Payment System Risks

Payment systems are potential sources of contagion which can transmit credit or liquidity problems directly and quickly from one system participant to the rest. Such domino effects are called systemic risks. Simulations of various failure scenarios in Kronos and the Sumclearing, the two central payment systems in Denmark, have been conducted. The simulations were carried out to quantify the extent to which the largest individual participant's problems can be transmitted to the other system participants. The simulations show a limited transmission to the other participants, but they also show that ample liquidity in the systems is a decisive factor.

Payment system risks

A payment system's exposure to systemic risks depends on its design and functioning. In principle, there is no spreading of credit risk in Kronos and only limited spreading of credit risk in the Sumclearing, but there is a risk of spreading liquidity risk^[1]. However, both systems are generally regarded as secure and stable. In an analysis of the systems' stability attempts have been made to quantify the extent to which liquidity risks can spread from one participant to the other participants in the Sumclearing and Kronos respectively. The method applied was simulations of various failure scenarios for the two systems^[2].

The Sumclearing and Kronos are the backbone of the settlement of retail and interbank payments respectively in Denmark. In 2001, the total transaction value in absolute terms in the Sumclearing was thus three times higher than the Gross Domestic Product (GDP) and in Kronos almost 30 times higher than GDP.

The simulations concern two different failure scenarios, both based on the largest system participant having problems^[3]. In the financial failure scenario the largest payment system participant is assumed to be subject to compulsory liquidation. Its removal from the system then results in inability to settle the participant's payments on own behalf and on behalf of others. In the operational failure scenario the largest participant is assumed to be unable to connect to the system for technical reasons for a given period.

These simulations confirm that Kronos and the Sumclearing are secure and stable payment systems. The removal of the largest participant has little effect on the rest. However, a prerequisite in the simulations is ample available liquidity such as the current level. It can thus be useful to monitor the liquidity in the payment systems.

When interpreting the simulation results it is important to bear in mind that the data reflect the participants' payment behaviour given the amount of liquidity actually available in the period under review. This behaviour would thus most likely be different in a failure scenario. It must also be emphasised that the simulations are isolated analyses which do not take into account the interaction between settlement of payments in Kronos, the Sumclearing and the VP settlement system.

Simulations of systemic risk in the Sumclearing

The Sumclearing is the Danish retail payment system for settlement of Dankort (debit card) payments, cheques, account-to-account payments and BetalingsService (direct debit) payments. The Sumclearing is a netting system. At fixed times of the day the 67 participants' multilateral net positions^[4] are calculated and subsequently settled via the participants' accounts with Danmarks Nationalbank^[5]. The payments of a participant with insufficient liquidity are removed from the settlement process.

The Sumclearing is subject to a potential systemic risk, since the removal of a participant could mean that other participants expecting to receive liquidity from the removed participant become unable to fulfil their own obligations. In the following this systemic risk is clarified by simulation of a financial and an operational failure scenario based on actual data for 22 settlement days in the period 21 December 2001-25 January 2002^[6]. A measure of systemic risk can be the number of participants unable to fulfil their settlement obligations as a consequence of compulsory liquidation of or operational problems for the largest participant. The analysis does not consider whether the participants are able to provide more liquidity in the settlement than they actually did on the days in question.

In the financial failure scenario the participant with the largest multilateral net position is assumed to be subject to compulsory liquidation. However, in the cases where Danmarks Nationalbank has the largest position the participant with the second-largest multilateral net position is assumed to be subject to compulsory liquidation. This is the case on 17 of the 22 days in the analysis. Danmarks Nationalbank's often very large positions in the Sumclearing reflect its role as banker for central-government payments. In the operational failure scenario outgoing central-government payments are assumed to be subject to operational problems^[7].

Financial failure scenario in the Sumclearing

The simulations of the financial failure scenario show that in general, compulsory liquidation of the largest participant does not lead to systemic effects in the Sumclearing. On only one of the 22 days in the analysis did problems spread from the liquidated participant to the other participants, and in this case the multilateral net position of only one participant exceeded the maximum drawing access. The analysis does not consider whether the participant in question could have provided extra liquidity for settlement.

The absence of systemic effects in the Sumclearing in the period under review indicates that there is ample liquidity in the Danish financial sector and that the participants provide considerably more liquidity for settlement in the Sumclearing than is strictly necessary. Chart 43 shows the excess liquidity in the actual settlement on the days in question.

Chart 43 Excess liquidity in the sumclearing

Chart 43 Excess liquidity in the sumclearing

The inherent risk in netting systems, which is also present in the Sumclearing, can be illustrated by a number of simulations in which the participants' liquidity is equal to their drawing requirements before compulsory liquidation of a participant^[8]. This shows how compulsory liquidation of the largest participant affects the liquidity of the other system participants.

The simulations with hypothetical and very limited maximum drawing access show that on average just over one third of the participants drop out of the settlement in the event of compulsory liquidation of the largest participant. Chart 44 shows how many of the 67 participants go through with the settlement and how large a share of the gross amounts is settled on the individual days, subject to the given assumptions.

Chart 44 Systemic effects at no excess liquidity

Chart 44 Systemic effects at no excess liquidity

With a view to assessment of the sensitivity of the results to changes in liquidity, simulations have been conducted with reduction of the excess liquidity in the system to 75, 50 and 25 per cent respectively of actual excess liquidity. The systemic risk in the Sumclearing turns out to be negligible even on reduction of excess liquidity to 25 per cent of the original level. On only 3 of the 22 days did one other participant encounter liquidity problems, and on average almost the entire part of the original gross amount which does not concern the participant subject to compulsory liquidation is settled. The correlation between liquidity and systemic risk is shown in Table 6.

Table 6 Financial failure scenario – liquidity and systemic risks

Excess liquidity	100 per cent	75 per cent	50 per cent	25 per cent	0 per cent
Number of days with systemic effects	1	1	1	3	22
Number of participants affected (average)	1	1	1	1	23

The conclusion of the analysis of the financial failure scenario is that in the present liquidity situation the Sumclearing is robust to compulsory liquidation of the largest participant. This also applies in the event of reduction of liquidity from the present level. However, in the absence of excess liquidity systemic effects occur on all 22 days.

Operational failure scenario in the Sumclearing

The operational failure scenario assumes operational problems in the settlement of central-government payments, excluding these payments from the Sumclearing. In this scenario, systemic effects occur on 2 out of the 22 days. On the one day the non-settlement of central-government payments means that two other participants have insufficient liquidity for settlement and on the other day 10 participants encounter problems. A reduction of the excess liquidity in the system entails an increase in the systemic risk so that problems will still only occur on a few days, but more participants will encounter liquidity problems. Table 7 shows the correlation between system liquidity and systemic risk.

Table 7 Operational failure scenario – liquidity and systemic risks

Excess liquidity	100 per cent	75 per cent	50 per cent	25 per cent	0 per cent
Number of days with systemic effects	2	2	2	2	22
Number of participants affected (average)	6	8	10	16	57

In the hypothetical situation where the maximum drawing access is equal to the participants' drawing requirements (without operational problems) systemic effects occur on all 22 days, as in the financial failure scenario. On 5 of the days in the analysis the settlement system collapses and no participants can complete their settlements. On average 57 participants will drop out of the settlement. The very considerable systemic effects in this scenario show that central-government payments are a central element of liquidity in the Sumclearing, since most of the participants typically receive payments from the central government.

The analysis of the operational failure scenario shows that the systemic effects in the Sumclearing generated by the inability to settle central-government payments are more pronounced than the systemic effects in the financial failure scenario.

Simulations of systemic risk in Kronos

Kronos is the Nationalbank's real-time gross settlement system (RTGS system) for payments in kroner and euro. RTGS systems allow account holders to transfer online payments from their own accounts to the accounts of other account holders. The individual payments are typically large and often time-critical and are settled immediately and finally via the accounts. An RTGS system in principle eliminates the credit risk for the recipient participant, but does not eliminate the liquidity risk. Payments entered to an RTGS system might therefore not be settled immediately, but take their place in a queue^[9].

A number of simulations have been carried out to assess the extent to which a financial and an operational failure for the largest participant generate queuing, which would contribute to delaying settlement of payments. The extent of delays can be regarded as a measure of the systemic risk in Kronos.

The participants are assumed to have access to the exact amount of liquidity required for settlement of all payments without delay under normal circumstances. The simulations are based on the participants' actual book entries over a continuous period of 64 banking days^[10].

In the financial failure scenario the largest participant in terms of value of payments is assumed to be subject to compulsory liquidation at the beginning of the day and is therefore unable to send or receive payments during the day. In the operational failure scenario it is assumed that the largest participant cannot connect to the system until 1 hour before the system closes. It is also assumed that the other participants are unaware of the problem and continue transmitting payments to the participant in question. An important difference between the two scenarios is that in the operational scenario some time elapses, depending on the reason for the technical problem, before the other participants become fully aware of the problem and act accordingly. On the other hand, the other participants are fully aware of a case of compulsory liquidation.

The effect of a failure on settlement delays

Table 8 shows how a financial and an operational failure for the largest individual participant affect settlement delays. The financial failure scenario showed very moderate delays on average, but significant delays on certain days. The delays are also relatively limited in the operational failure scenario, but somewhat longer than in the financial failure scenario. The reason is that a considerable proportion of liquidity is locked on the relevant participant's settlement account until 1 hour before the system closes.

Table 8 The effect of failure on settlement delays

Indicator of delay, r	Financial failure	Operational failure
Average delay	0.04	0.12
Maximum delay	0.26	0.31

Note.:

The delays in the settlement of payments are measured by the indicator, r. The indicator is zero if all payments are settled immediately and 1 if all payments are settled at the end of the day.

The system participants can themselves reduce the delays in the settlement of payments by countermeasures to deal with another participant's operational problems. The model therefore moderates the assumption that all the other participants are unaware of the relevant participant's problems by giving payments to this participant a low priority. However, the simulations indicate that the effect of such measures is very limited, since the indicator of average and maximum delay is 0.11 and 0.30, respectively.

In the financial failure scenario it was not possible to settle all payments during the day given the amount of liquidity available to the participants. This is due to the fact that the total liquidity requirement of the remaining participants at the end of the day was greater after the removal of the largest individual participant. On average, approximately 4.5 per cent of the total payment value was not settled, but on the worst day in the period under review, the share of unsettled payments was 27 per cent. The operational failure scenario showed only one case of unsettled payments at the end of a day^[11].

The simulations in Kronos were subject to relatively restrictive assumptions regarding the amount of available liquidity. The simulations showed that operational and financial failure have an effect on the extent of delays and unsettled payments, but that this effect is limited.

Payment system efficiency

The simulations of failure scenarios in the Sumclearing and Kronos focus on the systems' stability and security. These aspects of payment systems have traditionally been a focus area for central banks. However, in recent years, focus has also been on system efficiency, which is reflected e.g. in the latest international recommendations concerning payment and securities settlement systems^[12]. The recommendations include e.g. that settlement in payment systems should be practical for the participants and efficient for the economy.

The background to central banks' increased focus on both stability and efficiency is that a payment system can be so "secure" that the very high costs of using it will induce the participants to settle payments in less secure systems. The completely secure payment system is thus not a realistic option, and a balance must be struck between security/stability and efficiency.

Examples of balancing stability and efficiency

In the autumn of 2001 Danmarks Nationalbank entered into cooperation with the financial sector and the Danish Securities Centre (VP) to improve the flexibility of the provision of collateral for settlement in the VP settlement system, the Sumclearing and the future currency settlement system, CLS^[13]. Flexibility is enhanced by extending the scope of the automatic collateralisation agreement which is a particularly flexible type of provision of collateral. Currently, the automatic collateralisation agreement is used only in connection with securities settlement, but under the new system it will also be used in the other settlement systems. This enhances the efficiency of the systems as a whole. However, this solution could also reduce the stability, since it creates interdependence between the settlement systems. Delays in one settlement system could e.g. be transmitted to other settlements in so far as collateral is tied in one system until settlement in this system has been completed. However, the recommended model for extending the automatic collateralisation agreement achieves a balance between efficient provision of collateral and the security of payment systems. The decision to extend the automatic collateralisation agreement is thus an example of actively balancing efficiency and stability.

Another example shows that different participants strike different balances between stability and efficiency. There are two types of participants in the Sumclearing and in the VP settlement system: direct and indirect participants. The indirect participant settles payments/trades via the direct participant. This provides for easier settlement at lower cost for the indirect participant. On the other hand, the indirect participant could be exposed to a higher risk, since this procedure is for the indirect participant associated with a credit risk as well as a liquidity risk on the direct participant. Finally, the indirect participant could also be exposed to an operational risk, since the indirect participant risks losing access to settlement if the direct participant drops out.

Footnotes

[1]	Payment system risks are described in Financial Stability, Danmarks Nationalbank, Monetary Review, 2nd Quarter 2001, p. 73f.
[2]	The survey is an element of Danmarks Nationalbank's oversight of systemically important payment systems. Central banks' oversight of these systems is described in further detail in Tobias Thygesen, International Standards for Payment Systems, Danmarks Nationalbank, Monetary Review, 1st Quarter 2001. The Nationalbank's oversight is described in Danmarks Nationalbank, Monetary Review, 2nd Quarter 2001, p. 89f.
[3]	A survey of the implications of problems for the largest individual system participant is a frequently applied method. Principle no. 5 of "Core Principles for Systemically Important Payment Systems" prepared by the central banks of the G10 countries in 2000 requires timely completion of daily settlements in a netting system in the event of an inability to settle by the participant with the largest single settlement obligation.
[4]	Multilateral net positions: a figure for each participant stating how much the participant must pay or receive in overall or net terms in the settlement in question.
[5]	The settlement concept and account structure of Danmarks Nationalbank are described in further detail in "Financial Institutions' Accounts at and Pledging of Collateral to Danmarks Nationalbank", Danmarks Nationalbank, Monetary Review, 4th Quarter 2001.
[6]	The simulations have been conducted in cooperation with the Danish Bankers Association and PBS, respectively the owner and operator of the Sumclearing. The methodology and results will be described in further detail in a subsequent working paper to be published by Danmarks Nationalbank.
[7]	Settlement of central-government payments is undertaken by a private bank. However, central-government payments are kept separate from the bank's own payments, since central-government payments are processed in the Sumclearing together with Danmarks Nationalbank's payments. Central-government payments will thus be unaffected, should the handling bank be subject to compulsory liquidation, but it cannot be ruled out that operational problems can lead to inability to effect central-government payments. Operational problems in connection with settlement of central-government payments are simulated by removing Danmarks Nationalbank from the settlement process. Danmarks Nationalbank's own payments in the Sumclearing are negligible compared to central-government payments.
[8]	This is not a realistic scenario, since the participants in a netting system should always have a certain amount of buffer liquidity.
[9]	Kronos is described in further detail in Thomas Angelius and Astrid Henneberg Pedersen, Danmarks Nationalbank's New Payment System Kronos, Danmarks Nationalbank, Monetary Review, 1st Quarter 2002.
[10]	The simulations are described in further detail in E-money and Payment System Review, Central Banking Publications, 2002.
[11]	All these payments could be settled by applying a gridlock resolution mechanism in the operational failure scenario. In the financial failure scenario the gridlock resolution mechanism contributed less to reducing the number of unsettled payments. Gridlocks and gridlock resolution are described in further detail in Morten Linnemann Bech and Kimmo Soramäki, Gridlock Resolution in Payment Systems, Danmarks Nationalbank, Monetary Review, 4th Quarter 2001.
[12]	See "Core Principles for Systemically Important Payment Systems" published by the G-10 countries' central banks in 2000 and "Recommendations for Securities Settlement Systems" published by the G-10 countries' central banks in cooperation with IOSCO (International Organisation of Securities Commissions) in 2001. Both reports are available at www.bis.org.
[13]	CLS is described in further detail in Danmarks Nationalbank, Report and Accounts 2001, p. 73.

	Face - Index - Top/Bottom - Previous/Next
	Version 1.0 Maj 2002 Nationalbanken. Published by Danmarks Nationalbank Maj 2002, http://www.nationalbanken.dk/