At its core, settlement of a foreign exchange (FX) trade requires the payment of one currency and the receipt of another. In the absence of a settlement arrangement that ensures that the final transfer of one currency will occur if and only if the final transfer of the other currency also occurs, one party to an FX trade could pay out the currency it sold but not receive the currency it bought. This principal risk in the settlement of foreign exchange transactions is variously called foreign exchange settlement risk or cross-currency settlement risk. It is also referred to as Herstatt risk, although this is an inappropriate term given the differing circumstances, including those described below, in which this risk has materialised.

FX settlement risk clearly has a credit risk dimension: whenever a party cannot make its payment of the currency it sold conditional upon its final receipt of the currency it bought, it faces the possibility of losing the full principal value involved in the transaction. In this situation, a party's foreign exchange settlement exposure (the size of its credit exposure to its counterparty when settling an FX trade) equals the full amount of the purchased currency. As will be described in Section 3, many banks currently face sizable FX settlement exposures overnight, and indeed for longer periods.

FX settlement risk also has a liquidity risk dimension: if a party did not receive the currency it purchased when due, it would need to cover and to finance this shortfall until its counterparty honoured its obligation. In fact, this liquidity risk is present even if, in this circumstance, a party could withhold its payment of the currency it sold (i.e. liquidity risk can be present even in the absence of credit risk). Thus, whether viewed from a credit or a liquidity perspective, the amount potentially at risk in settling an FX trade equals the full value of the purchased currency.

To be sure, FX trading poses many other forms of risk, including market risk (the risk of loss from an unfavourable exchange rate movement), replacement risk (the risk of having to replace, at current exchange rates, an unsettled yet profitable FX transaction with a failed counterparty) and operational risk (the risk of incurring interest charges or other penalties for misdirecting or otherwise failing to make FX settlement payments on time owing to an error or technical failure). FX market participants must recognise and manage appropriately each of these risks. Nevertheless, since the associated amounts at risk represent only a fraction of the underlying value of each transaction, they are dwarfed by the size of foreign exchange settlement exposures.

2.2 Factors leading to central bank concerns

If these risks were to crystallise in a disorderly manner, they would be likely first to affect the domestic payments systems of the world's major currencies, since a significant share of their daily flows is accounted for by the settlement of foreign exchange transactions. Secure and well-functioning payments systems are necessary for the attainment of central banks' monetary, macroprudential, supervisory and other policy objectives. They are also essential mechanisms in the management by individual commercial banks of their assets and liabilities, and in the settlement of their own transactions as well as those of their customers. It is therefore appropriate that central banks should be concerned that the settlement arrangements in the foreign exchange markets should be structured so as to minimise systemic risk (the risk that the failure of one market participant to meet its required FX settlement or other obligations when due may cause significant liquidity or credit problems for other participants, and so may threaten the stability of the financial markets).

Major commercial banks also now accept that there is an international dimension to the domestic payments system of every major currency. These payments systems are interdependent, given the extent to which banks from a range of different countries are participants, directly or through a local correspondent, in each of them, and therefore have a direct interest in the efficiency and robustness of their settlement arrangements. The market, and in particular the major correspondent banks in each country, now realise that every individual commercial bank and banking sector (however defined) is vulnerable to unexpected endogenous or exogenous events, which could occur on a sufficient scale to cause one or more banks to be unable to settle their foreign exchange trading obligations on any one day.

The scale of these potential settlement problems is demonstrated by the latest survey of FX market turnover. The BIS estimates the average daily turnover of global exchange markets in spot, outright forward and foreign exchange swap contracts at US$ 1,230 billion in April 1995. Since each trade could involve two or more payments, daily settlement flows are likely to amount, in aggregate, to a multiple of this figure, although no comprehensive data are available.

Given the serious domestic and international repercussions that a significant FX settlement disruption could have in a market of this size, a bank might believe that public authorities in some countries would not close a major FX market participant during the day or permit it to default unexpectedly and cause significant losses during the settlement process. This belief might make a bank unwilling to reduce its present settlement exposures, or even increase its willingness to take on even greater settlement exposures with its counterparties. To the extent that this belief is widely held in the market, it has already produced an unacceptable level of risk in the financial system.

Moreover, the extent of this risk is in reality substantially greater than is suggested by estimates of market turnover and settlement flows. The definition of and methodology for measuring FX settlement exposure, as set out in this report, make it clear that it is not just an intraday phenomenon: in practice, FX settlement exposure typically represents overnight risk; it can last for several business days; and it will therefore be present over weekends and public holidays. Furthermore, at any point in time a bank's FX settlement exposure can greatly exceed its capital.

It is also the case that the market's belief that a major FX market participant will not be closed during the day is ill-founded. There is in fact no time, during a weekday, at which the large-value payments systems of every major currency are closed. To the extent that commercial banks maintain this belief, an unnecessary and avoidable element of risk remains in the market.

Set out below are brief summaries of five case studies that demonstrate the ways in which a settlement problem can arise. They also demonstrate that despite the steps that have been taken since 1974 to improve coordination between banking supervisors and to begin to introduce settlement risk control measures in the major financial centres, the possibility of a bank failing or being closed during the business day remains, and any collapse will almost inevitably occur during the business day of one financial centre or another. While the timing of the withdrawal of a banking authorisation may in some circumstances be controllable so as to minimise shocks to the markets, there will be other cases in which a banking supervisor may have little choice as to the timing of its actions. If, for example, a banking supervisor becomes aware that a bank has sustained major losses, sufficient to seriously impair its capital base, it may need to take immediate action of some sort to protect depositors. The timing of this action may also be influenced by the need to ensure that a bank does not continue to trade while insolvent, and by the need in such circumstances to act quickly lest the fact that the institution is in difficulty becomes publicly known, precipitating a "run" on the bank. In some countries it is not legally possible to put a bank into liquidation outside the business hours of the court that must appoint the liquidator.

2.2.1 The failure of Bankhaus Herstatt (1974)

On 26th June 1974 the Bundesaufsichtsamt für das Kreditwesen withdrew the banking licence of Bankhaus Herstatt, a small bank in Cologne active in the FX market, and ordered it into liquidation during the banking day but after the close of the interbank payments system in Germany. Prior to the announcement of Herstatt's closure, several of its counterparties had, through their branches or correspondents, irrevocably paid Deutsche Mark to Herstatt on that day through the German payments system against anticipated receipts of US dollars later the same day in New York in respect of maturing spot and forward transactions.

Upon the termination of Herstatt's business at 10.30 a.m. New York time on 26th June (3.30 p.m. in Frankfurt), Herstatt's New York correspondent bank suspended outgoing US dollar payments from Herstatt's account. This action left Herstatt's counterparty banks exposed for the full value of the Deutsche Mark deliveries made (credit risk and liquidity risk). Moreover, banks which had entered into forward trades with Herstatt not yet due for settlement lost money in replacing the contracts in the market (replacement risk), and others had deposits with Herstatt (traditional counterparty credit risk).

2.2.2 Drexel Burnham Lambert (1990)

In February 1990 the Drexel Burnham Lambert (DBL) group collapsed, the initial cause being severe liquidity problems. The Bank of England had to intervene, as a facilitator, to minimise the impact of DBL's problems on the counterparties of one of its London subsidiaries, Drexel Burnham Lambert Trading (DBLT), which traded as a principal in the foreign exchange and gold markets.

As market awareness grew in February of the extent of the problems in the DBL group, DBLT's counterparties became progressively less willing to incur intraday exposures to it in the settlement of their FX deals. At the same time DBLT was unwilling to pay the amounts it owed on maturing deals, because of concerns that the counterparties might decline to pay the other currency involved and instead set off the receipts from DBLT against amounts due to them from other companies in the DBL group.

After intensive discussions with DBLT, which was required to produce evidence of its solvency, the Bank of England put in place a settlement facility, which remained open for a full week, to resolve this developing gridlock. Under this facility, DBLT's counterparties were invited to pay amounts due into accounts held in the Bank of England's name with the Bank's correspondent bank (in almost all cases the central bank) in each country concerned. Once the Bank had received confirmation that funds had been credited to these accounts it informed DBLT. DBLT then made irrevocable payments of countervalue to each counterparty directly, using funds made available for the purpose by its immediate parent company. Upon receipt of these payments the respective counterparty was asked to confirm to the Bank of England that it was prepared for the Bank to release the relevant deposit to DBLT.

Several key factors were present in the DBLT case which might not all be present in other cases of FX market gridlock. Crucially, DBLT itself was solvent, and it had a relatively small FX book, almost flat in non-dollar terms, and with relatively few forward deals. Its immediate parent in the DBL group was willing to provide the initial liquidity needed to enable DBLT to settle all amounts due. Finally, the Bank of England was in a position to act as a neutral facilitator, acceptable to all parties, and was able to work with the management and staff of DBLT, who remained in place for the whole of that week.

2.2.3 BCCI (1991)

The appointment of a liquidator to BCCI SA on 5th July 1991 caused a principal loss to UK and Japanese foreign exchange counterparties of the failed institution.

An institution in London was due to settle on 5th July 1991 a dollar/sterling foreign exchange transaction into which it had entered two days previously with BCCI SA, London. The sterling payment was duly made in London on 5th July. BCCI had sent a message to its New York correspondent on 4th July (a public holiday in the United States) to make the corresponding US dollar payment for value on 5th July. The payment message was delayed beyond the time of the correspondent bank's initial release of payments (at 7 a.m.) by the operation of a bilateral credit limit placed on BCCI's correspondent by the recipient CHIPS member. The payment remained in the queue until shortly before 4 p.m. (New York time), when it was cancelled by BCCI's correspondent, shortly after the correspondent had received a message from BCCI's provisional liquidators in London on the subject of the action it should take with regard to payment instructions from BCCI London. In this way, BCCI's counterparty lost the principal amount of the contract.

A major Japanese bank also suffered a principal loss in respect of a dollar/yen deal due for settlement on 5th July, since yen had been paid to BCCI SA Tokyo that day, through the Foreign Exchange Yen Clearing System, and the assets of BCCI SA in New York State were frozen before settlement of the US dollar leg of the transaction took place.

The UK institution's loss illustrates a particular aspect of the difficulties which face the private sector under current circumstances in any attempt to coordinate the timing of payments; in this instance, the loss would almost certainly not have occurred but for the measures in place to reduce risk domestically within CHIPS. Moreover, the closure of BCCI by the banking supervisors illustrates that it is generally not possible to close a bank which is active in the foreign exchange market at a time when all the relevant payments systems have settled all its transactions due on a given day. In this case, the closure required the Luxembourg Court to appoint a liquidator, an action which under Luxembourg law can take place only within the normal business day of the Court.

2.2.4 The attempted Soviet coup d'état (1991)

The short-lived coup d'état in Moscow in August 1991 led to uncertainty about the status and possible actions of certain financial institutions based in, or owned by institutions in, the then Soviet Union. For a few days the uncertainty had a disruptive effect on settlement in the foreign exchange market, in which these institutions were active traders. Some of their market counterparties were unwilling, given the political climate, to expose themselves to what they saw as potentially very acute principal risk in settling their maturing FX contracts. They instead pressed for the receipt of countervalue (or a guarantee from an acceptable third party) in advance of releasing funds. As a result, some deals were not settled when due.

There were also some instances of unwillingness on the part of the Soviet-based institutions' correspondent banks to release funds even when countervalue had been received, including at least one attempt by a correspondent to withhold funds it was due to pay out to its customer on one day to cover an amount it was due to receive from the same customer the next day.

The effect of these measures was to protect the western counterparties from principal risk, but to expose the Soviet institutions to an immediate liquidity risk, at a time when money market participants were increasingly reluctant to deal with them. Fortunately, no widespread systemic problems developed, partly because some counterparties were able to come to bilateral understandings - in some cases with the help of public authorities - which enabled deals to be settled. This type of situation could, however, in different circumstances, have had more wide-ranging and serious consequences.

2.2.5 The Barings crisis (1995)

The unforeseen collapse of Baring Brothers at the end of February 1995 caused a problem in the ECU clearing. On Friday, 24th February one clearing bank had sent an ECU payment instruction addressed to Barings' correspondent for a relatively small amount for value on Monday, 27th February. After the appointment of an administrator to Barings on 26th February the sending bank sought to cancel the instruction but it found that the rules of the ECU clearing did not permit this; moreover, the receiving bank was legally unable to reverse the transaction. As it turned out, the sending bank happened to find itself in an overall net debit position in the clearing at the end of the day. Under pressure of time the bank agreed to cover that position by borrowing from a long bank, so enabling the settlement of more than ECU 50 billion in payments between the 45 banks participating in the clearing eventually to be completed on the due date.

This demonstrates the potential problems which can be caused when banks do not have a thorough understanding of the rules of the clearing systems through which they will pay or receive the currencies of their market transactions. If the sending bank had not eventually agreed to borrow in order to cover its payment, the end-of-day settlement would have been frustrated. The clearing would have had to be unwound, so that no payments between any of the 45 ECU clearing banks would have been settled on the due day, even though less than 1% of those payments had anything to do with Barings. The failure to settle could have had very serious consequences for the banks, and for their customers, in the ECU market and more widely.

2.3 Defining and measuring foreign exchange settlement exposure

To contain the systemic risk inherent in current arrangements for settling foreign exchange transactions, it is first necessary to develop a realistic understanding of the nature and scope of FX settlement exposures. On the basis of discussions with market participants, the CPSS has adopted the following definition of foreign exchange settlement exposure:

A bank's actual exposure - the amount at risk - when settling a foreign exchange trade equals the full amount of the currency purchased and lasts from the time a payment instruction for the currency sold can no longer be cancelled unilaterally until the time the currency purchased is received with finality.

It is important to note that this definition is designed to address the size and duration of the credit exposure that can arise during the FX settlement process. It says nothing about the probability of the occurrence of an actual loss.

The definition also does not specifically address the ability of a bank to measure and to control its FX settlement exposure at a particular moment. To develop a practical methodology for measuring current and future FX settlement exposures in a manner consistent with the above definition, a bank would need to recognise the changing status - and, hence, the changing potential settlement exposure - of each of its trades during the settlement process. Although settling a trade involves numerous steps, from a settlement risk perspective a trade's status can be classified according to five broad categories:

• Status R: Revocable. The bank's payment instruction for the sold currency either has not been issued or may be unilaterally cancelled without the consent of the bank's counterparty or any other intermediary. The bank faces no current settlement exposure for this trade.

• Status I: Irrevocable. The bank's payment instruction for the sold currency can no longer be cancelled unilaterally either because it has been finally processed by the relevant payments system or because some other factor (e.g. internal procedures, correspondent banking arrangements, local payments system rules, laws) makes cancellation dependent upon the consent of the counterparty or another intermediary; the final receipt of the bought currency is not yet due. In this case, the bought amount is clearly at risk.

• Status U: Uncertain. The bank's payment instruction for the sold currency can no longer be cancelled unilaterally; receipt of the bought currency is due, but the bank does not yet know whether it has received these funds with finality. In normal circumstances, the bank expects to have received the funds on time. However, since it is possible that the bought currency was not received when due (e.g. owing to an error or to a technical or financial failure of the counterparty or some other intermediary), the bought amount might, in fact, still be at risk.

• Status F: Fail. The bank has established that it did not receive the bought currency from its counterparty. In this case the bought amount is overdue and remains clearly at risk.

• Status S: Settled. The bank knows that it has received the bought currency with finality. From a settlement risk perspective the trade is considered settled and the bought amount is no longer at risk.

The diagram below illustrates this simplified description of the FX settlement process. To classify its trades according to the categories indicated, a bank would need to know the following three critical times for each currency it trades:

1. its unilateral payment cancellation deadline;
2. when it is due to receive with finality the currency it bought; and
3. when it identifies final and failed receipts.

FOREIGN EXCHANGE SETTLEMENT PROCESS: CHANGING STATUS OF A TRADE

As described in Appendix 1, these times depend on the characteristics of the relevant payments systems as well as on the individual bank's internal settlement practices and correspondent banking arrangements. Nevertheless, once a bank determines these times and appropriately classifies the status of each of its trades, it is a straightforward calculation to measure its FX settlement exposure even in the absence of real-time information. Banks that always identify their final and failed receipts of bought currencies as soon as they are due can determine their exposures exactly. For these banks, current exposure equals the sum of their Status I and F trades. In contrast, banks that do not immediately identify their final and failed receipts cannot pinpoint the exact size of their FX settlement exposures. The uncertainty they face reflects their inability to know which of their Status U trades have or have not actually settled (i.e. they do not know the amount of bought currencies that should - but might not - have been received on time).

Faced with this uncertainty, a bank should be aware of both its minimum and maximum FX settlement exposure. For instance, a bank that only measures and controls its minimum exposure could, in adverse circumstances, experience a much larger unexpected and undesirable actual exposure. On the other hand, a bank that only monitors its maximum exposure might, if it believes that its actual exposure usually falls well short of this amount, set up excessively accommodative internal controls that would not prevent an unexpected and undesirable jump in its actual exposure.

Recognising the uncertainty that might surround its actual FX settlement exposure, a bank can use the following general guidelines to measure its minimum and maximum exposure on the basis of the current status of its unsettled trades:

Minimum exposure: Sum of Status I and F trades. This is the value of the trades for which a bank can no longer unilaterally "stop payment" of the sold currency but has not yet received the bought currency.

Maximum exposure: Sum of Status I, F and U trades. This equals a bank's minimum exposure plus the amount of bought currencies that should - but might not - have been received.

A bank can also project its FX settlement exposure using its knowledge that each of its trades will go through a predictable change in status, based on its current settlement practices. Appendix 1 describes in detail this methodology for measuring a bank's current and future FX settlement exposure.