Interbank systems are of great importance to the economy and the
financial system. Using simulations based on real data from Norges
Bank's settlement system, this article illustrates trade-offs
between delayed payments and liquidity usage in interbank settlement
systems. The simulations demonstrate, for example, that the speed with
which payments are settled may be affected by changes in the liquidity
available to settlement participants. The effect of optimisation
routines in the settlement system is also simulated.
1. Introduction
Banks are linked together by interbank systems, through technical
systems and agreements for clearing and settling money transfers between
banks. Norwegian interbank systems comprise of several systems with
different clearing and settlement procedures for retail payments,
securities trading and individual large-value transactions. Gross
turnover in the Norwegian Interbank Clearing System (NICS), which is the
largest system, is on average NOK 200 billion per day. The bulk of these
transactions is settled over banks' accounts in Norges Bank's
settlement system (NBO). The average daily value of settlements in NBO
is over NOK 150 billion. Most large-value payments in NBO are settled in
NICS-SWIFT (2) gross settlements. Chart 1 shows that these settlements
also account for the bulk of turnover in NBO.
Settlement systems for large-value payments are central to the
financial infrastructure, due to the size of the payment transactions
and the fact that it is important that they are executed correctly and
at the right time. Smoothly functioning systems for large-value payments
are thus crucial to the efficiency of the financial markets, the
stability of the financial system and the implementation of monetary
policy in a country. As they are typically regarded as systemically
important, central banks and supervisory authorities have a particular
interest in how these systems are organised and operated (see separate
box).
[GRAPHIC 1 OMITTED]
In an efficient payment and settlement system, payments are carried
out cost efficiently and with low risk. For participants in the
financial sector, the cost of carrying out payment transactions includes
the cost of producing payment services, the cost of any payment delays
and the cost of payment system participants having to keep a different
asset portfolio than they might otherwise have done, in order to execute
payments. This may, for example, take the form of deposits in the
settlement bank and securities that provide borrowing rights for
carrying out settlement.
Berger, Hancock and Marquardt (1996) present a theoretical
framework for analysing the trade-off between risk (e.g. delayed
completion of payment) and costs in the payment system (e.g. liquidity
costs). A payment system is deemed to be technically efficient if costs
are minimised at a given risk level and risk is minimised at a given
cost level. The simplified illustration given in Chart 2 shows risk
(settlement delays) and costs (liquidity usage), where the curve, FF,
represents a set of technically efficient points. The curve also shows
that risk rises at an increasing rate as costs are reduced (convexity).
Innovations in the payment system, for example, technical developments
that make it possible to carry out payment faster at a given liquidity
level, shift the curve inwards (towards F'F'). Where the
outcome on this line occurs depends on the preferences of participants
in the payment system, represented by curve II. All points on II are in
principle equal for all participants. The curve's concave form
reflects the assumption of a decreasing marginal utility of risk
reduction, in other words, that participants are less willing to pay for
risk reduction from a starting point of low risk, than for a similar
reduction from a high risk level. A number of such curves can be drawn
inside and outside II, where participants are more satisfied the closer
the curve is to the origin in the chart, i.e. the lower risk and costs
are. Point A is the outcome here, given the participants' trade-off
between delays and liquidity usage and the technical possibilities
represented by F'F':
Using a simulation-based approach, this article will illustrate the
trade-offs that exist between payment delays and liquidity usage in
interbank settlement systems. A number of key concepts and features of
settlement systems are introduced in the next section.
2. Features of settlement systems
a) Gross and net settlement
Large-value payments can either be settled individually in gross
systems or included in a clearing that is then settled in a net system.
Other solutions (hybrid systems) also exist. Gross and net systems
entail different risks and costs for settlement participants. Three key
risk/cost elements in a settlement system are liquidity, risk of delay
and credit risk.
In a net system, participants settle the result from an earlier
clearing of incoming and outgoing payments at designated times. Given
the interval that elapses between the time that transactions are
submitted for clearing and the final settlement of the clearing, banks
receiving funds in the settlement implicitly provide credit to other
participants for this period. If a bank that owes money in the clearing
experiences solvency problems after the transactions have been submitted
for clearing, but before final settlement, other banks will be exposed
to credit risk in relation to that bank. In this way, the settlement
system may cause the spread of solvency problems from one bank to
others. This is often called systemic risk and is potentially a danger
to the stability of the financial system. (3) In gross systems, or RTGS
(Real Time Gross Settlement), positions between banks are settled on an
individual basis continuously throughout the day, as soon as the payment
transaction enters the system. A payment transaction can only be settled
if the participant has cover (sufficient liquidity) in their account in
the settlement bank. When this account is debited, the payment is
completed with final effect. The continuous settlement of transactions
entails no credit risk in these systems. Settlement systems for
large-value payments have increasingly been based on RTGS (see box). (4)
[GRAPHIC 2 OMITTED]
From a risk/cost perspective, there are different advantages and
disadvantages attached to gross and net settlement systems. Net
settlement economises on liquidity, as participants only require the
amount needed to cover the results of the clearing. However, as
settlement is delayed, net settlement does expose participants to
potential credit risk. RTGS settlement is carried out swiftly and does
not involve credit risk, but requires more liquidity, as payment
transactions are settled individually. Efficient liquidity management
throughout the day is therefore important for participants in such
settlement systems.
b) Intraday liquidity and transaction cycles
Banks are expected to settle their obligations to customers and
other banks in time. They therefore need liquidity, i.e. funds that can
be used as means of payment in the settlement system. If a bank does not
have sufficient liquidity to fulfil its obligations, payment
transactions from that bank are delayed. For the other participants in
the system, delayed settlement constitutes the risk of an unexpected
need to refinance and possible further delays. In order to avoid delays
in settlement, banks have to manage their liquidity so that payment
obligations can be settled at the right time in the course of the day.
If banks carry out settlement in a central bank, intraday liquidity
typically takes the form of deposits and borrowing facilities against
pledged securities. In NBO, borrowing facilities generally account for
the bulk of banks' disposable liquidity through the day; see Chart
3, which shows disposable liquidity in NBO in the form of borrowing
rights and deposits at start-of-day.
As long as the settlement bank does not provide unlimited,
unsecured and free credit to banks participating in the settlement,
banks will incur costs in connection with acquiring and maintaining
liquidity in order to fulfil their payment obligations. These costs are
linked to the acquisition of liquid funds (direct funding costs) and
also the maintenance of deposits in a settlement account in the central
bank (alternative cost in the form of lost interest income). The fact
that participating banks have to pledge securities as collateral for
borrowing rights in the central bank does entail alternative costs to
the extent that it influences banks' choice of securities
portfolio. Banks also incur direct costs in connection with liquidity
management.
In addition, costs accrue if settlement is delayed or transactions
are not settled at all. As the payments that are transferred in
interbank systems are often large or time-critical, the costs incurred
for banks and their customers may be substantial if transactions are not
settled at the anticipated time. The fact that costs accrue in
connection with maintaining liquidity and in the event of delayed
settlement, banks have incentives to reduce their liquidity costs, but
without it resulting in delays. The trade-off between liquidity costs
and the cost of settlement delay is thus an important consideration for
banks when adjusting their liquidity levels. Different banks may have
different preferences as regards this trade-off, and these can change
over time. If a bank's costs in connection with settlement delays
are substantial in relation to liquidity costs (e.g. because many of the
transactions are time-critical), it will probably choose to hold more
liquidity in order to avoid delays in the settlement.
Several conditions affect participants' liquidity requirements
in a RTGS system. Incoming payments from other settlement participants
are one important source of liquidity. The structure of the banking and
settlement system and how payments flow through the day will also be of
importance to participants' liquidity requirements (i.e. how evenly
incoming and outgoing payments are distributed). A bank can influence
its liquidity needs if outgoing payments are managed to coincide with
incoming payments. Coordination between settlement system participants
may help to reduce the liquidity need and the risk of delays. This can
be achieved through the use of shared information systems and the
general agreement and regulatory framework for the settlement system in
question, including any arrangements for coordinating the exchange of
transactions over the course of the day. Such an arrangement may help to
prevent situations where individual participants intentionally wait for
liquidity from incoming payments before placing their own transactions
in the system (free-riding). (5) In order to economise on liquidity
usage, banks in Norway have coordinated the exchange of gross
transactions in NBO.
A well-designed settlement system can help to make liquidity usage
more efficient, which is particularly pertinent in situations where
there is not much liquidity and payments are queued. If the settlement
system includes elements of both gross and net settlement, improved
recycling of liquidity can be achieved. In RTGS systems, for example, it
is usual to have queue management mechanisms, where transactions that do
not have sufficient liquidity to be settled are placed in a queue in the
settlement system. These transactions then await new liquidity from
later payments and are settled according to more detailed rules for
settlement prioritisation and sequencing. Systems may also include
netting procedures for transactions in the queue, where the netting
effect is bilateral between two participants and/or multilateral between
several participants. Features such as queue management and netting
procedures minimise participants' liquidity need at the same time
as reducing the delays in settlement. The result is a better trade-off
between liquidity and settlement speed than might otherwise be achieved
in a purely gross or net settlement.
The article will use a simulation-based approach to illustrate
trade-offs between different levels of bank liquidity and payment delays
in NBO settlement. But first of all, the data and methodology on which
the analysis is based will be presented.
[GRAPHIC 3 OMITTED]
3. Data and method
A simulation tool developed by the Finnish central bank makes it
possible to carry out simulations based on actual settlement data. (6)
The simulator can be used to analyse the effect of changes in the
liquidity available to settlement participants and/or the introduction
of new settlement procedures. The effect on variables such as the
liquidity requirements, payment delays and the settlement ratio can then
be studied.
The simulations presented in this article were carried out using
settlement data from the RTGS system in NBO (NICS-SWIFT gross
transactions) and by generating systems data in the simulator. The
relevant data from NBO includes participating banks, transactions
between participants (time, sender/receiver and amount) and their
available liquidity (balance in settlement accounts and borrowing
facilities). The settlement procedures and rules are defined in the
simulating tool, including the system's opening hours, how
transactions are settled and any optimisation routines (queuing
function, netting procedures, etc.).
The analysis is based on settlement data from 10 days in October
2005. The days can be characterised as relatively normal, with
transactions for a value of NOK 160 billion on average being settled in
NICS-SWIFT gross settlements per day. This accounts for around 87 per
cent of total turnover in NBO in the period. On average, there were 558
transactions per day and the average size of transactions was NOK 287
million. A maximum of 20 banks participated in the settlement. The
settlement volume is relatively concentrated. The five largest banks
accounted for over 88 per cent of the transaction value. As the banks
coordinate the exchange of transactions, the bulk of the settlement is
carried out between 12.30pm and 1.30pm (69 per cent of the turnover
value). When presenting the results, the average for the days in the
period has been used.
Two types of simulation have been carried out. In section 4 a) the
theoretical reference points for the amount of liquidity needed to
finalise a given flow of gross transactions are calculated. In sections
4 b) and c) the effects of varying liquidity on payment delays and the
volume of unsettled transactions are studied. NBO's features have
been imitated as closely as possible here, based on the following
settlement procedures: when a settlement participant places a payment
transaction in the system, the transaction will be settled immediately
if there is sufficient liquidity (a positive balance and/or borrowing
rights). If the participant lacks liquidity, the transaction is placed
in a queue until there is sufficient cover for settlement. The
transaction will be settled if and when incoming transactions from other
participants can supply sufficient new liquidity or if the
participant's borrowing facilities are increased. Transactions that
are waiting in a queue are managed according to the FIFO principle
("first in, first out"), which means that transactions are
released from the queue for settlement in the order in which they joined
the queue ("longest in the queue, processed first"). When a
queue starts to form, a gridlock mechanism will also try to offset the
transactions between participants both bilaterally and multilaterally.
(7) If a participant still lacks liquidity at end-of-day, the
transaction will not be settled that day.
In the simulations in sections 4 b) and c) information about
participating banks' actual liquidity in NBO has been used. The
simulations were made by changing the level of available liquidity, by
adjusting participants' balances and borrowing facilities by the
same percentage. What determines participants' liquidity needs is
the actual transaction flows through the day and the settlement system
properties. This entails an assumption that banks' behaviour
remains unchanged even though their liquidity level varies. There is,
however, reason to assume that participants' transaction patterns
will also change when liquidity in the settlement is changed.
Furthermore, liquidity is in practice not just used for NICS-SWIFT gross
settlement, but also for other settlements in NBO. On any given day,
other settlements will thus be able to supply or draw in liquidity for
the participant in question. The simulations do not take account of the
fact that some transactions are time-critical.
(8) The results must be evaluated in light of this.
4. Simulation results
a) Theoretical reference points for liquidity requirements
Liquidity requirements in a RTGS system will, among other things,
depend on whether payment transactions are settled immediately or
whether they are placed in a queue for settlement later. This means that
a trade-off between liquidity need and settlement delays has to be
considered. This trade-off can be illustrated by calculating the
reference points for liquidity needs. The concepts of upper bound (UB)
and lower bound (LB) for liquidity requirements are relevant in this
context (Koponen & Soramaki, 1998). UB shows how much liquidity a
participant in a RTGS system needs to ensure that all outgoing payments
are settled immediately when they enter the system (without waiting in
the queue). LB shows the minimum amount of liquidity a bank needs to
cover its net obligations at end-of-day, for all its transactions
through the day. (9) When assessing the trade-off between liquidity
needs and the speed with which payments are settled, UB can be seen as a
situation where the liquidity requirement and settlement speed are both
maximised. LB minimises the liquidity requirement, but it also minimises
the speed of settlement, as all transactions are settled at end-of-day.
The reference points for liquidity needs can be illustrated by
looking at the liquidity cycle throughout the day for a hypothetical
bank in a RTGS system, as shown in Chart 4. The bank starts the day with
a positive balance in its account in the settlement bank. Transactions
through the morning are largely outgoing, so the bank reaches a position
where it has to draw on its borrowing rights in the settlement bank. In
the afternoon, the bank has substantial incoming payments and ends up in
a net deposit position. The bank's largest negative position in the
course of the day is thus the banks" upper bound (UB) or maximum
liquidity requirement (to ensure that its transactions are settled
immediately). For the bank in question, this was at around 8am, when it
used NOK 7 billion of its borrowing facility. The bank's net
payment obligations through the day equal the differential between its
liquidity position at the start and end of the day. This is the same as
the bank's lower bound (LB) liquidity requirement.
The simulations are based on actual gross transactions in NBO. UB
is calculated by simulating a RTGS system where participants have
unlimited borrowing rights to settle outgoing payments immediately. The
individual bank's greatest intraday negative balance is then, as
mentioned, its maximum liquidity requirement. LB is simulated by
carrying out a net settlement at end-of-day. The system's liquidity
need is then the sum of participants' liquidity requirements.
Table 1 shows UB and LB as a share of total turnover in NICS-SWIFT
gross settlements. The result of the simulations was that, on average,
there was a liquidity requirement equivalent to 5 per cent of the
turnover value in order to carry out one net settlement at end-of-day
(LB). If participants had had unlimited borrowing rights, the average
liquidity requirement would be 27 per cent of the turnover value (UB).
As the table indicates, there is some variation in liquidity
requirements over the period.
[GRAPHIC 4 OMITTED]
It is important to emphasise that these limits are theoretical
measures. In order for a liquidity level equivalent to UB to actually
result in maximised settlement speed, it is assumed, among other things,
that the liquidity in the system is optimally divided between
participants at all times. No consideration is taken of the fact that a
number of transactions may be time-critical or that it is possible for
participants to reprioritise transactions that are placed in a queue. In
the event of time-critical transactions, LB will, for example, be too
low as participating banks have to secure liquidity in order to carry
out such transactions at a given time. (10)
b) Liquidity and settlement delays
The starting point for the following simulations was to study how
changes in actual liquidity available to settlement participants
influence the speed with which payments can be settled, or if they can
be settled at all. The settlement procedures imitate NBO as closely as
possible (cf. section 3). Transaction flows are the same as actual
NICS-SWIFT gross settlement in NBO.
Settlement delays as a result of varying liquidity availability can
be measured in several ways. One way to express the overall level of
delays is with an indicator introduced by Leinonen & Soramaki
(1999). The level of delays is measured by [rho]:
[rho] can assume values between 0 and 1. k is an index for each
separate payment, s is the value of each payment, t' is point at
which the payment enters the system, t the point at which the payment is
settled and T is end-of-day. The indicator is based on the time each
separate payment spends in the queue compared with its maximum possible
time in the queue. This gives a value-weighted average of the relative
delay for all payments. If all payments are settled immediately on entry
into the system, [rho] = 0. If [rho] = 1, all payments have remained in
the queue from the time they were placed in the system until end-of-day.
Chart 5 shows the effect of changing the level of available
liquidity on the extent of delayed settlement, measured by the indictor
[rho]. A liquidity level of say 40 percent means that participants are
allocated liquidity equivalent to 40 per cent of their actual balance
and borrowing rights. The chart shows that generally liquidity must be
reduced substantially for the value of the indicator [rho] to rise
noticeably. At a liquidity level of 50 per cent, [rho] has a value of
0.05. If the level is reduced to 20 per cent, [rho] more than doubles to
0.11. With a 5 per cent liquidity level, [rho] increases markedly to
0.34. One key observation is the shape of the curve. The curve is
generally convex, which means that the more the settlement
participants' liquidity is reduced, the greater the delay in
settlement. Or, in other words, at low liquidity levels, a small
injection of liquidity can substantially reduce delays.
Table 2 shows the effects of varying liquidity on some other
indicators for delays and for settlement in NICS-SWIFT gross settlement.
When available liquidity was halved, the average settlement time was,
for example, four minutes." If liquidity was reduced to 20 per
cent,
[rho] = [summation over k] ([t.sub.k] -
[t'.sub.k])[S.sub.k]/[summation over k] ([t.sub.k] -
[t'.sub.k])[S.sub.k]
the average settlement time was 19 minutes. The simulation results
also show that the value of transactions that remained unsettled at this
liquidity level was NOK 7 billion. With a liquidity level of 10 per
cent, this figure rose to NOK 16 billion.
c) Effect of optimisation features
NBO contains features for managing queues and gridlock situations.
The gridlock mechanism attempts, as the name indicates, to resolve
gridlocks, in other words, situations where there is little liquidity
and the queue formation means that several payment transactions between
banks are awaiting settlement. None of the transactions in the queue can
be settled if they are viewed in isolation. If several transactions are
taken for settlement at the same time, they could, however, be settled.
Thus the gridlock mechanism makes the use of liquidity more efficient by
netting the transactions that are waiting in the queue. The netting can
be both bilateral (between two participants) and multilateral (between
several participants). The purpose is to reduce delays in settlement and
the number of transactions that cannot be settled at end-of-day. If
there is still insufficient liquidity to settle the payments after this
procedure, the situation is characterised as 'deadlock'. Only
the supply of new liquidity will then be able to prevent transactions
from remaining unsettled.
[GRAPHIC 5 OMITTED]
Simulations can be used to illustrate the effect of such
optimisation routines. As in the last section, the simulations are based
on a RTGS system with a FIFO queuing function where a gridlock mechanism
attempts to achieve netting effects between participants. Two reference
scenarios are made. In the first, a pure RTGS system without queuing and
gridlock functions is simulated. In the second, a RTGS system with a
FIFO queuing function, but no gridlock mechanism, is simulated. The
effect on the settlement ratio in the different scenarios is then
compared.
Chart 6 shows the value of total unsettled payment transactions as
a share of total turnover. The results show that optimisation routines
ensure a considerably higher settlement ratio than in a system without
such features. The difference is relatively small between RTGS with a
queuing function and RTGS with a queuing and gridlock function at
liquidity levels down to 20 per cent. But below this level, the
difference increases. One reason that the gridlock mechanism is
efficient is that a substantial share of transactions are between a
small number of larger participants and bilateral netting effects can
thus be achieved. At very low liquidity levels, however, the settlement
ratio is reduced noticeably even when there is a gridlock function. This
is because the system increasingly experiences "deadlocks", in
other words, only the supply of new liquidity will increase the level of
settlement. At a 5 per cent liquidity level, more than 27 per cent of
the transaction value was unsettled at end-of-day.
5. Conclusion
Using a simulation-based approach, this article has illustrated
relationships between settlement delays and liquidity usage.
The banks that participate in NBO generally hold liquidity levels
that entail little delay in payment settlements. The simulations
indicate that liquidity must be reduced substantially before
considerable settlement delays occur. It must be emphasised that the
analysis is based on data from a period with relatively normal
transaction volumes and liquidity levels. However, even though the level
of liquidity is sufficient in normal situations, the extent of delays
and unsettled transactions may become significant when a
"critical" liquidity level has been reached. The simulations
regarding the effect of optimisation routines show that these do
contribute to a higher payment settlement ratio.
The liquidity levels that participants in NBO have chosen, may
indicate that the costs of delays in the Norwegian settlement system are
deemed to be relatively high compared with liquidity costs. If the
relative costs of liquidity increase, banks might adapt to new levels of
liquidity and/or adjust their transaction pattern. However, further
analyses would be needed in order to establish whether this might result
in a greater number of delays and unsettled transactions.
[GRAPHIC 6 OMITTED]
Organisation and operation of settlement systems
It is normal that the central bank operates the most important
settlement system in a country. The manner in which the authorities deal
with interbank systems that are not operated by the central bank varies
from country to country, but they are often subject to some form of
public regulation and supervision. In Norway, for example, there is
specific legislation governing this area (the Act relating to Payment
Systems) that vests Norges Bank with responsibility for authorising and
supervising interbank systems. The purpose of the Act is to ensure that
interbank systems are organised in such a way that the consideration of
financial stability is upheld.
"The RTGS revolution"
Net systems with end-of-day settlement were replaced by RTGS
systems with continuous settlement throughout the day in a number of
countries in the 1990s. Technological developments and central
banks' focus on systemic risk were important reasons for this
transition (BIS 2005). In Norway, RTGS was introduced in 1999 in
connection with the modernisation of the settlement system in Norges
Bank (NBO).
References
Berger, Allen, Diana Hancock and Jeffrey Marquardt (1996): "A
Framework for Analyzing Efficiency, Risks, Costs, and Innovations in the
Payments System". Journal of Money, Credit and Banking 28 (4), pp.
696-732.
Bank for International Settlements (2005): "New developments
in Large-Value Payments Systems". May, BIS.
Bank for International Settlements (1997): "Real-Time Gross
Settlement Systems". March, BIS.
Gronvik, Gunnvald and Eline Vedel (1999): "Oppgjorssystemer i
et internasjonalt perspektiv", Sentralbanken i Forandringens
Tegn--Festskrift til Kjell Storvik, Norges Bank Working Papers no. 28,
pp. 72-89 (Norwegian only)
Leinonen, Harry and Kimmo Soramaki (2003): "Simulating
interbank payment and securities settlement mechanisms with the BoF-PSS2
simulator". Bank of Finland Discussion Papers, no. 23/2003.
Leinonen, Harry and Kimmo Soramaki (1999): "Optimising
Liquidity Usage and Settlement Speed in Payments Systems". Bank of
Finland Discussion Papers, no. 16/1999.
Koponen, Risto and Kimmo Soramaki (1998): "Intraday Liquidity
Needs in a Modern Interbank Payment System--A Simulation Approach".
Bank of Finland Studies, no. E: 14/1998.
McAndrews, James and John Trundle (2001): "New Payments
Systems Designs: Causes and Consequences". Bank of England
Financial Stability Review, December, pp. 127-136.
Norges Bank (2005): Annual Report on the Payment System 2004.
(1) The authors would like to extend a special thanks to Steinar
Guribye and Bent Vale for their useful comments.
(2) SWIFT is an acronym for Society of Worldwide Interbank
Financial Telecommunication. In this context, SWIFT means a standard
messaging format for settlement trans actions.
(3) Net systems can, however, be organised in such a way that
credit risk is managed, for example, with deferred customer crediting,
limits on counterparty exposure (caps). loss sharing agreements, etc.
(4) For a more detailed description of RTGS systems, see BIS
(1997). For the introduction of RTGS in Norway, see Gronvik and Vedel
(1999).
(5) For a discussion on payment flows and ways in which to
influence transaction patterns, see Trundle & McAndrews (2001), pp.
131-133.
(6) For a description of the simulator ("BoF-PSS2"), see
Leinonen & Soramaki (2003) or Bank of Finland:
http://www.bof.fi/eng/3_rahooitusmarkkinat/3.4_Maksujarjestelma/
3.4.3_Kehittaminen/3.4.3.3_Bof-pss2/
(7) This procedure has been simplified somewhat in the simulations
in relation Io the actual system.
(8) In the simulations, payments in connection with the foreign
exchange settlement system. CLS. are treated such that payments from CLS
are executed irrespective of banks' pay-ins. Payments in connection
with Scandinavian Cash Pool (SCP) are not included in the simulations.
(9) In situations where liquidity is below LB, transactions will
remain unsettled, whereas a liquidity level below UB means that payments
cannot be settled immediately and have to be placed in a queue.
(10) For more detailed discussion of upper and lower bounds, see
Koponen & Soramaki (1998).
Asbjorn Enge, senior advisor, and Frode Overli, advisor, Payment
Systems Department (1)
Table 1 Liquidity requirement limits as a percentage
of total turnover
Lower bound Upper bound
Average 5% 27%
Highest value 14% 33%
Lowest value 2% 23%
Source. Norges Bank
Table 2 Selected indicators at different liquidity levels (1)
Liquidity level in % of
actual liquidity
5% 10% 20% 30%
Settlement delay
indicator (p) 0.34 0.20 0.11 0.08
Average settlement
time (min) 107 56 19 12
Value of unsettled
payments (NOK billions) 43 16 7 5
Liquidity level in % of
actual liquidity
40% 50% 100%
Settlement delay
indicator (p) 0.06 0.05 0.02
Average settlement
time (min) 7 4 1
Value of unsettled
payments (NOK billions) 3 1 0
Average daily turnover value in period: NOK 160 billion
(1) Results are presented as an average of daily
observations in the sample that describes the settlement.
Source: Norges Bank