... have clearly-defined procedures for the management of credit risks and liquidity risks which specify the respective responsibilities of the netting provider and the participants. These procedures should also ensure that all parties have both the incentives and the capabilities to manage and contain each of the risks they bear and that limits are placed on the maximum level of credit exposure that can be produced by each participant.

3.1 In netting schemes that are strictly bilateral the counterparties are necessarily responsible for all of the risks they bear as a result of their own activities. If a bilateral scheme has a sound legal basis, and both parties have a clear understanding of the nature and extent of the risks they bear, each party should have sufficient incentives to manage and contain those risks.

3.2 Multilateral netting systems which provide for the netting of financial contracts or payment orders among a number of participants and produce legally binding net positions necessarily have a bilateral element. This is the relationship between each participant and the clearinghouse or central counterparty, the expression of which is the multilateral net -or "net-net" -position of each participant. But all multilateral systems also necessarily involve a multiplicity of relationships which make possible the shifting of risks both among participants and between the participants and the central counterparty or netting provider. In particular, multilateral systems which rely on the substitution of a central counterparty involve the additional risk that the central counterparty itself could fail, imposing losses on all of the system's participants. Thus, in order to contain systemic risks, all multilateral netting systems need to have procedures that identify, quantify and allocate each of the risks and that clearly define the responsibility for managing these risks.

3.3 The most basic issue to be addressed is the division of risk-management responsibility between the clearinghouse, central counterparty, or netting provider and the participants. A range of approaches to this issue can be thought of as reflecting different degrees of centralisation of risk allocation and risk management. In a fully centralised system, the netting provider would directly take on all risks associated with the system and undertake all aspects of risk management. It may alternatively be possible for some responsibilities to be decentralised so that the participants would have both the incentives and the ability to take an active part in risk management. However, in any risk-management structure, it is critical that there be a strong link between the ability to contain risks and the incentives to do so.

3.4 Multilateral systems should also be able to ensure that the credit exposures of the central counterparty, produced by the activities of each participant, are kept well within each participant's financial resources. In systems with centralised risk-management, this will be undertaken by the netting provider in the management of the system's direct exposure to each participant. But even in decentralised systems, where participants have responsibilities for the management of counterparty credit exposures, it will be necessary for the netting provider to place some form of overall limits on participants' exposures in order to ensure that their total exposures remain within reasonable levels.

Centralised Risk Management

Contract netting systems.

3.5 Systems for the multilateral netting of forward financial obligations, such as over-the-counter foreign exchange contracts, could be designed on a centralised model similar to that typically employed by the clearing organisations associated with options and futures markets. The central counterparty or clearinghouse would become the counterparty on each contract executed and confirmed by any two participants. Each original counterparty would then bear an exposure to the central counterparty and vice versa but would no longer bear any direct exposure to other participants. To limit its own exposures, the central counterparty could require each participant to past collateral in an amount at least equal to the full extent of the central counterparty's exposure to that participant.

3.6 In doing so, the central counterparty would, in effect, place a ceiling on each participant's exposures equal to the collateral posted. The collateral, in turn, would need to be greater than or equal to the sum of (a) the current cost to the central counterparty of replacing all of its outstanding contracts with the participant; (b) a cushion to cover potential increases in the net-replacement cost; and (c) Herstatt exposures at settlement arising from commitments by the central counterparty to make payments in some currencies prior to receipt from that participant of payments in other currencies for the same value date. The central counterparty would not accept a contract for netting if doing so would increase its exposure to a participant, measured as a sum of these components, to an amount in excess of the value of the participant's posted collateral. Even so, movements in exchange or interest rates, subsequent to acceptance of a contract, might leave the central counterparty with an uncollateralised exposure before it could require the participant to post further collateral. In the event of a default, the central counterparty could promptly close out all of the defaulting participant's outstanding positions by entering into offsetting contracts with other market participants. Collateral posted by the defaulting participant would be liquidated as necessary.

3.7 Under this sort of arrangement, all responsibilities for risk management reside with the central counterparty who limits the risks through the imposition of collateral requirements. Each participant "prepays" for the risk of its own default by posting collateral. But this cannot ensure that the central counterparty would never incur a loss in excess of a defaulting participant's posted collateral. In a period of large movements in exchange rates immediately followed by failure of a participant, for example, there might be a loss to the central counterparty beyond that covered by the participant's collateral.

3.8 In a centralised system, to the extent that such excess losses could not be borne by the central counterparty's own capital or reserves, they would need to be allocated to the surviving participants and charged against their collateral contributions. If the loss sharing allocation were based on participants' overall level of business with the system, these further losses would, in effect, be "mutualised" among the surviving participants. In such a system, participants would have no direct means of controlling the level of their contingent loss-sharing obligations associated with the default of particular other participants. As a consequence, the prudence and viability of this approach depends upon the central counterparty's ability to place binding limits upon participants' exposures to keep them within the amount of posted collateral.

Payment netting systems.

3.9 Systems for the multilateral netting of payment orders in a given currency can also be designed on the basis of centralised risk management. Such systems could conceivably be structured on the same basis as centralised contract netting systems, with each participant posting collateral equal to the size of its daily settlement position. It is unlikely, however, that all participants' exposures in payment netting systems could be fully collateralised because of their large size. In practice, cross-border payment netting systems have tended to develop as an extension of the traditional correspondent banking service of providing book transfers between accounts. In these systems, participants deliver instructions to debit their accounts and credit a counterparty's account. Where risk management is completely centralised, the central counterparty would guarantee the settlement of the end-of-day net positions.

3.10 Under such netting-provider guarantee arrangements, the only exposures requiring management are the bilateral ones between the central counterparty and the individual participants. The management of these exposures would necessarily be undertaken by the central counterparty. In the event of a net-debtor participant's inability to settle its end-of-day position, the central counterparty would take on both the liquidity exposure and the credit losses, satisfying the full amount of funds owed to participants in net-credit positions from its own credit and liquidity resources. Individual participants would have no need, and thus no incentive, to manage or limit the risks.

3.11 The viability of such systems would depend crucially upon the financial condition of the central counterparty which would need to have sufficient resources at its disposal, in the form of its independent credit standing (or, possibly, assets pledged by participants), to ensure settlement. The central counterparty would also need the ability to place binding limits on the net-debit positions which participants could incur so that participants' net-settlement obligations would remain well within their financial resources. In the absence of such an ability to set limits on net-debit positions the central counterparty would be unable to manage and contain the exposures it bears.

Decentralised Risk Management

Contract netting systems.

3.12 Multilateral systems for the netting of financial obligations, such as foreign exchange contracts, could also be designed to require a degree of ongoing risk-management responsibility directly by the participants. Individual participants would retain significant responsibilities for credit decision-making. Bilateral credit limits, rather than collateral requirements, would be the basic mechanism for limiting counterparty credit exposures. Such decentralisation of risk-management responsibilities among the participants could supplement, but should not completely replace, the need for the central counterparty to manage and contain the level of each participant's obligations to the system.

3.13 In this type of system the fundamental netting process would remain unchanged. Pairs of participants would originate transactions. Once confirmed, the central counterparty would be substituted as the counterparty on the transaction to each of the original trading parties. But while the central counterparty would bear the direct exposure on each transaction, and thus also maintain a running net position with each participant, in the event of a participant's default any resulting credit losses associated with its net-position against the central counterparty would be allocated among surviving participants on the basis of their bilateral dealings with the defaulting party. The central counterparty would not need to impose collateral requirements as a mechanism to limit direct exposures. Instead, each participant would have a strong incentive to set separate bilateral credit limits for every other participant.

3.14 Instead of participants "prepaying" for the possibility of their own default, as in a centralised , collateral -based system , survivors would have a contingent obligation to bear the credit losses from a counterparty's failure. Multilateral netting would ensure that, in the event of one participant's failure, the loss to be allocated among the surviving participants would be no greater than, and generally would be less than, the sum of the bilateral net exposures to the defaulting participant. Thus, the loss to be allocated to a surviving participant generally would be less than, and would not exceed, its bilateral net exposure to the defaulting participant.

3.15 However, in addition to the bilateral credit limits on counterparty exposures, a degree of centralised risk management would also be necessary. In particular, because the central counterparty would be relying upon the surviving participants to cover credit losses in the event of a default, it would need to assure itself that the surviving participants are capable of satisfying their contingent obligations under the loss-sharing agreement. To do so, the central counterparty would need to set limits on the level of participants' contingent obligations that are consistent with their financial resources.

3.16 The central counterparty will also need to assure itself that each participant has the financial resources to satisfy the sum of both its contingent obligations and its direct obligations. Although participants would be expected ultimately to bear the losses from a counterparty's default, and therefore be expected to limit and contain the level of their own counterparty exposures, the central counterparty would also need to place an upper limit on participants' total obligations -both direct and contingent. If acceptance of an additional contract by the central counterparty would create net exposures in excess of this limit, the central counterparty would need to be able to reject or delay acceptance of the contract in order to keep exposures within appropriate limits. Alternatively, the central counterparty could require participants to post collateral to cover the amount of any exposure that would exceed their limits.

3.17 In setting overall limits on the combination of participants' direct and contingent obligations, the central counterparty would need to decide whether to anticipate the default of only a single participant. At a minimum, each participant should be expected to be able to satisfy its own direct obligations plus its contingent obligations in the event of the default of the participant to whom it has its largest counterparty exposure. Alternatively, the overall limit could apply to each participant's direct obligations plus the contingent obligations it would incur in the event of the default of two or more other participants.

3.18 However, if there were any serious doubt as to the individual participants' abilities to satisfy promptly the full amount of their total obligations (up to their overall limit) in the event of a crisis, consideration would need to be given to the appropriateness of requiring collateral to support participants' contingent obligations. Such a requirement could still be consistent with a decentralised approach to risk management provided that it did not also entail a collateralisation of participants' direct obligations.

3.19 Moreover, whatever the level of the participants' limits, all systems should have specific procedures for the allocation of losses in the event of multiple defaults. As in the case of a single participant's failure, such loss-allocation rules should take into account the effects on the participants' incentives to manage their exposures. Various approaches are possible. The sequential application of rules that allocate losses on the basis of bilateral dealings would tend to provide the strongest incentives. Rules that distribute losses from multiple defaults on the basis of the surviving participants' overall activity would tend to mutualise these losses and provide weaker incentives for participants to manage their exposures.

Payment netting systems.

3.20 A similar sharing of both risk and risk-management responsibilities may also be possible in multilateral systems for the netting of payment orders in a single currency. In the event of a participant's inability to settle its net-debit position, the resulting short-fall could ultimately be borne by the other participants on the basis of a pro-rata distribution. However, if such losses were to be allocated to participants -thereby providing them with an incentive to manage these risks -then, in contrast to contract netting systems, the system itself would need to give participants the technical capability to limit these exposures.

3.21 In payment netting systems participants are exposed to settlement risks with respect to payment orders sent to them by other participants. If a participant defaults on its settlement obligation, and if this exposure is to be allocated among the surviving participants on the basis of their bilateral positions with the failed participant, the remaining participants' contingent obligations would be represented by their share of the payment orders sent by the failed participant. It is the activity of other participants' sending payment orders which would give rise to contingent obligations under the loss-sharing agreement and it is this activity that receiving participants would need to be able to limit.

3.22 In contract netting systems, however, which typically involve two-sided trade confirmation procedures, counterparty exposures only arise when two participants enter into a trade. Participants can limit their exposures to a given counterparty by ceasing to enter transactions with that party. As a result, although it may be desirable, it is not absolutely necessary for contract netting systems to provide participants with the technical capacity to limit bilateral exposures on a real-time basis. But this is not the case in payment systems.

3.23 Payment orders may be in settlement of bilateral dealings between the counterparties or they may be payments for customer business. Moreover, payments received may not even be for the account of a bank's own customer but for a correspondent's customer. For participants in payment netting systems to be able to limit the level of their exposures, the system itself will need to provide them with the technical capability to set real-time limits on the level of payments they receive from other participants.

3.24 In principle, the strongest incentives to control exposures are provided by allocating losses in strict proportion to bilateral exposures with a defaulting party. But participants in payment systems are incapable of directly controlling the actual level of bilateral exposure which is determined by the level of payments sent by the other party. Thus, as a practical matter, a loss-sharing formula based upon the bilateral limits which participants set for one another will provide the most effective link between incentives and capabilities to manage risk in payment netting systems.

3.25 Provided that the system gives participants the direct capacity to limit their bilateral exposures within the netting process, decentralised arrangements for risk management in payment netting systems should be quite similar to those necessary in multilateral contract netting systems. Once again, because the netting provider will be relying upon the surviving participants to absorb the short-fall caused by a participant's default on its settlement obligation, the netting provider will need to assure itself that each participant has the financial resources to satisfy the sum of its direct settlement obligation plus its contingent obligations. To do this, the netting provider will need to establish binding overall limits on participants' direct and contingent obligations.

3.26 Both the timing and the nature of the settlement and the failure-to-settle procedures are particularly important for payment netting systems which settle on a same-day basis. Moreover, the process of completing daily settlements and of managing the liquidity risks associated with settlement, in conjunction with the procedures for the allocation of any losses , are vitally important to the soundness of any multilateral netting system . Because of this, it is necessary for the design and operation of both centralised and decentralised approaches to risk-management to take into account the necessary procedures for ensuring the timely completion of settlement.