The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
3 I m p l ic a t io n s o f N e ttin g A r r a n g e m e n t s f o r B a n k R is k in F o r e i g n E x ch a n g e T ra n sa ctio n s 1 7 I n s t i t u t i o n a l D e v e l o p m e n t s in t h e G lo b a liz a t io n o f S e c u r i t i e s a n d F u tu re s M a rk e ts 3 1 D a t a E n v e l o p m e n t A n a ly s is a n d C o m m e rcia l B a n k P e r fo r m a n c e : A P r i m e r W i t h A p p lic a t io n s to M isso u ri B a n k s THE FEDERAL A RESERVE RANK of A r ST.IjOIIIS 1 F e d e r a l R e s e r v e B a n k o f St. L o u is R e v ie w January/February 1992 In T h is Issu e . . . In the first article of this Review, “Implications of Netting A rrange m ents for Bank Risk in Foreign Exchange T ransactions,” R. Alton Gilbert describes th e risks assum ed by banks in settling foreign exchange tran s actions w ith other banks. The risks involve default by the other parties to the transactions. As the author notes, the volume of transactions in the foreign exchange m arket is very high, and banks commonly engage in transactions w ith counterparties headquartered in other countries. Thus, a default by a m ajor participant in the foreign exchange m arket could affect the operation of paym ents systems in several countries. Central banks have interest in the design of any arrangem ents am ong banks th at m ight reduce their risk in settling foreign exchange transac tions. One way that banks may be able to reduce transaction costs and risks is through netting arrangem ents. The central banks of 10 deve loped countries recently issued a report on netting arrangem ents, w hich included a list of guidelines for their design. Gilbert examines the impli cations of netting arrangem ents for risk assum ed by banks in settling foreign exchange transactions and indicates w hy some of the guidelines are im portant if netting arrangem ents are to reduce risk. *** Financial transactions, like the buying and selling of securities, com modities, foreign exchange and bonds, have increasingly involved in dividuals and firm s from different countries. In the second article of the Review, "Institutional Developm ents in the Globalization of Securities and Futures M arkets,” Jodi G. Scarlata describes recent institutional develop m ents in this globalization and discusses the advantages and disadvan ta g e s of th e s e c h a n g e s . S u b s ta n tia l b e n e fits , s h e n o te s , are o c c u r r in g because of these developm ents. At the same time, dom estic rules and regulations are not sufficient safeguards for m any international trades. In particular, she stresses how some risks at various stages of the clear ing and settlem ent process are m ore im portant in an international set ting than in a strictly dom estic setting. W eaknesses in the clearing and settlem ent system have prom pted w orld financial leaders to w ork tow ard global coordination. Scarlata concludes th at significant steps rem ain in integrating the w orld’s grow ing securities and futures m arkets. *** In the third article in this issue, “Data Envelopm ent Analysis and Com m ercial Bank Perform ance: A Prim er w ith Applications to Missouri Banks,” Piyu Yue discusses a relatively new m ethodology for evaluating JANUARY/FEBRUARY 1992 2 the technical or productive efficiency of business enterprises. The m ethodology is called Data Envelopm ent Analysis or DEA. After discuss ing the distinction betw een technical and economic efficiency, the author explains w hat DEA is and how it can be used to partition a group of firm s into those that are DEA-efficient and those th at are not. She illustrates the usefulness of the technique w ith data for 60 Missouri com m ercial banks for the period 1984-90. *** FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 3 R. Alton Gilbert R. Alton Gilbert is an assistant vice president at the Federal Reserve Bank of St. Louis. Richard I. Jako provided research assistance. Implications of Netting Arrangements for Bank Risk in Foreign Exchange Trans actions H HE MAJOR FINANCIAL institutions of m any nations are active participants in the m arket for foreign exchange. The exchanges of currencies that take place through this m arket facilitate in ternational trade and the international flow of capital for investm ents. The volume of transactions in the foreign ex change m arket—already very large—has grow n rapidly in recent years. As of April 1989, the date of the last international survey, foreign ex change transactions had an average value of $640 billion per business day. W ith dollar am ounts in this lofty range, p ar ticipants in the foreign exchange m arket could incur substantial losses if the other parties to their transactions w ere to default on the pay m ents required to settle their side of the tran s 'Netting agreements between pairs of banks may apply to payments in settlement of transactions other than foreign exchange. This paper, however, limits analysis to the net ting of foreign exchange transactions. All participants in the foreign exchange market are called banks to simplify exposition. In some markets, the important participants in clude firms that are not banks. See Federal Reserve Bank actions. To reduce the costs of transactions and limit the size of these possible losses, some banks engage in bilateral netting of their for eign exchange transactions .1 In bilateral netting, tw o banks exchange daily only the net units of currencies in the transactions betw een them . Some groups of banks have also studied the possibility of m ultilateral arrangem ents for net ting foreign exchange transactions, though none are in operation at this tim e .2 M em bers of a m ultilateral netting arrangem ent w ould settle transactions w ith each other by m aking pay m ents to a clearing house for th eir net position in each currency w ith the other m em bers. As part of their responsibility to avoid disrup tions in the operation of paym ent systems, cen tral banks have a strong interest in such netting of New York (1989) and Bank of England (1989). See glos sary on page 14 for definition of netting and other terms used in this paper. 2See Deeg (1990), Duncan (1991), Luthringhausen (1990) and Polo (1990). JANUARY/FEBRUARY 1992 4 Table 1 Minimum Standards for the Design and Operation of CrossBorder and Multi-Currency Netting and Settlement Schemes I. Netting schemes should have a well-founded legal basis under all relevant jurisdictions. II. Netting scheme participants should have a clear understanding of the impact of the particular scheme on each of the financial risks affected by the netting process. III. Multilateral netting systems should 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. IV. Multilateral netting systems should, at a minimum, be capable of ensuring the timely completion of daily settlements in the event of an inability to settle by the participant with the largest single net-debit position. V. Multilateral netting systems should have objective and publicly disclosed criteria for admission which permit fair and open access. VI. All netting schemes should ensure the operational reliability of technical systems and the availa bility of back-up facilities capable of completing daily processing requirements. SOURCE: Bank for International Settlements (1990c). arrangem ents .3 Since foreign exchange transac tions often involve parties headquartered in different countries, a default by one participant is likely to affect those in other countries. Banks adversely affected by such defaults typically w ould tu rn to their central banks for assistance in coping w ith liquidity problem s. In recent years, prom oters of interbank net ting arrangem ents have requested the views of central banks individually on projects that ap peared to have implications for a num ber of countries. The central banks of 10 m ajor indus trialized countries recently issued a joint state m ent, through the Bank for International Set tlem ents, about the netting of foreign exchange transactions. This is commonly called the "Lamfalussy Beport,” nam ed after the com m ittee chairm an w ho drafted the report. The com m it tee expressed concern about the risks involved in settling foreign exchange transactions and discussed the potential benefits and draw backs of netting arrangem ents. The central bankers listed m inim um standards for the design of net 3See Summers (1991) for a discussion of the role of central banks in the operation of payment systems. 4Bank for International Settlements (1990c). 5Cody (1990) also provides an introduction to the risk in settling foreign exchange transactions and the implications FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis ting arrangem ents for bankers w ho may develop them (see table l ).4 This paper illustrates the risk in settling for eign exchange transactions and the risk implica tions of netting, using a hypothetical example of transactions am ong three banks. This exercise illustrates how netting m ay reduce risk, if net ting arrangem ents conform to the guidelines in the Lamfalussy R eport .5 THE MARKET FOR FOREIGN EX CHANGE A foreign exchange transaction is an agree m ent by tw o parties (generally large banks) to exchange currencies on a given date, called the value date of the transaction. The m ost com m on type of transaction betw een participants in the foreign exchange m arket, a spot transaction, is an agreem ent betw een tw o parties to exchange units of currencies tw o business days from the date the transaction is negotiated. A transaction w ith a value date m ore than tw o days after the of netting. See Juncker, Summers and Young (1991) for a general discussion of the issues raised by netting ar rangements. 5 date of negotiation is called a forw ard transac tion. Some forw ard transactions have value dates m ore than a year into the future, but m ost call for settlem ent w ithin a m onth. Several other types of transactions, including futures contracts, options and swaps, have been deve loped to m ore effectively limit the effects of changes in exchange rates on the w ealth of banks and their custom ers .6 Large com m ercial banks are the m ajor p ar ticipants in the foreign exchange m arket. The latest international survey of foreign exchange activity, in April 1989, indicates that the three m ost active centers are London, New York and Tokyo (table 2). The value of foreign exchange transactions has been grow ing faster than inter national trade in goods and services (table 3). Such grow th reflects m ore than the grow th of international trade; it also reflects international capital flows and transactions by banks and their custom ers to m anage exchange rate risk. Transactions in the foreign exchange m arket link the m ajor financial institutions of the world. In the London m arket, for instance, 80 percent of the value of foreign exchange transactions in April 1989 was by firm s w ith headquarters out side of England .7 In the survey of foreign ex change m arket activity in New York, 40 percent of the value of transactions was reported by offices of foreign banks .8 Thus, one of the im portant ways in w hich a m ajor financial institu tion can affect institutions in other countries is by defaulting on foreign exchange transactions. THE CONFIRMATION AND SET TLEMENT OF FOREIGN EX CHANGE TRANSACTIONS The process of confirm ation and settlem ent begins after traders at tw o banks agree on the term s of a transaction. Each bank sends the other a m essage specifying the term s of the transaction, using a variety of m ethods, includ ing telephone calls. If the details of the m es sages m atch, the transaction is considered confirm ed. The next step depends on the value date of the transaction. If it is a forw ard transaction, w ith a value date several weeks or m onths into the future, the inform ation is stored for future 6For a more detailed discussion of the foreign exchange market, see Chrystal (1984). Table 2 Foreign Exchange Market Activity in April 1989 (billions of U.S. dollars)1 Countries and items United Kingdom United States Japan Switzerland3 [85°/o] Singapore Hong Kong Australia France3 [95%] Canada Netherlands Denmark3 [90°/o] Sweden Belgium3 [90%] Italy3 [75%] Other countries4 Value of transactions per day $ 1872 1292 115 57 55 49 30 262 15 132 13 13 10 10 22 744 Total Adjustment for cross-border double-counting -204 Total reported net turnover 540 Estimated gaps in reporting 100 Estimated global turnover $ 640 ’ Value of transactions in currencies other than U.S. dol lar converted to dollars at prevailing exchange rates. The figures for individual countries indicate turnover net of double-counting arising from local interbank busi ness. The totals at the foot of the table are estimates of turnover net of double-counting arising from both local and cross-border interbank business. 2Based on estimates of domestic and cross-border inter bank business arranged through brokers. 3No adjustment for less than full coverage; estimated market coverage is given in square brackets. 4Bahrain, Finland, Greece, Ireland, Norway, Portugal and Spain. SOURCE: Bank for International Settlements (1990a). settlem ent. On th e value date, banks transm it inform ation to initiate paym ent. The steps to in itiate paym ent depend on the paym ent system used in the country issuing the currency and the relationship of the paying bank to th at pay8Federal Reserve Bank of New York (1989). 7Bank of England (1989). JANUARY/FEBRUARY 1992 6 Table 3 Growth of Foreign Exchange Market Transactions, Foreign Trade and International Banking Activity Countries Value of foreign exchange transactions: percentage change between March 1986 to April 1989 net turnover Exports and imports of goods and services: percentage change from 1/1986 to 1/1989 United Kingdom 108% 62% United States 120 44 Japan 140 82 58 44 116 56 Canada Total SOURCE: Bank for International Settlements (1990a). m ent system. For a bank paying in a currency other than th at of its hom e country, paym ent generally is m ade by a correspondent headquar tered in the foreign country. A correspondent is a bank th at holds deposits and provides services for other banks. The paying bank commonly sends a m essage over SWIFT, instructing its cor respondent to m ake paym ent to the counter party in the foreign exchange transaction .9 Suppose, for instance, th at a bank headquar tered in the United States m ust pay Germ an m arks to a counterparty to settle a foreign ex change transaction. The U.S. bank instructs its Germ an correspondent to m ake paym ent to the counterparty (or the counterparty’s Germ an cor respondent). The G erm an correspondent debits the account of the U.S. bank denom inated in m arks and transfers the m arks to the counter party. Suppose a U.S. bank is obligated to pay dollars. It w ould send a m essage over CHIPS to m ake paym ent to the counterparty, either di rectly if it is a m em ber of CHIPS, or through a correspondent in New York w ho is a m em ber of CHIPS.10 9SWIFT (Society for Worldwide Interbank Financial Telecommunication) is an electronic system, located in Brussels, Belgium, for sending messages among the world’s major banks. 10See Bank for International Settlement (1990b) for a description of payments systems in various countries. CHIPS (Clearing House for Interbank Payments System) is an electronic payments system operated by the New York Clearing House Association. CHIPS participants (131 as of the end of 1990) exchange payment messages during each business day and settle for the net amounts at day- FEDERAL RESERVE BANK OF ST. LOUIS THE RISKS INVOLVED IN SETTLING FOREIGN EXCHANGE TRANSACTIONS: AN ILLUS TRATION Banks assum e the risk that their counterpar ties will default on paym ents on their side of foreign exchange transactions. Effects on coun terparties of default on settlem ent obligations depend on the financial condition of the bank that defaults. A solvent bank m ay default for a variety of reasons. O perating problem s (for ex ample, com puter failure) may prevent them from executing their paym ent instructions. A solvent counterparty may not have funds in the proper currency on the value date, or simply may forget to send paym ent orders to settle some of their transactions. Defaults by solvent banks on settlem ent obli gations may have systemic effects, preventing other banks from settling their obligations. These banks may tu rn to their central banks for short-term loans denom inated in the currenend with transfers of reserve balances at the Federal Reserve. See Federal Reserve Bank of New York (1991). A large share of CHIPS messages involve payment for the dollar side of foreign exchange transactions. Given that most foreign exchange transactions involve the U.S. dollar, CHIPS has a major role in the settlement of foreign ex change transactions. See Federal Reserve Bank of New York (1987). 7 Table 4 Payments in Settlement of Foreign Exchange Transactions under Gross Settlement and Bilateral Netting_________________________________________ Bilateral netting Gross settlement Counterparties Transaction number Direction of payment 1 Bank A to Bank C £ 100 Bank C to Bank A $170 (Profit of $5.00 for Bank A) Bank B to Bank C £ 150 Bank C to Bank B $262.50 (Profit of $15.00 for Bank B) Bank B to Bank C $85.00 Bank C to Bank B £ 50 (Profit of - $2.50 for Bank B) cies necessary to settle their obligations. Thus, central banks have a collective interest in mini mizing the chances of such liquidity problems. Most liquidity problem s are often only tem porary. Bankruptcy and liquidation of a p ar ticipant in the foreign exchange m arket, how ever, pose a m ore serious threat to individual counterparties and create the potential for sys temic disruptions in the paym ent system (default by one bank causing default by others). U nder a general definition of bankruptcy, the value of liabilities exceeds the value of assets. Some large bankrupt banks have been reorganized w ith as sistance of their hom e governm ents. The reo r ganized banks continue to operate as going con cerns, making paym ents in settlem ent of their obligations. Such reorganizations impose no loss es on their counterparties. In other cases, however, bankrupt banks cease to operate as going concerns. The courts appoint receivers to liquidate the bankrupt Bank A to Bank B £ 50 Bank B to Bank A $90 (Profit of $7.50 for Bank A) Bank A to Bank C $262.50 Bank C to Bank A £ 150 (Profit of -$ 1 5 .0 0 for Bank A) 1 Units of currencies Bank A to Bank B $85 Bank B to Bank A £ 50 (Profit of -$ 2 .5 0 for Bank A) 2 Bank B and Bank C Bank A to Bank B £ 100 Bank B to Bank A $175 (Profit of $10.00 for Bank A) 2 Bank A and Bank C 1 2 Bank A and Bank B Direction of payment Units of currencies Bank A to Bank C $92.50 Bank C to Bank A £ 50 (Profit of -$ 1 0 .0 0 for Bank A) Bank B to Bank C £ 100 Bank C to Bank B $177.50 (Profit of $12.50 for Bank B) banks’ assets and m ake paym ents to their credi tors. The receivers may impose losses on other banks that w ere counterparties to foreign ex change transactions. Such losses depend on the legal principles followed by bankruptcy courts and the nature of netting agreem ents betw een counterparties. The effects of the liquidation of a participant in the foreign exchange m arket on its counter parties are illustrated below. Legal assum ptions are specified along the way as the example raises questions about the principles followed by bankruptcy courts. In each case in w hich a bank is assum ed to go bankrupt, it is also as sum ed to be liquidated by a court-appointed receiver. The Exam ple Suppose three banks (A, B and C) engage in foreign exchange transactions in tw o currencies: the U.S. dollar and the British pound. Each bank has foreign exchange transactions w ith the other two. Table 4 lists the transactions between JANUARY/FEBRUARY 1992 8 the counterparties to be settled on the same value date. Each pair of banks has tw o transac tions to settle. In one transaction, a bank pays dollars in exchange for pounds; in the other, a bank pays pounds in exchange for dollars. The exchange rate on the value date is $1.65 per British pound. Transactions to be settled on the value date w ere negotiated a few days earli er w hen the exchange rate was higher: some transactions w ere negotiated w ith an exchange rate of $1.70; others, w ith an exchange rate of $1.75. Transactions are of varying size, creating im balances in the flows of currencies betw een counterparties. The exam ple is designed to be as simple as possible and yet illustrate the risk involved in netting arrangem ents. T here m ust be at least tw o transactions betw een a pair of banks if bilateral netting is to reduce the volume of pay m ents and settlem ent risk. T hree is the mini m um num ber of banks for m ultilateral netting. The E ffects o f Bilateral Netting on the N um ber and Value o f Trans actions Figure 1 illustrates how bilateral netting af fects the flows of currencies betw een Banks A and B in settling the transactions listed in table 4. U nder gross settlem ent, banks m ake paym ents to each other to settle each transaction betw een them . To settle transaction num ber 1, Bank A pays £ 100 to Bank B, receiving $175 in turn. Since the exchange rate is $1.65 on the value date, this exchange of currencies yields a profit of $10 to Bank A. (Bank A receives $175, w hereas the £ 100 paid by Bank A has a value of $165 on the value date). Bank A pays $85 to Bank B in settlem ent of transaction num ber 2 , receiving £ 50. This exchange yields a loss of $2.50 for Bank A on the value date. Banks A and B can econom ize on transactions by netting their paym ents flows. As illustrated in the bottom half of figure 1, Bank A could pay £ 50 to Bank B and receive $90 from Bank B. Bilateral netting reduces the num ber of pay m ents from four to tw o and the value of pay m ents, converted to dollars at the exchange rate of $1.65, from $507.50 to $172.50. "B a n k for International Settlements (1989), pp. 13-14. FEDERAL RESERVE BANK OF ST. LOUIS The R isk in Settling Foreign Ex change Transactions w ithout a Netting A rrangem ent To illustrate how netting arrangem ents affect risk, one m ust first understand the risk that banks assum e w ithout a netting agreem ent. — This section specifies several assum ptions about the legal principles that the bankruptcy court follows w hen banks settle their transactions w ithout netting arran ge m ents. W hile these principles are not applied in all cases, they are com m on and they simplify the analysis. L e g a l A s s u m p tio n s One assum ption concerns the application of le gal rights o f set-off perm itted by the court. Un der the legal rights of set-off, the counterparty of a failed bank may settle its obligations w ith the receiver by paying the net am ount of the transactions betw een them . If on net the failed bank owes a solvent counterparty, the counter party is a general creditor of the failed bank for the net am ount. Applying the rights of set-off to the foreign exchange transactions betw een a pair of banks yields the same loss to the solvent counterparty as it w ould un d er bilateral netting. Applying the legal rights of set-off, how ever, is uncertain and varies am ong the courts of differ ent countries .11 In this paper, rights of set-off are assum ed not to apply in bankruptcy. Each transaction is treated separately, not linked to other transactions betw een the same parties. The court w ith jurisdiction in a bankruptcy case is assum ed to appoint a receiver. In m ak ing paym ents to settle foreign exchange transac tions or defaulting on transactions, the receiver acts to maximize the retu rn to all creditors of the failed bank, w ithout regard for the counter parties to foreign exchange transactions as a particular group of creditors. A final issue concerns the status of claims against a bankrupt bank th at result from its default on foreign exchange transactions. Sol vent counterparties are assum ed to have the status of general creditors. In our example, loss es are calculated under the assum ption that general creditors receive nothing. All proceeds from the liquidation of assets go to creditors w ith m ore senior claims. 9 Figure 1 Flow of Currencies Between Banks Under Gross Settlement and Bilateral Netting Gross Settlement £100 £ 50 $85 Transaction #2 A Profit to Bank A: -$2.50 Number of Payments: 4 Dollar Value: $507.50 Bilateral Netting £ 50 Profit to Bank A: $7.50 $90 Number of Payments: 2 Dollar Value: $172.50 JANUARY/FEBRUARY 1992 10 These legal assum ptions yield the maximum losses to counterparties. Thus, the losses calcu lated in particular cases can be viewed as the m aximum, not necessarily the most likely, losses. Ex p o s u r e — Suppose Banks A and B agree to set tle their transactions as illustrated in the top half of figure 1 , the gross settlem ent m ethod. Suppose also that Bank A goes bankrupt before the four paym ents are executed on the value date. The possible loss to Bank B depends on the timing of the bankruptcy of Bank A. In one situation, called pre-settlement failure, Bank A goes bankrupt before the value date, and Bank B knows about this event by the opening of business on the value date. In the other situation, called settlement failure, a bank m akes paym ent on the value date for its side of foreign exchange transactions bu t does not re ceive paym ent from a counterparty. One feature of the foreign exchange m arket that makes banks vulnerable to settlem ent fail ure is the difference in the tim e zones of cen tral banks. The failure of the H erstatt Bank in 1974 illustrates the relationship betw een time zones and settlem ent failure. On June 26, 1974, Germ an banking authorities closed H erstatt as of the close of business in Germany. H erstatt had received paym ent in m arks during Germ an banking hours for its foreign exchange transac tions w ith th at value date. It w as closed, how ever, before the tim e for m aking paym ents in dollars in New York. C ounterparties of H erstatt w ere left w ithout the dollars they expected, af ter paying m arks to H erstatt earlier in the day.lz O ur example of settlem ent failure in this paper reflects the implications of differences in time zones. One bank is assum ed to go bankrupt af ter the time for paym ents in pounds but before the tim e for m aking paym ents in dollars. P re-S ettlem en t F ailure — Suppose Bank A goes ban krup t before the value date. W ithout a netting agreem ent betw een Banks A and B, the legal obligations of each bank are those speci fied in the individual transactions betw een them . W ith an exchange rate of $1.65 on the value date, transaction num ber 1 is profitable to Bank A. The receiver of Bank A will pay £ 100 to Bank B to settle transaction num ber 1. Bank B is T im in g o f B a n k r u p t c y a n d S ize o f 12Moore (1974). FEDERAL RESERVE BANK OF ST. LOUIS obligated to pay $175 to Bank A to settle this transaction. Since transaction num ber 2 is not profitable to Bank A on the value date, the re ceiver will default on transaction num ber 2 . Bank B anticipated a profit of $2.50 on the value date from transaction num ber 2. Thus, the bank ruptcy of Bank A im poses a loss of $2.50 on Bank B. Table 5 shows the loss to each bank due to the bankruptcy of its counterparty be fore the value date, under both gross settlem ent and netting arrangem ents. S ettlem en t F ailure — Suppose Bank A goes bankrupt on the value date after paym ent in pounds bu t before paym ent in dollars. Bank A defaults on its paym ent of $85 to Bank B on the value date. U nder gross settlem ent of transac tions, how ever, Bank B is obligated to pay the $175 to Bank A. Bank B becom es a general cre ditor of Bank A for $85. The m aximum loss to Bank B, as table 6 indicates, is $85. Settlem ent failure can create liquidity problem s for the counterparties of a failed bank. Suppose Bank B pays the $175 to Bank A before discov ering that Bank A is bankrupt. The cash bal ances of Bank B denom inated in dollars will be $85 below the level it had projected for the value date. Bank B m ight request a discount w indow loan from the Federal Reserve to cover the $85 shortfall in its reserve account. H ow B ilateral Netting A ffects the Losses in Settlem ent o f Foreign Ex change Transactions If Banks A and B engage in bilateral netting, the effects of the bankruptcy of Bank A on Bank B depend on w hether paying the net am ount dis charges the obligations betw een counterparties. L e g a l A s s u m p tio n s — U nder one type of agreem ent called position netting, tw o banks agree to net their paym ents to reduce transac tions costs, but the agreem ent has no effect on their legal obligations. U nder the legal assum p tions in this paper, the position netting agree m ent w ould not prevent the receiver from m aking paym ents in settlem ent of some transac tions but defaulting on others w ith the same counterparty. The bankruptcy court w ould treat the paym ent obligation of Banks A and B as though they had no netting agreem ent. The bankruptcy of one party has the same implica- 11 Table 5 Bank Losses from Pre-Settlement Failure Losses to Failure of Bank A Bank B Gross settlement Bilateral netting Multilateral netting 5.00 0.00 0.00 $15.00 10.00 2.50 $10.00 7.50 0.00 Bank C Gross settlement Bilateral netting Multilateral netting BankC $ 2.50 0.00 0.00 Bank A Gross settlement Bilateral netting Multilateral netting Bank B 2.50 0.00 0.00 15.00 12.50 2.50 Table 6 Bank Losses from Settlement Failure Losses to Failure of Bank A Bank B Gross settlement Bilateral netting Multilateral netting 170.00 0.00 0.00 $262.50 92.50 2.50 $175.00 90.00 0.00 Bank C Gross settlement Bilateral netting Multilateral netting Bank C $ 85.00 0.00 0.00 Bank A Gross settlement Bilateral netting Multilateral netting Bank B 85.00 0.00 0.00 262.50 177.50 85.00 tions for the counterparty as if they settled transactions using the gross settlem ent m ethod. 13Bank for International Settlements (1989), p. 13. 14One firm that offers legal advice and a communications network for bilateral netting by novation is FXNET. The netting contract drafted by FXNET includes netting by no vation and closeout. See Bartko (1990). For further refer ence to FXNET, see Scarlata (1992), this Review. Plans Netting agreem ents th at reduce this exposure to loss m andate th at banks discharge th eir obli gations by paying the net am ount of the tran s actions betw een them . The legal language for such agreem ents is netting by novation. This paper assum es th at bankruptcy courts recognize a contract for netting by novation as the only contract betw een counterparties for settlem ent of foreign exchange transactions. A provision of bilateral netting contracts that reduces risk is called closeout, w hich becom es effective w hen a receiver or liquidator is ap pointed after a bank declares bankruptcy .13 A netting agreem ent includes a form ula that con verts all outstanding transactions betw een a pair of counterparties, for all value dates, into one am ount payable immediately. The closeout provision prohibits the receiver of a bankrupt bank from m aking paym ents in settlem ent for transactions w ith some value dates b u t default ing on transactions w ith other value dates .14 B ankruptcy courts are assum ed to recognize closeout provisions as valid parts of netting a r rangem ents. P re - s e ttle m e n t F a ilu re — As the bottom half of figure 1 illustrates, th e one contract be tw een Banks A and B under netting by novation calls for Bank A to pay £ 50 and receive $90. At the exchange rate of $1.65 on the value date, this contract is profitable for Bank A. Thus, the receiver of Bank A w ould pay the £ 50 to Bank B to settle the contract. The bankruptcy of Bank A prior to the value date w ould impose no loss on Bank B, since Bank B had anticipated honor ing its contract w ith Bank A before discovering th at Bank A was bankrupt. In each case of presettlem ent failure illustrated in table 5, the loss es are sm aller un d er bilateral netting by nova tion than under gross settlem ent. S e ttle m e n t F a ilu re — The bankruptcy of Bank A after paym ents in pounds but before paym ents in dollars imposes no loss on Bank B since, under the netting agreem ent, Bank A had no obligation to pay dollars to Bank B. As table 6 indicates, in settlem ent failure, the loss to a bank from the bankruptcy of its counterparty is for multilateral netting include similar closeout provisions in contracts between individual members and the clearing houses that would act as paying agents for the netting ar rangements. See Duncan (1991). These closeout provi sions limit the losses of solvent banks resulting from default by counterparties. JANUARY/FEBRUARY 1992 12 sm aller under bilateral netting by novation than u n d er gross settlem ent for each com bination of failed bank and counterparty. The Im portance o f the Legal Basis f o r Netting A greem ents The assum ptions in this paper concerning the principles that bankruptcy courts follow yield the m aximum reductions in losses from netting. These reductions in losses could be sm aller un der alternative assum ptions. The Lamfalussy Report indicates that bilateral netting could increase risk in settling foreign ex change transactions if netting arrangem ents do not have a sound legal basis. If netting "obscures the level of exposures, then netting arrange m ents have the potential to contribute to an in crease in systemic risk .”15 The argum ent that bilateral netting may pose greater risks is based on assum ptions about how banks that are active in the foreign exchange m arket set credit limits w ith counterparties. Banks w ith bilateral netting agreem ents may set credit limits w ith each other based on their net positions rath er than the gross value of the underlying transactions be tw een them . If a bankruptcy court requires paym ents by a solvent counterparty based on the value of the underlying transactions rather than the netting agreem ent, the exposure of the solvent counterparty w ould be larger th an ex pected. This point indicates w hy the Lamfalussy Report em phasizes the legal basis for netting ar rangem ents (table 1 ). M ultilateral Netting Banks m ay be able to fu rth er reduce their transaction costs and their exposure to loss by engaging in m ultilateral netting. No m ultilateral netting arrangem ents are in operation at this time. This section examines the implications of a m ultilateral netting arrangem ent modeled after a draft of the plans of the ECHO NETTING sys tem in London .16 L e g a l A s s u m p tio n s — In th e contract for m ultilateral netting, m em bers of a netting ar rangem ent establish a clearing house, which receives and pays out currencies in settlem ent of foreign exchange transactions. The clearing house is the counterparty for each transaction betw een m em bers of the m ultilateral netting ar rangem ent. Each m em ber settles its legal obliga 15Bank for International Settlements (1990c), p. 3. FEDERAL RESERVE BANK OF ST. LOUIS tions w ith the others by m aking paym ents to the clearing house. The clearing house assum es responsibility for paying all net am ounts due to m em bers, even if a m em ber defaults on its pay m ents to the clearing house. The contract in a m ultilateral netting arrange m ent is assum ed to include a closeout provision. If a m em ber of the clearing house goes bank rupt, its receiver has only one decision to m ake about the foreign exchange transactions that the failed bank negotiated w ith other m em bers: m ake the paym ents to settle the one contract w ith the clearing house or default. P a y m e n t s F lo w s a n d L o s s S h a rin g — Figure 2 presents the paym ents betw een m em bers of the netting arrangem ent and the clear ing house, derived from paym ents that w ould be m ade un d er bilateral netting in table 4. The calculation of the num bers in figure 2 is illus trated for Bank A. U nder bilateral netting, Bank A pays the other banks £ 50 (Bank B) and $92.50 (Bank C) and receives $90 (Bank B) and £ 50 (Bank C). U nder m ultilateral netting, therefore, Bank A owes the clearing house $2.50 and the clearing house owes Bank A nothing on the value date. Figure 2 also indicates the paym ents betw een the clearing house and Banks B and C. Any clearing house losses resulting from the default of a m em ber are allocated to the other m em bers in proportion to the losses they would have in curred under bilateral netting. This for mula gives each m em ber of the arrangem ent an incentive to avoid transactions w ith m em bers it considers to be in danger of going bankrupt. P r e -S e ttle m e n t F a i lu r e — If Bank A goes bankrupt before the value date, its receiver will default on the paym ent of $2.50 to the clearing house. The loss of $2.50 is allocated to Bank C, since only Bank C w ould have a loss under bilateral netting. If Bank B goes ban krup t before the value date, its receiver will m ake the paym ent of £ 50 to settle the contract w ith the clearing house, since it yields a profit of $5 to Bank B. As table 5 indicates, the bankruptcy of Bank B before the value date imposes no loss on the other banks. The bankruptcy of Bank C imposes a loss of $2.50 on Bank B. In each case in table 5, the 16Duncan (1991). 13 Figure 2 Payments Between Members of a Multilateral Netting Arrangement and the Clearing House BANK A BANK B Profit to Bank A: -$2.50 Profit to Bank B: $5.00 CLEARING HOUSE Profit to Bank C: -$2.50 N um ber of Payments: 5 Dollar Value: $340 JANUARY/FEBRUARY 1992 14 A Glossary of Terms B i l a t e r a l n e ttin g C le a r in g h o u s e C o u n te rp a rty F o r e ig n e x c h a n g e tra n s a c tio n G ro ss s e ttle m e n t L egal rig h ts of s e t-o ff M u ltila te r a l n e t t i n g N e ttin g N e ttin g b y n o v a tio n P o s i t i o n n e ttin g P re - s e t t l e m e n t f a ilu r e S e ttle m e n t S e ttle m e n t f a ilu r e S y s te m ic r i s k V a lu e d a te An arrangem ent betw een tw o parties in w hich they exchange only the net units of the currencies specified in the transactions betw een them , ra th e r than the currencies for each transaction individually. An institution established by a group of banks to facilitate the settle m ent of obligations am ong them selves. Each bank settles its obliga tions w ith the others by m aking paym ent to the clearing house for the net am ount ow ed the other m em bers. The other party in a transaction. In a foreign exchange transaction, one party agrees to m ake paym ent in one currency and its counter party agrees to pay in another currency. An agreem ent by tw o parties (generally large banks) to exchange cu r rencies on a given date. A m ethod of m aking paym ents betw een a pair of parties in w hich each party m akes a separate paym ent in settlem ent of each transac tion betw een them . U nder bankruptcy law, a right to net obligations w ith a bankrupt counterparty. An arrangem ent betw een th ree or m ore parties in w hich each m em b er m akes paym ents to a clearing house for the net paym ents due to the other m em bers and receives from the clearing house th e net am ounts due from the other m em bers. An arrangem ent by w hich parties w ith m ore th an one transaction to settle on a given date exchange only the net am ounts of the transac tions betw een them . The replacem ent of tw o existing contracts betw een tw o parties for delivery of a specified currency on the sam e date by a single net contract for that date, so that the original contracts are satisfied and discarded. The netting of paym ent obligations betw een tw o or m ore parties w hich neither satisfies no r discharges the original obligations that w ere netted. Bankruptcy of a bank prior to the value date of transactions w ith a counterparty. Completion of a paym ent betw een tw o parties discharging an obli gation. Default by a bank on paym ent in one currency after the bank and its counterparty had m ade paym ents in the other currency. The risk th at the inability of one institution w ithin a paym ent system to m eet its obligations w hen due will cause other participants to be unable to m eet th eir obligations w hen due. The date on w hich banks exchange currencies in settlem ent of a for eign exchange transaction. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 15 loss un der m ultilateral netting is either zero or sm aller than under bilateral netting. S e ttle m e n t F a i l u r e — Suppose Bank A goes bankrupt after the paym ent of pounds but be fore the paym ent of dollars. The loss to be borne by m em bers of the clearing house is $2.50, the paym ent obligation of Bank A. This loss is im posed on Bank C, w hich w ould have a loss of $92.50 under bilateral netting (table 6). Bank B has no obligation to pay dollars to the clearing house. Thus, the bankruptcy of Bank B after the paym ent of pounds but before the paym ent of dollars imposes no losses on other m em bers of the clearing house. The bankruptcy of Bank C after paym ent in pounds imposes a loss of $85 on Bank B. The loss in each case u n der m ultilateral netting in table 6 is either zero or less than the loss under bilateral netting .17 L iquidity R equirem ents o f the Clearing H ouse One of the concerns central bankers have about m ultilateral netting is w hether the clear ing house w ould have access to sufficient li quidity to m ake paym ents to other m em bers if one of them defaults. The Lamfalussy Report in dicates th at a clearing house should "be capable of ensuring the timely com pletion of daily settle m ents in the event of an inability to settle by the participant w ith the largest single net-debit position.” In figure 2, Banks A and C each have net debit positions of $2.50. The clearing house w ould need access to at least $2.50 to m eet the minim um liquidity requirem ent of the Lamfal ussy Report. This requirem ent is a cost of oper ating the clearing house, either as the oppor tunity cost of liquid assets held by the clearing house or the cost of credit lines. Bilateral net ting, in contrast, involves no such costs. CONCLUSIONS Banks assum e risk in settling foreign exchange transactions. This paper exam ines the implica tions of netting by using a hypothetical exam 17The generality of the result that netting reduces losses to solvent counterparties can be investigated by simulating the losses resulting from default in an example with more banks and more transactions and with some terms of the transactions chosen at random. Our simulation includes 10 banks. Each has 10 transactions to settle with each of the other nine banks. The size of the transactions and ex change rates are chosen at random. The multilateral net- ple. The exam ple shows how netting schem es can reduce the size of losses to counterparties w hen a bank goes bankrupt and is liquidated. A com m ittee of central bankers from the m a jor developed countries recently exam ined the implications of netting arrangem ents for risk. The com m ittee's report indicates that netting ar rangem ents may either increase or decrease risk, depending on w hether they m eet certain m inim um standards listed in the report. Bilateral netting could reduce the loss w hen a counterparty defaults, if the bankruptcy court w ould recognize the paym ent of the net am ount betw een the counterparties as a settlem ent of the transactions betw een them . It could increase risk in settlem ent of foreign exchange transac tions, however, if counterparties set credit limits based on their net exposures but the court requires paym ent in settlem ent for each underlying transaction betw een counterparties. M ultilateral netting can reduce the losses re sulting from default even m ore than bilateral netting, if the clearing house created to settle transactions has access to the liquid assets ne cessary to com plete the settlem ent. Lack of sufficient liquidity for the clearing house could create a m ajor disruption in the operation of the paym ent system. REFERENCES Bank for International Settlements. Report on Netting Schemes (BIS, February 1989). _______ . Survey of Foreign Exchange Market Activity (BIS, F eb ru ary 1 9 9 0 a ). _______ . Large-Value Funds Transfer Systems in the Group of Ten Countries (BIS, May 1990b). _______ . Report of the Committee on Interbank Netting Schemes of the Central Banks of the Group of Ten Coun tries (BIS, November 1990c). Bank of England. “ The Market in Foreign Exchange in Lon don,” Bank of England Quarterly Bulletin (November 1989), pp. 531-35. ting arrangement uses the loss-sharing formula described above. Each of the 10 banks is assumed to go bankrupt, imposing losses on nine counterparties. In each of the 90 cases of losses to counterparties, losses are smaller under bilateral netting than under the gross settlement method, and losses under multilateral netting are either zero or smaller than under bilateral netting. JANUARY/FEBRUARY 1992 16 Bartko, Peter. "Foreign Exchange and Netting by Novation,” Payment Systems Worldwide (Spring 1990), pp. 48-51. Chrystal, K. Alec. “A Guide to Foreign Exchange Markets,” this Review (March 1984), pp. 5-18. Cody, Brian J. “ Reducing the Costs and Risks of Trading Foreign Exchange,” Federal Reserve Bank of Philadelphia Business Review (November/December 1990), pp. 13-23. Deeg, Ernst. “A Proposal for a Multi-Currency Netting Sys tem,” Payment Systems Worldwide (Spring 1990), pp. 40-45. Duncan, G. M. "ECHO NETTING: Multilateral FX Netting in Europe,” mimeo, Barclays Bank (March 1991). Federal Reserve Bank of New York. “A Study of Large-Dollar Flows Through CHIPS and Fedwire,” (December 1987). _______ . “ Summary of Results of U.S. Foreign Exchange Market Survey Conducted in April 1989” (September 13, 1989). FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis _______ . The Clearing House Interbank Payments System (January 1991). Juncker, George R., Bruce J. Summers and Florence M. Young. “A Primer on the Settlement of Payments in the United States,” Federal Reserve Bulletin (November 1991), pp. 847-58. Luthringshausen, Wayne P. “ Characteristics of a Compre hensive Clearing System,” Payment Systems Worldwide (Spring 1990), pp. 38-39. Moore, A. E. “ Foreign Exchange,” Bankers' Magazine (August 1974), pp. 49-51. Polo, Renato. “A Perspective on the Euronetting Project,” Payment Systems Worldwide (Spring 1990), pp. 46-47. Scarlata, Jodi G. “ Institutional Development in the Globaliza tion of Securities and Futures Markets,” this Review (Janu ary/February 1992), pp. 17-30. Summers, Bruce J. “ Clearing and Payment Systems: The Role of the Central Bank,” Federal Reserve Bulletin (Febru ary 1991), pp. 81-91. 17 Jodi G. Scarlata Jodi G. Scarlata was a visiting scholar at the Federal Reserve Bank of St. Louis when this paper was written. David H. Kelly and Lynn Dietrich provided research assistance. Institutional Developments in the Globalization of Securities and Futures Markets r [NANCIAL TRANSACTIONS such as the buying and selling of securities, commodities, foreign exchange and bonds, have increasingly involved individuals and firm s from different countries. For example, a Japanese resident m ight purchase U.S. dollars w ith Japanese yen (a foreign ex change transaction) to buy shares of IBM on the New York Stock Exchange (a securities transac tion). To accom m odate such transactions, futures and securities exchanges have expanded the services they offer their users, adding num erous financial instrum ents, engaging in cooperative efforts across exchanges and introducing computer-based technologies. The globalization of w orld m arkets provides significant benefits, including greater opportuni ties for investors to diversify risk, and access to broader m arkets for dem anders of funds. Inter national trading in financial instrum ents, how ever, does pose risks, some of w hich can be m itigated by coordination betw een global finan cial m arkets. This paper describes recent institutional de velopm ents in the globalization of financial m ar kets and discusses the advantages and disadvan tages of these innovations. The paper opens w ith a brief overview of the various transn a 1Abken (1991), p. 3. tional developm ents th at are occurring in world securities and comm odities m arkets. It then ad dresses both the benefits of expanding financial m arkets and the costs that accom pany the move. Risk factors and standardization of procedures are highlighted as issues of concern as financial centers globalize. The paper closes w ith a dis cussion of the Group of T hirty proposal for the coordination of clearing and settlem ent in w orld securities m arkets. INTERNATIONAL DEVELOP MENTS IN FINANCIAL AND COMMODITY MARKETS The tren d tow ard internationalization of finan cial m arkets can be illustrated by highlighting the rapid increases in transactions in a few m arkets. For example, cross-country activity, w hen m easured as the volume of foreign tran s actions in securities of U.S. firm s (aggregate purchases and sales), grew from $75.3 billion in 1980 to $361.4 billion in 1990.1 Similarly, U.S. transactions in securities of foreign firm s (aggre gate purchases and sales) grew from $17.85 bil lion to $253.4 billion betw een 1980 and 1990.2 In futures and options m arkets, 20 new exchanges 2lbid. JANUARY/FEBRUARY 1992 18 w ere established w orldw ide betw een 1985 and 1989, bringing the total to 72.3 Likewise, nearly 40 million futures and options contracts w ere traded w orldw ide in 1988, an increase of ap proxim ately 186 percent since 1983.4 Eurodollar interest rate futures saw an especially large change, increasing alm ost 70 percent annually betw een 1983 and 1988.5 ing the position book betw een time zones is ac tually to transfer the handling instructions be tw een traders. An exam ple is a New York currency trad er w ho instructs the trad er at his Singapore office to track the price of a currency during evening hours in New York. W hen the m arket reaches a particular price, the Singapore trad er will buy or sell, depending on instructions from New York. The globalization of financial and com m odity m arkets involves num erous activities and institu tional developm ents th at facilitate access to for eign m arkets, w h eth er by a trad er or a security. One of these activities is the cross-listing of securities in several countries. Cross-listing sim ply m eans, for example, that a firm in the United States lists its stocks on a London exchange. In 1990, the International Stock Exchange (ISE) of London had one of the highest percentages (23 percent) of foreign com pany stock listings .6 A nother tren d is cross-country hedging and portfolio diversification. A U.S. trader, for exam ple, can diversify a portfolio com posed of U.S. stocks by buying stocks of a U.K. firm in Lon don through a London broker. Globalization can also m ean holding m em bership in another coun try ’s exchanges. For example, after “The Big Bang” of 1986 in London, m any U.S. securities firm s and banks applied to buy seats on London exchanges .7 A th ird tren d in the internationalization of financial m arkets is called “passing the book,” w hereby control of trading is passed betw een traders at exchanges around th e globe. This enables 24-hour trading of a financial instrum ent. An exam ple of this w ould be a U.S. investm ent firm trading from New York during U.S. and Japanese hours and from its London desk during U.K. hours. The m ore com m on practice of pass One tren d that does not involve actual trading is the underw riting of corporate securities through offices outside the hom e country. An u n derw riter is a firm th at buys an issue of securities from a com pany, th en resells it to in vestors. For the com pany issuing the securities, underw riting provides a guarantee th at a certain am ount of m oney will be derived from the sale of the securities that can be used for capital ex penditure. A large stock issue may have un d er w riters from several countries, for example, to com pensate for a country w hose capital m arket does not have the depth to handle large securi ties offerings .8 The distribution of underw riters across several countries provides the issuing firm w ith a w ider access to funding .9 Investors, on the other hand, obtain a broader selection of securities. Illustrations o f G lobalization 3Baer, Evanoff and Pavel (1991), p. 11. ‘'Ibid. E uro do lla r deposits are dollar-denominated deposits out side the United States. Eurodollar interest rate futures con tracts are futures contracts on the interest rates on these deposits. The figures are from Baer, Evanoff and Pavel (1991), p. 11. 6For the New York Stock Exchange, the figure was 3.7 per cent. U.S. Congress, Office of Technology Assessment (1990), p. 29. TThe Big Bang was a deregulation effort for British financial markets which began on October 27, 1986. Examples of changes to the London equity markets are the end of fixed commission rates; barriers between order-taking brokers and risk-taking market makers were broken down; and, the self-regulating Securities and Investment Board (SIB) was FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis D evelopm ents in A utom ated Trad ing S ystem s M ore recently, the developm ent of autom ated trading has received substantial attention. Auto m ated or electronic trading system s allow agents to m ake trades via com puter, w ithout the "open outcry” or pit auction system .10 Interestingly, the developm ent of m uch of the cu rren t auto m ation is an extension of technological innova tions originally developed for dom estic m arkets. It is clear, how ever, that this autom ation has afestablished. The SIB is a non-governmental version of the Securities and Exchange Commission (SEC) in the United States. Khoury (1990), p. 129. 8Depth means that there are enough buyers and sellers in a market that a large transaction will not affect the price. 9An example is the privatization of French companies in 1986, where the value of these newly privatized companies was approximately $30 billion, but the total value of list ings on the Paris Bourse was only about $80 billion. U.S. Congress, Office of Technology Assessment (1990), p. 34. 10Open outcry occurs on an organized exchange when ord ers between buyers and sellers are traded between third parties in anonymity. The buyer/seller enters into a con tract with the exchange or its representative clearinghouse. 19 fected the globalization of financial m arkets con siderably. The National Association of Securities Dealers Autom ated Quotations (Nasdaq) was one of the earliest developm ents in financial m arket com puterization, beginning in 1971. Nasdaq p ro vides com puter listings of price inform ation for several thousand companies. By 1982, the Na tional Association of Securities Dealers (NASD) had produced the National M arket System, w hich provided investors w ith inform ation as sales w ere com pleted, and by 1991, had developed the Private Offerings, Resales, and Trading through Autom ated Linkages (PORTAL) system .11 From a com puter term inal, PORTAL enables users to trade in unregistered domestic and for eign debt and equity securities .12 Nasdaq has established com puter telephone linkage as well as autom ated trade execution and international clearing and settlem ent with the International Stock Exchange of London and the Stock Exchange of Singapore. Nasdaq has since becom e a significant m arket for the listing of foreign securities, trading approxim ately $6 billion in foreign securities as of 1991, up from the $2.6 billion in 1985.13 Thus, Nasdaq provides the cross-listing of securities, together w ith the rapid trade execution of an autom ated system. The grow th of international trading has also affected futures and options exchanges in the United States. The fact th at traders could access instrum ents and overseas m arkets after norm al U.S. trading hours had ended, provided a m otiva tion for m any of the extended-hour and 24-hour trading initiatives (see shaded insert for examples of extended trading hours and table 1 for auto m ated trading systems). A significant portion of U.S. financial in stru m ents, futures and options is traded at exchanges throughout the world. That is, foreigners do not have to use the Chicago M ercantile Exchange to trade Eurodollar contracts, a CME staple. For "P O R T A L uses a book entry settlement system with no physical delivery, eliminating the problem of unmatched trades. PORTAL is currently the only fully automated clear ing and settlement system in the United States. Clearing and settlement issues, book entry and matching trades will be discussed in more detail later in the text. 12Unregistered securities are not registered with the Securi ties and Exchange Commission (SEC). They are issued in a limited volume or by small companies. example, in 1989, a th ird of th e trade in con tracts offered by the CME originated outside of N orth Am erica .14 In 1989, 10 percent of the CME’s daily volume w as transacted overnight in an overseas exchange while the CME was closed .15 F urtherm ore, in 1985, the CME and Chicago Board of T rade together accounted for 83 percent of all futures volume. By 1990, the figure had fallen to 55 percent .16 The attem pt to regain m arket share instigated such CME expan sions as extended trading hours and autom ated trading systems. Perhaps the m ost am bitious project in auto m ated trading is Globex, an attem pt to create a 24-hour trading m arket originally proposed in 1987 by the CME .17 Globex is an electronic trade execution system w hereby traders enter buy and sell orders th at are m atched autom atically according to price and tim e priority .18 Originat ing as a strictly off-hours trading system, the purpose of Globex is to enable continued active trading beyond the CME's regular trading hours. The CME intends to use Globex to access m arkets after its ow n close of business and re gain some of the m arket share lost to foreign exchanges. POTENTIAL GAINS AND RISKS OF INTERNATIONALIZATION The significant changes accom panying the in ternationalization of financial trading systems al low for th e realization of substantial gains; at the same time, globalization also exacerbates the risks already present in financial trading. The m ost significant of these gains and risks are described below. B enefits o f G lobalization One of the m ost im portant areas of progress is the speed w ith w hich inform ation is processed and dissem inated to m arket participants. In creased flow of m arket data provides greater 15lbid. 16Chesler-Marsh (1991), p. 33. 17The Chicago Board of Trade has since become a par ticipant with the CME in the Globex project. 18For detailed reading on automated trading systems, see Domowitz (1990). 13See Nasdaq (1991), p. 16, for the 1991 figure, and NASD (1991), p. 15, for the 1985 figure. 14Hansell (1989), p. 187. JANUARY/FEBRUARY 1992 20 “After-Hours" Trading The following is a listing of some of the de velopm ents in trading after regular operating hours. N ew York S tock Exchange Has tw o after-hours trading sessions, begin ning June 13, 1991. One 45-m inute, orderm atching session and a one-hour-and-15-minute session for crossing baskets of stocks. A m erican Stock Exchange Plans for an after-hours session betw een 4:15 p.m. and 5 p.m. w here buyers and sellers can trade at regular session’s closing price. C urrently, talks w ith R euters Holdings PLC, the Chicago Board Options Exchange and the Cincinnati Stock Exchange for a global after-hours trading system for stocks and their options and derivatives such as stock-index w arrants. N asdaq Has filed w ith the SEC to start trading on Nasdaq International at 3:30 a.m. EDT. P acific S tock Exchange Has filed plans w ith the SEC to extend tra d ing in listed stocks in a regular auction m ar ket lasting until 4:50 p.m. EDT. Currently, PSE closes at 4:30 p.m. EDT. Will m atch buy and sell orders in a 5 p.m. EDT session at the NYSE closing price. Philadelphia S tock Exchange T rades currency options and futures for 20.5 hours a day. Has filed plans w ith the SEC for an after-hours order m atching sys tem similar to the Big Board. Also plans to fill custom er orders based on after-hours activity on the Big Board. B oston S tock Exchange Will not have its ow n session, but will fill custom er orders based on after-hours activity on the Big Board. Chicago B oard O ptions Exchange M idw est Stock Exchange Has no cu rren t plans to trade stock options before or after norm al trading hours. Involved w ith Amex in electronic system. Will not have its ow n after-hours session, but will fill custom er orders based on afterhours activity on the New York Stock Ex change. An electronic trading system ow ned by Reuters Holdings, operates up to 16 hours a day. Instinet ’ “ Big Board After-Hours Trading” (1991). accessibility to foreign m arkets. In turn, the larger the num ber of participants using a m arket, the greater the liquidity of the m arket and, thus, its desirability for investors .19 The level of m arket activity and the transm is sion of data on prices, m arket supply and de m and conditions are a few examples of inform a tion relevant to traders. Yet, this inform ation technology has less to do w ith im proving the ef ficiency of forecasting techniques than it does to shaving seconds off the receipt of up-to-the19Liquidity is the depth of the market (for example, securities or futures) and its ability to absorb sudden shifts in supply and demand without excessive price fluctuation. FEDERAL RESERVE BANK OF ST. LOUIS m inute m arket events. Inform ation like the an nouncem ent of a com m odity quota or a cor porate m erger provides the im petus for the rapid decision-making th at characterizes financial trading. Access to foreign stocks provides investors w ith opportunities for diversification; investors need inform ation about the foreign firm (for ex ample, its financial stability or successful m anage ment), how ever, in order to m ake an investm ent decision. The rapid rise of inform ation technolo- 21 gy increases the fam iliarity of foreign corpora tions and their operations. This spread of infor m ation reduces one of the traditional obstacles to foreign investm ent and opens up both savings and investm ent opportunities for firm s and in dividuals. The payoff is a m ore efficient alloca tion of capital and, thus, a stim ulus to production and real output. Likewise, the advent of alm ost im m ediate transfer of inform ation around the globe reduces the inform ational discrepancies betw een m arket participants. A rbitrageurs, of course, attem pt to profit from price discrepancies. The m ore people have access to the same inform ation, how ever, the m ore likely price discrepancies will be spot ted and acted upon. The divergence of prices from th eir no-arbitrage relationship (which p ro vides profit potentials), will be quickly arbitraged away as both the quantity and speed of infor m ation transfer is enhanced .20 A nother benefit of internationalization is ac cess to m arkets otherw ise inaccessible. As m en tioned earlier, the Chicago M ercantile Exchange will be accessible after regular trading hours through Globex. Not only will U.S. traders now be able to operate after U.S. trading hours, but foreign traders can also use Globex to operate during th eir ow n regular trading hours. New m arkets enable investors to introduce diversity into their portfolios in both the type of in stru m ent and the country from w hich it is issued. Just as com puterized systems, such as Globex, facilitate diversification and accessibility, so too can other m ethods, such as cross-exchange list ings and cross-m em berships. R isks in G lobalization Trading in financial assets, w h eth er done domestically or across national boundaries, in volves risk. Some of these risks are m ore im por tan t in an international setting than a strictly dom estic setting. They occur prim arily at vari ous stages of the clearing and settlem ent process. Unlike risks comm only associated w ith price uncertainty, the risks in clearing and set tlem ent procedures involve uncertainty about the timely paym ent of funds and the transfer of assets in financial trades. 20An example of a no-arbitrage condition is that the differ ence between the cash and futures price of a storable commodity, at any point in time, should reflect carrying costs of storing the commodity until maturity. If the price differential exceeds carrying costs, then there exists an in- An exam ple of a typical securities transaction can provide a clear illustration. Once a securities trade is executed, the m em ber firm s involved subm it the trade inform ation for confirm ation to the clearing agent. The trade is th en com pared and m atched by com puter for accuracy and the inform ation on the trade is sent to the relevant m em bers on either the day of the trade or the day after. If both parties concur w ith the conditions of the trade, the trade is ready for settlem ent. At present, settlem ent in securities occurs five days after the trade in the United States. Using this example, the next section will briefly discuss the concepts and institutions in clearing and settlem ent procedures before introducing the specific risks of globalization. Clearing and Settlem ent P rocedu res “Clearing” a trade involves th e confirm ation of the type and quantity of the financial in stru m ent being traded, th e transaction date and price, and the identification of th e buyer and seller. “Settlem ent” m eans the fulfillm ent of the obligations of the transaction. In equities and bonds, for example, settlem ent m eans paym ent to the seller and delivery of the security certifi cate or transfer of ow nership to the buyer. The clearing and settlem ent process depends on the institutions th at facilitate transactions. Commodities and securities exchanges provide the facilities for trad ers to conduct their busi ness, establish and enforce trading rules, collect and distribute m arket and economic inform ation about prices, and provide an institutional fram e w ork for arbitrating and settling disputes. A nother institution—the clearinghouse— com pares trades betw een parties and can re move risk from the settlem ent process .21 A clearinghouse places itself betw een the buyer and seller, ensuring th at the buyer receives the instrum ent purchased and the seller receives paym ent. T hat is, by becom ing the counterparty to all trades, the clearinghouse guarantees every trade. Each participant has a net obligation w ith the clearinghouse to buy or sell th e security centive to enter the market, that is, to buy today and sell at the higher futures price. 21Trade comparison involves confirming and matching the terms of the trade to ensure accuracy. JANUARY/FEBRUARY 1992 22 Table 1 Automated Trading Systems System’ Sponsor Purpose Instruments Access New York Mercantile Exchange (NYMEX) Computerized screen trading system to automatically match trade on a first-in-first-out basis. Allows traders to select a standing bid or offer (but are blind to the counterparty chosen). Energy futures and futuresoptions for crude oil, heating oil, gasoline, propane and natural gas. Automated Pit Trading (APT) London International Financial Futures Exchange (LIFFE) Intended to extend trading hours to cover European trading day. Its aim is to copy the life of the trading floor on to a computer screen. FT-SE 100 index futures and most of LIFFE's main contracts. EJV (Electronic Joint Venture) Collaboration between First Boston, Goldman Sachs, Morgan Stanley, Salomon Brothers, Shearson Lehman and Citibank Trading system that allows dealers to buy and sell securities electronically using voice-activated computer technology. Also provides price and analytic services. Currently restricted to Treasury bills and notes with maturity of less than three years. Once established, it expects to extend coverage to all maturities. Euroquote European Community (EC) national stock exchanges A European-wide share trading system. Will combine price infor mation from 12 EC exchanges into an electronic feed for sub scribers. Eventually, may be come a full trading system and integrate Euroquote with a set tlement system to decrease the cost and difficulties of settling cross-country transactions. (This proposal has since been aban doned by the chairmen of Eu rope’s national stock exchanges.) EC stocks Globex Chicago Mercantile Exchange (CME), Reuters and Chicago Board of Trade (CBOT) An automated trading system with anonymous buy and sell orders that are matched by price and time. (See text for details.) Traded instruments will be introduced in three separate waves: 1) financial futures and options, e.g., Eurodollar futures, futures and options on Eurodollar currencies— Deutschemark, yen, pound sterling, Canadian and Aus tralian dollars, Swiss franc, LIBOR, and U.S. Treasury bond and note futures and options; 2) equity-related products; and 3) agriculturerelated futures. FEDERAL RESERVE BANK OF ST. LOUIS 23 Table 1 (continued) Automated Trading Systems System1 Sponsor Purpose Instruments Quotron System Inc. Current testing of the prototype involves The Bank of America, Barclays Bank, Chase Manhattan Bank, Chemical Bank, Citibank, Credit Suisse, Lloyds Bank, Midland Bank, Morgan Guaranty, National Westminster, Swiss Bank Corp. and Union Bank of Switzerland. Joint project to develop a computer system that automatically matches and executes foreign exchange trades. Trades in foreign exchange. Swiss Options and Financial Futures Exchange (SOFFEX) Owned by Switzerland’s three leading stock exchanges and five largest banks. Fully automated trading and clearing system, where quotes and orders are recorded, sorted and matched automatically. The computer screen is com posed of five segments, each with different types of informa tion and automatically updated throughout the day. Futures and options on the 13 underlying stocks and the Swiss Market Index (SMI), a basket of Switzerland's 24 leading stocks. 10 f these systems, only LIFFE’s APT system and SOFFEX are currently in operation. NOTE: For a more extensive survey of automated systems, see Peter A. Abken, “ Globalization of Stock, Futures, and Options Markets,” Federal Reserve Bank of Atlanta Economic Review (July/August 1991). based on h er net position w ith other participants in the clearinghouse. A third institution—the depository—is an or ganization (not necessarily p art of an exchange) that holds stocks and bonds for safekeeping on behalf of their ow ners. It has a com puterized accounting system to record and transfer ow ner ship of securities betw een participants by in tegrating a book-entry system w ith a m oney transfer system .22 The procedures for clearing and settlem ent vary across countries. At present, there are th ree com m on m ethods of clearing and settle m ent. Each involves various com binations of the th ree central institutions involved in futures and securities m arkets. 22A book-entry system means a credit or debit to a cus tomer’s account will transfer securities between buyer and seller. A money transfer system transfers the funds be tween the parties to the trade, such as a wire transfer. The first model is exemplified by the United Kingdom’s equities m arket. In this model, there is neither a central depository nor a separate clearinghouse. Instead, the stock exchange itself is responsible for trade m atching and confirm a tion as well as providing a location for the deliv ery and receipt of securities and paym ents be tw een trad ers .23 The second model, exem plified by G erm any’s D eutscher Kassenverein depository system, has no independent clearinghouse, b u t does have a centralized depository and a stock exchange that provides th e m atching and confirm ation of transactions. Once m atched and confirm ed, the trade inform ation is sent to the depository for settlem ent .24 23U.S. Congress, Office of Technology Assessment (1990), p. 58. 24lbid. JANUARY/FEBRUARY 1992 24 The th ird model, as seen in the U.S. equities m arket, contains all th ree institutions: a stock m arket, a central depository and an indepen dent clearinghouse. For example, the National Securities Clearing Corporation (NSCC), w hich processes 95 percent of all equities trades in the United States, is jointly ow ned by the New York Stock Exchange, the Am erican Stock Exchange, and NASD .25 The m ajority of the securities for NSCC m em bers, in turn, are held by the Deposi tory T rust Company. The stock m arket and clear inghouse together m atch and confirm transac tions. The clearinghouse also places itself be tw een counterparties to trades, then passes trade inform ation to the depository .26 R isks in Clearing and Settlem ent Credit (or counterparty) risk occurs w hen one side of the transaction does not settle in full, either w hen due or on a later date. The exis tence of counterparty risk, w hich is of minimal significance in m any U.S. m arkets because of a clearinghouse, can be critical in an international transaction. The clearinghouse, generally well capitalized, guarantees th at all trades will be honnored. In m any international transactions, how ever, no clearinghouse exists. Thus, a trad er lacks inform ation about the counterparty’s relia bility. Varying regulations on foreign trading m ay m ake it even m ore difficult to ascertain the safeguards available to a trad er in that m arket. Closely related to credit risk is liquidity risk, w hich is the risk that trades will not be settled at the appointed time, bu t at some undeterm ined tim e th ereafter .27 At settlem ent, counterparties are exposed to both credit and liquidity risks. Liquidity risk occurs because settlem ent may not occur on the specified date; credit risk occurs because the other party may not deliver at all. Thus, at settlem ent, th e parties m ay not know w h eth er the problem will be one of liquidity or credit. The settlem ent of international trades 25U.S. Congress, Office of Technology Assessment (1990), p. 81. 26For related readings on clearance and settlement systems, see the monographs prepared by the Payment System Studies Staff of the Federal Reserve Bank of New York in the references to this paper. 27A temporary inability to convert assets into cash is often associated with liquidity risk while bankruptcy of a counter party is associated with credit risk. For a more detailed description, see Federal Reserve Bank of New York (February 1989). 28For further reading on market risks, see Baer and Evanoff (1990). FEDERAL RESERVE BANK OF ST. LOUIS can exacerbate the problem of sim ultaneously exchanging securities for paym ent because of tim e zone differences .28 A nother risk, replacem ent cost risk, occurs w hen the price of th e security changes betw een trade and settlem ent. W hen one party has de faulted and the price of the instrum ent changes, then one of the parties involved would be ad versely affected by th e price change and suffer a loss in replacing the transaction. In foreign m arkets, the potential for adverse changes in the exchange rate can exacerbate this risk. O perational risk occurs because of the possible failure of com puter systems, telecom m unications or institutionalized procedures during trading. Given the heavy reliance on technology in ac cessing financial m arkets abroad, this issue is ex trem ely im portant in determ ining the success or failure of new trading systems. The precautions taken by th e CME for its Globex system —an im portant p art of its initial proposal to the Com modity Futures Trading Commission, a govern m ent regulator of futures exchanges—are a good example of this .29 Yet another risk, especially w orrisom e to regu lators, is systemic risk. Systemic risk occurs w hen credit risks stem m ing from operational or financial problem s result in agents exiting the m arket, which, in tu rn , threatens the industry. The inability of one financial institution to m ake its paym ents can cause other participants to be unable to m eet their financial obligations in a timely m anner. In the banking sector, this is typified as a ru n from deposits to currency. In futures and options, it occurs w hen agents no longer trade through standard channels like an exchange. For example, if m em bers of an ex change begin trading elsew here, the financial stability of the exchange is threatened as m em bers w ithdraw their financial collateral .30 29Examples of CME precautions include measures to pre vent unauthorized individuals from accessing the system, such as four different identification codes; termination of a computer operator’s session if nonstandard instructions are entered; and, in the failure of the central computer, recovery would involve automatic switchover to a back-up mainframe, taking approximately 60-90 seconds. See CFTC (1989), pp. 125-32. 30For further reading on systemic risks, see OECD (1991). 25 GLOBAL COORDINATION The O ctober 1987 stock m arket crash, w ith w orldw ide repercussions, revealed w eaknesses in the clearing and settlem ent system. Many feared th at the default of a m ajor m arket player could threaten the financial systems of m any countries. This prom pted w orld financial lead ers to w ork tow ard global coordination. The clearing and settlem ent of trades w as consi dered one of the m ost crucial aspects of this coordination. In 1989, the Group of Thirty issued a report, Clearing and Settlement System s in the W orld’s Securities M arkets.31 Based on its examination, five critical deficiencies in the clearing and set tlem ent systems across countries w ere iden tified: [1] Absence of com patible trade confirm ation and m atching systems for both dom estic and international trades; [2] Varying settlem ent periods across the differ ent m arkets; [3] Absence of delivery versus paym ent in some m arkets; [4] Absence of standardized trade guarantees; [5] Absence of book entry processing for settle m ent of securities transactions in several m arkets. Trade C onfirm ation and M atching T rade confirm ation and m atching, also know n as trade com parison, is the process of confirm ing and m atching the term s of a trade to ensure accuracy (for example, the issue, price, quantity and counterparties) and is usually done by a clearinghouse (although som etim es by an ex change or by the parties them selves, in the for w ard foreign exchange m arket). If not confirm ed and m atched, a chain reaction of failed trades is possible as subsequent trades are m ade on the assum ption that earlier trades will be success fully completed. Rapid trade com parison shortens the am ount of time betw een w hen the trade is m ade and w hen it is successfully m atched. This reduces credit risk by reducing the am ount of tim e an 31The Group of Thirty is a private sector organization that takes its membership from financial sectors such as ex changes, banks and investment houses. agent has to opt for defaulting on a trade. In the international context, delays of hours in a dom estic m arket may result in a delay of days for international trades. Requiring all investors to obtain m em bership in a trade com parison system and achieving a com patible system across international m arkets can reduce the delays and credit risks involved in diverse systems. Settlem ent P eriods The second deficiency is unequal settlem ent periods, w hich can increase settlem ent risk and potential default. Settlem ent risk occurs w hen there are gaps in the timing of paym ents and receipts on settlem ent date .32 The harm of dif ferent settlem ent periods is that, as m entioned earlier, traders or investors w ho are active p ar ticipants in the m arket m ake later trades contin gent on th e assum ption of the successful settle m ent of earlier trades. Hence, the harm is tw o fold—the default of an earlier counterparty and the dependence on this trade that could jeopardize subsequent trades. As w ith m any trade issues th at require timeliness, delays in settlem ent can be exacerbated if spread across different trading hours and tim e zones. While this is costly in a dom estic m arket, the investm ent of a U.S. agent dealing in interna tional m arkets can be even m ore costly because it is also subject to the econom ic conditions of foreign countries and exchange rates. Adverse changes in the exchange rate can tu rn a m inor loss into a significant one in the presence of currency risk. Thus, for agents moving betw een international m arkets, an uncertain settlem ent period com bined w ith an uncertain exchange rate can increase financial losses. The grow ing volume of trades has led to a num ber of techniques w here, to reduce the num ber of settlem ent transactions, trades are not processed one at a time. “Netting” is a sys tem w hereby transactions are aggregated, so th at debit and credit positions offset each other, leaving a participant w ith one final position in the m arket of owing or being owed. Netting greatly increases the liquidity of the m arket and the trad er’s flexibility because, rath er than post ing collateral for every trade, the trad er is responsible only for th e net settlem ent debit .33 33Board of Governors of the Federal Reserve System and the Federal Reserve Bank of New York (1990), p. 40. 32Settlement risk encompasses both liquidity risk and credit risk. JANUARY/FEBRUARY 1992 26 T here are th ree m ain choices for a netting system. The first is bilateral netting, w hereby all trades in the same security and betw een the same parties to the trade are netted to one final delivery versus paym ent (DVP).34 For example, if Ralph sells 100 shares of British M ohair to Sam, th en buys back 75 shares of British M ohair from Sam, the net position is that Ralph m ust deliver 25 shares to Sam. This is the narrow est of the th ree netting options. The second is m ultilateral netting (or daily netting), which, unlike bilateral netting, allows for different counterparties in the netting schem e. In this instance, all trades in the same security are netted to a final debit or credit po sition for each participant. The last option is continuous net settlem ent, w hereby all trades in a particular security are pooled by issue to a final debit or credit posi tion for the day and any unsettled trades are carried over and offset against the next day’s trades. In practice, the clearing corporation sub stitutes as the counterparty to the trade in con tinuous net settlem ent. The type of netting system im plem ented de pends on the volume of the m arket. Establishing a m ultilateral or continuous net settlem ent sys tem is a costly procedure, requiring a risksharing arrangem ent am ong m em bers, a clear ing corporation (as w ith continuous net settle ment) and pow erful com puter systems to handle the volume of trades. The costs of such a sys tem may exceed the costs of operating w ith only a bilateral system. This is especially tru e in low-volume m arkets w here bilateral netting can be a feasible and less costly alternative. A pro posal for a m ultilateral netting system in the high-volume foreign exchange m arket w as exa m ined in 1988 by m em bers of FXNET, a bilater al netting system .35 R epresentatives of leading international banks, responding to FXNET’s ques tionnaire, felt that a m ajor benefit would be to reduce processing costs .36 This is especially relevant in m arkets expand ing their foreign m em bership. If netting is desir able because it reduces the num ber of trades to s^DVP is a payment system whereby the debits and credits of a trade are applied to the parties’ accounts simul taneously. 35For further reading on the netting of foreign exchange transactions, see Gilbert (1992). 36Minutes of FXNET Multilateral Netting Steering Committee, (1989). FEDERAL RESERVE BANK OF ST. LOUIS process, it becom es even m ore so as m arkets service no longer just domestic, bu t a grow ing num ber of foreign clients. In the FXNET ques tionnaire, respondents stated that, "Cross-border aspects of m ultilateral netting should be consi dered early in the process, as they will be m ore im portant th an w ith bilateral netting .”37 W ith the addition of cross-border trad ers increasing the transactions volume a m arket handles, a netting system would simplify the re peated paym ents th at w ould be introduced. W hichever netting system is chosen, the desir able settlem ent tim e fram e is a rolling settle m ent system. In such a system, trades settle on all business days of the week, scheduled the same num ber of days after the trad e .38 Thus, the presence of a standardized settlem ent period and a netting system is a crucial aspect to mov ing betw een international m arkets w ith security of settlem ent. D elivery Vs. P aym ent The th ird finding by the Group of Thirty is the absence of delivery versus paym ent (DVP) in some m arkets. DVP is a tw o-sided paym ent sys tem th at sim ultaneously debits or credits the cash account of one m em ber and m akes the corresponding entry on the securities side of the transaction. This reduces th e settlem ent risk that occurs w hen th ere is a discrepancy be tw een the timing of paym ents and receipts on settlem ent date. The Group of Thirty, arguing the need for prom pt two-sided paym ents, has recom m ended interim procedures: risk can be reduced by de livering securities only against a certified check or by employing a m echanism w hereby delivery and paym ent are done sim ultaneously although through different systems. In either case, net settlem ent of cash and securities is com pleted by the end of the day. Even w ithout a form alized DVP, m ethods can be developed to minimize settlem ent risk by having both parties to a trade settle their ac counts sim ultaneously. W ith m arkets in different tim e zones and, thus, different operating hours, allowing each side of a trade to settle at a differ37lbid. 38The Group of Thirty recommends the implementation of a rolling settling system by 1992 so that final settlement oc curs three days after the trade. 27 ent time could result in a next-day paym ent, not one w ithin the hours of the first settlem ent. S tan dardized Trade Guarantees The fourth deficiency is the absence of stan dardized trade guarantees. A trade guarantee ensures th at all com pared or netted trades will be settled, based on the conditions on w hich they w ere com pared, even in the event of coun terparty default. To assure trade guarantees, each m em ber of the com parison and netting systems assum es the default risk of the system. A standard m ethod of providing a guarantee is to establish a general clearing fund based on m em ber contributions. W hen a default occurs, the losses are first extracted from the defaulting party’s clearing fund contribution. If that contri bution does not m eet the full am ount of the loss, the rem ainder is charged against the clearing corporation’s general clearing fund. The international environm ent adds an extra facet to these guarantees. Since m em bership is becom ing increasingly international, a m ajor financial loss can strain the capacity of the cor poration to handle the failure immediately. Ob taining perm ission for access to additional funding, for example, could cause unnecessary delays. Thus, the m aintenance of additional sources of funds, like m em ber deposits or ac cess to bank lines, becom es crucial in an inter national setting. To ensure the integrity of the corporation and, thus, the m arket, trade guaran tees provide a m easure of security and stability in the face of potential failures. B ook E ntry P rocessing The last issue to be addressed in global coor dination is the absence of book entry processing for settling securities transactions in several m arkets. Before addressing book entry, however, other institutions surrounding this process should be introduced. The first of these is a central securities depository (CSD).39 The prim ary activity of a CSD is to immobilize and dem aterialize securi ties so th at they can be processed in the m ore efficient book entry m ethod. Immobilization of 39The strict definition of a CSD requires that a country should have only one depository. In practice, however, more than one may exist. This type of system can be ef fective as long as there is linkage between the entities to coordinate trade information. The United States has sever al depositories. securities m eans that the physical docum ents (for example, share certificates) are stored at the depository, eliminating th eir actual m ove m ent w hen ow nership changes. Dem aterializa tion m eans th at no physical securities w ith title of ow nership are issued. Securities exist solely as com puter records. T ransfers of certificates are done by book en try, w here a simple credit and a balancing debit to custom ers’ com puterized accounts on the books of the CSD will transfer securities from one account to another. Immobilization and de m aterialization replace the m ore risky and timeconsum ing process of transferring the securities in paper form w henever a transaction is made. T ransfers of stocks trade-by-trade introduce a needless com plication to the clearing and settle m ent system, w hich becom es even m ore compli cated if it involves delivering them to investors w orldw ide. B ecom m endations b y the Group o f Thirty The Group of T hirty has proposed nine recom m endations found in table 2 to correct the preceding five deficiencies. The status of the Group of Thirty recom m endations are listed in table 3. This table depicts the extent to w hich 21 countries have m ade progress on these recom m endations. While th e United States has accom plished m ore th an m ost of the countries surveyed, the fact th at so m any countries have not finalized these policies, and m ay not by 1992, has implications for the eventual tim etable of global coordination .40 Currently, th ere is no well-defined regulatory structure for the global m arketplace. While regulatory authorities exist in specific countries— for example, the Securities and Exchange Com mission has the regulatory authority, oversight and arbitration of securities disputes in the U.S. stock m arket—the international arena has no similar agency to govern global financial rela tions. In its absence, voluntary coordination of clearing and settlem ent system s can help reduce the risks th at lead to defaults, failures and po tential disputes betw een legal and regulatory authorities. Thus, th ere are potential gains if 40ln addition to the Group of Thirty proposal, other groups, such as the Working Group on Financial Markets have studied clearing and settlement issues. JANUARY/FEBRUARY 1992 28 Table 2 Group of Thirty Recommendations________________________ The following are nine recommendations put forth by the Group of Thirty to correct the deficiencies it finds in the coordination of clearing and settlement systems.'The numbers in brackets are the rele vant deficiencies presented in the section, Global Coordination, that these recommendations address. • By 1990, all comparisons of trades between direct market participants (i.e., brokers, brok er/dealers and other exchange members) should be accomplished by T+1, (the first day after the trade). [1] • Indirect market participants (such as institutional investors or any trading counterparties that are not broker/dealers) should, by 1992, be members of a trade comparison system that achieves positive affirmation of trade details. [1] • Each country should have an effective and fully developed central securities depository, or ganized and managed to encourage the broadest possible industry participation (directly and indirectly) by 1992. [3, 5] • Each country should study its market volume and participation to determine whether a trade netting system would be beneficial in terms of reducing risk and promoting efficiency. If a netting system would be appropriate, it should be implemented by 1992. [2] • Delivery versus payment (DVP) should be employed as the method for settling all securities transactions. A DVP system should be in place by 1992. [3, 5] • Payments associated with the settlement of securities transactions and the servicing of securities portfolios should be made consistent across all instruments and markets by adopt ing the “ same day” funds convention. [2] • A “ Rolling Settlement” system should be adopted by all markets. Final settlement should oc cur on T+3 by 1992. As an interim target, final settlement should occur on T+5 by 1990 at the latest, save only where it hinders the achievement of T+3 by 1992. [2] • Securities lending and borrowing should be encouraged as a method of expediting the set tlement of securities transactions. Existing regulatory and taxation barriers that inhibit the practice of lending securities should be removed by 1990.2 [4] • Each country should adopt the standard for securities messages developed by the Interna tional Organization for Standardization [ISO Standard 7775]. In particular, countries should adopt the ISIN numbering system for securities issues as defined in the ISO Standard 6166, at least for cross-border transactions. These standards should be universally applied by 1992. ’ Group of Thirty (March 1989). 2For information on securities lending, see Paul C. Lipson, Bradley K. Sabel, and Frank Keane, “ Secu rities Lending” (Federal Reserve Bank of New York, August 1989). the clearing and settlem ent systems operating in dom estic m arkets are coordinated am ong global financial m arkets. CONCLUDING REMARKS This discussion has attem pted to present an overview of issues th at are currently of interest in the globalization of financial m arkets. The linkage am ong international m arkets is of in terest both to private investors and to national governm ents, w ho desire stable dom estic and international financial sectors. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis International com petition am ong financial m arkets is grow ing rapidly and has produced benefits such as new financial instrum ents, new m arkets and extended trading hours. These changes, how ever, are not w ithout costs. Domes tic rules and regulation are not sufficient safeguards for a system th at operates in an in creasingly international environm ent. Financial and governm ental com m unities are addressing the need to integrate international expansion to facilitate the continued safe and profitable grow th of financial instrum ents and the im por tan t functions these m arkets serve. It is clear that m uch w ork rem ains. 29 Table 3 Current Status of the Group of Thirty’s Recommendations for International Settlement—Equities 1 2 3 4 5 6 Institutional Recommendation No. Central Delivery Rolling Comparison Comparison Securities Country Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Italy Japan Korea Netherlands Norway Singapore Spain Sweden Switzerland Thailand United Kingdom United States 7 8 9 Securities versus Settlement Same-Day on T+11 System Depository Netting Payment on T+5* Funds ISO/ISM Securities Lending Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes No No No Yes No No No No No No No No No No No No No No No Yes Yes No Yes Yes Yes No No Yes Yes No Yes No Yes Yes No No No Yes Yes Yes No Yes No No No Yes No No No No No No Yes No Yes No No No No No Yes Yes Yes Yes No Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes No Yes Open Weekly Fortnightly T+5 T+3 T+5 Monthly T+2 T+1 Monthly T+3 T+2 T+5 T+6 T+5 Weekly T+5 T+3 T+4 Fortnightly T+5 No Yes Yes Yes Yes No Yes Yes No Yes No No No Yes No No No Yes Yes No No No No No No No No No No No No No No No No No No No No No No No Yes No Yes Yes Yes No Yes No Limited Limited Yes No Yes No Yes Limited Yes Yes No Limited Yes SOURCE: Updated by the Office of Technology Assessment, July 1990, from A Comparative View: The Group of Thirty’s Recommendations and the Current U.S. National Clearance and Settlement System (Morgan Stanley & Co., June 1989). 1T+1 means the first day after the trade. 2T+5 means five days after the trade. REFERENCES Abken, Peter A. “ The Globalization of Stock, Futures, and Options Markets.” Federal Reserve Bank of Atlanta Eco nomic Review (July/August, 1991), pp. 1-22. Baer, Herbert L., and Douglas D. Evanoff. “ Payments Sys tem Risk Issues in a Global Economy,” Federal Reserve Bank of Chicago Working Paper 90-12 (August 1990). Baer, Herbert L., Douglas D. Evanoff, and Christine A.Pavel. “ Payments System Issues in a 24-Hour Global Economy,” Research in Financial Services: Public and Private Policy (vol. 3) (JAI Press, Inc., 1991). “ Big Board After-Hours Trading May Lead to a Two-Tier Market,” Wall Street Journal, June 13, 1991. Board of Governors of the Federal Reserve System and the Federal Reserve Bank of New York. Clearance and Settle ment in U.S. Securities Markets, Prepared for the Commit tee on Payment and Settlement Systems, Bank for International Settlements. (Basle, Switzerland: December 5-6, 1990). Chesler-Marsh, Caren. “ Globex Countdown,” Euromoney (March 1991), pp. 33-35. Chicago Mercantile Exchange. Globex Report: An Update for the Global Trader (November 14, 1990). Chicago Mercantile Exchange, Business Development Group. The Challenges of the 24-Hour Financial Marketplace (May 1989). Commodity Futures Trading Commission, Division of Trading and Markets. Chicago Mercantile Exchange’s Proposed Glo bex Trading System (February 2, 1989). Diamond, Barbara B., and Mark P. Kollar. 24-Hour Trading: The Global Network of Futures and Options Markets (John Wiley & Sons, 1989). Domowitz, Ian. “ The Mechanics of Automated Trade Execu tion Systems,” Journal of Financial Intermediation (June 1990), pp. 167-94. Federal Reserve Bank of New York. "An Overview of the Operations of the Options Clearing Corporation” (April 1989). JANUARY/FEBRUARY 1992 30 ________“ Clearing and Settlement Through the Board of Trade Clearing Corporation” (February 1990). Hansell, Saul. “ The Computer that ate Chicago,” Institutional Investor (February 1989), pp. 181-88. ________“ Clearing and Settling the Euro-Securities Market: Euro-Clear and Cedel” (March 1989). Khoury, Sarkis J. The Deregulation of the World Financial Markets: Myths, Realities, and Impact (Greenwood Press, Inc., 1990). _______ .“ Exchanges and Clearinghouses for Financial Fu tures and Options in the United Kingdom” (March 1989). Nasdaq Stock Market, Business Development. Nasdaq: The Stock Market for the Next 100 Years (Washington, D.C., 1991). _______ .“ The Clearing House Interbank Payments System" (January 1991). ________Bank for International Settlements. Report on Net ting Schemes (February 1989). FXNET Multilateral Netting Steering Committee. Documenta tion (February 1, 1989). Gilbert, R. Alton. “ Implications of Netting Arrangements for Bank Risk in Foreign Exchange Transactions,” this Review (January/February 1992), pp. 3-16. National Association of Securities Dealers (NASD). Nasdaq Fact Book & Company Directory (Washington, D.C., 1991). Organization for Economic Cooperation and Development (OECD). Systemic Risks in Securities Markets (Paris: OECD Publications Service, 1991). Soffex. Soffex Swiss Options and Financial Futures Exchange Ag, reprinted from a Supplement to Euromoney (June 1990). Group of Experts on Payment Systems, Bank for Internation al Settlements (BIS). Report on Netting Schemes (Basle, Switzerland, February 1989). U.S. Congress, Office of Technology Assessment. Trading Around the Clock: Global Securities Markets and Informa tion Technology-Background Paper, OTA-BP-CIT-66 (Govern ment Printing Office, July 1990). Group of Thirty. Clearance and Settlement Systems in the World’s Securities Markets (New York and London, March 1989). Working Group on Financial Markets. Clearance and Settle ment Reform: The Stock, Options, and Futures Markets are Still at Risk (Government Printing Office, April 1990). FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis 31 Piyu Yue Piyu Vue, a research associate at the 1C2 Institute, University of Texas at Austin, was a visiting scholar at the Federal Reserve Bank of St. Louis when this article was written. Lynn Dietrich provided research assistance. The author would like to thank A. Charnes, Rolf Fare and Shawna Grosskopf for their constructive comments and useful suggestions. Their DEA computer code led to a significant improvement of the paper. Data Envelopment Analysis and Commercial Bank Performance: A Primer With Applications to Missouri Banks C -JO M M E R C IA L BANKS PLAY a vital role in the econom y for tw o reasons: they provide a m ajor source of financial interm ediation and their check able deposit liabilities represent the bulk of the nation’s m oney stock. Evaluating th eir overall perform ance and m onitoring th eir financial condi tion is im portant to depositors, ow ners, potential investors, m anagers and, of course, regulators. Currently, financial ratios are often used to m easure the overall financial soundness of a bank and the quality of its m anagem ent. Bank regu lators, for example, use financial ratios to help evaluate a bank’s perform ance as part of the CAMEL system .1 Evaluating the economic p erfo r m ance of banks, how ever, is a com plicated process. Often a num ber of criteria such as 'For more details, see Booker (1983), Korobow (1983) and Putnam (1983). 2The name DEA is attributed to Charnes, Cooper and Rhodes (1978), for the development of DEA, see Charnes, et al.(1985) and Charnes, et al. (1978); for some applications of DEA, see Banker, et al. (1984), Charnes, et al. (1990) and Sherman and Gold (1985). profits, liquidity, asset quality, attitude tow ard risk, and m anagem ent strategies m ust be consi dered. The changing nature of the banking industry has m ade such evaluations even m ore difficult, increasing the need for m ore flexible alternative form s of financial analysis. This paper describes a particular methodology called Data Envelopm ent Analysis (DEA), that has been used previously to analyze the relative effi ciencies of industrial firms, universities, hospitals, m ilitary operations, baseball players and, m ore recently, com m ercial banks .2 The use of DEA is dem onstrated by evaluating the m anagem ent of 60 M issouri com m ercial banks for the period from 1984 to 1990.3 Ehlen (1983), Korobow (1983), Putnam (1983), Wall (1983) and Watro (1989)), actual banking efficiency has received limited attention. Recently, a few publications have used DEA or a similar approach to study the technical and scale efficien cies of commercial banks (e.g., Sherman and Gold (1985), Charnes et al. (1990), Rangan et al. (1988), Aly et al. (1990), and Elyasiani and Mehdian (1990)). 3Although there is vast literature analyzing competition and performance in the U.S. banking industry (e.g., Gilbert (1984), JANUARY/FEBRUARY 1992 32 DATA ENVELOPMENT ANALYSIS: SOME BASICS DEA represents a m athem atical program m ing m ethodology that can be applied to assess the effi ciency of a variety of institutions using a variety of data. This section provides an intuitive explana tion of the DEA approach. A form al m athem atical presentation of DEA is described in appendix A; a slightly different nonparam etric approach is described in appendix B. The DEA Standard f o r Efficiency DEA is based on a concept of efficiency th at is widely used in engineering and the natural sciences. Engineering efficiency is defined as the ratio of the am ount of w ork perform ed by a m achine to the am ount of energy consum ed in the process. Since m achines m ust be operated according to the law of conservation of energy, their efficiency ratios are always less than or equal to unity. This concept of engineering efficiency is not im m ediately applicable to economic production because the value of output is expected to exceed the value of inputs due to the "value added” in production. Nevertheless, under certain circum stances, an economic efficiency standard—similar to the engineering standard—can be defined and used to com pare the relative efficiencies of econom ic entities. For example, a firm can be said to be efficient relative to another if it produces either the sa m e level of output w ith few er inputs or m ore output w ith th e same or few er inputs. A single firm is considered "technically efficient” if it cannot increase any output or reduce any input w ithout reducing other outputs or increasing other inputs .4 Consequently, this concept of tech nical efficiency is similar to the engineering concept. The som ew hat b roader concept of "economic efficiency,” on the other hand, is achieved w hen firm s find the com bination of inputs th at enable them to produce the desired level of output at m inim um cost .5 4See Koopmans (1951). 5This is also named “ allocative efficiency” because a profit maximizing firm must allocate its resources such that the technical rate of substitution is equal to the ratio of the prices of the resources. Theoretical considerations of allocative effi ciency can be found in the articles by Banker (1984) and Banker and Maindiratta (1988). 6lt is common to estimate production functions using regres sion analysis. When cross-section data are used, the esti- FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis DEA and Technical Efficiency The discussion of the DEA approach will be undertaken in the context of technical efficiency in the m icroeconom ic theory of production. In microeconomics, the production possibility set consists of the feasible input and output com bina tions th at arise from available production tech nology. The production function (or production transform ation as it is called in the case of m ultiple outputs) is a m athem atical expression for a process th at transform s inputs into output. In so doing, it defines the frontier of the production possibility set. For example, consider the wellknow n Cobb-Douglas production function: (1) Y = AKaLll_al, w here Y is the m axim um output for given quanti ties of tw o inputs: capital (K) and labor (L). Even if all firm s produce the same good (Y) w ith the same technology defined by equation 1 , they may still use different com binations of labor and capital to produce different levels of output. Nonetheless, all firm s w hose input-output com binations lie on the surface (frontier) of the production relationship defined by equation 1 are said to be technologi cally efficient. Similarly, firm s w ith input-output com binations located inside the frontier are tech nologically inefficient. DEA provides a similar notion of efficiency. The principal difference is th at the DEA production frontier is not determ ined by some specific equa tion like th at show n in equation 1 ; instead, it is generated from the actual data for the evaluated firm s (which in DEA term inology are typically called decision-making units or DMUs ).6 Conse quently, the DEA efficiency score for a specific firm is not defined by an absolute standard like equation 1. Rather, it is defined relative to the other firm s under consideration. And, similar to engi neering efficiency m easures, DEA establishes a "benchm ark” efficiency score of unity th at no individual firm ’s score can exceed. Consequently, efficient firm s receive efficiency scores of unity, while inefficient firm s receive DEA scores of less than unity. mated production function represents the average behavior of firms in the sample. Hence, the estimated production function depends upon the data for both efficient and inefficient firms. By imposing suitable constraints, these statistical procedures can be modified to orient the estimates toward frontiers. In this manner, the frontier of the production set can be esti mated econometrically. 33 In m icroeconom ic analysis, efficient production is defined by technological relationships w ith the assum ption th at firm s are operated efficiently. W hether or not firm s have access to the same technology, it is assum ed th at they operate on the frontier of their relevant production possibilities set; hence, they are technically efficient by defini tion. As a result, m uch of m icroeconom ic theory ignores issues concerning technological ineffi ciencies. DEA assum es that all firm s face the same unspecified technology w hich defines their production possibilities set. The objective of DEA is to determ ine w hich firm s operate on their effi ciency frontier and w hich firm s do not. T hat is, DEA partitions the inputs and outputs of all firm s into efficient and inefficient com binations. The efficient input-output com binations yield an implicit production frontier against w hich each firm ’s input and output com bination is evaluated. If the firm ’s input-output com bination lies on the DEA frontier, the firm m ight be considered effi cient; if the firm ’s input-output com bination lies inside the DEA frontier, the firm is considered inefficient. An advantage of DEA is that it uses actual sam ple data to derive the efficiency frontier against w hich each firm in the sample can be evaluated .7 As a result, no explicit functional form for the production function has to be specified in advance. Instead, the production frontier is gener ated by a m athem atical program m ing algorithm w hich also calculates the optimal DEA efficiency score for each firm. To illustrate the relationship betw een DEA and economic production in its simplest form, consider the example shown in figure i , in w hich firm s use a single input to produce a single output. In this example, there are six firm s w hose inputs are denoted as Xj and w hose outputs are denoted 7DEA has two theoretical properties that are especially use ful for its implementation. One is that the DEA model is mathematically related to a multi-objective optimization problem in which all inputs and outputs are defined as multiple objectives such that all inputs are minimized and all outputs are maximized simultaneously under the tech nology constraints. Thus, DEA-efficient DMUs represent Pareto optimal solutions to the multi-objective optimization problem, while the Pareto optimal solution does not neces sarily imply DEA efficiency. Another important property is that DEA efficiency scores are independent of the units in which inputs and outputs are measured, as long as these units are the same for all DMUs. These characteristics make the DEA methodology highly flexible. The only constraint set originally in the CCR model is that the values of inputs and outputs must be strictly positive. a s y (i = 1 , 2 ,..., 6); their input-output com binations are labeled by Fs(s =1,2,..., 6). While the produc tion frontier is generated by th e input-output com binations for the firm s labeled F1( F3, F5 and F6, the efficient portion of the production frontier is shown by the connected line segments. F2 and F 4 are clearly DEA inefficient because they lie inside the frontier; F6 is DEA inefficient because the same output can be produced w ith less input. The Im portance o f Facets in DEA “Facets” are an im portant concept used to evaluate a firm ’s efficiency in DEA. The efficiency m easure in DEA is concerned w ith w hether a firm can increase its output using the same inputs or produce the same output w ith few er inputs. Consequently, only p art of the entire efficiency frontier is relevant w hen evaluating the efficiency of a specific firm. The relevant portion of the effi ciency frontier is called a facet. For example, in figure 1, only the facet from F, to F3 is relevant for evaluating the efficiency of the firm designated by F2. Similarly, only the facet from F3 to F5 is used to evaluate the firm denoted by F 4.8 The use of facets w ith DEA enables analysts to identify inefficient firm s and, through com parison w ith efficient firm s on relevant facets, to suggest ways in w hich the inefficient firm s m ight im prove their perform ance. As illustrated in figure 1, F2 can becom e efficient by rising to some point on the Fi-F3 facet. In particular, it could move to A by simply using less input, to B by producing m ore output or to C by both reducing input and increasing output. Of course, in this example, the analysis is obvious and the recom m endation trivial. In m ore com plicated, m ultiple inputm ultiple output cases, how ever, the appropriate efficiency recom m endations w ould be m uch m ore difficult to discover w ithout the DEA m ethodology .9 This constraint, however, has been abandoned in the new additive DEA formulation. As a consequence, the additive DEA model is used to compute reservation prices for new and disappearing commodities in the construction of price indexes by Lovell and Zieschang (1990). 8ln a multiple dimensional space, the efficiency frontier forms a polyhedron. In geometry, a portion of the surface of a polyhedron is called a facet; this is why the same term is used in DEA. These facets have important implica tions in empirical studies, such as identification of compe titors and strategic groups in an industry. See Day, Lewin, Salazar and Li (1989). 9For alternative measures of efficiency, see appendix B. JANUARY/FEBRUARY 1992 34 F igu re 1 Production Frontier and Efficiency Subset Output Y tionate decrease in th eir input and o utput places them inside the production frontier. A propor tionate increase in th eir input and output is im pos sible because it w ould move them outside of the frontier. Constant retu rn s to scale occur if all p rop or tionate increases or decreases in inputs and outputs move the firm either along or above the production frontier. In figure 2, for example, F3 exhibits constant retu rn s to scale because propor tionate increases or decreases w ould place it outside the production frontier. Since the facets are generated by efficient firms, the scale efficiency of these firm s is determ ined by the properties of th eir particular facet. Scale effi ciencies for inefficient firm s are determ ined by their respective reference facets as well. Thus, F2 and F4 in figure 1 exhibit increasing and decreasing retu rn s to scale, respectively. DEA and Econom ic E fficiency Scale E fficiency In addition to m easuring technological effi ciency, DEA also provides inform ation about scale efficiencies in production. Because the m easure of scale efficiency in DEA analysis varies from model to model, care m ust be exercised. The scale effi ciency m easured for the DEA m odel used in this study, how ever, corresponds fairly closely to the m icroeconom ic definition of econom ics of scale in th e classical theory of production .10 To illustrate, consider the F!-F3 facet in figure 2. Firms located on this facet exhibit increasing retu rn s to scale because a proportionate rise in their input and output places them inside the production frontier. A proportionate decrease in th eir input and output is impossible because it w ould move them outside of the frontier. This is illustrated by a ray from the origin th at passes through the F,-F3 facet at F’2. Firms located on the F3-Fs facet exhibit decreasing retu rn s to scale because a prop or 10See Fare, Grosskopf and Lovell (1985). Different DEA models employ different measures of scale efficiency. See appendixes A and B for details. FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis While the discussion of DEA in the context of technological efficiency of production is useful for illustrative purposes, it is far too narrow and limiting. DEA is frequently applied to questions and data th at transcend the n arro w focus of tech nical efficiency in production. For example, DEA is frequently applied to financial data w hen addressing questions of economic efficiency. In this regard, its application is som ew hat m ore problem atic. For example, w hen firm s face different m arginal costs of production due to regional or local wage differentials, one firm may a p p e a r inefficient r e la tiv e to a n o t h e r . G iv en th e potential differences in relative costs th at a firm may face, how ever, it m ight be equally efficient. Alternatively, differences th at appear to be due to econom ic inefficiencies m ay in fact be due to cost differences directly attributable to the non hom ogeneity of products. Because of problem s like these, DEA m ust be applied judiciously. DEA W indow Analysis To this point, the discussion of DEA has been concerned w ith evaluating the relative efficiency of different firm s at the same time. Those w ho use DEA, how ever, frequently em ploy a type of sensi tivity analysis called “w indow analysis.” The perform ance of one firm or its reference firm s 35 F igure 2 An Illustration of Scale Efficiencies Output Y for a firm over time. Of course, com parisons of DEA efficiency scores over extended periods may be misleading (or w orse) because of significant changes in technology and the underlying economic structure. APPLYING DEA TO BANKING: AN EVALUATION OF 6 0 MISSOURI COMMERCIAL BANKS To dem onstrate DEA’s use, it is applied to evaluate relative efficiency in banking. Financial data for 60 of the largest M issouri com m ercial banks for 1984 (determ ined by th eir total assets in 1990) are used. Initially, the relative efficiency of these banks is exam ined using tw o alternative DEA models: the CCR m odel and the additive DEA model. A discussion of these alternative DEA models appears in appendix A. In extending the discussion and analysis, how ever, we focus solely on the CCR model. M easuring Inputs and Outputs may be particularly “good” or "bad” at a given tim e because of factors th at are external to the firm ’s relative efficiency. In addition, the num ber of firm s th at can be analyzed using the DEA m odel is virtually unlimited. T herefore, data on firm s in different periods can be incorporated into the analysis by simply treating them as if they represen t different firms. In this way, a given firm at a given tim e can com pare its perform ance at different tim es and w ith the perform ance of other firm s at the same and at different times. Through a sequence of such "w indows,” the sensitivity of a firm ’s efficiency score can be derived for a partic ular year according to changing conditions and a changing set of reference firm s . 11 A firm th at is DEA efficient in a given year, regardless of the window, is likely to be truly efficient relative to other firms. Conversely, a firm th at is only DEA efficient in a particular w indow may be efficient solely because of extraneous circum stances. In addition, w indow analysis provides some evidence of the short-run evolution of efficiency Perhaps the m ost im portant step in using DEA to exam ine the relative efficiency of any type of firm is the selection of appropriate inputs and outputs. This is partially tru e for banks because there is considerable disagreem ent over the appropriate inputs and outputs for banks. Previous applica tions of DEA to banks generally have adopted one of tw o approaches to justify their choice of inputs and outputs .12 The first ‘‘interm ediary approach” views banks as financial interm ediaries w hose prim ary busi ness is to borrow funds from depositors and lend those funds to others for profit. In these studies, the banks’ outputs are loans (m easured in dollars) and their inputs are the various costs of these funds (including interest expense, labor, capital and operating costs). A second approach views banks as institutions that use capital and labor to produce loans and deposit account services. In these studies, the banks’ outputs are th eir accounts and transac tions, while their inputs are th eir labor, capital and operating costs; th e banks’ interest expenses are excluded in these studies. ’ ’ This is called “ panel data analysis” in econometrics. 12Some studies have adopted the simple rule that if it produces revenue, it is an output; if it requires a net ex penditure, it is an input. For example, see Hancock (1989). JANUARY/FEBRUARY 1992 36 O ur analysis of 60 Missouri banks uses a variant of the interm ediary approach. The banks’ outputs are interest incom e (IC), non-interest incom e (NIC) and total loans (TL). Interest incom e includes interest and fee incom e on loans, incom e from lease-financing receivables, interest and dividend incom e on securities, and other income. Non interest incom e includes service charges on deposit accounts, incom e from fiduciary activities and other non-interest income. Total loans consist of loans and leases net of unearned income. These outputs represen t the banks’ revenues and major business activities. The banks' inputs are interest expenses (IE), non-interest expenses (NIE), transaction deposits (TD), and non-transaction deposits (NTD). Interest expenses include expenses for federal funds and the purchase and sale of securities, and the inter est on dem and notes and other borrow ed money. Non-interest expenses include salaries, expenses associated w ith prem ises and fixed assets, taxes and other expenses. Bank deposits are disaggre gated into transaction and non-transaction depos its because they have different turnover and cost structures. These inputs represen t m easures for the banks’ labor, capital and operating costs. De posits and funds purchased (m easured by their interest expense) are the source of loanable funds to be invested in assets .13 Evaluation o f M issouri Bank M anagem ent P erf orm ance in 1984 The DEA scores and retu rn s to scale m easures resulting from applying the CCR and additive DEA m odels are presented in table l .14 Although the overall results are similar across the tw o models, th ere are m inor differences in the individual effi ciency scores th at may provide inform ation about the relative efficiency of these banks. The tw o models differ fundam entally in their definition of the efficiency frontier. In particular, the CCR model assum es constant retu rn s to scale, while the additive m odel allows for the possibility of constant (C), increasing (I) or decreasing (D) 13This is controversial, however. Some researchers specify deposits as outputs, arguing that treating deposits as inputs makes banks that depend on purchased money look artifi cially efficient (see Berg et al., 1990). 14The results from solving the DEA model also include informa tion about DEA scale efficiencies, the efficient projection on the efficiency frontier, slack variables sr+ andS;- and the dual variables iu, and v:. The “ dual” variables represent “ shadow prices” for each input and output. That is, they represent the FEDERAL RESERVE BANK OF ST. LOUIS returns. Because of this, banks th at are efficient in the CCR m odel m ust also be efficient in the addi tive model. As table 1 illustrates for our Missouri banks, the converse, how ever, is not true. The overall efficiency score is com posed of "pure” technical and "scale” efficiencies. In the CCR model, a firm w hich is technologically effi cient also uses the m ost efficient scale of opera tion. In the additive model, how ever, the score represents only “p u re” technical efficiency. By com paring the results of the CCR and additive models, w e can see th at while five of our M issouri banks w ere technologically efficient, they w ere not operating at the m ost efficient scale of opera tion. The reader is cautioned, how ever, th at this analysis excludes a num ber of factors (such as dem ographic characteristics of the m arkets in w hich they operate) th at m ay be im portant in determ ining the m ost economically efficient scale of operation. Since the efficiency scores are defined differ ently in the CCR and the additive DEA models, it is not possible to generate a m easure of scale ineffi ciency using the results in table 1. Nevertheless, the fact th at the efficiency scores from the tw o models are quite similar suggests that the scale inefficiency is not a m ajor source of overall ineffi ciency for these banks. It appears th at the ineffi cient banks simply used too m any inputs or produced too few outputs rath er than chose the incorrect scale for production .15 A Further A nalysis o f the CCB M odel An illustration of the use of DEA analysis can be obtained by considering the data for the bank w ith the lowest efficiency score, bank 59. The results for this bank are sum m arized in table 2 . The reference banks m aking up the facet to w hich bank 59 is com pared and "lam bda,” a m easure of the relative im portance of each reference bank in the facet, are given. The table shows th at three reference banks com pose the facet for bank 59. Banks 51 and 39 play the m ajor role and the other bank is relatively unim portant. marginal effects of the input and output variables on the bank’s DEA efficiency score. See appendix A for details. 15Similar results of insignificant scale-inefficiency of U.S. banks have been reported by Aly et al. (1990). 37 Table 1 Overall Performance of 60 Missouri Commercial Banks Evaluated by the CCR and Additive DEA Models (1984)________ ________ Efficiency Ratio________ ________ Efficiency Ratio________ Bank no. CCR model Additive model Type of scale1 Bank no. CCR model Additive model Type of scale 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 .8545 .9228 .9033 .8588 1.0000 .8766 .8709 .8841 .8735 .8115 .9086 .7852 .8338 .9739 .8937 .8292 .8705 .9684 .8439 .9527 .9746 .8681 .9744 .9003 1.0000 .8714 1.0000 1.0000 .8753 .9003 .8825 1.0000 .9129 .9498 1.0000 .9042 .9144 .9323 .9857 .9116 .9856 .8388 .9927 .9024 .9829 .8492 .8211 .9783 1.0000 .9930 1.0000 .8888 .9642 .9646 1.0000 .8406 1.0000 1.0000 .9351 .9319 D I I I C I I I I I I I I I I I I I D I I I I I C I c c I D 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 .8568 .9305 .8509 .8392 .8596 1.0000 .8712 .8707 1.0000 1.0000 .8500 .8867 .8220 .8254 1.0000 .9124 1.0000 1.0000 .9507 1.0000 1.0000 1.0000 .8992 .9443 .9303 .8889 .8434 1.0000 .7600 .8614 .9310 .9537 .8642 .9554 .8986 1.0000 .9813 .9150 1.0000 1.0000 .9453 .9656 .8965 .9069 1.0000 .9889 1.0000 1.0000 .9890 1.0000 1.0000 1.0000 .9705 1.0000 .9931 1.0000 .9338 1.0000 .7824 .9541 D D D I I C I I C C I I I I c I c c I c c c I I I D I c I I Scale efficiency is measured by the CCR model. C = constant returns to scale I = increasing returns to scale D = decreasing returns to scale ’ Determined by the CCR model. The value m easure in the first colum n in the low er half of the table gives the value of the outputs and the inputs for bank 59 in 1984. The second colum n gives the value m easure th at bank 59 w ould have to achieve in order to be DEA effi cient. The difference betw een these num bers is presented in the third colum n .16 Bank 59 should increase its total loans by 143 percent and its non interest incom e by 6 percent. Bank 59 should reduce its four inputs by 26.6 percent of interest expenses and by 24 percent of the other inputs. Table 2 also presents a m easure for bank 59 denoted as the “dual.” This m easure is im portant because the ratio of th e duals for outputs and inputs shows the tradeoff of increm ents or decre m ents in inputs and outputs to DEA efficiency. This is w ith the assum ption th at the bank is free to vary all of its inputs and outputs. The fact th at the dual for NIE is large relative to the others suggests th at the biggest efficiency gains for bank 59 will come from decreasing non-interest expenses. A similar analysis can be conducted for each ineffi- 16ln the case of outputs, this difference is a measure of “ slack.” In the case of inputs, however, the slack variable is more complicated. JANUARY/FEBRUARY 1992 38 cient bank to determ ine its reference banks and the w ay in w hich it can becom e DEA efficient. A W indow A nalysis Table 2 Detailed Results for Bank 59 The available data cover a seven-year span from 1984 through 1990. A three-year period w as chosen to allow five windows. The w indow s and the periods they cover are as follows: window 1 1984 1985 1986 window 2 1985 1986 1987 window 3 1986 1987 1988 window 4 1987 1988 1989 window 5 1988 1989 1990 In each w indow, the num ber of banks is tripled because each bank at a different year is treated as an independent firm . Repeating the procedure discussed above for each window, inform ation about the evolutions of DEA efficiencies of every bank during the seven-year period w as obtained. Table 3 lists the DEA scores of th ree banks by year in each w indow. The average of th e 15 DEA effi ciency scores is presented in the colum n denoted “m ean.” The colum n labeled GD indicates the greatest difference in a bank’s DEA scores in the same year b u t in different w indows. The colum n labeled TGD denotes the greatest difference in a bank's DEA scores for the entire period. A bank can receive a different DEA efficiency score for the same year in different w indows. This variation in the DEA scores of each bank reflects both the perform ance of that bank over tim e as Efficiency Score = .7600 Facet 51 39 27 Lambda = .315 .188 .037 Value if efficient Difference 9,627.0 371.9 54,599.8 .0 21.9 32,157.8 .7895E-04 1000E-08 .3704E-10 7,887.0 2,182.0 19,915.0 77,005.0 1C NIC TL Value measures 9,627.0 350.0 22,442.0 Outputs 5,784.3 1,658.4 15,136.0 58,526.1 2,102.7 523.6 4,779.0 18,478.9 .4762E-09 .2277E-03 .2780E-05 .5815E-05 Dual Inputs IE NIE TD NTD well as th at of other banks. The distribution of banks by th eir average efficiency over the five window s is presented in table 4. Bank 48 w as the only one th at w as efficient for every year in every w indow over the 1984-90 period. Its average efficiency of 1.00 indicates th at bank 48 w as a superb bank in the sample DEA evaluation. Bank 41, on the other hand, began in the first w indow w ith scores of 0.84 in 1984, 0.85 in 1985 and 0.89 in 1986. In the second w indow, bank 41 had scores of 0.86 in 1985, 0.90 in 1986 and 0.94 in 1987. Although all of its efficiency scores fluctu- Table 3 DEA Window Analysis Efficiency Scores Bank 48 41 59 YR84 YR85 YR86 1.00 1.00 1.00 1.00 1.00 1.00 0.84 0.76 0.85 0.86 0.68 0.70 0.89 0.90 0.90 0.60 0.60 0.59 FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis YR87 Summary Measures 0.67 0.70 0.71 0.96 0.98 0.75 0.76 GD TGD 0.00 0.05 0.14 0.68 0.63 0.63 0.65 0.91 0.94 0.96 1.00 1.00 YR90 0.00 0.92 0.94 0.94 0.96 1.00 1.00 1.00 YR89 MEAN 1.00 1.00 1.00 1.00 YR88 0.04 0.18 1.00 0.98 0.77 39 Table 4 Distribution of Average DEA Scores (1984-1990) Model CCR Five-year average DEA score Number of banks 1.00 0.98 — 0.99 0.96 — 0.97 0.93 — 0.95 0.91 — 0.92 0.90 0.88 — 0.89 0.86 — 0.87 0.83 — 0.85 0.80 — 0.82 0.79 0.68 1 8 4 13 7 3 4 10 5 3 1 1 ated slightly in the other th ree w indows, they tended to increase. W ith a gradual im provem ent in its DEA efficiency over the seven years, bank 41 w as alm ost fully efficient in the last year, w ith a DEA score of 0.98. However, its average-efficiency score of 0.92 does not put it am ong the top 13 banks for the period. In contrast to the banks previously discussed, bank 59 displayed relatively erratic and inefficient behavior over the entire seven-year period. Its average DEA score of 0.68 w as the lowest of the 60 M issouri banks analyzed. The w indow analysis enables us to identify the best and the w orst banks in a relative sense, as well as the m ost stable and m ost variable banks in term s of th eir seven-year average DEA scores. CONCLUDING REMARKS The DEA m ethodology discussed in this article has the potential to provide crucial inform ation about banks’ financial conditions and m anagem ent perform ance for the benefit of bank regulators, m anagers and bank stock investors. The DEA fram ew ork is extrem ely general, perm itting multiple criteria for evaluation purposes. M oreover, DEA requires only data on the quantity of inputs and outputs; no price data are necessary. This is especially appealing in the analysis of banking because of the difficulties inherent in defining and m easuring the prices of banks’ inputs and outputs. In addition, the DEA m ethod is highly flexible. In particular, the selection of inputs and outputs has considerably few er limitations th an alternative econom etric approaches. Nevertheless, if the anal ysis is to be useful, care m ust be exercised in the selection of inputs and outputs. REFERENCES Ahn, T., A. Charnes, and W. W. Cooper. “ Some Statistical and DEA Evaluations of Relative Efficiencies of Public and Private Institutions of Higher Learning,” Socio-Economic Planning Sciences, Vol. 22, No. 6,1988, pp. 259-69. ________“ Efficiency Characterizations in Different DEA Models,” Socio-Economic Planning Sciences, Vol.22, No. 6, 1988, pp. 253-57. Aly, Hassan Y., Richard Grabowski, Carl Pasurka, and Nanda Rangan. “ Technical, Scale, and Allocative Efficiencies in U.S. Banking: An Empirical Investigation,” Review of Economics and Statistics (May 1990), pp. 211 -18. Amel, D., and L. Froeb. “ Do Firms Differ Much?” Finance & Economics Discussion Series, Federal Reserve Board, #87 August 1989. Banker, Rajiv D. “ Estimating Most Productive Scale Size Using Data Envelopment Analysis,” European Journal of Opera tional Research 217(1984), pp. 35-40. Banker, Rajiv D., A. Charnes and W. W. Cooper. “ Models for Estimating Technical and Scale Efficiencies,” Management Science, Vol. 30, (1984), pp. 1078-92. Banker, Rajiv D., R. F. Conrad and R. P. Strauss.“ A Compara tive Application of DEA and Translog Methods: An Illustrative Study of Hospital Production,” Management Science Vol. 36 (1986), pp. 30-34. Banker, Rajiv D., and Ajay Maindiratta. “ Nonparametric Anal ysis of Technical and Allocative Efficiencies in Production,” Econometrica (November 1988), pp. 1315-32. Berg, S. A., F. R. Forsund, and E. S. Jansen. “ Deregulation and Productivity Growth in Norwegian Banking 1980-1988: A Non-parametric Frontier Approach,” (Bank of Norway, 1990). Booker, Irene O. “ Tracking Banks from Afar: A Risk Monitoring System,” Federal Reserve Bank of Atlanta Economic Review (November 1983), pp. 36-41. Bovenzi, John F., James A. Marino, and Frank E. McFadden. “ Commercial Bank Failure Prediction Models,” Federal Reserve Bank of Atlanta Economic Review (November 1983), pp. 14-26. Charnes, A., W. W. Cooper, B. Golany, L. Seiford and J. Stutz. “ Foundations of Data Envelopment Analysis for ParetoKoopmans Efficient Empirical Production Functions,” Journal of Econometrics (November 1985), pp. 91-107. Charnes, A., W. W. Cooper, Z. M. Huang and D.B. Sun. “ Poly hedral Cone-Ratio DEA Models with An Illustrative Applica tion To Large Commercial Banks,” Journal of Econometrics (October/November 1990), pp. 73-91. Charnes, A., W. W. Cooper and E. Rhodes. “ Measuring Effi ciency of Decision Making Units,” European Journal of Oper ational Research Vol. 1 (1978), pp. 429-44. Day, D. L., A. Y. Lewin, R. J. Salazar, and H. Li. “ Strategic Leaders in the U.S. Brewing Industry: A Longitudinal Anal ysis of Outliers,” presented at the conference on New Uses of DEA, Austin, Texas, September 27-29,1989. Ehlen, James G. Jr. “ A Review of Bank Capital and its Adequacy,” Federal Reserve Bank of Atlanta Economic Review (November 1983), pp. 54-60. JANUARY/FEBRUARY 1992 40 Elyasiani, Elyas, and Seyed M. Mehdian. “ A Nonparametric Approach to Measurement of Efficiency and Technological Change: The Case of Large U.S. Commercial Banks,” Journal of Financial Services Research (J u ly 1990), pp. 157-68. Congress of the Econometric Society, Barcelona, Spain, August 21-28,1990. Fare, Rolf, Shawna Grosskopf, and C. A. K. Lovell. The Meas urement of Efficiency of Production (Kluwer-Nijhoff, 1985). Putnam, Barron H. “ Concepts of Financial Monitoring,” Federal Reserve Bank of Atlanta Economic Review (November 1983), pp. 6-13. Fare, Rolf, and W. Hunsaker. “ Notions of Efficiency and Their Reference Sets,” Management Science Vol. 32 (February 1986), pp. 237-43. Gilbert, R. Alton. ‘‘Bank Market Structure and Competition, A Survey,” Journal of Money, Credit, and Banking (November 1984, Part 2), pp. 617-45. Grosskopf, Shawna. ‘‘The Role of the Reference Technology in Measuring Productive Efficiency,” The Economic Journal (June 1986), pp. 499-513. Hancock, Diana. “ Bank Profitability, Deregulation, and the Production of Financial Services,” Research Working Paper 89-16, Federal Reserve Bank of Kansas City (December 1989). Koopmans, T. C. “ An Analysis of Production as an Efficient Combination of Activities,” in T. C. Koopmans, ed., Activity Analysis of Production and Allocation, Cowles Commission for Research in Economics, Monograph No. 13 (John Wiley and Sons, Inc., 1951). Korobow, Leon, and David P. Stuhr. “ The Relevance of Peer Groups in Early Warning Analysis,” Federal Reserve Bank of Atlanta Economic Review (November 1983), pp. 27-34. Lovell, C. A. K „ and K. D. Zieschang. “ A DEA Approach to the Problem of New and Disappearing Commodities in the Construction of Price Indexes,” presented at the Sixth World Noonan, John H., and Susan Kay Fetner. “ Capital and Capital Standards,” Federal Reserve Bank of Atlanta Economic Review (November 1983), pp.50-53. Rangan, Nanda, Richard Grabowski, Hassan Y. Aly, and Carl Pasurka. “ The Technical Efficiency of U.S. Banks,” Economics Letters Vol. 28, No. 2 (1988), pp. 169-75. Sherman, H. David, and Franklin Gold. “ Bank Branch Oper ating Efficiency: Evaluation with Data Envelopment Anal ysis,” Journal of Banking and Finance (June 1985), pp. 297-315. Thrall, R. M. “ Overview and Recent Development in DEA: The Mathematical Programming Approach,” paper presented at IC2 Institute, Conference Proceedings, University of Texas at Austin, October 1989. Wall, L. “ Why Are Some Banks More Profitable Than Others?” Working Paper Series No. 12, Federal Reserve Bank of Atlanta (November 1983). Watro, Paul R. “ Have the Characteristics of High-Earning Banks Changed? Evidence From Ohio,” Economic Commen tary, Federal Reserve Bank of Cleveland (September 1, 1989). Whalen, Gary. “ Concentration and Profitability in Non-MSA Banking Markets,” Federal Reserve Bank of Cleveland Economic Review (1:1987), pp. 2-9. Zukhovitskiy, S. I., and L. I. Avdeyeva. Linear and Convex Programming (W.B. Saunders Company, 1966). Appendix A A Comparison of the CCR The CCR R atio M odel The m ost im portant characteristics of the DEA m ethodology can be presented w ith the CCR Ratio Model. Consider a general situation w here n deci sion m aking units, DMUs, convert the same m inputs into the same s outputs. The quantities of these outputs can be different for each DMU. In m ore precise notation, the j-th DMU uses a m-dimensional input vector, Xj, (i = 1 ,2 ,...,m), to produce an s-dimensional output vector, yrj (r = 1,2,..., s). The particular DMU being evaluated is identified by subscript 0; all others are denoted by subscript j. The following optim ization problem is form ed for each DMU: s m r= 1 i= 1 Max h 0 = I u ry r0 / I V;Xi0 subject to the constraints: s 1 r = 1 m uryrj I I VjXji < 1 , u r > 0, Vj > 0 i= 1 for i = 1 ,2 ,..., m; r = 1 ,2 ,..., s; j = 1 ,2 ,..., n. FEDERAL RESERVE BANK OF ST. LOUIS w here the output weights denoted by u r (r = l , 2 ,...,s) and the input w eights denoted by V; (i = 1 ,2 ,...,m) are required to be non-negative (i.e., u r, v, > 0 for r = 1 , 2 , ..., s; i = 1 , 2 , ..., m). The “virtual output” is the sum ( 1 u ryri) and the s r= 1 m "virtual input” is the sum ( 1. VjxJ. The objective function is defined by h„, that is, the ratio of virtual output to virtual input. The solution is a set of optimal input and output w eights. The maximum of the objective function is the DEA effi ciency score assigned to DMU0. The first set of inequality constraints guarantees th at the effi ciency ratios of other DMUs (com puted by using the same w eights ur and Vj) are not greater than unity. The rem aining inequality constraints simply require all input and output w eights to be positive. Since every DMU can be DMU0, this optimization problem is well-defined for every DMU. Because the weights (vi; u r) and the observations of inputs i= 1 41 and outputs (xii; yrj) are all positive and the con straints m ust be satisfied by D M U 0, the maximum value of h 0can only be a positive num ber less than or equal to unity. If the efficiency score h 0 = 1, D M U 0 satisfies the necessary condition to be D EA efficient; otherw ise, it is D EA inefficient. The above problem cannot be solved as stated because of difficulties associated w ith nonlinear (fractional) m athem atical program m ing. Charnes and Cooper, how ever, have developed a m athe matical transform ation (the so-called “CC transfor m ation”) w hich converts the above nonlinear program m ing problem into a linear one. Existing duality theory and simplex algorithm s in linear program m ing are used to solve the transform ed problem .1 For a linear program m ing problem , th ere exists a pair of expressions w hich are "dual” to each other. The CCR ratio m odel is form ed by problem 1 and problem 2 below: Problem 1: M inh 0 = 0O- e( iZ1 s ] + rI=1 s+ r) = subject to m s n 0 0 ^io — n I j= i ^ ^ij j= l _ ^ i yrj A, - s+ = yr0, A, > 0 , r n 2 i-i yri A, and 0oxio > n 2 i-i X„ Aj; for r = 1 ,..., s; i = 1 ,..., m. 2. Technical efficiency will be achieved if, and only if, all of the following conditions are satisfied: 60 = 1 and s+ = 0, = 0 r for i = 1 ,.., m; r = 1 ,.., s. The condition 90 1 ensures that D M U 0 is located on the production frontier; the conditions s+ = 0 r and s] = 0 exclude situations such as F6 in figure 1 of the text. 3. The constant retu rn s to scale condition for = s] > 0, s+ > r n DMU0 occurs if 1 A = 1, otherwise; X A > 1 : j 0; s r= 1 s y rtl < n — for i = 1,.., m; r = 1,.., s; j = 1,.., n. Problem 2: Max Y0 = S MrYro subject to m to ensure th at all of the observed inputs and outputs have positive values or shadow prices and th at the optimal value h 0is not affected by the values assigned to the so-called "slack variables" ( s + or S7). 2 r The main conclusions from the CCR model are sum m arized as follows: 1. The optim al values of s+, S7, and A, via problem 1 m ust be positive. The following inequal ities should then be satisfied: j= i j= i n implies decreasing retu rn s to scale; iI- i A ,< 1 implies increasing retu rn s to scale. 4. An adjustm ent can be m ade in order to move (or project) inefficient D M U 0 onto the efficiency frontier. The projection (x', y ‘) in the CC R m odel is form ed by the following form ulas: m I V xi0 = 1 , 1 M - 2 ^x,, < 0, ; 1= 1 r =1 .-yr, 1=1 Mr >£. Vj > £ for i = 1 ,.., m; r = 1 ,.., s; j = 1 ,.., n. As before, the subscript 0 represents the D M U being evaluated, x(j denotes input i, yrj denotes output r of DM U j, and /ur and represent the weights for outputs and inputs, respectively. An arbitrarily small positive num ber, £, is introduced x i0‘ = 0 n X io yro‘ = y ro + “ S i S i = !< r r = 1. •- m S. The differences (xi0 - xi0'), i = 1 ,.., m, represent am ounts of inputs to be reduced; (yro' - yr0), r = l,..,s, represen t th e am ounts of outputs to be increased in order to move D M U 0 onto the effi ciency frontier. Hence, these differences can provide diagnostic inform ation about the ineffi ciency of D M U 0. 1This also opens the way for many different DEA models which are refined, more flexible or more convenient for computations. These DEA models (BCC model, additive DEA model, cone ratio DEA model, CCW model) and their mathematical characteristics are beyond this paper. 2For the e-Method, see Zukhovitskiy et al. (1966), pp. 46-51. JANUARY/FEBRUARY 1992 42 5. Problem 1 is defined as the "prim al” problem while problem 2 is the "dual.” The dual variables have the economic interpretation of “shadow prices.” The value of vi indicates the m arginal effect of input x i0on the DEA efficiency score. The value of fu, indicates the m arginal effect of output yr on the DEA efficiency score. A com parison of these dual variables provides inform ation on the relative im portance of inputs and outputs in the DEA evaluation. 6 . In the CCR model, problem 1 (or problem 2) is solved for each DMU. Theoretically, th ere is no limitation on how m any DMUs can enter the DEA model. Hence, the DEA model can perform an effi ciency diagnosis for m any DMUs. W hy is this approach referred to as data envelopm ent analysis? The tw o inequalities in conclusion 1 , n n yro < i-i yrjAj and 0oxio > j-i x,^, Z I for r = 1 ,..., s; i = 1 ,..., m are constraints to be satisfied for the optimal solu tion. The first inequality implies th at the output of DMU 0should not exceed the linear com bination of all observed output yrj; thus, the optimal solutions will create a hyperplane to envelop the output of DMU 0from above. Similarly, the second constraint can be in terpreted such th at the optimal solutions create another hyperplane w hich envelops the input of DMU 0from below. Since both outputs and inputs of the DMU evaluated are enveloped from above and below, the nam e DEA exactly m atches the geom etric interpretation of the procedure. To see how this w orks, assum e that th ere is a group of DMUs th at produces the same outputs using the same inputs, bu t in varying am ounts. In ranking th eir efficiencies of DMUs, DEA assigns weights to the outputs and inputs of each DMU. These w eights are neither predeterm ined nor based on p rior inform ation or preferences of the decision m akers. Instead, each DMU receives a set of "optim al” weights th at are determ ined by solving the above m athem atical program m ing problem . This procedure generates a DEA effi ciency score for the DMU evaluated based on the solution value for the input and output weights. A set of constraints guarantees th at no DMU, including the one evaluated, can obtain an effi 3See Charnes et al. (1985). FEDERAL RESERVE BANK OF ST. LOUIS ciency score th at exceeds unity. In this way, DEA derives a m easure of the relative efficiency rating for each DMU in the cases of m ultiple input and output. The A dditive M odel Among DEA models, the additive m odel has been im portant in applications. The additive m odel can be form alized as the following tw o problem s, w hich are dual to each o th er .3 Problem 3: m s i= 1 r= 1 Max I st / |x j + I s \ l |y j subject to n n xi0 - i-i Xjj Aj — s 7 = 0, i-i yri Aj - s+ = y*,, I I r 2 A, = 1 , A, >0, sj >0, s+ > 0, for i = 1 ,.., m; r = 1,.., s; j = 1,.., n. Problem 4: s m r=1 i~l Min I /iry r0 + I VjX.o + u 0 subject to: s I r= 1 m M + 1 ,-yn 1= 1 VjXjj + u 0 > 0, V > 1 / |xi0|, H < i / l y j , i r for i = 1 ,.., m; r = 1 ,.., s; j = 1 ,.., n. Com pared w ith the CCR model, the additive m odel has introduced another constraint I I = 1 and a new variable u 0. The new i-i constraint in problem 3 ensures th at the efficiency frontier is constructed by the convex com bina tions of original data points rath er than a convex cone as in the CCR model. The new variable u 0in problem 4 is used to identify retu rn s to scale. The other variables in the additive m odel have interpretations similar to the CCR model. In addition, th ere is a difference in the w ay the additive model and the CCR ratio model locate the efficient reference point on the facet. In figure A .l, an output isoquant consists of input com bina tions for five firm s (F„ F2, F3, F4and F.) in the case of one-output (y) and tw o-input (xt and x2). Point F5 represents an inefficient DMU w hich uses m ore of x, and x 2to produce the same am ount of output as n 43 Figure A.1 The Difference Between CCR and Additive DEA Models Inpu t X 2 to the intersection point B divided by the length from the origin to Fs. In the additive model, how ever, the reference efficient point on facet F2-F3is denoted by A, w hich is determ ined by maximizing the sum of the slacks, s, + s2. Geom etri cally, the slack variables are expressed by the horizontal line starting from Fs and the vertical line extending to the facet F 2-F3. Point A is selected such th at the sum of the lengths of the horizontal and vertical lines are maximized. The DEA effi ciency score in th e additive m odel th at we used is com puted by the following form ula: m s m s s 1*1 r - 1 i«1 r= 1 r = 1 ( I x(o + I y'J /( I x l0 + I y r t + I 2 s;). w here x -0and y r0are corresponding inputs and ' outputs of the efficient reference point, such as point A. The DEA scale efficiency in the additive model is identified by a variable u 0in problem 4 in accor dance w ith the following criteria: O V .________________________________________________ In pu t X j its efficient reference DMUs, F2and F3. By the CCR ratio model, the efficiency score is determ ined via a value h0, w hich can be interpreted in term s of the ray from the origin to F5. That is, h 0 is ex pressed by the length of the ray from the origin If u 0 = 0, DMU 0has constant retu rn s to scale; otherw ise, u 0 > 0 implies decreasing retu rn s to scale; u 0 < 0 implies increasing retu rn s to scale. The value of variable u 0is part of an optimal solu tion of the additive model and is produced by the com puter code such th at facet rate = - u 0. Appendix B Data Envelopment Analysis: An Alternative Approach In m easuring and evaluating technical and scale efficiencies th ere are tw o basic approaches: the DEA technique developed by Charnes, Cooper and others in operations research and the approach developed by Farrell, Fare and Grosskopf, am ong others, in econom ics .1 The latter approach is based upon a set of axioms on production tech nology to define the concept of efficiency. Some connections of the tw o approaches have been investigated by Banker, Charnes and Cooper (1984) and by Fare and H unsaker (1986). Both approaches share the characteristics that th ere is no need to specify a production function or cost function and to estim ate the param eters. T herefore, they are nonparam etric, nonstochastic techniques th at can be used to construct a m ultiproduct frontier relative to w hich the effi ciency m easures of the entities in the sam ple are calculated. Because the frontier in these approaches is generated by data and all observa tions are enveloped by the frontier, both approaches can be viewed as Data Envelopm ent 'See Fare and Hunsaker (1986); Fare, Grosskopf and Lovell (1985). JANUARY/FEBRUARY 1992 44 Analysis. In this appendix, some of the differences and similarities am ong the CCR and the additive models and the Farrell or Russell models are discussed. The choice of efficiency reference on the rele vant frontier is a m ajor difference am ong these DEA models. In the Farrell or Russell models, th ree m easures of technical efficiency can be defined: input, output and graph efficiency m easures. Using the input efficiency m easure, the ob served output vector is fixed and the search for efficient reference is constrained to proportion ally reducing inputs until the efficient frontier is reached. The “ratio of contraction,” as it is called, is the ratio of the particular input to be efficient to the cu rren t level of inputs (in the Farrell input model). Using the output efficiency m easure, the ob served input vector is fixed and the outputs p ro portionally expanded until th e efficient frontier is reached. The "stretch ratio” of the output, as it is called, is th e ratio of efficient output to the cu rren t level of output (in the Farrell output model). For the graph efficiency m easure, both input and output vectors are varied. Inputs are reduced and outputs are expanded, both proportionally, w ith the input ratio reciprocal to the output ratio. In the case of figure 1 in the text, A is the refer ence point for the input efficiency m easure, B is the reference point for the output efficiency m easure and C m ight be the reference point for the graph efficiency m easure. These th ree effi ciency m easures can be classified as radial because proportional changes of inputs and/or outputs are used in defining them . To illustrate the input efficiency m easure, ray OF3in figure 1 of the text is used to represen t the optimal scale that would be generated by long-run com petitive equilibrium . The overall input effi ciency m easure is defined w ith respect to the ray OF3, while the input pure technical efficiency is defined w ith respect to the line segm ent connect ing F1( F 3 and F5. The m easure of input overall technical efficiency, KD/KF,, can be decom posed into the m easure of pure technical input efficiency given by the ratio KA/KF2 and the m easure of input scale efficiency given by the ratio KD/KA. W hen 2See Grosskopf (1986). FEDERAL http://fraser.stlouisfed.org/ RESERVE BANK OF ST. LOUIS Federal Reserve Bank of St. Louis the scale efficiency equals unity, the constant re tu rn s to scale occur; otherw ise non-increasing or varying retu rn s to scale hold. It is clear from these examples that, in general, these radial efficiency m easures will be different. M oreover, th ere is nothing to guarantee th at a firm th at is output efficient by this m easure is also input efficient or vice versa. For example, the firm denoted by F6in figure 1 of the text is output effi cient by the output efficiency m easure, but is not input efficient (see Fare, Grosskopf and Lovell (1985)). However, the Farrell input efficiency m easure is reciprocal to the Farrell output effi ciency m easure, if and only if, the technology is hom ogeneous degree one. Because this condition is satisfied by constant retu rn s to scale tech nology, the Farrell input and output efficiency m easures are "identical” in this case. For models w ith other technologies, simple relationships betw een input and output efficiency m easures do not hold. An im provem ent of the Farrell or Russell models over the others is the use of non-radial efficiency m easures. The use of proportional changes of inputs and/or outputs in searching for efficient reference is abandoned. M oreover, different piecewise linear technology can be accom m odated in both Farrell and Russell models to m eet the needs of various users. For example, to m easure scale efficiency w e can use constant retu rn s to scale, non-increasing retu rn s to scale or varying retu rn s to scale technologies. These technology constraints can be easily im posed by corresponding restrictions on the "intensity param eters” in the Farrell or Russell models. In the CCR or additive DEA m odel discussed in appendix A, how ever, only one efficiency m easure is defined: the CCR m odel uses the radial m easure of efficiency while the additive m odel uses the non-radial m easure. Geometrically, the efficiency frontier w ith cons tant retu rn s to scale technology is a convex cone, bu t it is a convex hull in cases of both non-increasing and varying retu rn s to scale. In general, these constraints on technology form a chain such that one efficiency frontier is enveloped by another. Consequently, the associated efficiency m easures are com patible and nested .2 45 As is presented in appendix A, the CCR model has a convex cone efficiency frontier that implies technology w ith constant retu rn s to scale. The additive m odel uses a convex hull as its efficiency frontier th at is associated w ith the varying returns to scale. Even though the efficiency frontier of the additive model is enveloped by the efficiency frontier of the CCR model, the efficiency scores given by both models are not com patible because one uses a radial m easure while the other uses a non-radial m easure. The efficiency ratio of the CCR m odel is identical to the Farrell input effi ciency m easure (or reciprocal output efficiency m easure) w ith constant retu rn s to scale technol ogy. Although both additive and Russell models define non-radial efficiency m easures, the defini tions are not identical. Hence, the efficiency m ea sures given by these m odels are not compatible. W ith our 1984 data of 60 M issouri com m ercial banks, we used the Farrell m odel w ith input and output efficiency m easures and different tech nology constraints. The overall technical efficien cies and scale efficiencies are presented in table B.l. The reported results are based upon the input m easure of efficiency. Com paring table B.l w ith table 1 in the text, we can see that the CCR model and the Farrell input m odel give identical technical efficiency m easures and classification of retu rn s to scale. Farrell input scale efficiency m easures in table B.l indicate that the scale inefficiency was not a m ajor source of technical inefficiency in this group of banks. For a few of the banks in the sample, how ever, the scale inefficiency might be a problem . Table B.1 Farrell Technical and Scale Efficiencies of 60 Missouri Commercial Banks (Input Efficiency Measure)_______ Bank no. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Efficiency measure Scale measure Type of scale1 Bank no. Efficiency measure Scale measure Type of scale1 .8545 .9228 .9033 .8588 1.0000 .8766 .8709 .8841 .8735 .8115 .9086 .7852 .8338 .9739 .8937 .8292 .8705 .9684 .8439 .9527 .9746 .8681 .9744 .9003 1.0000 .8714 1.0000 1.0000 .8753 .9003 .8556 .9228 .9177 .9852 1.0000 .9983 .9308 .9980 .9731 .9943 .9962 .9740 .9572 .9994 .9550 .9938 .9714 .9939 .8439 .9867 .9747 .9306 .9843 .9877 1.0000 .9930 1.0000 1.0000 .9622 .9538 D I I I C I I I I I I I I I I I I I D I I I I I C I c c I D 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 .8568 .9305 .8509 .8392 .8596 1.0000 .8712 .8707 1.0000 1.0000 .8500 .8867 .8220 .8254 1.0000 .9124 1.0000 1.0000 .9507 1.0000 1.0000 1.0000 .8992 .9443 .9303 .8889 .8434 1.0000 .7600 .8614 .9813 .9964 .8777 .9738 .9849 1.0000 .9870 .9316 1.0000 1.0000 .9853 .9637 .9836 .9887 1.0000 .9769 1.0000 1.0000 .9983 1.0000 1.0000 1.0000 .9758 .9443 .9762 .8889 .9427 1.0000 .9565 .9830 D D D I I C I I C C I I I I C I C C I C C C I I I D I C I I Where C = constant returns to scale I = increasing returns to scale D = decreasing returns to scale 1Determined by the Farrell input measure. JANUARY/FEBRUARY 1992 The Federal Reserve Bank of St. Louis 1991 J Research W orking Paper Series Number Author(s) Title 91-001A Mark D. Flood Market Structure and Inefficiency in the Foreign Exchange Market 91-002A James Bullard and Alison Butler Nonlinearity and Chaos in Economic Models: Implications for Policy Decisions 91-003A James Bullard Collapsing Exchange Rate Regimes: A Re-interpretation 91-004A James Bullard Learning Equilibria 91-005A David C. Wheelock and Subal C. Kumbhaker Which Banks Choose Deposit Insurance? Evidence of Adverse Selection and Moral Hazard in a Voluntary Insurance System 91-006A David C. Wheelock Regulation and Bank Failures: New Evidence from the Agricultural Col lapse of the 1920s Note: Single copies of Research papers are available upon request by writing to: Research and Public Information Dept. Federal Reserve Bank of St. Louis P.O. Box 442 St. Louis, Missouri 63166-0442 F e d e ra l R e s e rv e B an k o f St. L o u is Post Office Box 442 St. Louis, Missouri 63166 The R e v ie w is pu b lish ed six tim es p e r y e a r b y the R esearch an d P ublic In form ation D epartm en t o f the F ederal R eserve R ank o f St. Louis. Single-copy su b scrip tio n s are available to the pu b lic f r e e o f charge. Mail requ ests f o r su bscrip tio n s, b a ck issues, o r a d dress changes to: R esearch an d P ublic In form ation D epartm ent, F ederal R eserve Rank o f St. Louis, P.O. Rox 442, St. Louis, M issouri 63166. The view s ex p ressed are th ose o f the individual au thors an d do n ot n ecessarily reflect officia l p o sition s o f the F ederal R eserve R ank o f St. Louis o r the F ederal R eserve System . A rticles herein m ay b e reprin ted p ro v id e d the sou rce is credited . P lease p ro vid e the R an k’s R esearch an d P ublic In form ation D epartm en t w ith a co p y o f reprin ted m aterial.