Abstract

Akinari Horii

Akinari Horii

Bruce J. Summers

Large-value transfer systems supporting the interbank markets are the main arteries of a nation’s payment system. The safe and efficient operation of the money and capital markets hinges upon the smooth functioning of these systems. Moreover, rapid and reliable settlement is essential for technical efficiency and innovation in financial markets. The safe and efficient operation of large-value transfer systems has a bearing not only on the markets they directly serve but on a nation’s whole financial system. In addition, large-value transfer systems have an international role to play, as they, in combination, provide the ultimate settlement vehicle for important cross-border markets in multiple currencies. The international goods and financial markets depend critically on national large-value transfer systems to settle obligations in the currencies in which trading is conducted.

For these reasons, the design and operation of large-value transfer systems are major concerns for policymakers and banking practitioners. In developed market economies, attention has recently been focused on strengthening these systems. Establishing at least a rudimentary large-value payment capability is a priority in developing market economies because it is needed to support emerging financial markets and will help to create conditions for improved execution of monetary policy by the central bank.

This chapter makes some practical distinctions to help single out large-value transfer systems from among the plethora of payment mechanisms in use, and presents three general models of such systems that embody the basic distinctions among the major systems that are currently operational. It examines actual large-value transfer systems, including the Swiss Interbank Clearing System (SIC), Fedwire, The Bank of Japan Financial Network System (BOJ-NET), and the Clearing House Interbank Payments System (CHIPS), that illustrate the models. It identifies issues relating to the use of these systems for settlement of cross-border payments.

Distinctions Between Large-Value Transfer Systems

Different payment mechanisms can be distinguished by the businesses they support and the customers they serve, as reflected in the value of the payments processed. As a result of the wide variation in the value of payments, payment mechanisms have become quite highly specialized. Although there is no clear-cut, quantitative demarcation between small-and large-value payments, some systems have specialized in handling payments that are typically very large.

In general, the interbank, securities, and business-to-business, or wholesale markets, give rise to payments whose large size and critical timing place them in the category of large-value payments. Participants in these markets naturally seek bank payment services and payment mechanisms that can meet their needs for reliability, security, accuracy, and timeliness. To meet these needs, specialized large-value transfer systems have evolved.

The average size of the payments handled by a system is a useful practical indicator of the system’s uses. Table 1 shows, for 1992, the average size of funds transfers made over the four large-value transfer systems discussed in this chapter. Although there is generally no minimum size restriction on transfers over these systems, the typical transfer is a very large sum. For example, the average size of funds transfers over Fedwire and BOJ-NET are about $3 million and $33 million, respectively.

Table 1.

Value of Funds Transfers Over Four Large-Value Transfer Systems, 1992

article image

In contrast to other large-value transfer systems, SIC is used heavily to process smaller-value transactions.

The cost of providing the superior levels of service required by the users of such systems is high compared with other systems that handle low values, in part owing to the greater security, reliability, and timeliness required in large-value payments. Accordingly, competitive pricing of payment services, together with the requirements placed by systems on their users to comply with minimum operational and security standards, have thus far limited the use of such systems for small-value transfers.

In addition to average transaction size, the total value of payments handled by a system during a normal business day is also a relevant indicator of the underlying purpose for which the system is used. Table 1 also shows the daily average value of funds transfers over the four systems treated here. The daily flows of funds over these systems is huge in relative terms as well, on average equaling the value of annual GDP every 2.6 and 2.8 days in Switzerland and Japan, respectively, and every 3.4 days in the United States for Fedwire and CHIPS combined.

Three Models of Large-Value Transfer Systems

The framework for analysis of payment systems presented in Chapter 3 introduced a number of concepts that are useful building blocks in constmcting models of payment systems. Three general models of large-value transfer systems are developed here using the following concepts: (1) operator of the system, that is, the central bank or a private organization such as a clearinghouse; (2) type of settlement, that is, gross or net; and (3) credit facilities, in particular, whether the system provides intraday credit and whether operational controls are in place to help manage such credit extensions.

The first general model of a large-value transfer system is a gross settlement system operated by the central bank without intraday credit. As explained in Chapter 2, intraday loans have a duration shorter than one business day; say, a few minutes or hours. In a gross settlement system operated by the central bank, agreement to honor a payment order when the funds in the account of the paying bank are insufficient to settle the payment results in an extension of credit. This is so regardless of whether the paying bank would fully fund the payment before the end of the business day, that is, repay the intraday loan. In the general model where the central bank does not agree to provide credit, a payment order will be honored only if funds are on deposit at the time the payment order is made. Otherwise, the payment order is returned to the originator (rejected) or held until covering funds become available during the day (pended or queued). This type of system implies real-time computer processing and operational controls that permit the central bank to prevent use of intraday credit. An example of such a system is the Swiss Interbank Clearing System.

The second general model of a large-value transfer system is a gross settlement system operated by the central bank with intraday credit. In this model, the central bank will honor payment orders during the day even if an ordering bank’s account does not contain sufficient funds to settle the transfer. Intraday credit is generally provided with the expectation that the covering funds will be deposited in the account before the end of the business day. The central bank’s willingness to extend intraday credit, however, is not unlimited. Financial and operational controls will be employed to govern the amount of intraday credit extended. An example of such a system is the Fedwire funds transfer system in the United States.

The third general model of a large-value transfer system is a deferred net settlement system. In such a system, settlement does not occur payment-by-payment, but at designated times during the day. Between—or at—designated settlement times, payments exchanged between banks are multilaterally netted, resulting in one net obligation for each net debtor bank that is due at settlement time.

Netting systems act to reduce, perhaps significantly, the intraday liquidity needed to settle large payments. In a netting system, these liquidity needs are met by the de facto extension of credit among participants in the system. However, this credit is extended by the originators and receivers of payments over the system, not by the operator of the system. Some deferred net settlement systems are operated by the central bank, whereas others are operated by the private sector. An example of the former type of system is BOJ-NET1 in Japan. An example of a privately operated system is CHIPS in the United States.

The netting principles that underlie the operation of deferred net settlement systems are the same, regardless of whether the system operator is a central bank or a private entity. In the latter, day-to-day management of the system is in the hands of the private sector. Nonetheless, the central bank almost certainly will exercise some oversight of the privately operated system, for example, by examining its operations and reviewing and approving rule changes before they are made. An important characteristic of all deferred net settlement systems is that netted obligations arising from payment activity between deferred settlement times are finally settled by transferring balances held in accounts with the central bank, that is, final settlement is made in central bank money. For this reason alone, a central bank should be extremely confident about the appropriateness of the risk management controls that such systems employ before it grants use of its net settlement services. The minimum standards that have come into use for judging the adequacy of the design and operation of cross-border and multicurrency netting systems are the so-called six Lamfalussy standards published by the Bank for International Settlements, most of which are applicable to large-value netting schemes in general.2

Examples of Large-Value Transfer Systems

This section describes the operation of four large-value transfer systems that typify the models presented above. Table 2 is a profile of the four systems, which comprise SIC, Fedwire, BOJ-NET, and CHIPS. Table 2 is intended as a quick reference to show the basic features of these systems. The appendices to this book provide detailed descriptions of each of the four systems. The essential operating features of these four systems are described below using numerical examples.

Table 2.

Summary Profile of Four Large-Value Transfer Systems

article image
article image

BOJ-NET also offers a gross real-time settlement capability without credit, which operates in a manner similar to SIC.

All times are local times.

The Board of Governors of the Federal Reserve System has announced that Fedwire will open at 0:30 beginning in early 1997.

Swiss Interbank Clearing System (SIC)

SIC is the best example of a gross settlement system providing final settlement in central bank money without any extension of intraday credit whatsoever. From its inception in 1987, it was designed as a no-overdraft system. It operates on the principle that all payment orders will be processed only if they can be fully funded from a bank’s account held at the Swiss National Bank. If funding is not available, the payment order will be queued and held until covering funds become available, up to the end of the operating day. At the end of the operating day, payment orders in the queues are canceled.3

SIC operates virtually around the clock. The beginning of the SIC operating day is 18:00 Zurich time (17:00 GMT) on day t and the end of the operating day is 16:15 Zurich time (15:15 GMT) on day t+1. Payment orders can be originated throughout this period and are processed and settled virtually immediately if funds are available in the account of the originating bank.

Table 3 is a simple numerical example of how SIC functions. Assume an opening of business balance in the account of the bank of 10 monetary units. Assume further that the first transaction of the day is the receipt of a payment order equal to 20 monetary units. Because SIC is a gross settlement system, the bank’s current account balance held at the Swiss National Bank is immediately increased to 30 monetary units. The second transaction is an order to pay 30. Because there are sufficient balances in the account at the Swiss National Bank, this payment order is accepted and settled immediately, reducing the current account balance with the Swiss National Bank to zero. The third transaction is an order to pay 10 monetary units. In this instance, the amount settled is zero, as the Swiss National Bank will not agree to process the transaction because there are insufficient funds in the account to settle the payment order. Instead, the transaction is placed in queue, resulting in an unsettled balance of minus 10. Finally, the fourth and last transaction of the operating day is the receipt of 20 monetary units. The receipt of 20 monetary units is settled immediately and results in a positive account balance of 20, which triggers release of the one payment order in the queue and elimination of the unsettled balance of 10. The end-of-day result of this activity is a current account balance of 10 and an unsettled balance of zero.

Table 3.

Numerical Example of Operation of Swiss Interbank Clearing System

article image

As described in Appendix 1, the SIC system is designed to process as many payment orders as possible following the “first-in, first-out” principle. The amount of funds held in the account of the originating institution must be sufficient to cover the payment that is at the head of the queue of unfunded payment orders, should such a queue exist. If there are unsettled balances outstanding at the end of the SIC business day, the payments in the queue will be purged, and the institution will be forced to resubmit the orders on the following day. An institution is, however, permitted to manage the payment orders held in its SIC queue by canceling orders in the queue, which would advance other payment orders to the head of the queue.4 A new feature will become available in 1994 that allows participants to attach a priority to a payment order. Processing will then be governed by the principle “by priority level and first-in, first-out for a given priority level.”

One important feature of SIC is that the institution designated as the intended receiver is notified of the amount of pending receipts. Further notification is received when a payment order is settled. In the numerical example discussed above, therefore, the intended receiving institution is informed of the amount of the third transaction, an order to pay 10, at the time the order is made. Only after the fourth transaction of the day, however, does the receiving institution receive funds and notification that the payment order is settled. Perhaps as important, institutions can use a real-time inquiry feature to monitor the current status of all payment messages.

Fedwire

Table 4 illustrates the operation of Fedwire, the gross settlement system operated by the Federal Reserve Banks since 1918. Fedwire began as a simple telegraph system that was used to transfer balances between accounts held at Federal Reserve Banks. Settlement is final when the Federal Reserve Bank holding the account of the originating institution agrees to process a payment order. The Federal Reserve permits daylight overdrafts over Fedwire, within limits. As described in Appendix 2, institutions are expected to perform a self-assessment of their creditworthiness and operational capabilities and to establish a Fedwire cap, which is based on a multiple of their tier 1 capital, if they have overdrafts that are large in relation to their capital.

Table 4.

Numerical Example of Operation of Fedwire Funds Transfer System1

article image

Sender debit cap equals 12.

The normal Fedwire operating day is 8:30 to 18:30 New York time (13:30 to 23:30 GMT). Fedwire is sometimes opened earlier than 8:30 or closed later than 18:30 to meet special needs. For example, Fedwire has been opened earlier than 8:30 in response to unusually volatile financial market conditions resulting in large price swings and consequently unusually large settlement obligations on the part of market participants. By opening Fedwire early, the Federal Reserve has given the settlement banks for the financial exchanges the ability to transfer early in the morning payments that represent settlement obligations of their members. Such settlement obligations might arise, for example, in connection with end-of-day variation margin calls on the futures exchanges. Early opening of Fedwire permits settlement of such end-of-day obligations before the beginning of trading on the subsequent day. Further, there have been extensions to the Fedwire operating day to give more time for CHIPS to settle. In addition, Fedwire operating hours can be extended if a participant with a large value of transfers has operating problems that result in its needing a longer operating day to make its payments.

In the numerical example illustrated in Table 4, the opening of business balance held at the Federal Reserve Bank is again 10 monetary units. The first transaction is the receipt of a Fedwire transfer of 20 monetary units, which is immediately settled and reflected in the account balance at the Federal Reserve Bank, which is increased to 30. The second transaction of the Fedwire day is a payment order of 30, which draws the account balance down to zero.

This example assumes that the originating bank’s debit cap is 12 monetary units. Consequently, the third transaction of the day, which is a payment order of 10, is processed by the Federal Reserve and settled immediately, even though this payment order results in a negative account balance of minus 10. The cap is not binding because the amount of the payment order does not exceed the debit cap. By agreeing to process the payment order, the Federal Reserve essentially agrees to lend intraday funds to the bank originating the payment so that the transaction can be settled.

The fourth transaction of the day is the receipt of 20 monetary units, which rebuilds the account balance to 10 and extinguishes the intraday loan. The fifth transaction of the day is a payment order of 25. In this case, the positive current balance in the Federal Reserve Bank account (10 monetary units) combined with the institution’s cap (12 monetary units) results in total capacity of 22 monetary units, an amount insufficient to settle the transaction, which is valued at 25 monetary units. Assuming, for the moment, that the Federal Reserve Bank monitors the institution in real time, the transaction will not be settled but will either be (a) rejected back to the sender or (b) pended for subsequent processing once the account is funded. Because the payment order of 25 cannot be funded, unsettled balances in the system (assuming that the transaction is pended) equal minus 25 monetary units.

Finally, the Fedwire day ends with the sixth transaction, which is the receipt of 5 monetary units. The positive account balance increases to 15, which, combined with the cap, increases capacity to 27 monetary units, an amount sufficient to fund the pended transaction of 25 even though the result is a negative balance of minus 2. If the Fedwire day ends and the institution’s balance is negative, as in this example, the implication is that the institution was unable to raise funds in the market to meet its Fedwire payment obligations for the day. The institution must then obtain a discount window loan from the Federal Reserve to bring its account into balance.

The Board of Governors of the Federal Reserve System has decided to price intraday overdrafts incurred by depository institutions using Fedwire. Pricing of intraday overdrafts is an extremely complex subject and raises a host of legal, operational, and monetary control issues.5 Pricing became effective in April 1994.

The operation of SIC and Fedwire highlights an important contrast between systems that do and do not provide credit. In particular, a “no-overdraft” system like SIC imposes tighter liquidity management constraints on banks than does Fedwire, through which the central bank provides intraday credit. Banks and other financial market participants that place a premium on timely settlement are likely to describe a no-overdraft system as being less efficient than a system like Fedwire, which permits intraday overdrafts. Their intraday credit needs cannot currently be efficiently met except through the payment mechanism they use. In short, the SIC system does not “lubricate” the payment system with intraday credit, and consequently many transactions are queued. In contrast, the Federal Reserve has historically been a generous provider of intraday credit over Fedwire, and Fedwire caps have historically not been binding.

Although Fedwire may be more efficient in terms of the timeliness of settlement for interbank transactions, this enhanced efficiency comes at a cost. The cost takes the form of the increased credit risk absorbed by the Federal Reserve in operating Fedwire. Moreover, because the abundant intraday credit has been provided free, banks have overused intraday overdrafts provided by the Federal Reserve, resulting in the absorption by the central bank of a certain amount of credit risk that should more appropriately be shouldered by the private sector.

BOJ-NET

The Bank of Japan is somewhat special among central bank operators of large-value transfer systems because it supports two distinctively different systems, both operating under the name BOJ-NET. On the one hand, BOJ-NET offers, like SIC, a gross real-time settlement service without intraday overdrafts. BOJ-NET is, however, different from SIC in that, if sufficient funds are not available in the account to settle the obligation, the payment order is automatically rejected, rather than queued.

The gross, real-time, no-overdraft service provided over BOJ-NET is not heavily used in comparison with the other services provided by the Bank of Japan. More heavily used by banks and other financial firms holding accounts at the Bank of Japan is the BOJ-NET designated-time net settlement system, which is estimated to handle 50 times the transfer volume that the gross real-time BOJ-NET system handles.

The BOJ-NET designated-time net settlement system accepts payment orders between 9:00 and 17:00 Tokyo time (00:00 to 8:00 GMT) for same-day settlement. The cutoff time for postdated transactions is 17:20. Although settlement obligations are calculated based on multilateral netting, in strict legal terms this BOJ-NET service is a designated-time gross settlement. Each payment order originated before or during the settlement day can be designated to settle at one of four designated settlement times during the day, namely, 9:00, 13:00, or 15:00 or 17:00. Table 5 is a numerical example illustrating the operation of this system.

Table 5.

Numerical Example of Operations of BOJ-NET Designated-Time Net Settlement System

article image

Balance held in the settlement account at the Bank of Japan.

Assume again that an institution, this time using the BOJ-NET designated-time settlement, has an opening of business balance in its account at the Bank of Japan of 10 monetary units. The first transaction of the day is the receipt of 20 monetary units and, because this is a designated-time system, the amount settled is zero—the account balance remains unchanged at 10, and total unsettled balances increase from zero to 20 monetary units. Next, assume that a payment order valued at 30 is originated. Again, there are no changes in the account balance, there is no final settlement associated with the transaction, but the unsettled net balance of the institution in question falls to minus 10 monetary units. The process is repeated for a payment order of 10 with the unsettled net balance falling further to minus 20 monetary units. Finally, the last transaction before the designated settlement time is the receipt of a payment valued at 25 monetary units, which increases the net unsettled balance to 5. Accordingly, at the designated settlement time, the total net amount to be settled for the institution is 5, which increases its balance in the settlement account of the Bank of Japan to 15, while unsettled balances fall to zero.

There are no formal procedures currently in place for the BOJ-NET designated-time net settlement system to address an institution’s inability to meet its settlement obligation. The Bank of Japan has discretion to provide the institution in question with credit or to delete the institution’s payment orders from the settlement. Although this room for discretion gives the central bank flexibility in forestalling systemic disruption of financial markets, it may invite moral hazard among participating institutions if central bank credits are expected to be granted easily.

Clearing House Interbank Payments System (CHIPS)

The Clearing House Interbank Payments System is operated by the New York Clearing House and processes primarily international payments. The numerical example shown in Table 5 could apply to CHIPS, as CHIPS operates under a netting arrangement similar to the BOJ-NET designated-time net settlement system. CHIPS differs from BOJ-NET in that there is one end-of-day settlement, not a series of designated settlement times during the day. Moreover, CHIPS is a privately operated payment system in which final settlement is achieved by funds transfers on Fedwire. CHIPS operations are governed by a set of risk controls that have been adopted by its members.6 In particular, as summarized in Table 2 and described more fully in Appendix 4, CHIPS participants have adopted a system of bilateral credit limits and sender net debit caps that limit both individual participant exposure and the entire system’s vulnerability to credit risk.

Under CHIPS bilateral credit limits, each participant establishes the maximum net amount it is willing to receive from another participant and this limit is enforced automatically, in real time, by the CHIPS computer system. Further, there is a sender net debit cap in place that limits the amount that any one participant can owe to the entire CHIPS system. Each participant’s sender net debit cap is equal to 5 percent of the sum of the bilateral credit limits established by each of its counterparties in CHIPS. Essentially, then, participants are able to limit their exposures bilaterally to participants they judge to be in questionable financial condition and, in the process, the entire CHIPS system’s exposure to that participant is limited.

All CHIPS participants agree to participate in a scheme for guaranteeing the daily settlement, if a participant with a large settlement obligation ever fails to meet that obligation. The settlement guarantee is combined with a loss-sharing arrangement to govern the distribution of the burden of funding a failure to settle among the members of CHIPS. CHIPS maintains significant liquidity to permit the mobilization of cash on very short notice to allow the system to settle in a timely fashion if a participant unexpectedly fails to meet its net debit obligation. As described in Appendix 4, the liquidity arrangements include a pool of U.S. Government securities collateral held in escrow at the Federal Reserve Bank of New York.

Cross-Border Large-Value Payments

As noted earlier, domestic large-value transfer systems are often used in the interbank markets to settle the local currency component of cross-border, multicurrency transactions. The value of such transactions is substantial, reflecting growing internationalization in the goods and financial markets. In fact, the largest cross-border financial flows in the world result from trading in the international interbank markets, the largest of which is the foreign exchange market.

Settlement of payment obligations that arise in these cross-border markets depends substantively on the international correspondent banking system. Correspondent banks provide the clearing, settlement, and credit services necessary to allow payments to be made efficiently. The correspondent banks, in turn, rely on either book transfers or domestic large-value transfer systems to settle their interbank obligations arising in such markets.

A special type of risk that must be managed by banks settling cross-border obligations is the temporal risk arising when the two sides of a settlement are separated in time owing to international time-zone differences. A mechanism for the simultaneous final settlement of interbank money positions in different currencies using large-value transfer systems does not currently exist for every currency pair. For example, time-zone differences and the current differences in the hours of operation of domestic large-value transfer systems and domestic money markets introduce temporal risk in the settlement of the three most important currency pairs on the foreign exchange markets—the U.S. dollar/deutsche mark, U.S. dollar/yen, and U.S. dollar/pound sterling.7

Several private arrangements to improve settlement particularly in the foreign exchange markets have been initiated. In the foreign exchange markets, two different groups of bankers, one in Europe and the other in North America, have explored establishing multilateral netting arrangements on behalf of their members. These arrangements are the Exchange Clearing House Organization and MultiNet, respectively. Their objective is to design and implement safe and efficient final settlement capabilities for foreign exchange dealing by attempting to synchronize their multicurrency settlements across time zones to the extent possible.8 At least one other service offered by a private bank is designed to support simultaneous settlement in multiple currencies through the exchange of shares in a mutual fund that is invested in highly liquid government securities.9

It is likely that the participants in any cross-border, multicurrency multilateral settlements that are established will wish to rely on domestic large-value transfer systems to achieve final settlement. Participants in such arrangements may also wish to achieve greater simultaneity in the settlement of their respective positions in various currencies. As long as there are needs in the market for settlement arrangements to reduce cross-border settlement risks, large-value transfer systems will over time likely provide such services.

Conclusions

A rich variety of large-value transfer systems exist that serve developed financial markets around the world. The operation of these systems can entail risks both for the participants transferring funds and for the operators of the systems. These risks must be allocated and controlled so that the participants in the systems and the system operators are reasonably protected. At the same time, when the systems are based on multilateral netting and result in a mutualization of risk, risk control mechanisms must be in place to protect against adverse systemic consequences for the financial system at large. These systems can be expected to continue to adapt over time to changing market needs, including the needs of cross-border markets for safer multicurrency settlement.

There is a trade-off between efficiency and risk in the design and operation of large-value transfer systems. At one extreme, a gross settlement system that does not provide for intraday credit to participants minimizes the concentration of credit risk in the large-value transfer system itself, but at the same time may severely constrain the flexibility with which payments can be made. Such a system is more likely to result in delayed payments or the accumulation by banks of large, probably sterile clearing balances. At the other extreme, a gross payment mechanism that provides liberal quantities of credit, either through the central bank or implicitly through a multilateral netting arrangement, brings significant risks both for participants and for the financial system. The models presented here highlight some of the trade-offs between the two different approaches and the types of controls that have been employed to protect against risk.

In the end, no single model of a large-value transfer system is necessarily best for a particular situation. More than one large-value transfer system can serve the same economy, meeting the needs of different types of markets and customer requirements. In the United States, for example, Fedwire and CIHPS operate side by side and have, over time, adapted to meet the varying requirements of the U.S. and international financial systems. Similarly, in Japan, there is a clearing system for yen settlement for foreign exchange and cross-border transactions—the foreign exchange yen settlement system (FEYSS)—as well as BOJ-NET. The FEYSS and CHIPS netting systems, however, depend critically on the final payment capabilities of the Bank of Japan and the Federal Reserve, respectively. A mix of public and private arrangements may represent the optimal solution to the needs of participants in markets that give rise to demands for large-value transfer services.

1

BOJ-NET also offers gross settlement without intraday credit.

2

Bank for International Settlements, Report of the Committee on lnterbank Netting Schemes of the Central Banks of the Group of Ten Countries, prepared by the Committee on Interbank Netting Schemes chaired by MA. Lamfalussy (Basle, November 1990).

3

Much of this section is based on Christian Vital, “Swiss Interbank Clearing—Further Experience With a Queuing Mechanism to Control Payment System Risk,” a paper presented in April 1990 at a conference sponsored by the Institute for international Research.

4

Cancellation before the first cutoff time of 3:00 p.m. can be done by the sending bank unilaterally. After the first cutoff time, cancellation requires the consent of the receiving bank. Otherwise, the sender will be charged a penalty equivalent to 5 percentage points over the prevailing market rate.

5

For an analysis of these issues, see the attachments to the Federal Reserve press release dated October 7, 1992.

6

See New York Clearing House Association, CHIPS Settlement finality Rules and Documents, April J990.

7

See Bank for International Settlements, Central Bank Survey of foreign Exchange Market Activity in April 1992 (Basle, March 1993).

8

Peter B. Smith, “Foreign-Exchange Netting Needed to Reduce Enormous Exposures,” Financier, Vol. 15 (January 1991), pp. 22–24.

9

See “A Fund-Shifting System That’s Open All Night,” New York Times, November 27, 1992. p. D12.