policy,’ Federal Reserve Bulletin, 81(12): 1065–1077.
Roberds, W. (1999) ‘The incentive effects of settlement systems: a comparison of gross
settlement, net settlement, and gross settlement with queuing,’ Bank of Japan Institute
for Monetary and Economic Studies Discussion Paper No. 99-E-25.
Rochet, J.-C. and Tirole, J.T. (1996) ‘Controlling risk in payment systems,’ Journal of
Money Credit and Banking, 28: 832–862.
Schoenmaker, D. (1995) ‘A comparison of alternative interbank settlement systems,’
London School of Economics Financial Markets Group Discussion Paper No. 204.
Scott, H.S. (1990) ‘Commentary,’ in D.B. Humphrey (ed.) The US payment system: 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
Introduction
In this chapter we present a model of a Real-Time Gross Settlement (RTGS)
payment system with tiered membership where settlement is facilitated by 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
payments.3
Where significant costs of obtaining intraday credit remain, agents might
choose not to participate directly in an RTGS system at all. The chapter by
Rochet in this volume argues that to the extent that an agent chooses to by-pass
an RTGS system, by entering into bilateral agreements with other agents or by
shifting flows to a competing DNS system, systemic risk may be increased. Such
alternative arrangements typically include recourse to a correspondent bank,
who processes payments on behalf of indirect system participants. This 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
collateral.
Where payments made by second-tier participants are not internalized, but
rather are effected by the correspondent over central bank settlement accounts,
any collateral-posting requirements at the central bank would apply. But here
too, tiering may reduce the deadweight social costs of collateral if agents with
high opportunity costs of posting collateral are able to take advantage of lower
collateral-posting costs enjoyed by their correspondent. The opportunity cost of
posting collateral may be proxied by the reverse-repo spread (i.e. the spread
between secured and unsecured borrowing costs). This may vary across agents,
according to differences in credit-worthiness which affect the unsecured cost of
borrowing. Also, market imperfections and regulatory policy may influence
agents’ relative opportunity cost of collateral. For example, in the United
Kingdom, banks subject to the Stock Liquidity Requirement (SLR), a prudential
liquidity regime, are able to meet intraday collateral requirements using assets
that they have to hold, in any case, to meet their prudential requirement at
the end of the day; hence they have a very low opportunity cost of posting
collateral intraday.4
Central bank intraday collateral policy 141
The foregoing discussion highlights the potential benefits of tiering.
However, tiering can also introduce additional risks to the system.
First, a correspondent bank might not have sufficient incentive to monitor
because it does not fully internalize the potential systemic consequences of a
customer’s default which triggers liquidity or solvency problems of its own.
Even if sufficient incentives were to exist, monitoring of second-tier agents by a
correspondent might not be accurate, causing the correspondent to incur credit
and liquidity exposures that might in turn lead to wider contagion. Harrison et
al. (2005) analyse the credit risk implications of the highly tiered structure of the
UK large value payment system, concluding that this channel might not impose
significant risks on the system as a whole, except in extreme circumstances.
However, other risks exist. For instance, internalized payments might be subject
to greater legal risks as they are not likely to be covered by provisions providing
protection against bankruptcy law (such as the Settlement Finality Directive in
the European Union). Tiering increases the risk that operational or financial
problems at a settlement bank lead to disruption of payments in a large part of
the system. Another important concern is that, in response to liquidity problems
among second-tier banks, settlement banks might decide to cut or restrict 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.
A B A B
C C
C has an obligation to A: C has an obligation to B:
Probability (1e) Probability e
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.
Time
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’ 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.
(Z,D): 0
(F,D): 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
M
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
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.
0.0020
(Z,D) Social
0.0018
0.0016
0.0014
Social/private cost
0.0020
X
0.0018
(F,D) Social Private
0.0016
0.0014
Social/private cost
W
(F,I ) Social Private 0.0012
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.
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
become:
(Z,D): 0
(F,D): C
Central bank intraday collateral policy 151
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
0.0020
(Z,D) Social
0.0018
0.0016
0.0014
Social/private cost
(F,D) Social Private
X 0.0012
Figure 9.5 The impact of imperfect monitoring (C = A = 0.0015; S = 1.1; = 0.0015;
e = 0.5; = 0.00075; = 0.0005).
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:
0.0020
(F,I ) Social 0.0019
(F,D) Y
0.0018
Social Private
0.0017
Social/private cost
X
0.0016
0.0015
0.0014
0.0013
(F,I ) Private
0.0012
0.0011
0.0010
0 0.0001 0.0002 0.0003 0.0004 0.0005
Monitoring cost (M)
Figure 9.6 The impact of tiering risk (C = A = 0.0015; S = 1.1; = 0.0015; e = 0.5;
= 0.00075; = 0.0005; T = 1.4).
154 J.P. Jackson and M.J. Manning
Indeed, until now we have worked with a value of C’s payment obligation of
unity. Generalizing this, by allowing C’s payment to take the value, P, we can
capture the policy implications associated with payment system participants of
different ‘sizes’. In particular, we find that the interval of monitoring costs over
which social and private incentives under full collateralization are misaligned is
increasing in the value of payments. That is, for a given error in monitoring, ,
and a given degree of spillover, T, higher payment values lead to higher 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
upwards.
Bank C:
(F,D): PC
Central Bank:
(F,D): PC
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
0.0090
(F,I ) Social Y
(F,D) Social Private 0.0085
X 0.0080
Social/private cost
0.0075
0.0070
0.0065
(F,I ) Private
0.0060
0.0055
0.0050
0 0.0003 0.0006 0.0009 0.0012 0.0015
Monitoring cost (M)
Figure 9.7 The impact of tiering risk with high payment value (C = A = 0.0015;
S = 1.1; = 0.0015; e = 0.5; = 0.00075; = 0.0005; T = 1.4; P = 5).
Conclusions
The model developed in this chapter can be used to examine the key factors
influencing both an agent’s decision over whether to participate directly in a
payment system and a central bank’s decision as to whether to require 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
concentration.
The model’s predictions are, prima facie, consistent with UK experience,
with full collateralization, a high degree of indirect participation, and significant
concentration in correspondent banking all key features of the UK landscape.
Although the model does not capture legal and operational risks, and hence the
full implications of internalization, our model can go some way towards offering
some policy guidance as to whether such a profile of participation is desirable.
Given that the agent’s and central bank’s decisions are taken sequentially, we
show that, under certain circumstances, the first-best outcome might not be
achievable. In particular, we find that zero collateralization and indirect 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.
Notes
1 The views expressed in this chapter are those of the authors, and not necessarily those
of the Bank of England. We would like to thank the following for helpful comments
during the preparation of this work: Victoria Saporta, Steve Millard, Matthew 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-
plification.
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.
References
Bank for International Settlements (BIS) (2005) New developments in large-value
payment systems, Committee on Payment and Settlement Systems Publication No. 67.
Bech, M. and Garratt, R. (2003) ‘The intraday liquidity management game’, Journal of
Economic Theory, 109(2): 198–219.
Chakravorti, S. (2000) ‘Analysis of systemic risk in multilateral net settlement systems’,
Journal of International Financial Markets, Institutions and Money, 10: 9–30.
Central bank intraday collateral policy 159
Furfine, C.H. and Stehm, J. (1998) ‘Analyzing alternative intraday credit policies in 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
Introduction
Central banks lend central bank balances and accept deposits on terms designed
to ensure that the overnight market interest rate is close to its ‘policy’ rate,
which it sets to meet its monetary policy objectives. This rate anchors market
interest rates for longer maturities. However, central banks also typically lend
central bank balances and accept deposits at low or zero interest rates intraday.
This anchors the intraday market interest rate at or close to zero, in so far as
money is traded intraday at all. Dale and Rossi (1996) show that the central bank
can set a low or zero intraday interest rate and ensure the overnight rate is close
to its (higher) policy rate because all intraday lending is repaid by the end of the
day; there is no spillover overnight.
The existence of a distinction between intraday and overnight interest rates
has its origins in the move from settling large-value payments on a deferred net
settlement basis to settling them on a gross basis in real time. (A description of
the diffusion of RTGS across the world’s large-value payment systems can be
found in the chapter by Bech in this volume.) When banks exchanged central
bank money at the end of each day there was simply no need for banks or the
central bank to lend at maturities of less than one day because money was not
being exchanged on a more frequent basis than once per day.2 But in a 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-
ments.
Payment delays are not desirable because they can imply the following risks
and inefficiencies in the payment system:
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
Policy rate
• payments on behalf of customers where the customer does not express any
preference for which time of day he wants his payment made; it seems
reasonable to assume that end of day would remain the ‘default’ for these
customers;
• payments which the bank has a particular reason for sending in a specific
half of the day including customer payments, where the customer has
requested the payment be sent in a particular half of the day, and some 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
system.
There will be an increase in operational risk, regardless of the mix of the
two kinds of payments. If a bank submits a payment to the payment system
close to the deadline by which it should settle, and at that time there is an
operational problem either with the bank or with the system itself, there is less
scope to overcome the problem and allow the payment to be settled by the
deadline. So as a general principle, as the time between a bank submitting a
payment and the deadline for that payment to settle falls, operational risk
increases.
Payments of the first kind are systematically delayed until the second half of
the day, reducing the average amount of time available for them to settle. 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
stability.
There could be further benefits for financial stability if the change in the
central bank’s interest-charging convention affects how intensely banks monitor
exposures to their customers. When the central bank only charges and 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.
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
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.
Table 10.3 Pay-offs when central bank charges and remunerates at midday and at end of
day
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
Further issues
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
Conclusion
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.
Notes
1 The views expressed in this chapter are those of the authors, and do not necessarily
reflect those of the Bank of England, the Federal Reserve Bank of New York or the
Federal Reserve System.
Banks’ interest calculating conventions 173
2 In the case of Japan, where there were a series of net settlements through the day, the
central bank did not extend intraday credit (see Hayashi, 2001).
3 The Swiss National Bank was an exception until 1999 since it provided no intraday
credit to banks operating in its RTGS payment system (see Heller et al., 2000).
4 We assume throughout that the policy rate consistent with the central bank’s monetary
policy objectives is strictly positive. Of course, if the policy rate were zero, there
would be no discontinuity between intraperiod and interperiod interest rates.
5 We assume that the central bank’s new regime is passed through to all banks, whether
they are direct settlement banks or customers of settlement banks. If this were not the
case, we would have a more complicated situation in the money market.
6 This is a common assumption in the literature, e.g. Bech and Garratt (2003).
7 In sterling; conventions differ slightly across currencies.
8 An alternative contract type would leave open in which half of the day payment should
occur, but with the sum paid adjusting to reflect interest foregone by the payee. If the
interest were calculated at a pre-agreed fixed rate, then this contract has an option
value for the payer, adding to its complexity. If the interest rate were tied to the market
rate, it might still not accurately reflect the opportunity cost for the payee. So the 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
afternoon.
References
Angelini, P. (1998) ‘An analysis of competitive externalities in gross settlement systems’,
Journal of Banking and Finance, 22: 1–18.
Bech, M.L. and Garratt, R. (2003) ‘The intraday liquidity management game’, Journal of
Economic Theory, 109: 198–219.
Coleman, S.P. (2002) ‘The evolution of the Federal Reserve’s intraday credit policies’,
Federal Reserve Bulletin.
Dale, S. and Rossi, M. (1996) ‘A market for intraday funds: does it have implications for
monetary policy?’, Bank of England Working Paper No. 46.
Freeman, S. (1996) ‘The payments system, liquidity and rediscounting’, American 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’
liquidity?
Jean-Charles Rochet1
Introduction
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-
cumstances.
The plan of the rest of this chapter is the following: the next section examines
the sources of liquidity risk for banks. We then list the main instruments that can
be used by banks for managing their liquidity risks before briefly discussing the
possible market failures in the provision of liquidity that may justify public
intervention in the regulation and provision of liquidity to banks. We then
explain why regulation of banks’ liquidity may be justified and discuss the way
in which these regulations could be set.
How should we regulate banks’ liquidity? 177
Sources of liquidity risk for banks
Like any other firm, a bank has to manage carefully its liquidity in order to be
able to cover mismatches between future cash outflows and cash inflows.
However the degree of uncertainty about these mismatches is clearly much
higher in the banking sector. This has several sources, which we now examine in
turn.
Assets
There is large uncertainty about the volume of new requests for loans (or
renewal of old loans) that a bank will receive in the future. Of course the bank
could refuse to grant these new loans but this would in general lead to the loss of
profit opportunities. This would also be detrimental to the firm if it is 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.
Liabilities
There is also large uncertainty about the amount of withdrawals of deposits
(including wholesale) or the renewal of rolled-over interbank loans. This is 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.
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
warranted.
Securities settlement
A related issue concerns the industrial organization of the financial sector. If
banks are allowed to merge with central security depositories or CSDs (forming
what is known as International Central Security Depositories or ICSDs), addi-
tional liquidity requirements may be needed since CSDs can be viewed as essen-
tial infrastructures enabling security trading.
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
bank.
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).
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.
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.
Conclusion
Like solvency regulations, liquidity regulations for banks can be justified by two
different motives: one is to limit the risk and the impact of individual bank 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
Notes
1 I acknowledge useful comments by Victoria Saporta. I am solely responsible for any
remaining mistakes.
2 See Bank of England (2003).
3 This is criticized in the chapter by Selgin reprinted in this volume which disputes the
suggestion that DNS systems are intrinsically subject to systemic risk, at least in the
absence of government intervention.
4 However the composition of payments in the two systems is different: the average
payment on Fedwire is much bigger than on CHIPS. (I thank Victoria Saporta for
providing these figures.)
5 See http://bodurtha.georgetown.edu/enron/derivatives_events.htm for a list of other
derivatives ‘disasters’.
6 However, Allen and Gale (2004) show that liquidity requirements for banks may be
needed when the financial markets for transferring aggregate risks among agents are
incomplete.
7 See the chapters by Bech, McAndrews and Lester et al. in this volume for more detail
on the evolution of different large-value interbank payment systems and on the 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.
References
Allen, F. and Gale, D. (2000) ‘Financial contagion’, Journal of Political Economy,
108(1): 1–33.
Allen, F. and Gale, D. (2004) ‘Financial intermediaries and markets’, Econometrica, 72:
1023–61.
Bank of England (2003) ‘Strengthening financial infrastructure’, Bank of England 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):
611–38.
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):
1116–47.
Tirole, J. (2005) Corporate Finance, Princeton, NJ: Princeton University Press.
Part IV
Introduction
Payments are an integral part of a modern market economy as most transactions
involve the use of cash, checks or electronic transfers. Commercial banks and
the central bank provide in a symbiotic relationship the infrastructure through
which payments flow. An efficient payment system is a prerequisite for a 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
systems.1
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
methodology.
At the apex of the financial system are a number of critical financial markets
that provide the means for agents to allocate capital and manage their risk 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
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.
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
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
fact.
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
150
80
125
Number of central banks
Adoption (%)
60 100
75
40
50
20
25
0 0
1970 1980 1990 2000 2010 2020
Year
Conclusion
Our analysis shows that RTGS has been very successful over the last three
decades and has become the de facto standard in terms of settlement 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
Notes
1 See, for example, Bank for International Settlements (1990).
2 These markets include overnight interbank money, foreign exchange, commercial paper,
government and agency securities, corporate debt, equity securities and derivatives.
3 The events of September 11, 2001 underscored the importance of a resilient interbank
payment system. See McAndrews and Potter (2002) and Lacker (2004).
4 These systems include the Depository Trust & Clearing Corporation (DTCC), the
Clearing House Interbank Payments System (CHIPS) and Continuous Linked 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
obligations).
12 Keller (2004).
13 See, for example, Griliches (1957) and Mansfield (1968).
14 The classification reflects the fact that for a normal distribution 68 percent of the
observations are within one standard deviation from the mean (or median) and
95 percent are within two standard deviations.
15 The members are the central banks of France, Germany, Belgium, Italy, Japan, the
Netherlands, Sweden, the United Kingdom, the United States, Canada and 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.
References
Bank for International Settlements (1990) Report of the Committee on Interbank Netting
Schemes of the central banks of the Group of Ten countries (Lamfalussy Report), 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
Introduction
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
days.4
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
internet.
The most dramatic change should be seen in interconnectivity. All companies
and most private customers are likely to have PCs and servers connected to the
internet and other networks. Everyone could be addressable and reachable over
the network. The next generation of mobile phones will have broadband inter-
faces and processing capacities, which are sufficient for all kinds of payment
processing.
Efficient electronic straight-through-processing interfaces between banks and
their customers could automate the processing of payments, not only for 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
norm.
The electronic payment services should be both reliable and secure. There
have been some security and reliability problems with the internet and some of
the first electronic payment implementations. However, the situation has
improved over recent years and there are already solutions and standards in
place to provide secure quality service to customers.
This modern payment scenario can already be found in some national
payment systems, especially in the Nordic and Baltic countries, where more than
50 per cent of customers use network-based e-payment solutions and over 90 per
cent of bank payments are introduced via paperless means.5 The pressure on the
banking sector to improve customer service in payments will increase in the
coming years especially with e/m-payments developments. It is in the interest of
the authorities to further development in order to ensure efficient payment ser-
vices for the general public.
Bank 2
Payment/
account
Bank 1 Payment server
Payment/
account TCP/IP
server Internet
network Payment
Payment
Payment
Bank n
Bank 3 Payment/
Payment/ account
account server
server
(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.
3
Bank-to-bank
4
Customer- Bank-to-
to-bank customer Statement
E-banking receipts
2
Customer-to-customer
1 Beneficiary
Payer
E-billing
Figure 13.3 E-settlement is part of the credit transfer circle, which provides efficient
electronic communications between participants in a payment.
cheques. In credit transfers, the payer’s bank identifies its customer, checks the
payment instruction and debits the payer’s account; the beneficiary’s bank
checks the settlement and credits the beneficiary’s account. In the future 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.
• 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
payee.
212 H. Leinonen
Clearing in decentralised infrastructures
The network environment is decentralised and telecommunication builds bridges
between the different independent but interoperable entities in the network. As
regards payments, these bridges must transfer payment messages as well as
interbank settlements. This is the main difference as compared with other 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.
Payment envelope
End of payment
Figure 13.4 The digital e-settlement stamp is part of the payment message.
Payment
Interbank
Bank 1 Bank 2
network
Digital
CB cover
Figure 13.5 The digital encrypted stamp with central bank cover will follow the payment
message through the network.
fer central bank money from payer’s bank to beneficiary’s bank. Final settlement
is part of the payment message, in the form of electronic central bank money for
interbank settlement purposes. The electronic stamp will accompany the
payment on its route through the interbank payment network to the receiving
bank as described in Figure 13.5. The electronic stamp can be seen as a modern
version of a central bank draft. It is the cover in central bank money of the
payment(s) it accompanies.
The stamp is protected by very strong and modern cryptographic technology
(e.g. Public Key Infrastructure: PKI14). These stamps are produced and decoded
by e-settlement modules situated close to banks’ payment systems, as shown in
Figure 13.6.
The e-settlement modules are tamper-resistant devices provided by central
banks to each bank. These are closely integrated with banks’ payment systems,
e.g. directly integrated with the SWIFTnet access platform (CBT). This makes
settlement transfers a highly automated part of payment processing. Integrating
the new settlement process will be quite straightforward, given that it will be
done on the access platform (e.g. SWIFT CBT) level. In traditional RTGS
systems, banks’ settlement accounts are located in the centralised RTGS system.
In the e-settlement system, each bank’s settlement account is distributed to the
bank’s own processing site in a central bank-controlled e-settlement module.
This module should be regarded as a completely automated central bank branch,
serving one customer with one account. Each bank has access to its own
214 H. Leinonen
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
Interbank
network
System
Central
administration
bank
site
Figure 13.7 A dedicated interbank network connects all banks and the central bank with
each other for payment processing.
Immediate liquidity
For settlement purposes, banks need liquidity. Liquidity is transferred by the
central bank to the system (e-settlement modules) at the start of the day. It can
be increased during the day by the central bank via liquidity transfers or 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.
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
sector.
The cost-advantage of the e-settlement system is in the low processing costs
of adding the e-settlement stamp that enables instant final settlement in central
bank money. The extra processing cost of adding the e-settlement stamp will be
practically nil. It will be an integral part of the payment process itself. Banks
need only invest in low cost equipment. The very low transaction costs of
e-settlement will also enable banks to transfer payment flows from centralised
processing centre systems to more efficient decentralised network-based
communications.
The bottlenecks created by centralised resources will disappear and even the
dependence on critical centralised resources will be dramatically reduced.
E-settlement could offer a solution for integrating the euro-area payment
systems, and a multi-currency version could serve an even larger area. In order
to achieve large-scale benefits via the e-settlement model, the number of 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.
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.
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.
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.
Costs
Current level
Possible level
Time
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
volumes.
This leaves the authority push as the most viable solution. The authority push
can be delivered using three general methods (separately or in combination)
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.
Conclusion
When will e-settlement become a reality? New technology always eventually
pushes obsolete, inefficient alternatives out of the market. There are three basic
alternative scenarios for the development of network and e-settlement based
development. First, the current market players may start to use modern 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.
Notes
1 The views expressed are those of the author and do not necessarily reflect those of the
Bank of Finland.
2 For examples from Finland and some of the new EU member countries, which have
built new banking systems based on real-time processing, see ECB Blue Book and
www.pankkiyhdistys.fi.
3 See www.paypal.com.
4 The ECB Blue Book gives details on European systems. For example in Belgium,
Finland and the Netherlands real-time or near real-time interbank express transfers
already exist. In the Netherlands and Norway normal payments are also cleared the
same day. In most other countries clearing and settling real-time payments takes at
least one day but often more. According to EU Commission studies on cross-border
payments, the average delivery time was 3.3 days in 2001.
5 The self-service automation level reached in Nordic and Baltic countries was, in
2002, already over 80 per cent and in the most efficient countries about 95 per cent.
Customers prefer to send payments over the internet using e-banking instead of using
branch services (see ECB Blue Book statistics). The growth has been very strong
since the end of 1990s when the internet became generally popular.
6 Modern technology has made it possible to administrate and handle large network
structures. SWIFTnet is a good example of this. The message-based network 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
participants.
7 Information on IBAN can be found on the web-page of the European Committee for
Banking Standards (ECBS): www.ecbs.org.
8 However, there are also competing options, e.g. the international card number used by
Visa, Mastercard, etc., the e-mail addresses used by PayPal and phone numbers used
by mobile payment systems. Competing account numbers make the structures more
complex.
228 H. Leinonen
9 For IBAN-information, see the ECBS web-page (www.ecbs.org) and the Thomson
Financial Publishing web-page (www.tfp.com/payment.shtml).
10 See Leinonen (2000) for details.
11 See, for example, www.swift.com, www.fba.fi, www.ecbs.org, www.twiststandards.
com, www.rosettanet.org, www.x12.org and www.ifxforum.org.
12 See info on XML www.w3schools.com/xml/default.asp and www.w3schools.com/
schema/default.asp.
13 See Leinonen et al. (2002).
14 See Schneier (1996).
15 Visa and MasterCard have developed the chip card standards for credit cards, which
have caught the interest of other credit card companies. See www.mastercard.com
and www.international.visa.com.
16 Kemppainen (2003) and McAndrews (1995, 1997).
17 McAndrews (1995, 1997) and Rochet and Tirole (2002).
18 Wright (2004), Schmalensee (2002) and Rochet and Tirole (2003).
19 Bank for International Settlements (2001, 2003).
20 Humphrey et al. (2001) and Guiborg and Segendorf (2002).
References
Bank for International Settlements (2001) Core principles for systemically important
payment systems, Committee on Payment and Settlement Systems Publication
No. 43.
Bank for International Settlements (2003) Policy issues for central banks in retail 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
2007).
European Central Bank (2004) Towards a Single Euro Payments Area – Third progress
report, Frankfurt am Main: European Central Bank. Online, available at:
www.ecb.int/pub/pdf/other/singleeuropaymentsarea200412en.pdf (accessed 22 January
2007).
Guibourg, G. and Segendorf, B. (2002) ‘Do prices reflect costs? A study of the price and
cost structure of retail payment services in the Swedish banking sector 2002’, Sveriges
Riksbank Working Paper No. 172.
Humphrey, D., Mosche, K. and Vale, B. (2001) ‘Realizing the gains from electronic pay-
ments: costs, pricing and payment choice’, Journal of Money, Credit and Banking, 33:
216–34.
Kemppainen, K. (2003) ‘Competition and regulation in European retail payment
systems’, Bank of Finland Discussion Paper No. 16/2003.
E-settlement 229
Leinonen, H. (2000) ‘Re-engineering payment systems for the e-world’, Bank of Finland
Discussion Paper No. 17/2000.
Leinonen, H., Lumiala, V.-M. and Sarlin, R. ‘Settlement in modern network-based
payment infrastructures: description and prototype of the e-settlement model’, Bank of
Finland Discussion Paper No. 23/2002.
McAndrews, J. (1995) ‘Antitrust issues in payment systems: bottlenecks, access and
essential facilities’, Federal Reserve Bank of Philadelphia Working Paper No. 96–19.
McAndrews, J. (1997) ‘Network issues and payment systems’, Federal Reserve Bank of
Philadelphia Business Review.
McCreevy, C. ‘The wind has changed’, speech at EUROFI-Banking and Finance in
Europe Annual Conference 2005.
Rochet, J.-C. and Tirole, J. (2002) ‘Competition among competitors: the economics of
payment card associations’, Rand Journal of Economics, 33: 549–70.
Rochet, J.-C. and Tirole, J. (2003) ‘Platform competition in two-sided markets,’ Journal
of the European Economic Association, 1: 990–1029.
Schmalensee, R. (2002) ‘Payment systems and interchange fees’, Journal of Industrial
Economics, 50: 103–22.
Schneier, B. (1996) Applied Cryptography Second Edition: Protocols, Algorithms, and
Source Code in C, Mississauga, Ontario: John Wiley & Sons.
Tumpel-Gugerell, G. (2004) ‘Time to act: clear objectives and a convincing roadmap for
the Single Euro Payments Area’, speech at the Coordination Committee of the 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
El
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
al ic g tra
r t ive mu l e
CDiCG
ex at sy di
ve at m a
n
ch ion st ng
O er n co tion
an s em
i
Al onic rad
u
n
ge ne s
tr T
s tw
London
s
ST
Clearing
er
t
ec
or
P
id
House
ks
El
ov
Pr
Pr
ov
id
P
er
ST
CDS SWIFT-RTGS
MEPS
CLS VPS 2nd wave
CLS currency
CDP RITS RTGS
BoJ-Net
Chess HKSCC
JasDec
ST
s
er
P
id
ov
Pr
s
El
ng s r k
ec A Ov
Pr
ov
di m o
ra te tw
tro lte er
id
te s ns s
Tr c c tiv e co
un ing tio ge
er
ST
r t ys ne
ad om e
ni r na th
Austraclear CCAS
s
co d a an
iti m tra nte
on u di r
h
he t n xc
al nic ng tra
r t ive mu l e
c
ex at sy di
ve at m a
i
JSCC
ch ion st ng
O rn o n
te c itio
a
an s em
r
Al onic rad
u
ge ne s
tr T
s tw
ec
or
ks
El
Note
1 For simplicity the many hundreds of users, SWIFT and custodial links are not shown.
15 Will central banking survive
electronic money?
Stefan W. Schmitz
Notes
1 See inter alia Browne and Cronin (1996), Centi and Bougi (2003), Cohen (2002),
Costa Storti and De Grauwe (2003), Crede (1995), England (1996), Freedman (2000),
Friedman (1999, 2000), Goodhart (2000), Henckel et al. (1999), King (1999), Kobrin
(1997), Krozner (2003) and Palley (2002). Schmitz (2006a) takes a critical view of
the theoretical coherence of these models, arguing that they collapse to either a 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
analysed.
7 This section builds on Schmitz and Wood (2006).
8 Some examples of policy initiatives are the CPSS Core Principles, the Single Euro
Payment Area, the EU New Legal Framework, revisions to Federal Reserve policy on
Will central banking survive electronic money? 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 proposed Community Act in its fields of com-
petence; and by national authorities regarding any draft legislative provision in its
fields of competence.
38 This section builds on Schmitz (2002b).
39 Inter alia, King (1999), page 26, illustrates the focus on technology in reasoning
about institutional change in the payment system:
The key to such developments [final settlement by the transfer of real wealth] is
the ability of computers to communicate in real time to permit instantaneous veri-
fication of the credit worthiness of counterparties, thereby enabling private sector
real-time gross settlement to occur with finality. Any securities for which elec-
tronic markets exist could be used as part of the settlement process.
40 The argument does not provide a rationale for legal barriers against potential currency
competition, though. Berentsen (2006) shows that the private provision of fiat-type
currency would be feasible under very restrictive assumptions. It is still considered
suboptimal in his model.
41 Selgin and White (2002) argue that monetary policy becomes even more effective as
the elimination of currency would reduce the variability of the money multiplier and,
thus, increase the predictability of the relationship between central bank money and
nominal spending. Furthermore, the ratio of central bank money to broad money is
reduced so that each unit change becomes more effective at the margin.
42 Goodhart (1989), page 271.
43 See Freedman (2000), Bindseil and Würtz (2006), White (2006), Sellon and Weiner
(1997) and Woodford (2002).
44 Schmitz (2006b) provides a more detailed presentation and derivation of the ideas
included in this section.
45 Irrespective of the fact that in extended netting systems private clearing and settle-
ment institutions allow for the extension of settlement and the exchange of debt
instruments (often highly liquid government bonds) as collateral in net payment
systems to economise on central bank reserves, final settlement takes place in the
generally accepted medium of exchange, eventually.
46 Schmitz (2006b).
References
Bank for International Settlements (1990) Report of the Committee on Interbank Netting
Schemes of the Central Banks of the Group of Ten Countries (Lamfalussy Report),
Committee on Payment and Settlement Systems Publication No. 4.
Bank for International Settlements (2001) Core principles for systemically important
payment systems, Committee on Payment and Settlement Systems Publication No. 43.
Bank for International Settlements (2003a) Policy issues for central banks in retail pay-
ments, Committee on Payment and Settlement Systems Publication No. 52.
Bank for International Settlements (2003b) The role of central bank money in payment
systems, Committee on Payment and Settlement Systems Publication No. 55.
Bank for International Settlements (2005) New developments in large-value payment
systems, Committee on Payment and Settlement Systems Publication No. 67.
Bank for International Settlements (2006) Statistics on payment and settlement systems in
selected countries – figures for 2004, Committee on Payment and Settlement Systems
Publication No. 74.
252 S.W. Schmitz
Berentsen, A. (2006) ‘On the private provision of fiat currency’, European Economic
Review, 50: 1683–98.
Bindseil, U. and Würtz, F. (2006) ‘Payment systems from the monetary policy implemen-
tation perspective’, in S.W. Schmitz and G.E. Wood (eds) Institutional Change in the
Payments System and Monetary Policy, London: Routledge.
Browne, F.X. and Cronin, D. (1996) ‘Payment technologies, financial innovation, and
laissez-faire banking: a further discussion of the issues’, in J.A. Dorn (ed.) The Future
of Money in the Information Age, Washington, DC: Cato Institute.
Centi, J.P. and Bougi, G. (2003) ‘The possible economic consequences of electronic
money’, in J. Birner and P. Garrouste (eds) Austrian Perspectives on the New
Economy, London: Routledge.
Cohen, B.J. (2002) ‘Monetary instability: are national currencies becoming obsolete?’, in
J. Busumtwi-Sam, M. Griffin Cohen, L. Dobuzinskis and S. McBride (eds) Turbulence
and New Directions in Global Political Economy, London: Palgrave Macmillan.
Costa Storti, C. and De Grauwe, P. (2003) ‘Monetary policy in a cashless society’, in M.
Balling, F. Lierman and A. Mullineux (eds) Technology and Finance, Challenges for
Financial Markets, Business Strategies and Policy Makers, London: Routledge, 241–60.
Crede, A. (1995) ‘Electronic commerce and the banking industry: the requirement and
opportunities for new payment systems using the internet’, Journal of Computer Medi-
ated Communication, 1(3). Online, available at: http://jcmc.indiana.edu/vol1/issue3/
crede.html (accessed 1 November 2006).
Einaudi, L. (1953) ‘The theory of imaginary money from Charlemagne to the French
Revolution’, in F.C. Lane and J.C. Riemersma (eds), Enterprise and Secular Change:
Reading in Economic History, London: Allen & Unwin.
England, C. (1996) ‘The future of currency competition’, in J.A. Dorn (ed.) The Future of
Money in the Information Age, Washington, DC: Cato Institute.
European Central Bank (2006a) The Implementation of Monetary Policy in the Euro
Area, Frankfurt am Main: European Central Bank. Online, available at:
www.ecb.int/pub/pdf/other/gendoc2006en.pdf (accessed 1 November 2006).
European Central Bank (2006b) Payment and Securities Settlement Systems in the Euro-
pean Union and in the Acceding Countries: Addendum Incorporating 2004 Data,
Frankfurt am Main: European Central Bank. Online, available at: www.ecb.int/
pub/pdf/other/bluebook2006addenden.pdf (accessed 1 November 2006).
European Commission (2003) Communication from the Commission to the Council and
the European Parliament Concerning a New Legal Framework for Payments in the
Internal Market (Consultative Document), COM (2003) 718.
Federal Reserve System (2002) Alternative Instruments for Open Markets and Discount
Window Operations, Washington, DC: Federal Reserve System.
Federal Reserve System (2004) The 2004 Federal Reserve System Payments Study,
Washington, DC: Federal Reserve System.
Federal Reserve System (2006) Consultation Paper on Intraday Liquidity Management
and Payment System Risk Policy, Washington, DC: Federal Reserve System.
Fratianni, M. and Spinelli, F. (2006) ‘Did Genoa and Venice kick a financial revolution in
the Quattrocento?’, Oesterreichische Nationalbank Working Paper No. 212.
Freedman, C. (2000) ‘Monetary policy implementation: past, present, and future: will the
advent of electronic money lead to the demise of central banking?’, International
Finance, 3: 211–27.
Friedman, B. (1999) ‘The future of monetary policy: the central bank as an army with
only a signaling corps?’, International Finance, 2: 321–38.
Will central banking survive electronic money? 253
Friedman, B. (2000) ‘Decoupling at the margin: the threat to monetary policy from the
electronic revolution in banking’, International Finance, 3: 261–72.
Fry, M. (1999) ‘Risk, costs and liquidity in alternative payment systems’, Bank of
England Quarterly Bulletin, 39(1): 78–86.
Goodhart, C.A.E. (1988) The Evolution of Central Banks, Cambridge, MA: MIT Press.
Goodhart, C.A.E. (1989) Money, Information, and Uncertainty, London: Macmillan.
Goodhart, C.A.E. (2000) ‘Can central banking survive the IT revolution’, International
Finance, 3: 189–209.
Henckel, T., Ise, A. and Kovanen, A. (1999) ‘Central banking without central bank
money’, IMF Working Paper No. 99/92.
Humphrey, D.B., Sato, S., Tsurumi, M. and Vesala, J.M. (1996), ‘The evolution of pay-
ments in Europe, Japan, and the United States’, IMF Policy Research Working Paper
No. 1,676.
Kindleberger, C.P. (1984) A Financial History of Western Europe, London: Allen &
Unwin.
King, M. (1999) ‘Challenges for monetary policy: old and new’, Bank of England Quar-
terly Bulletin, 39(4): 397–415.
Kobrin, S.J. (1997) ‘Electronic cash and the end of national markets’, Foreign Policy,
107: 65–77.
Kohn, M. (1999) ‘The Origins of Western Economic Success: Commerce, Finance, and
Government in Pre-Industrial Europe’, unpublished thesis, Dartmouth University.
Online, available at: www.dartmouth.edu/~mkohn/ (accessed 1 November 2006).
Krozner, R.S. (2003) ‘Currency competition in the digital age’, in D. Altig and B. Smith (eds)
Evolution and Procedures in Central Banking, New York: Cambridge University Press.
Krüger, M. (1999) ‘Towards a moneyless world?’, University of Durham Department of
Economics & Finance Working Paper No. 99–16.
Lacker, J.M., Walker, J.D. and Weinberg, J.A. (1999) ‘The Fed’s entry into check clear-
ing reconsidered’, Federal Reserve Bank of Richmond Economic Quarterly, 85: 1–31.
McAndrews, J. and Trundle, J. (2001) ‘New payment system designs: causes and con-
sequences’, Bank of England Financial Stability Review, 11: 127–36.
Menger, C. (1909) ‘Geld’, Handwörterbuch der Staatswissenschaften, Jena; trans. (2002)
‘Money’, in M. Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Pay-
ments Systems: From Barter to Electronic Money, Cheltenham: Edward Elgar.
Palley, T.I. (2002) ‘The e-money revolution: challenges and implications for monetary
policy’, Journal of Post Keynesian Economics, 24: 217–33.
Payments Risk Committee (2003) Managing Payment Liquidity in Global Markets: Risk
Issues and Solutions, New York: Payments Risk Committee. Online, available at:
www.newyorkfed.org/prc/manage.pdf (accessed 1 November 2006).
Prescott, E.S. and Weinberg, J.A. (2003) ‘Incentives, communication, and payment
instruments’, Journal of Monetary Economics, 50: 433–54.
Rip, A. and Kemp, R. (1998) ‘Technological change’, in S. Rayner and E.L. Malone
(eds) Human Choice and Climate Change, Columbus, OH: Batelle Press.
Schmitz, S.W. (2002a) ‘Carl Menger’s “Money” and the current neoclassical models of
money’, in M. Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Pay-
ments Systems: From Barter to Electronic Money, Cheltenham: Edward Elgar.
Schmitz, S.W. (2002b) ‘The institutional character of electronic money schemes’, in M.
Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Payments Systems:
From Barter to Electronic Money, Cheltenham: Edward Elgar.
Schmitz, S.W. (2006a) ‘Models of worlds without money: a critical assessment of the
254 S.W. Schmitz
literature’, in S.W. Schmitz and G.E. Wood (eds) Institutional Change in the Payments
System and Monetary Policy, London: Routledge.
Schmitz, S.W. (2006b) ‘Monetary policy in a world without central bank money’, in
S.W. Schmitz and G.E. Wood (eds) Institutional Change in the Payments System and
Monetary Policy, London: Routledge.
Schmitz, S.W. and Wood, G.E. (2006) ‘Institutional change in payment systems and
monetary policy: an introduction’, in S.W. Schmitz and G.E. Wood (eds) Institutional
Change in the Payments System and Monetary Policy, London: Routledge.
Selgin, G.A. and White, L.H. (2002) ‘Mengerian perspectives on the future of money’, in
M. Latzer and S.W. Schmitz (eds) Carl Menger and the Evolution of Payments
Systems: From Barter to Electronic Money, Cheltenham: Edward Elgar.
Sellon, G.H. and Weiner, S.E. (1997) ‘Monetary policy without reserve requirements:
case studies and options for the United States’, Federal Reserve Bank of Kansas City
Economic Review, Second Quarter: 6–30.
White, L.H. (2006) ‘Payment system innovations in the United States since 1945 and
their implications for monetary policy’, in S.W. Schmitz and G.E. Wood (eds) Institu-
tional Change in the Payments System and Monetary Policy, London: Routledge.
Woodford, M. (2002) ‘Financial markets efficiency and the effectiveness of monetary
policy’, Federal Reserve Bank of New York Economic Policy Review, 8(1): 85–94.
16 Payment systems and central
banks
Where are we now and where will
e-payments take us?
Charles Freedman1
This chapter focuses on two major issues. The first relates to the trade-off
between risk containment, on the one hand, and cost or efficiency, on the other
hand, in the design of large-value clearing and settlement systems for payments,
securities, and foreign exchange. In the course of the discussion of this issue, we
touch upon the role of central banks in the oversight or regulation of such
systems and of retail payment systems. The second issue involves the implica-
tions for central banks and monetary policy of a world of electronic money.
The ability of central banks to achieve the desired level of very short-term
interest rates should not be hampered in a significant way by the spread of
e-money. However, techniques to adjust the supply of reserves or settlement
balances might have to be modified if the amount of bank notes on the
central bank balance sheet declined very substantially. In particular, central
banks might have to issue other kinds of liabilities, such as central bank
bills, to give them the resources to purchase securities for use in open-
market transactions.
The chapter by Schmitz in this volume uses the traditional model in which
the overnight interest rate is determined by the intersection of the demand curve
and the supply curve for reserves or base money. While appropriate for some
systems, such as those in place in the United States and in the Euro area, it is not
particularly relevant for the corridor-type system that is now in place in coun-
tries such as Canada, Australia, and New Zealand that have eliminated reserve
requirements. In the Canadian system, for example, as part of its standing liquid-
ity facilities, the central bank stands ready to make whatever loans are needed by
banks short of funds as a result of payment system outcomes at an interest rate
of 25 basis points above the target policy interest rate, and to accept all deposits
that banks wish to hold with it at an interest rate of 25 basis points below the
target policy interest rate. Effectively, the infinitely elastic supply of loans and
the infinitely elastic demand for deposits by the Bank of Canada at the pre-
scribed interest rates force the overnight interest rate to fall within the corridor
or band.
There are (at least) two important aspects of this type of system. First, the
supply and demand for base money (or settlement balances) becomes a sec-
ondary feature of the system, with the setting of the corridor interest rates the
primary mechanism of central bank influence. Second, in some versions of the
system, the total supply of settlement balances can be set equal to zero without,
in any way, lessening the ability of the central bank to control the overnight
260 C. Freedman
interest rate. Indeed, in the case of the Bank of Canada mechanism, the overall
supply of settlement balances was for some period of time actually set equal to
zero. More recently, for reasons of operational efficiency, the banks have asked
the Bank of Canada to supply a small amount ($50 million) of settlement bal-
ances at the end of each day.
The question of whether a central bank can continue to influence the very
short-term interest rate in a case where its balance sheet shrinks to zero because
its liabilities are replaced by some form of e-money, as the means of final settle-
ment in payment systems, has been the subject of some discussion and disagree-
ment.4 I would argue that even in such unlikely circumstances the central bank
could continue to influence the very short-term interest rate, although the
mechanisms it could use might have an artificial element about them. One possi-
bility would involve the central bank insisting on settlement of its own transac-
tions on its own balance sheet and refusing to settle through alternative
mechanisms. This type of arrangement would be strengthened if the central bank
continued to act as banker for the government. A second possibility would
involve a situation in which the central bank continued to establish a corridor for
the overnight interest rate by providing standby facilities in which it was pre-
pared to accept overnight deposits and to extend overnight loans at prescribed
rates of interest. If the overnight rate of interest in the market were tending to
decline below the central bank’s deposit rate, market participants would choose
to hold overnight deposits with the central bank. Conversely, if the overnight
rate were tending to increase above the top of the band, borrowing institutions
would turn to the central bank for funds. The central bank would fund these
loans by issuing its own liabilities (either marketable paper or deposits on its
books).
More fundamentally, as Woodford (2000) argues, the unit of account in a
purely fiat system is defined in terms of the liabilities of the central bank.
Because of this, the central bank can clearly define the nominal yield on
overnight deposits in settlement accounts as it chooses. The special feature of
central banks in this view is that they are entities, the liabilities of which happen
to be used to define the unit of account in a wide range of contracts that people
exchange with one another. Woodford goes on to argue that the question about
such an arrangement is how much central banks’ monetary policies would
matter. And that would depend on how many people still choose to contract in
terms of the currencies, the values of which central banks continued to deter-
mine. That in turn would depend on the cost of transacting in different types of
units of account.
While all this is theoretically fascinating, it is very unlikely to become a prac-
tical issue in any foreseeable future. As long as central banks provide stable
value for their currencies, it is likely that the public will continue to transact in
those currencies, and alternative units of account with equally low transaction
costs are not likely to develop.
Payment systems and central banks 261
Notes
1 I would like to thank Clyde Goodlet of the Bank of Canada for helpful comments on an
earlier draft of these remarks.
2 It is worth noting that some central banks were critical of the LVTS arrangements
since the LVTS differed in structure from an RTGS. This may have reflected a lesser
concern with collateral costs than was the case in Canada.
3 Among the challenges to the spread of electronic money were the high initial fixed
costs that would have been needed to establish the networks and the fact that the
pricing of competing instruments might have made it more difficult to charge for
e-money services (such as reloading cards, home readers, etc.).
4 The debate about the influence on interest rates of a central bank in a world in which
e-money replaces central bank liabilities can be found in Friedman (1999) and the
symposium in the July 2000 issue of International Finance.
References
Bank for International Settlements (1996) Implications for central banks of the develop-
ment of electronic money, Basel: Bank for International Settlements.
Friedman, B.M. (1999) ‘The future of monetary policy: the central bank as an army with
only a signal corps?’, International Finance, 2: 321–38.
Woodford, M. (2000) ‘Monetary policy in a world without money’, International
Finance, 3: 229–60.
Index
Wallace, N. 35, 48, 58, 77; and Cavalcanti, zero collateralization 146, 147, 151
R. 48, 58, 75–85; and Kocherlakota, N. zero intraday interest rate 163, 166–8, 172
84 zero marginal costs 244, 247
weakly implementable allocations 78–80 zero unwind risk 129
Weinberg, J.A. and Prescott, E.S. 239 zero-pricing convention 220
welfare 98; social 96–8, 148 Zhou, R. 56, 65, 87, 163