The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of New York Quarterly Review W inter 1993-94 Volume 18 Number 4 1 Causes and Consequences of the 1989-92 Credit Slowdown: Overview and Perspective 24 International Interest Rate Convergence: A Survey of the Issues and Evidence 38 Debt Reduction and Market Reentry under the Brady Plan 63 Index Amortizing Rate Swaps 71 The Pricing and Hedging of Index Amortizing Rate Swaps 75 Recent Trends in Commercial Bank Loan Sales 79 Treasury and Federal Reserve Foreign Exchange Operations Federal Reserve Bank of New York Quarterly Review Winter 1993-94 Volume 18 Number 4 Table of Contents 1 Causes and Consequences of the 1989-92 Credit Slowdown: Overview and Perspective M.A. Akhtar This article is the overview essay for a volume, Studies on Causes and Con sequences of the 1989-92 Credit Slowdown, published by the Federal Reserve Bank of New York. The twelve papers in the volume cover a broad range of issues concerning the credit slowdown, including the importance of credit demand relative to credit supply factors, the role of bank and nonbank credit sources, and the impact of credit supply shifts on the economy. This essay provides a conceptual framework for the analysis of credit issues, reviews the evidence presented in the volume, and offers a perspective on the implications for monetary policy. 24 International Interest Rate Convergence: A Survey of the Issues and Evidence Charles A. Pigott World financial markets have become substantially integrated over the last two decades. Contrary to widespread expectations, however, this integration has not led to any greater convergence of interest rates across countries. This article examines why convergence has not occurred and more generally what financial integration does— and does not— mean for international inter est rate relations. 38 Debt Reduction and Market Reentry under the Brady Plan John Clark In March 1989, U.S. Treasury Secretary Brady proposed a new approach to resolving the developing country debt problem and restoring the creditworthi ness of restructuring countries. The Brady Plan encouraged market-based reductions in debt and debt service for countries implementing economic reforms. This article analyzes the structure of the financial packages that fol lowed this change in approach and considers their impact on countries and their creditors. Table of Contents 63 Index Amortizing Rate Swaps Lisa N. Galaif As short-term interest rates have declined over the past several years, investors have increasingly sought higher yielding investment vehicles. The index amortizing rate (IAR) swap is one of several new instruments that have been developed in response to this investor demand for yield enhance ment. This article explains the structure and pricing of IAR swaps, some of the risks associated with the product, and the uses and growth prospects of the market. 71 The Pricing and Hedging of Index Amortizing Rate Swaps Julia D. Fernald IAR swaps have proved difficult to price because of the complexity of their embedded options. Since these options depend on the path of interest rates, pricing requires a model of interest rate movements. This article uses a sim ple interest rate model to illustrate the pricing and hedging of an IAR swap. 75 Recent Trends in Commercial Bank Loan Sales Rebecca Demsetz The dollar volume of commercial bank loan sales rose rapidly in the mid1980s but has declined equally rapidly over the past few years. This article provides insight into these loan sales trends by looking beyond the aggregate data and separately examining the sales activities of the largest loan sellers and those of all other banks. 79 Treasury and Federal Reserve Foreign Exchange Operations 84 List of Recent Research Papers Causes and Consequences of the 1989-92 Credit Slowdown: Overview and Perspective by M. A. Akhtar This article is the overview essay for a volume, Studies on Causes and Consequences of the 1989-92 Credit Slowdown, published by the Federal Reserve Bank of New York. In addition to the present essay, the volume contains twelve papers dealing with a broad range of issues con cerning the credit slowdown, including the importance of credit demand relative to credit supply factors, the role of bank and nonbank credit sources, the impact of credit supply shifts on the economy, and the implications of those shifts for monetary policy. The volume is available from the Public Information Department of the Federal Reserve Bank York. Purchase information appears on page 85 of this issue of the Quarterly Review. Between early 1989 and late 1992, U.S. econom ic growth averaged less than 1 percent, well below the long-run trend growth of the economy. This sluggish pattern of growth per sisted in the face of substantial easing in m onetary policy. Indeed, the econom y failed to recover significantly after the 1990-91 d o w n tu rn . A p p a re n tly the fa vo ra b le e ffe c ts of m onetary easing were not sufficient to overcom e numerous factors depressing the economy: lower defense spending, com m ercial real estate depression, relatively tig h t fiscal policy, global com petition, corporate restructuring, histori cally low levels of consum er confidence, and the overex tended financial positions of households, businesses, and financial institutions. T h e s lu g g is h re a l g ro w th w a s a c c o m p a n ie d by an u n p re ce d e n te d ly sharp slow dow n in c re d it grow th over 1989-92. Many observers have identified high debt service burdens of the nonfinancial sectors and w idespread bal ance sheet problem s of borrowers and lenders as crucial elem ents underlying both the credit slowdown and the per sistent w eakness of the econom y. Others have attributed the sluggish econom ic perform ance to supply-side factors underlying the credit slowdown, w hich resulted in a pro longed period of substantially reduced credit availability to b u s in e s s e s and h o u s e h o ld s . M ore re c e n tly , c o n c e rn s about cred it a va ila b ility appear to have eased as credit growth has shown some signs of recovery. A g a in s t the b a ckg ro u n d of th e se d e ve lo p m e n ts, this overview provides a broad perspective on the causes and consequences of the 1989-92 credit slowdown. It begins by presenting a general conceptual fram ew ork for the analysis and then reviews the evidence from the collection of stud ies on the credit slowdown. The article also discusses im pli cations of the evidence for m onetary policy and offers some tentative general observations on the recent credit slow down experience. Overall, studies reviewed here provide substantial evi dence of credit supply problems, or a “credit crunch,” du r ing the 1989-92 period fo r both bank and nonbank credit sources. The evidence on the consequences of credit su p ply constraints is less com pelling, but the studies do indi ca te , at le a s t c o lle c tiv e ly , th a t c re d it c o n s tra in ts have played som e role in w e a ke nin g e co n o m ic a c tiv ity . The depressing effects of the credit crunch appear not to have been the prim ary or dom inant cause of the econom ic slow down, however. As for the im plications for m onetary policy, credit supply problem s have clearly contributed to reducing the effectiveness of m onetary policy, although it is difficult to isolate their effects from those of other factors disrupting or altering the channels of policy influence to the econom y. Credit slowdown vs. credit crunch: A general framework There is no generally accepted definition of the term “credit FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 1 crunch,” but it is usually taken to mean a sharp reduction in the supply or availability of credit at any given level of inter est rates. To clarify term inology and to provide a broad co n text for the issues involved in identifying a credit crunch, we begin with the more encom passing notion of credit slo w down or decline. At the broadest level, an observed slow down or decline in credit may result from either the demand side or the supply side. A t a given lending rate or price of credit, the dem and fo r cre d it m ay fall because of oth e r (no nprice) d e te rm in an ts of cre d it dem and. In the usual graphical supply-dem and fram ew ork, the demand sched ule for credit may shift down and to the left. This is shown in Chart 1, panel 1, under very sim plistic market conditions, where the price of credit includes both the loan rate and n o n ra te lo an te rm s , su ch as c o lla te ra l, m a tu rity , and covenants. From a m acroeconom ic perspective, this type of shift may occur because of lower credit dem and stem ming from either cyclical w eakness in econom ic activity or structural factors— such as changes in the tax code, inven tory techniques, or the borrow ers’ desired debt-to-incom e ratio— that reduce the perceived need fo r cre d it p e rm a nently. In general, shifts in credit demand induced by cycli cal w eakness in econom ic activity are relatively com m on place while credit dem and shifts due to structural changes are som ew hat less frequent but not unusual. A dow nward shift in credit dem and tends to put dow n ward pressures on loan rates and other loan term s and, given an unchanged supply schedule, leads to easier loan term s at the new credit m arket equilibrium . Moreover, if a dow nward credit dem and shift is caused by structural fa c tors, it may also be accom panied by a steepening (flatten ing) of the dem and schedule; the demand for credit may becom e less (m ore) responsive to changes in the price of credit (C hart 1, panel 1, D2 schedule). On the supply side, a credit slowdown or decline may reflect reduced w illing ne ss to lend at prevailing interest ra tes and d e m an d c o n d itio n s . F a cto rs th a t can ca u se reduced w illingness to lend include, among others, balance sheet difficulties of lenders (poor quality assets, high loan losses, and so forth), higher capital requirem ents and regu latory constraints on lenders, and increases in actual or perceived riskiness of bo rro w e rs’ credit quality. The last factor is intended to capture credit supply shifts resulting from changes in a bo rro w e r’s balance sheet conditions. Specifically, a deterioration in the quality of a borrow er’s balance sh e e t re fle c tin g , fo r e xa m p le , a d ro p in a sse t prices, w eakens his a bility to repay existing debts or to borrow new fu n d s.1 The decline in creditw orthiness of the borrower, in turn, may reduce the lender’s w illingness to 1 More generally, the deterioration in the quality of the borrower’s balance sheet (and the associated decline in creditworthiness) may result either from a cyclical decline or from noncyclical shocks (economy-wide or partial) such as an asset price drop in one or more sectors. As explained below, it is very difficult to separate credit supply effects from demand effects of general cyclical shocks to the economy. 2 forFRBNY Digitized FRASERQuarterly R eview /W inter 19 9 3 -9 4 extend a loan, causing a decline in the supply of credit. In this situation, the supply shift reflects reduced credit a vail ability to borrowers whose credit quality has been impaired, Chart 1 Credit Demand and Supply Price of credit i Panel 1 S(k,g,z) D * \ \ * \ i \ \ \ \ *, v * N V / S V / \ \ \ \ ^ ijP i n 1 \ ^ % i \ A 1 ----------------------------------------------1------C *, C* D (y,x) ' D, Quantity of credit Price of credit D= initial credit demand schedule, where all determinants other than the price of credit (loan rate and nonrate loan terms) are held constant (y is a proxy for aggregate demand in the economy and x represents all other variables) S = initial credit supply schedule, where all determinants other than theprice of credit are held constant (k is a proxy for capital, g is a proxy for regulatory and supervisory constraints, and z represents all other variables) C *= initial equilibrium credit r * = initial equilibrium price of credit D, and Dz represent the credit demand schedule after shifts, while S, and S2 represent the credit supply schedule after shifts. C*,, C*2, r *„ and r*2 represent new equilibrium values for credit and the price of credit. but there is no change in the lender’s desire to lend to those borrowers whose creditworthiness has remained un changed. Note that the drop in borrowers’ creditworthiness could be treated, in principle, as a drop in credit demand by borrowers of given risk characteristics (unchanged credit worthiness) in that there are fewer such borrowers. Nonetheless, at a practical level, it is more convenient to look at the effect of changes in borrowers’ credit quality— especially those resulting from noncyclical shocks—on the willingness of lenders to supply credit. In any event, the reduced willingness to lend may show up as a leftward shift in the credit supply schedule (Chart 1, panel 2). In this case, borrowing is rationed by price as loan rates and nonrate loan terms tend to tighten and the new credit market equilibrium is attained at higher interest rates and generally more restrictive loan terms, other things equal. The reduced willingness to lend may not show up as a simple leftward shift of credit supply envisaged in the con text of a market-clearing environment, however. Instead, lenders may resort to increased nonprice credit rationing; that is, loans are rationed by quantity rather than by varia tions in prices (interest rates and nonrate loan terms). In this case, lenders do not feel that they can protect them selves against risk by charging higher credit prices. Put another way, the credit supply schedule is not fully opera tive; in the extreme case, the schedule shifts leftward and becomes vertical, with the supply of credit becoming com pletely insensitive to interest rates (Chart 1, panel 2, S2 schedule). In practice, the existence of nonprice credit rationing does not preclude the role of interest rates and other loan terms; some borrowings may be rationed by price and others by quantity or by both. Nonprice credit rationing may take many different forms: some borrowers obtain loans while other borrowers with identical creditwor thiness do not; loans for certain types of borrowing or to certain classes of borrowers are unavailable; some appar ently creditworthy borrowers are denied loans at prevailing interest rates because lenders do not perceive them to be creditworthy.2 The papers in this volume deal with both demand and supply factors in the credit slowdown since 1989, but the emphasis is on sorting out the role of supply-side factors and their implications for nonfinancial economic activity. Accordingly, the term credit crunch as used here refers to a slowdown or decline in the supply of credit, whether rationed by price or nonprice mechanisms, or simply to credit supply problems. This definition is clearly much broader than the narrow use of that term to describe situa tions of nonprice credit rationing. It is also broader than another frequently mentioned definition of credit crunch: “ a widespread, sudden, sharp, indiscriminate, and rather brief 2 See Jaffee and Stiglitz (1990) for a detailed survey of various aspects of credit rationing. credit shutdown” (Wojnilower 1993).3 In a macroeconomic context, the existence of credit sup ply problems implies that the observed credit slowdown or reduction cannot be fully explained by cyclical develop ments in aggregate demand, except insofar as cyclical developments may have significant adverse effects on bor rowers’ creditworthiness as perceived by lenders. There are, of course, numerous identification problems in sorting out supply from demand factors in the credit slowdown. For example, a sharp reduction in the willingness to lend may lead to a decline in output, inducing a reduction in the demand for credit. In these circumstances, the credit slow down will be reported as reflecting lower demand for credit even though it was, in fact, caused by an initial shock to the supply of credit (Friedman 1993a, 1993b). More generally, with demand and supply factors operat ing simultaneously and interacting with each other, it is very difficult to distinguish shifts in the supply schedule from developments on the demand side. Lenders usually tend to tighten credit standards and terms for lending when the overall economy slips into a recession because, on aver age, business and household loans entail higher risks than before. But the extent of lenders’ response depends not only on the degree of perceived economic weakness and its effects on borrowers’ credit quality but also on the state of their own balance sheets. From the perspective of bor rowers, this situation would look like a contraction in credit supply, while lenders may believe this to be a response to developments in aggregate demand. Strictly speaking, there is no change in the lenders’ willingness to extend credit to borrowers of given circumstances (that is, un changed creditworthiness). At the same time, the reduced supply is not a response to lower demand for credit. The constriction in the supply of credit has clearly been caused by a decline in the willingness of lenders, albeit one that reflects the adverse effect of the weaker economy on the creditworthiness of borrowers and balance sheets of banks. Any sorting out of the demand and supply aspects in this case would be further complicated by the fact that the recession itself would reduce the demand for credit. Identifying demand and supply factors in the recent credit slowdown is particularly difficult because of the conjunction of the prolonged cyclical weakness in the economy with a correction of earlier credit excesses. Those credit ex cesses, as noted below, reflected the unusually rapid in creases in debt in the mid-1980s and became unsustain able over time as both borrowers and lenders experienced balance sheet and other difficulties, with cyclical develop3 For other perspectives on defining a credit crunch, see Peek and Rosengren (1992), Owens and Schreft (1992),and Wojnilower (1992a). For other perspectives on the current credit crunch, see Bernanke and Lown (1991), Cantor and Wenninger (1993), Jones (1993), Jordan (1992), Kaufman (1991), Kliesen and Tatom (1992), Peek and Rosengren (1992), Sinai (1993), Syron (1991), and Wojnilower (1993). For detailed analysis of earlier crunches, see Wojnilower (1980) and Wolfson (1986). FRBNY Quarterly Review/Winter 1993-94 3 ments reinforcing pressures for correction. In this highly “endogenous” process, the demand for credit is believed to have fallen simultaneously with reductions in banks’ capac ity and willingness to lend. Notwithstanding these difficulties, the twelve studies in this volume examine a broad range of issues concerning the 1989-92 credit slowdown. Five of these studies (Lown/ Wenninger, Cantor/Rodrigues, Johnson/Lee, Demsetz, Seth) look at various aspects of the role of bank and non bank credit sources in the slowdown of private nonfinancial debt, focusing on the importance of credit demand relative to credit supply factors. One study (Hamdani/Rodrigues/ Varvatsoulis) reviews survey data on credit tightening from lenders and borrowers, and another study (M osser/ Steindel) explores the role of economic activity and other “fundamentals” in explaining the recent credit slowdown. Three studies (Harris/Boldin/Flaherty, Mosser, Steindel/ Brauer) investigate the effects of credit supply problems on various aspects of nonfinancial economic activity. Finally, two studies (Hilton/Lown, Hickok/Osler) consider some special aspects of the credit slowdown: one attempts to assess the impact of credit supply shifts on the broadly defined money stock, M2, and the other provides a broad overview of the nature and extent of the credit slowdown abroad, largely based on the experience in France, Japan, and the United Kingdom. The remainder of this article reviews evidence from the twelve studies under four broad headings: the extent of the credit slowdown; factors behind the credit slowdown; con sequences of the credit crunch for nonfinancial economic activity; and implications of the credit crunch for monetary policy. The last section offers a few tentative concluding observations on the recent credit crunch experience. Extent of the recent credit slowdown Collectively, the studies in this volume show that the U.S. economy has experienced a broadly based and sharp credit slowdown in recent years. In documenting and describing the credit slowdown from the viewpoint of vari ous types of borrowers (business, household, real estate, small business) or lenders (banks, other depositories, finance companies, insurance companies, foreign banks, bond markets), most of the studies begin by examining the extent of credit slowdown in the recent period. Since the timing of the slowdown is not uniform across all borrowers and lenders, however, these studies do not target a com mon time period for the recent credit slowdown. Nor do they judge the recent credit slowdown against a common histor ical benchmark. Instead, each study provides a compre hensive look at relevant credit developments from its par ticular vantage point using whatever time periods make most sense. Nevertheless, it may be useful to provide a common time frame for summarizing the extent of the slowdown in private Digitized FRASER 4 forFRBNY Quarterly Review/Winter 1993-94 nonfinancial debt and its main components on both the lending and the borrowing sides. I use the flow of funds data to highlight the breadth and depth of credit slowdown over the three years from 1989-IV to 1992-IV, taken as a whole, relative to long-term trends in the periods 1960-82 and 1982-89. Because inflation was greater in the earlier periods than in the most recent period, comparisons of nominal credit growth rates may be misleading. I have, therefore, presented data in both nominal and real terms in many cases. For simplicity and convenience, however, I have used the GDP deflator to convert nominal dollars into real dollars rather than search for specific sectoral defla tors. (Sectoral deflators might change precise real dollar values but they are unlikely to alter the broader contours of constant dollar data obtained on the basis of the GDP deflator.) The points made here provide a broad overview of the extent of the credit slowdown to nonfinancial borrow ers from both bank and nonbank sources, and may be viewed as a summary of details in various studies. Private nonfinancial debt Using data on nominal and real debt and ratios of debt to GDP, I begin by looking at the extent of the slowdown in pri vate nonfinancial debt in terms of its three broad decompo sitions: business versus household debt, mortgage versus nonmortgage debt, and corporate versus noncorporate debt. As shown in Table 1, private nonfinancial debt growth declined sharply to about 3 percent, at an annual rate, over 1989-92 from long-term trend rates of 9 1/2 to 10 1/2 per cent. Both businesses and households experienced large debt slowdowns, but the rate of decline was much greater for the business sector. Nonfinancial business sector debt growth averaged less than 1 percent in the recent period, compared with a long-term trend rate of 10 percent, while household debt growth averaged 5.6 percent in the recent period, about one-half the average growth rate over 1982-89. In real terms, private nonfinancial debt actually declined somewhat over 1989-92 compared with trend rates of nearly 7 percent and 4 1/4 percent over 1982-89 and 196082, respectively. For both the business and household sec tors, real debt trend growth rates were significantly higher in the 1982-89 period than in the earlier period. Credit to the nonfinancial business sector declined by nearly 3 percent, on average, in real terms over 1989-92, following more than 6 percent average growth over 1982-89. The sharp declines in private and business debt growth in recent years have reversed the rising trends of ratios of private and business sector debt to GDP (Chart 2 and Table 1) despite a sustained period of weak growth of nominal GDP. With nonmortgage debt of both businesses and house holds slowing to about 2 percent at an annual rate over 1989-92, the greater decline in total business debt growth relative to household debt growth in recent years appears to be largely the result of differences in home and business m ortgage debt developm ents (Table 2). Home m ortgage debt advanced at a hefty 7 percent annual rate in the 198992 period, although its rate of growth decelerated substan tially from the historically high average growth rate over 1982-89. By contrast, business debt for real estate de ve l opm ent declined at an average annual rate of about 2 per cent during 1989-92, down from an average annual growth rate of close to 10 percent in the earlier period. In real terms, both m ortgage and nonmortgage com po nents of business debt declined significantly in the 1989-92 period. But businesses have experienced a much sharper decline in credit flow s for m ortgages than for other activity in recent years. Recent business debt developm ents have also differed significantly by the size of borrowers. As a group, large or corporate business borrow ers fared better than sm all or no ncorporate b o rro w ers in the recent cre d it slow dow n. Credit to corporate borrowers increased at an annual a ve r age rate of nearly 2 percent during the last three years, down from an 11.3 percent average increase over 1982-89 (Table 3). By con trast, n o n co rp o ra te bo rro w e rs e x p e ri enced an outright credit decline of 1.3 percent, at an annual rate, in the 1989-92 period, com pared with growth rates of about 11 percent in 1982-89. It is interesting to note that noncorporate borrowing is the only category among those reported here tha t show ed sig n ific a n tly low er re a l debt growth in the 1982-89 period than in the earlier period. broadly spread across depository (banks and thrifts) and nondepository credit sources (Table 4). Banks and thrifts, however, experienced a sharper decline in credit growth over 1989-92 than did o verall nondepository credit growth. Total depository credit actually declined at an annual rate of about 2 percent over 1989-92 following 9.3 percent aver- Chart 2 Ratios of Debt to GDP Ratio 1.4 Household and Nonfinancial Business Debt Bank a nd nonbank credit sources The slowdown in private nonfinancial debt growth was IS1 Table 1 Nonfina Fourth Qu Household debt/GDP -Fourth Quarter Percent Change, An Total Private Nonfinancial 8.6 11.0 5.2 9.6 10 6 3.1 10 .0 10 .1 0 .7 9.2 11.1 5.6 Constant 1987 Dollarst 1960-82 3.1 1982-89 7.2 1989-92 1.7 4.2 6.8 -0 .4 4 .5 6 .3 -2..8 3.8 7.3 2.0 Ratio of Debt to GDP 1960-82 0.2 1982-89 3.5 1989-92 0.3 1.2 3.1 -1 .8 1, 6 2. 6 -4..2 0.8 3.6 0.6 31..4 33.9 --------------------- .___ Current Dollars 1960-82 1982-89 1989-92 Nonfinancial business debt/GDP Memo: 1992-IV 100.0 current dollar share of total nonfina nonfinancial debt 65.3 P m ■ r n t GDP deflator was used to construct constant dollar series. 1.3 1960 Sources: Board of Governors of the Federal Reserve System, Flow of Funds Accounts; U.S. Department of Commerce. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 5 Table 2 Private Nonfinancial Debt Fourth Quarter-over-Fourth Quarter Percent Change, Annual Rate Mortgage Private Nonfinancial Total 9.6 10.6 3.1 Constant 1987 Dollars1 1960-82 4.2 1982-89 6.8 -0 .4 1989-92 Current Dollars 1960-82 1982-89 1989-92 Memo: 1992-IV current dollar share of private nonfinancial debt 100.0 Nonmortgage Business Home Mortgage Total Business Household 9.6 10.9 4.2 10.3 9.7 -1.9 9.3 11.5 7.1 9.6 10.3 2.0 9.9 10.3 1.9 9.1 10.3 2.3 4.2 7.1 0.7 4.8 5.9 -5 .4 3.8 7.8 3.6 4.2 6.5 -1.4 4.4 6.5 -1 .6 3.7 6.5 -1.1 50.2 14.3 35.9 49.8 33.8 16.0 + Based on GDP deflator. age growth over 1982-89, while total nondepository credit growth slowed to a 7 percent average rate in the recent period from about 12 percent in the preceding period. Both depository and nondepository credit growth rates are, of course, much lower on a constant dollar basis. At this level of aggregation, the bulk of the deceleration in private n onfi nancial credit growth over 1989-92 relative to the 1982-89 average rate is accounted fo r by depository sources, with both banks and thrifts making substantial contributions to the slowdown. The outright decline in total depository credit over 198992 reflects, to a considerable extent, the collapse of the savings and loan industry. In fact, the com m ercial bank credit com ponent— w hich represents about 70 percent of total depository credit— advanced at a 2 percent average annual rate over the 1989-92 period, com pared with a long- Table 3 Nonfinancial Business Debt Fourth Quarter-over-Fourth Quarter Percent Change, Annual Rate By Size of Borrower By Type of Borrowing Total1 Large* Smali§ Mortgage Other 8.7 11.3 1.8 14.1 10.9 -1 .3 10.3 9.7 -1 .9 9.9 10.3 1.9 dollars1 4.5 3.3 7.5 6.3 -1.7 -2.8 8.5 7.1 -4 .8 4.8 5.9 -5 .4 4,4 6.5 -1 .6 31.4 15.0 14.3 33.8 Current dollars 10.0 1960-82 1982-89 10.1 0.7 1989-92 1960-82 1982-89 1989-92 48.1 current dollar share of private nonfinancial debt t All corporate and noncorporate debt. * Corporate sector, excluding farm debt. 5 Nonfarm, noncorporate debt. 11 Based on GDP deflator. Digitized6for FRBNY FRASERQuarterly R eview /W inter 1 9 93-94 Table 4 Nonfinancial Private Credit Growth Fourth Quarter-over-Fourth Quarter Percent Change, Annual Rate Depository Nondepository Bank Depository Bank Credit Credit Credit Loans Loans Current dollars 1960-82 9.7 1982-89 9.3 1989-92 -2 .0 Constant 1987 dollars1 1960-82 4.2 1982-89 5.5 1989-92 -5.4 Memo: 1992-1V 39.9 current dollar share of private nonfinancial debt 9.6 11.8 7.0 10.1 10.1 2.0 9.7 9.0 -2 .7 10.3 9.9 1.1 4.1 8.0 3.5 4.7 6.3 -1 .5 4.3 5.3 -6.1 4.8 6.1 -2 .4 60.1 28.1 36.3 25.5 t GDP deflator was used to construct constant dollar series. term trend rate of around 10 percent. This m odest bank credit growth was more than fully offset, however, by a 45 percent (13 1/3 percent at an annual rate) decline in credit by savings and loan associations. W hile overall nondepository credit growth has held up better than overall depository or bank credit grow th, many co m po nents of n o n d e p o s ito ry cre d it did not fare m uch b e tte r th a n b a n k c re d it. A s e x p la in e d in th e C a n to r/ R odrigues study, credit grow th to businesses experienced roughly sim ilar slow dow ns in com m ercial paper, finance com pany lending, and bank loans in recent years relative to e arlier trends. Com paring the contribution of depository and nondeposi tory sources to business credit developm ents reveals that banks and thrifts accounted for about four-fifths of the fall in b u sin e ss m o rtg a g e d e b t g ro w th in 1989-92 re la tiv e to 1982-89 (Table 5). The slowdown in nonm ortgage business debt in the recent period relative to the earlier period was som ew hat more evenly divided betw een dep o sito ry and nondepository sources. For the nonfinancial business se c tor as a w hole, m ost of the deceleration in the average c re d it g ro w th from the 1 9 82-89 p e riod to the 1989-90 period reflected the slowdown in depository credit; banks accounted for som ew hat more than one-half of the deposi tory contribution. On the household side, the collapse of the savings and loan industry and the lending slow dow n by o th e r th rifts were responsible for m ost of the slowdown in home m ort gage debt growth in 1989-92 relative to 1982-89. The pace of com m ercial bank credit flow s for home m ortgages actu ally picked up som ew hat during the 1989-92 period. Banks, however, made the largest contribution to the slowdown in nonm ortgage household credit, accounting fo r more than half of the total slowdow n in that com ponent. S elected aspects o f bank business loans Data reported above clearly indicate that com m ercial banks have played a m ajor role in the 1989-92 credit slowdown for both b u s in e s s m o rtg a g e s and n o n m o rtg a g e b u s in e s s loans. For the nonfinancial business sector as a whole, the slowdown in bank loans accounted fo r more than one-third of the deceleration in average credit growth from 1982-89 to 1989-92. Both large (corporate) and sm all (noncorporate) b usi ness borrowers from banks experienced outright declines in bank loans over 1989-92, but the rate of decline was con siderably g re a te r fo r noncorporate borrow ers (Table 6). Specifically, over the 1989-92 period, nonm ortgage bank loans to noncorporate borrowers declined at a 4 1/2 per cent annual rate, more than tw ice the pace of decline for corporate borrowers. In the absence of bank loan sales, bank credit flow s to bu sin esse s w ould p ro b a b ly have been even w e a k e r in recent years. The study by Demsetz indicates, however, that adjustm ents for bank business loan sales to nonbanks and nonfinancial institutions over the 1986-92 period a ctu ally increase the severity of the recent slowdown in com mercial and industrial loans on banks’ books because busi ness loan sales have decreased in recent years. (Note that the flow of funds data fo r nonfinancial borrowers reported here already incorporate loan sale adjustm ents.) Even so, the liquidity provided by loan sales and securitization has most likely enabled banks to maintain higher levels of total loan origination than w ould have been the case otherwise. Cantor and Rodrigues point out in their study for this vo l ume that m ortgage-backed securities have grown about 70 . : 1 Table 5 Contributions to the Credit Slowdown mMmm From 1982-89 to 1989-92 |g|j - Business : '' WSB3M Household Mortgage Other Total Mortgage Other Total Decline in credit growth rate* 11.6 8.4 9.4 4.4 6.0 5.6 Percent of total decline contributed by: Depository sources 82.8 58.3 69.1 84.1 67.5 78.6 38.8 21.4 37.2 -6 .8 55.0 23.2 44.0 36.9 31.9 90.9 12.5 55.4 Nondepository sources 17.2 41.7 30.9 15.9 1 Annual average credit growth rate over 1982-89 minus annual average growth rate over 1989-92 # i 32.5 21.4 Banks Thrifts n FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 7 percent since 1988 and that securitization of business and consum er credit has proceeded even more rapidly over that period.4 Clearly, recent sharp advances in securitiza tion have, to some extent, cushioned the credit slowdown. As described in detail in the study by Lown and W en ninger, the bank credit slowdow n was spread fairly broadly across various regions of the country, but N ortheast (New England and M id-Atlantic) and Pacific regions experienced ve ry large o u trig h t d e clin e s in total and b u sin ess bank loans over 1989-92. O ther regions also experienced con tractions in com m ercial and industrial loans, although in some cases the rates of decline were relatively modest. W ithin the banking system , the bulk of the recent bank c re d it slo w d o w n is a ttrib u ta b le to d o m e s tic b a n k s as opposed to foreign banking offices in the U nited States (C hart 3). Total loans of U .S .-chartered banks show ed less than 1 percent annual average growth over 1989-92, and business loans actually declined outright at a 4.5 p e r cent annual rate. By contrast, total U.S. loans of foreign b a n k in g o ffic e s in th e U n ite d S ta te s a d v a n c e d at an annual rate of about 14 percent over the recent three-year period, only slightly below the average increase over the 1982-89 period. B usiness loans by foreign banking offices did register a significa n t slow dow n in the recent period, 4 Cantor and Demsetz (1993) show that over the two years to the second quarter of 1992, the growth in loans for home mortgages, consumers, and businesses inclusive of off-balance-sheet lending (securitization and loan sales) exceeded the growth in loans on the books of banks, thrifts, mortgage companies, and finance companies as a group. Table 6 Nonfinancial Business Loans by Banks but they continued to increase at a hefty annual pace of about 9 percent. These trends in foreign bank loans to U.S. borrowers are analyzed in more detail by Rama Seth in her study fo r this collection. She finds that as a group, foreign banks su p p o rte d to ta l U .S . c re d it g ro w th d u rin g th e re c e s s io n , although many foreign banks, especially those from Japan, Italy, and the United Kingdom, cut back on loans over that period. W hile Seth is unable to provide a full accounting of the continued strong loan grow th at fo re ig n banks, she notes that their desire to increase m arket share and their capital strength may have been im portant in m aintaining the relative strength of foreign bank lending. The differing patterns of loan developm ents for foreign relative to dom estic banks have substantially reduced the dom estic bank shares of total and business loans (C hart 3). Moreover, the flow of funds data used here understate the extent of foreign bank loans to U.S. residents because o ff shore foreign banks’ U.S. lending is excluded (M cCauley and Seth 1992). Adjusted for offshore data, the true shares of U .S .-c h a rte re d banks are c o n s id e ra b ly s m a lle r than shown in Chart 3. Factors behind the credit slowdown Studies in this volum e investigate dem and and supply fa c tors underlying the slowdown in private nonfinancial debt for both bank and nonbank sources of credit. The evidence includes descriptive and econom etric analysis and is based on hard data as well as survey m aterials for borrowers and lenders. On the dem and side, the studies look fo r both cyclical effects— the credit slowdown viewed as a by-prod uct of the econom ic slow dow n— and noncyclical demand influences. On the supply side, the evidence for both price and nonprice rationing of credit is considered. Fourth Quarter-over-Fourth Quarter Percent Change, Annual Rate Nonmortgage Business Loans Total Current dollars 1960-82 10.6 1982-89 9.9 1989-92 -1.7 Totalf Large Small Business* Business^ Mortgages 10.3 7.2 -2 .3 10.0 8.0 -2 .2 14.0 7.1 -4 .5 12.0 16.5 -0 .7 Constant 1987 dollars'1 1960-82 5.2 4.9 3.5 1982-89 6.1 -5 .7 1989-92 -5 .2 4.5 4.3 -5 .6 8.5 3.3 -7.9 6.5 12.7 -4 .2 8.7 6.9 1.4 5.0 Memo: 1992-IV 13.7 current dollar share of private nonfinancial debt I * 5 II All corporate and noncorporate business. Nonfarm corporate business. Nonfarm, noncorporate business. Based on GDP deflator. Digitized8for FRBNY FRASERQuarterly R eview / W inter 19 9 3 -9 4 C yclical a nd noncyclical dem and influences At an im pressionistic level, the recent credit slowdown can not be fully explained by the 1990-91 recession and the slow grow th period s u rro u n d in g the recession. S everal s tu d ie s in o u r c o lle c tio n — e s p e c ia lly th o s e by C a n to r/ Rodrigues, Low n/W enninger, and M osser/S teindel— p ro vide noneconom etric data analysis of cyclical effects on var ious debt or credit com ponents. The general thrust of the authors’ analysis of cyclical effects is captured by data in Table 7, although collectively these studies cover a much broader range of issues and detail. Briefly, the growth rate of private nonfinancial debt in nominal and real term s has been s u b s ta n tia lly lo w e r in the p e riod su rro u n d in g the recent recession than over com parable periods for the four earlier major recessions, on average, or considered individ ually. Broadly, this pattern holds for m ajor aggregate bor rowing com ponents and fo r both bank and nonbank credit. The only significant exception is the flow of home mortgage debt from both bank and nonbank sources, w hich has been age in the earlier cycles. Nevertheless, as pointed out by Low n/W enninger and others, the differences in the pace of activity do not fully explain the sharp credit slowdown in the current episode relative to the earlier episodes. Moreover, the credit w eakness itself may be responsible, in part, for the slower pace of econom ic activity in the current cycle. With changing relationships between credit flow s and eco nomic activity, it is very difficult to assess the contribution of w eaker than average grow th in the current cycle to the sig n ifica n tly stronge r in real term s over the period su r rounding the latest recession than around the last three major recessions since 1970. The com parison of credit flow s reported in Table 7 proba bly understates, to some extent, the contribution of cyclical developm ents to the private credit slowdown around the current recession relative to the earlier episodes. As shown in Chart 4, the pace of econom ic activity, nominal and real, was w eaker in the current cycle than it had been on a ve r Chart 3 Comparison of Domestic and Foreign Bank Loan Shares Percent Percent 18 102 ------------ r --------------------Share )f Loans to Total Biin k Loans r * “ 1" “ n V U.S. char tered loarIS/ bank loans - Scale i to ta l Loans Fourth Quarter-over-Fourth Quarter Percentage Change at an Annual Rate U.S. Banks Foreign Banks 1961-82 9.8 18.6 1983-89 1990-92 8.7 0.6 15.0 lc bank loa IS Scale----/ \ . / / 13.9 / / / Vv w- /• __ _________ x '- ' il l Llll 111111111111 LLiJ.Jll j j i i i i i l m i i i i l l l l m l m l m ............ l l l l l l l l l l l l l l l ,1,11 u l i n t Im l i i i l i i i i i i l j j . i L n l i i i I ii I lllllllllll Share o f Business Loans to Bank B usiness Loans U.S. chartered business loans; bank business loans ^ Foreign bank business loans/ bank business loans Scale------ ► ; a M t -«------- Scale B usiness Loans Fourth Quarter-over-Fourth Quarter Percentage Change at an Annual Rate U.S. Banks Foreign Banks 1961-82 9.7 23.6 1983-89 4.9 14.5 1990-92 -4.5 9.2 Ir H t r ri- 111111i 111111111111 1959 Source: Board of Governors of the Federal Reserve System, Flow of Funds Accounts. Note: Shaded areas indicate periods designated recessions by the National Bureau of Economic Research. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 9 severity of the credit slowdown. But one simple w ay to get a very rough sense of this contribution is to use the average relationship between real credit flow s and econom ic growth fo r the e a rlie r cycle s as a b e n ch m a rk to c a lc u la te the implied credit flow s associated with recent growth perfor mance. This type of exercise suggests that the w eaker than average pace of econom ic activity accounts fo r only about 35 percent of the gap between the private credit growth in the current cycle and the average private credit growth in the past four cycles. Some noncyclical or structural demand shifts may also have c o n trib u te d to re d u cin g the dem and fo r c re d it in recent years. Such shifts are “perm anent,” by definition, but their influence on dem and may be difficult to separate from that of cyclical forces. Some studies in this volum e note the relevance of structural dem and shifts in recent d e ve lop ments in credit flow s. In particular, the Low n/W enninger and M osser/Steindel papers discuss the influence of a pos sible dow nward shift in inventory demand relative to sales, especially in the m anufacturing sector, on the dem and for com m ercial and industrial loans. Because of ju st-in-tim e and other m anagem ent techniques, the amount of invento ries needed for a given level of sales and, therefore, the financing requirem ents for those inventories have declined in recent years. Even though such a shift is likely to have been gradual and to have started before the recent credit slowdown, a considerable portion of the unusual w eakness in com m ercial and industrial bank loans over the recent period may be explained, Lown and W enninger argue, by the need to finance a lower than normal level of inventories. Econom etric analysis yields results that are broadly con sistent with the less form al data analysis, namely, demand influences as reflected in standard m acroeconom ic v a ri ables are unable, by them selves, to explain adequately the recent credit slowdown. At the outset, it is worth noting that the estim ates discussed here generally do not distinguish between cyclical and noncyclical dem and influences. The estim ated equations sim ply attem pt to explain particular cre d it flo w s using a g g re g a te dem and co m p o n e n ts and other appropriate m acroeconom ic factors as explanatory variables. M ovem ents of explanatory variables, in this co n text, capture all relevant normal or long-run influences on credit flows. Using cash flow and income or aggregate dem and com ponents as e xp la n a to ry v a ria b le s, M osser and S teindel estim ate total loan equations fo r nonfinancial corporations, consum ers, hom e m ortgages, and business m ortgages. They find that swings in econom ic a ctivity-re la te d fu n d a m entals seem to account fo r only about one-quarter to onehalf of the slowdown in corporate and consum er borrow ings. In the case of consum er credit, the authors reestim ate equations by adding home equity lines to take account of shifts betw een co n su m e r cre d it and hom e e q u ity loans resulting from the Tax Reform Act of 1986; the results are roughly sim ilar to those w ithout the home equity variable. For business and home m ortgage com ponents, estim ates are unstable, although for home mortgages, the estim ated equations are able to explain the recent slowdown in loans. M osser and Steindel provide a particularly detailed a n aly sis of corporate and consum er loans, and argue that most of the p re d ictio n errors fo r those loans do not seem to reflect any exogenous shift in the relationships betw een credit dem and and explanatory variables. For bank loans, Lown and W enninger estim ate four sets of equations, one each for com m ercial and industrial loans, b u s in e s s m o rtg a g e s, hom e m o rtg a g e s, and c o n s u m e r loans. The equations are estim ated with vector autoregres sion m ethodology to approxim ate reduced-form relation- Table 7 Credit Growth over Various Business Cycles .’ V ■. V ; . ' -.^0% Private Business Nonfinancial Nonfinancial ....................................................... Average, current cycle Average, earlier cycles(A)t Average, earlier cycles(B)* Constant 1987 dollars§ Average, current cycle Average, earlier cycles(A)t Average, earlier cycles(B)* 3.1 8.8 9.0 -0.4 3.9 3.2 'E m . Mortgage Household Business Home Mortgage Nonmortgage 0.7 9.2 9.7 5.6 8.5 8.3 -1 .9 10.2 10.6 7.1 8.5 8.2 2.0 8.6 9.0 -2 .8 42 3.9 2.0 3.5 2.5 : -5.4 5.2 s-r- -" - ---j 4.7 > v -(llE i 3.6 3.6 2.4 -1 .4 3.7 32 pgp sNilPi Note: Business cycle periods cover four quarters before trough, trough quarter, and seven quarters after trough. f Average of the 1958, 1970, 1975, and 1982 cycles. * Average of the 1970, 1975, and 1982 cycles. 5 Based on GDP deflator. Digitized 10for FRASER FRBNY Quarterly R eview /W inter 1 9 93-94 llB B i ships, using a range of econom ic activity and interest rate variables. Broadly, the estim ated equations fo r business m ortgages and consum e r loans u n d e rp re d ict the cre d it slowdown, while those for home m ortgages more than fully account for the extent of the slowdown. For com m ercial and industrial loans, Lown and W enninger are unable to reach any firm conclusions because of unstable regressions. C antor and Rodrigues estim ate equations fo r total bank b u sin ess loans and fo r n o n b a n k b u sin ess c re d it using GDP, investm ent, and inventories as explanatory variables. The pre diction errors from both the bank and nonbank equations are large, indicating that m acroeconom ic activity Chart 4 Economic Activity in Various Business Cycles Four quarters before trough = 100 200- -3 -2 -1 Trough Quarters before trough 3 1 Source: U.S. Department of Commerce. 4 Quarters after trough variables do not provide an adequate explanation for the slowdown in either bank business lending or nonbank busi ness credit. In summ ary, aggregate dem and influences are unable to explain a substantial part of the recent slowdown or decline in nonfinancial business borrow ings from bank and non bank sources; this is true fo r both m ortgage and nonm ort gage bu sin ess b o rro w in g s. Dem and fa cto rs also fa il to account fo r the recent slowdown in consum er credit, and taking account of shifts between consum er credit and home equity loans does not significantly alter this result. Recent developm ents in total home mortgage debt and home mort g a g e b a n k lo a n s , h o w e v e r, a p p e a r to be a d e q u a te ly explained by the evolution of aggregate demand influences. Supply-side factors W ith a significant fraction of the credit slowdown left unex p la in e d by s ta n d a rd a g g re g a te dem and v a ria b le s , one m ust turn to the supply side. Indeed, the prediction errors or residuals from equations estim ated with dem and v a ri ables may be viewed as representing one measure of the supply-side influence on the credit slowdown. Of course, even if we could account for all of the recent credit slow down with the help of dem and variables, that result by itself would not necessarily im ply that supply-side factors did not contribute im portantly to the credit slowdown. Such a result m ight sim p ly reflect, fo r exam ple, the fa ct th a t dem and influences overw helm supply-side factors. More generally, with both credit dem and and supply falling, if the drop in credit dem and is larger, actual cre d it d evelopm ents will tend to be dom inated by dem and influences, making it d iffi cult to estim ate the net contribution of supply-side factors. Four studies in this collection— Low n/W enninger, C antor/ Rodrigues, Johnson/Lee, and H am dani/R odrigues/Varvatsoulis— have devoted considerable attention to the role of supply-side factors in the credit slowdown. T heir analysis covers bank and nonbank sources of cre d it and survey data. Overall, the evidence points to significant credit su p ply problem s fo r both bank and nonbank sources of credit. On the bank side, Lown and W enninger look at a number of su p p ly-sid e fa cto rs and provide both d e scrip tive and econom etric evidence on the role of those factors. They find that in the 1989-92 period, spreads between bank lend ing rates and bank funding costs for both corporate and consum er loans were at or above their previous record lev els. They also note that the percentages of short- and long term lo a n s re q u irin g c o lla te ra l in c re a s e d s h a rp ly o ve r 1989-92. Both indicators are consistent with a leftward shift in the bank loan supply schedule. O th e r n o n e c o n o m e tric e vid e n ce d iscu sse d by Low n/ W e n n in ge r and o thers sug g e sts th a t banks engaged in nonprice credit rationing or, more generally, experienced re d u c e d a b ility o r w illin g n e s s to le n d . B a n ks s h a rp ly increased their holdings of securities relative to loans, and FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 11 some of the increase appeared to be noncyclical.5 Survey data from banks indicate significant tightening in credit standards on mortgages and other business loans during 1989-92. Weakening bank capital positions— reflecting, in part, deteriorating bank loan quality and increasing charge-off rates— seem to have played a significant role in credit sup ply problems over 1989-92. Lown and Wenninger argue that poorly capitalized banks reduced their lending more sharply than well-capitalized banks during 1990-91. Draw ing on a more comprehensive examination of the relation ship between bank capital positions and bank credit, John son and Lee reach a somewhat stronger conclusion along the same lines. Specifically, the results indicate that banks with weak capital positions did less lending than banks with strong capital positions during the 1990-92 period. Lown and Wenninger also argue that the increased emphasis by the regulators on bank capital and the riski ness of bank loan portfolios may have contributed to the bank loan slowdown, although the role of the regulators and examiners is difficult to separate from other factors. While Lown/Wenninger and Johnson/Lee explore the effects of capital positions on bank lending, none of the studies in this volume explicitly investigate the role of regu lators and regulatory changes in the credit slowdown process.6 Using state-level data, Lown and Wenninger estimate cross-sectional regressions for bank loan growth with employment, capital, and loan-loss reserves as indepen dent variables; the latter two variables are intended to cap ture the effect of banking conditions (that is, supply-side factors) on loan growth. The results suggest that capital and/or loan-loss reserves contributed significantly to weak bank lending in 1990 and 1991 and that the effects of these supply-side factors were greatest for the New England region, followed by the Mid-Atlantic and the West South Central regions. By applying the cross-sectional regression coefficients to changes in the explanatory variables by region, Lown and Wenninger provide a quantitative sense of the contribution of supply-side factors to the overall bank credit slowdown. Specifically, they suggest that supply-side problems accounted for roughly 15 to 40 percent of the slowdown in bank lending from 1989 to 1990. Also using cross-sectional data, Demsetz estimated equations for bank loan sales with expected economic activity, assets, capital ratios, nonperforming loan ratios, and other bank characteristics as explanatory variables. 5 More formally, Rodrigues (1993) shows that weak economic activity cannot explain all of the recent run-up in securities holdings and that the sustained steepness in the term structure of interest rates and risk-based capital standards may have contributed to that run-up. 6 For various perspectives on the role of regulators/examiners and capital standards, see Greenspan (1992), Syron and Randall (1992), Peek and Rosengren (1992), LaWare (1992), and Wojnilower (1992b, 1993). FRBNY Quarterly Review/Winter 1993-94 Digitized 12 for FRASER She finds that both capital ratios and nonperforming loan ratios are significant in explaining loan sales but their con tribution to predictions of loan sales declines is modest and swamped by that of economic activity. Turning to nonbank credit sources, the Cantor/Rodrigues study offers evidence that supply-side forces were at work here as well. The authors’ econometric estimates for non bank business credit using GDP and its components as explanatory variables yield large prediction errors that sug gest a significant role for supply-side factors. The results also indicate that the timing of the credit slowdown for non bank sources was parallel to that for bank sources, with no evidence of a shift from bank to nonbank sources of funds. Cantor and Rodrigues also provide considerable descrip tive evidence on the role of supply-side factors in the slow down of credit from nonbank sources such as finance com panies, life insurance companies, and the commercial paper market. Business credit extended by finance compa nies advanced at a significantly slower pace starting in late 1989, when many finance companies were downgraded by the credit rating agencies because of major losses in com mercial lending and, more generally, weak balance sheet positions. With more credit downgrades during the reces sion and large amounts put up for loan loss provisions and net charge-offs, total finance company business credit became roughly flat over 1990-92. Cantor and Rodrigues note that credit downgrades probably had a significant effect on lending because finance companies raise most of their funds in short-term public credit markets. The authors also suggest that credit stringency at banks may have had adverse feedback effects on finance company credit avail ability as many finance companies, faced with problems in raising funds in the commercial paper market, increased their borrowings from bank backup credit lines, presumably at higher costs. Most of the problems of the life insurance industry, Can tor and Rodrigues argue, stemmed from commercial real estate lending, junk bond portfolios, and high rates on guar anteed investment contracts. Against the background of weak economic activity, these difficulties led to numerous credit downgrades, sharp declines in stock prices, and some outright failures in the life insurance industry. Life insurers became generally preoccupied with preserving liq uidity and avoiding a collapse. In this environment, the National Association of Insurance Commissioners in mid1990 adopted new rules establishing more stringent re serve and capital requirements for below-investment-grade bonds and private placements. These developments, Can tor and Rodrigues believe, have reduced the willingness of insurance companies to invest in below-investment-grade bonds and, more generally, have induced a shift toward low-risk assets. Nonfinancial business borrowers did not increase the rate of commercial paper issuance during the latest credit crunch, as they had done in earlier credit crunches. Because of numerous credit rating downgrades and fifteen defaults since 1989 (compared with only two defaults in the entire earlier history of the market), perceived credit risk in the commercial paper market increased greatly, leading investors, especially mutual fund investors, to lose confi dence. Meanwhile, to protect small investors and sustain confidence in the money market mutual fund industry, the Securities and Exchange Commission in July 1990 imposed strict limits on the amount of “second-tier” (low-quality) com mercial paper that mutual funds could hold. As a result of these developments, both the amount of second-tier com mercial paper issued and the mutual fund holdings of that paper dropped precipitously over 1990-92. Cantor and Rodrigues believe that the credit quality concerns are not fully reflected in the rate spread between the top-tier and second-tier paper because the second-tier issuers are often “rationed” out of the market before they drive up rates. Cantor and Rodrigues also discuss the public bond mar ket. The market for below-investment-grade public bonds (“junk bonds”) showed virtually no activity during 1990 and 1991 but recovered significantly in 1992. By contrast, the market for publicly placed investm ent-grade bonds remained quite strong, cushioning weakness in other credit markets to some extent.7 Survey evidence on supply-side factors Hamdani, Rodrigues, and Varvatsoulis examine survey data from bank lenders and nonfinancial borrowers on credit tightening in recent years. Using both the narrative approach and econometric estimates, they find evidence of significant credit tightening by lenders because of supplyside factors. By purging the NFIB (National Federation of Independent Business) Survey data of aggregate demand influences, they uncover particularly strong and consistent evidence of a credit crunch for small business borrowers that depend primarily on banks for their financing (about 90 percent of small business debt consists of bank loans).8 The results indicate that for small borrowers, the recent credit crunch was more severe than earlier crunches. A sig nificant part of this credit crunch appears to have taken the form of nonprice credit rationing or tightening of nonrate loan terms. Hamdani, Rodrigues, and Varvatsoulis also find consid erable evidence of credit supply constriction for large bor rowers. They conclude that overall, the extent of bank 7 The severity of credit supply reductions, as noted earlier, has also been moderated somewhat by rapid increases in off-balance-sheet lending (securitization and loan sales) in recent years. 8 In fact, the authors’ credit supply proxies, purged of aggregate demand influences, may understate the extent of credit supply shifts because they exclude supply shifts associated with movements of lending spreads and at least some of the effect of changes in borrowers’ quality on the willingness to lend. credit tightening for large businesses appears to have been greater than what can be explained by the general eco nomic slowdown. Using the SLO (Senior Loan Officer) sur vey data from banks, again purged of aggregate demand influences, the authors argue that the degree of credit strin gency during 1990-91 seems to have been similar to that in the 1974-75 episode. Finally, Hamdani, Rodrigues, and Varvatsoulis estimate loan growth models using standard demand variables and survey variables on loan availability for both the SLO and NFIB surveys. The results suggest that restrictive loan sup ply conditions as proxied by the survey supply variables have had a significant impact on commercial and industrial bank loan growth over 1989-92. Correction for the debt overhang of the 1980s As noted earlier, disentangling the supply and demand fac tors underlying the recent credit slowdown is particularly difficult because the economic downturn was superim posed on a process of balance-sheet corrections for debt excesses of the mid-1980s. This process of correction for earlier debt excesses is widely believed to have contributed significantly to the credit slowdown over 1989-92. During the last decade, a broad range of forces— includ ing financial deregulation and innovation, developments in information and data processing technology, commercial real estate development, and mergers, acquisitions, and leveraged buyouts— combined to increase greatly both the supply of and the demand for credit, resulting in enormous increases in the amount of debt.9 The upward march of debt was supported, in part, by speculative asset price increases, especially for real estate. Over time, the process of rapid debt increases led, per haps inevitably, to problems for both borrowers and lenders. By 1989 and 1990, households and businesses faced historically unprecedented and unsustainable debt and debt service burdens (Chart 5). With weakening eco nomic activity and declining real estate and other asset val ues, high debt burdens resulted in balance sheet difficulties for borrowers and loan quality problems for lenders. Not surprisingly, therefore, bank and nonbank lenders alike experienced a weakening of capital positions and increas ingly higher loan loss reserves, charge-offs, and delin quency rates. All these factors together, so the argument runs, explain the sharp credit slowdown in recent years. This account of the correction process is consistent with the view that the credit slowdown contained important sup ply-side elements although it was perhaps driven by demand forces. In particular, in the down-phase, balance sheet changes induced by declining real estate and other asset values led to weaker capital positions for banks and, 9 For a review of developments leading up to the credit crunch period, see Cantor and Wenninger (1993). For a broad perspective on the debt overhang of the 1980s, see Frydl (1991). FRBNY Quarterly Review/ Winter 1993-94 13 consequently, lower capacity and w illingness to lend over 1989-92, just as on the up-side, balance sheet changes had increased capacity and w illingness to lend in the earlier period. The le n d ers’ reduced w illingness to lend, in this case, re fle cte d not on ly ch a n g e s in th e ir ow n b a la nce sheets but also a shift in their attitude associated with the deterioration, actual or perceived, in the quality of borrow ers’ balance sheets and creditw orthiness. Perhaps even more im portant, according to this story, the correction process seem s to have been dom inated by m arket forces (both dem and and supply) as opposed to pol icy factors. In fact, m onetary policy had been easing since early 1989, and as a result, unlike earlier credit crunches, in te re st rates had d e clin e d s ig n ific a n tly before se rio u s credit supply problem s em erged. To be sure, tighter capital re q u ire m e n ts and re g u la to ry p re s s u re s ,s te m m in g from both legislative changes and more intensive supervisory oversight, contributed to the credit slowdown, in part by reinforcing and highlighting prudential concerns. Such p o l icy factors, however, appear not to have been the prim ary cause of the credit slowdown. In any event, any contribu tion of policy factors to the credit slowdown is likely to have been much sm aller than the role played by m arket forces; these forces, particularly evident in a reduced desire to bor row and hold or extend debt, caused a decline in both credit dem and and credit supply.10 Research w ork in this volum e does not provide any e sti m ates of the extent to which the credit slowdown is a ttrib u t able to the correction process for the debt overhang of the 1980s. W hile several studies discuss developm ents lead ing up to the credit slow dow n, quantitative assessm ents are generally aimed at sorting out demand from supply (or cyclical from noncyclical) factors using historical trends. The study by Johnson and Lee does address the related question of the linkage between the earlier credit excesses by banks and the recent bank credit slowdown. It finds that banks that indulged in “high-risk” activities during the 198588 period were obliged to curtail their lending more sharply than other banks during the three years to end-1992. But the study does not estim ate the extent of “excess d e b t” resulting from those earlier high-risk activities. Nevertheless, it may be useful to get a rough sense of the im pact of the correction for the debt overhang on the credit slowdown since 1989. S pecifically, I address the following question: W as the actual cum ulative expansion in private nonfinancial debt from end-1989 to end-1992 h ig h er or low er than w hat is co n siste n t w ith “norm al” or long-run trend credit growth adjusted fo r cyclical developm ents and for the debt overhang of the 1980s? Using the simple rela- tionship that the am ount of credit expansion in any given tim e period is made up of the credit expansion consistent with the normal or long-run trend rate adjusted fo r cyclical Chart 5 Debt Service Burdens Percent — ------------------- ------------------- ------------------- ... '... ----------H<>usehold Sec tor Debt Ser vice as a A P€‘rcentage of Disposable Personal Inco me Percent 8.0 1 7.5 L 7.0 / / \ 17.C Home mortgages . j SDale-----► J | Sm‘\.r 1 ! N ------ Scale 6.5 r / i l.i \ 5.5 A 15.! 1/ f 1 n / * 7' 15J 4.5 f, M 14.1 5.0 / ^ V, J tS 4.0 i i d w k k l i i ) u M u i i i i l m U lllllllllllllllll m i d 3.5 14.1 ii'f U lllllllllllllllll — ------------------- r- isiness Sect sr Net Interest Payments iis a o l rcentage of Cash Flow plu. Net Intere st Payments Pe 28 / 26 - k 24 22 A 20 / 18 6.0 \ r A J \\ \\ \ A n l [ J \ 16 s y uiluiliiMiiiiiliulwlmlmujI i.h1iii 1ih1j.il JuluJ 14'till mlmlmlmlm 1970 72 74 76 78 80 82 84 86 88 90 92 Source: U.S. Department of Commerce. 10 Incidentally, note that shifts in attitudes toward debt would normally be treated as “exogenous” in most macroeconomic models; the use of exogenous/endogenous in the current context, however, would appear to be inappropriate since such terms must be expressed relative to a specific model. Digitized14 for FRASER FRBNY Quarterly R eview /W inter 1 9 93-94 Notes: In the upper panel, debt service is an estimate of scheduled payments of principal and interest on home mortgage and consumer debt. In the lower panel, cash flow is defined as depreciation (book value) plus retained earnings (book value). and other shifts away from that trend, I attempt to measure the gap between the actual credit expansion over the threeyear period to the fourth quarter of 1992 and the amount of credit expansion implied by the adjusted long-run path under various assumptions for the relevant variables. If the actual credit expansion over 1989-92 falls short of the esti mated credit expansion for that period, the recent credit slowdown has been greater than what could be reasonably attributed to the combination of cyclical effects and the cor rection for earlier debt excesses. In this case, the correction process itself might have produced overshooting or shifts unrelated to the earlier credit excesses, and cyclical devel opments might have further depressed credit flows. Of course, a significant positive gap between the actual and the estimated credit expansion has the opposite implications. There is no obvious and definitive way to measure the “normal” or long-run credit expansion rate. The usual pro cedure is to use some measure of the historical trend rate. But with credit expansion rates much higher in the 1980s than in the preceding two decades, history does not offer a clear choice for the trend rate or the benchmark period. Perhaps more important, since long-run credit growth must be viewed in real terms, we need relevant prices. At an empirical level, however, the choice of the appropriate price measures needed to deflate various debt components is ambiguous. Similarly, the use of the debt-to-GDP ratio at the component level in figuring out the long-run or normal rate is quite problematical—the ratio of a particular debt component to GDP (or to broad sectoral income measures) need not be stable over time. Adjustment of the long-run trend to account for cyclical and noncyclical developments raises equally difficult questions: How should we measure cyclical effects? How much time should we allow for the correction of the debt overhang to be completed— as much time as it took to build up the problem, more time, or less? Using various alternatives for the long-run or normal trend credit expansion rate and adjustment factors, I calcu lated the cumulative amount of excess debt over 1982-89 and several measures of the gap between the level of actual credit expansion over 1989-92 and the amount of trend credit expansion, adjusted for the debt overhang and cyclical developments, during that period. One such exer cise is reported in Table 8. The long-run trend rates in this exercise are based on business and household data for mortgage and nonmortgage debt over the 1960-82 period, converted into constant 1987 dollars using the GDP defla tor.11 The cyclical effects are measured on the basis of dif ferences between the 1960-82 trend rates and the com 11 The use of a national price index instead of sectoral price indexes seems to be preferable for at least two reasons: appropriate component price measures are not always readily available, and even when they are available, their use would legitimatize credit excesses of the 1980s by incorporating any speculative price increases for particular sectors such as real estate. bined average growth rates for the periods surrounding the 1970,1975, and 1982 recessions. This exercise suggests that the decline in business credit over 1989-92 has gone far beyond what was necessary to correct the earlier debt excesses; only about 55 percent of the decline in business credit over 1989-92 relative to the long-run trend can be attributed to the need to correct the debt overhang. Combining the correction for the debt over hang with cyclical effects still accounts for only a part of the business credit slowdown. Even assuming complete adjustment over three years (1989-92) for the credit ex cesses that took place over seven years (1982-89), the actual business credit increase over 1989-92 fell short of the long-run trend expansion, adjusted for the debt over hang and cyclical effects, by about $246 billion in 1987 prices; the shortfall represents nearly 7 percent of total business credit at end-1992. Under partial adjustment, with three-sevenths of the excess debt eliminated over 1989-92, the debt shortfall from the trend expansion level increases to $461 billion, or about 12.5 percent of total business credit at end-1992. While both commercial mortgages and nonmortgage business debt declined more than implied by the estimated adjusted trend expansion levels under the two adjustment scenarios, the bulk of the shortfall reflects commercial mortgages. For the household sector, the correction for the earlier debt excesses and cyclical effects together more than fully account for the credit slowdown. In fact, actual household credit expansion over 1989-92 exceeded the amount of credit expansion consistent with the adjusted long-run trend, assuming complete adjustment over three years, by $665 billion in 1987 dollars; the excess is nearly 17 percent of total household debt at end-1992. About 90 percent of the excess debt is attributable to home mortgages. Under partial adjustment, the amount of household excess debt drops to less than half that under complete adjustment, but it is more than fully accounted for by home mortgages, with nonmortgage household debt actually showing a moderate shortfall relative to the estimated level. In sum, there has been no correction for the debt overhang for home mort gages. On the contrary, home mortgage debt over 1989-92 continued to advance at a faster rate than the long-run trend rate, apparently unaffected by cyclical developments and by the need to correct earlier debt excesses. Alternative measures of the long-run trend rate yield, in some cases, significantly larger or smaller estimates of the debt excess over 1982-89 and of the gap between actual and estimated debt changes over 1989-92. Two general messages of the results in Table 8 hold up, how ever. First, although the correction process for the debt overhang played a major role in the credit slowdown, it is difficult to explain all of the business credit slowdown by appealing to the need for correction. Second, home mort gage debt in recent years has remained immune to the FRBNY Quarterly Review/Winter 1993-94 15 co rre c tio n p ro ce ss fo r the e a rlie r debt e xce sse s. One im p lica tio n of the firs t p o in t is th a t som e c re d it supply shifts largely or com pletely unrelated to the m arket c o rre c tion process for the debt overhang may have played an im portant role in the cre d it slowdow n. Such supply shifts were presum ably caused by tig h te r capital standards and regulatory pressures. econom ic activity helped reduce the pace of credit growth in all three countries, but their role appears to have been relatively m odest in Japan and the United Kingdom. Finally, bank capital m ovem ents seem to be significant in e xplain ing credit m ovem ents in Japan and to a lesser extent in the United Kingdom , but they appear not to have made any contribution to credit developm ents in France. The credit slowdow n abroad A num ber of foreign countries have also experienced credit slowdowns, to varying extents, during the last three years or so. The H ickok/O sler study in this volume exam ines the foreign experience, focusing on Japan, France, and the United Kingdom. Since a single study cannot be expected to deal w ith all a sp e cts of th e fo re ig n e x p e rie n c e , the autho rs con sid er only the broad contours of the recent credit experience abroad and the common forces that may have driven that experience. Using both descriptive analysis and regression results, Hickok and O sier find that for all three countries, the w a n ing of the credit surge of the 1980s contributed im portantly to the credit slowdown during 1990-91. The broadly defined process of financial deregulation and innovation, working through expanded access to credit markets, asset va lu a tio n s, and o th e r chang e s, led to in cre a se s in both the dem and for and the supply of credit during the mid- and late 1980s. Subsequently, as actual credit changes adjusted to “perm anently” higher equilibrium levels, credit growth rates tended to return to more norm al levels. Hickok and O sier also find that for Japan and the United Kingdom , a reversal of the speculative factors played a considerable role in the credit slowdown. Developm ents in To the extent that the credit slowdown reflects the slow down in aggregate dem and or econom ic activity, it is a sym ptom and not a direct cause of the w eakness in the e conom y. A c c o rd in g ly , an y in v e s tig a tio n of the c o n s e quences of the credit slowdown for nonfinancial econom ic activity must focus on credit supply problems. In this vo l ume, three studies— Mosser, S teindel/Brauer, and H arris/ B oldin/Flaherty— deal with this subject. Overall, the three studies indicate that credit supply problem s have not been the prim ary or dom inant cause of the recent w eakness in econom ic activity. But collectively, the studies do suggest that credit constraints are likely to have m ade at least some contribution to the econom ic slowdown. Credit supply problems and economic activity A ggregate dem and M osser exam ines the effects of credit supply problem s on aggregate demand com ponents while attem pting to control for changes in credit demand. She estim ates reduced form equations for several dem and com ponents with and w ith out variables representing credit supply restraints. Four d if ferent proxies, all based on other studies in this volum e, are used for credit supply constraints: (1) regression residuals from various bank loan equations in W enninger/Low n, rep resenting part of the credit slow dow n not attrib u ta ble to Table 8 liv .' ..m m - Long-Run Trend and Actual Credit Expansion, 1989-92 Total Actual credit expansion Trend expansion Cyclical adjustment Correction for excess expansion over 1982-89 Adjusted trend credit expansion Excess/shortfall Partial adjustment Business tlrtl .n—.l~« 1 nuubenuiu a— i m M ortgage Home Mortgage Other Other Total Total Private -261.5 423.6 -62.9 -1 59.0 151.5 -2 .0 -1 0 2 .5 272.1 -60.9 191.6 284.9 -93.5 227.6 185.9 -67.8 -3 6 .0 98.9 -25.6 -6 9 .9 708.5 -156.4 -376.0 -1 5.3 -246.1 -461.0 -75.1 74.4 -2 33.4 -276.3 -301.0 -8 9 .8 -1 2 .8 -184.7 -665.1 -4 73.7 665.3 285.2 -479.9 -3 61.8 589.4 315.2 -185.2 -1 11.9 75.9 -29.9 -1041.1 -4 89.0 419.1 -175.8 Notes: Table reports changes in billions of 1987 dollars from 1989-IV to 1992-IV. Long-run trends are based on the 1960-82 growth rates of busi ness and household components. Cyclical adjustments are based on the differences between the 1960-82 trend rates and the com bined average growth rates for the periods surrounding the 1970, 1975, and 1982 recessions. Figures in the last row are estimated on the basis of partial correction (3/7) for the 1982-89 excess expansion over 1989-92. In current dollars, actual cumulative private credit expansion over 1989-92 was about $680 billion (11.0 percent of 1992 GDP). Sums may not add up precisely because of rounding. 16 FRBNY Quarterly R eview / W inter 19 9 3 -9 4 demand factors; (2) regression residuals from various sec toral loan equations in Mosser/Steindel, measuring the gap between actual credit flows and the estimates based on historical relationships between credit and aggregate demand variables; (3) residuals from regressions in Hamdani/Rodrigues/Varvatsoulis, capturing credit availability restraints for small business, purged of cyclical influences; and (4) interest rate spreads between market rates and loan rates on business and consumer lending. Using data for the 1980-92 period, Mosser performs some Granger-Causality tests to determine whether credit aggregates or credit supply proxies are statistically more significant predictors of future economic activity. Her results tend to favor credit supply proxies. For the more recent period, Mosser finds significant effects of credit sup ply problems on commercial real estate activity and produc ers’ durable equipment. In particular, the credit supply proxy for small business seems to account for a consider able part of the 1989-92 weakness in nonresidential con struction and producers’ durable equipment. Even so, Mosser argues that the weakness in these demand compo nents relative to predictions based on normal historical relationships cannot be fully explained by credit supply problems. Doubtless, the widespread sluggishness of eco nomic activity during 1989-92 reflected a broader set of fac tors than just credit supply problems. Construction activity Harris, Boldin, and Flaherty investigate the effects of credit supply problems on the real estate industry. Focusing on the three construction industry sectors— single family homes, multifamily housing, and nonresidential struc tures—they provide a comprehensive review of credit and noncredit factors underlying the recent decline in construc tion activity. Overall, their study finds that credit supply problems are likely to have played only a modest role in the real estate contraction. For single family housing, the authors begin by examin ing predictions of housing activity from several standard models that use mortgage rates, income, and other funda mentals as explanatory variables. Since these models are not able to predict the recent weakness in housing, the authors search for an explanation by focusing on “special” factors or other variables that have been left out of the mod els. They argue that of the missing variables, demand-side factors such as a generalized effort to reduce debt and an adverse shift in investor psychology rather than narrowly defined credit supply problems explain the bulk of unusual weakness in housing. This view is consistent with the fact that because of the mortgage-backed securities market and other financial innovations, credit supply for home mortgages has not experienced any significant problems. The supply of loans to homebuilders has been constrained significantly, but this appears not to have caused a perva sive housing shortage. Even so, credit supply problems may explain part of the recent weakness in housing activity since without credit constraints, the housing supply would have been larger and prices lower. More generally, given the weakness of both credit demand and credit supply, the identification problems make it difficult to rule out a signifi cant role for credit supply difficulties. Multifamily and nonresidential construction have de clined greatly since 1989 and have remained the two weak est sectors of the economy. According to the Harris/Boldin/ Flaherty study, overbuilding in the 1980s (together with the resulting excess capacity) dominates the credit crunch as an explanation for the collapse of activity in both sectors. The study recognizes, however, that these sectors have experienced credit supply problems and that the simultane ous weakness in (and interaction between) credit demand and credit supply makes it difficult to isolate the effect of credit supply constraints. It is likely that in the absence of credit supply constraints, the decline in the nonresidential and multifamily sectors would have been more moderate. Put differently, the credit crunch does not appear to be the dominant cause of the collapse in construction activity, but it may well have played some role in the timing and process of decline. Business activity excluding construction The Steindel/Brauer study explores the consequences of credit supply problems for business activity excluding con struction. Overall, this study provides only limited support for the view that credit supply problems impeded business activity over 1989-92. Steindel and Brauer consider five different types of evi dence. First, they review recent movements in corporate, noncorporate, and manufacturing activity, together with rel evant credit flows. The review suggests that the sharp slowdown in credit flows may have been a significant con tributing factor to weakness in small business activity and that such firms may have borne a disproportionate share of the shortfall in both output and debt. Second, the authors look at survey evidence on lending to smaller firms and the connection between credit supply proxies from other studies in this volume and noncorporate business output. This survey evidence does point to a sig nificant credit tightening which may have contributed to weakness in small business activity. Third, using detailed industry- and firm-level data, the study compares activity for small and large businesses and attempts to infer the role of credit in the recent weakness of small business activity. The focus is on manufacturing busi nesses, but the analysis does include some nonmanufac turing establishments as well. In most cases, small busi ness activity appears not to have shown any unusual weakness relative to large business activity, and so, by in ference, Steindel and Brauer do not find any more support FRBNY Quarterly Review/Winter 1993-94 17 for the effect of credit supply problems on small businesses than on large businesses. But with data on the relevant credit flows unavailable, this type of evidence is entirely indirect and does not necessarily contradict the view that credit supply problems may have contributed to the slow down in business activity over 1989-92. A fourth type of evidence considered by Steindel and Brauer focuses on indicators of financial strength. Again using industry- and firm-level data, the authors explore the role of financial factors in the recent weakness of business activity by examining various measures of real economic activity for financially “weak” and “strong” businesses. This evidence is also indirect and yields mixed results. Finally, using firm-level data, Steindel and Brauer per form formal regression tests to look for the effect of size and debt- to-asset ratios on employment, inventories, capi tal spending, and spending on research and development for various periods. Once again, the results are mixed. Implications for Monetary Policy In reviewing the implications of the credit crunch or credit supply problems for monetary policy, this section focuses on two related issues: implications of the credit crunch for the impact of monetary policy actions on economic activity, and consequences of credit supply problems for monetary policy guides, M2, and other financial variables. The sec tion begins with some background information on the main features of the recent credit crunch and on the channels of monetary policy influence on the economy. Overview of credit supply problems The evidence in this volume is consistent with the view that credit supply problems contributed importantly to the credit slowdown over 1989-92, although demand influences may have dominated overall credit movements. The nature and causes of the 1989-92 credit supply problems were signifi cantly dissimilar to those of most earlier credit crunches. The distinctive features of the most recent episode are summarized below. First, credit supply problems in the 1989-92 period were widely spread across both bank and nonbank sources of credit. As a result, unlike earlier credit crunches, nonfinan cial borrowers were not able to substitute nonbank credit freely for bank credit. In fact, finance companies, life insur ance companies, and commercial paper issuance seem to have experienced credit supply problems that were essen tially similar to those of banks. Together with a broadly based retrenchment in credit demand, credit supply prob lems led to a sharp slowdown in all major components of private debt flows. Second, credit restraints during 1989-92 took the form both of more stringent price terms— higher lending rates relative to funding costs and tighter nonrate loan terms— and of nonprice credit rationing. Although this phenomenon 18 FRBNY Quarterly Review/Winter 1993-94 is probably fairly typical of earlier credit crunches, the per vasiveness of nonprice rationing and tighter loan terms over an extended period of time in the recent credit crunch is unusual. Earlier credit crunches were generally short lived; the 1989-92 crunch period was characterized by per sistently high spreads between lending rates and funding costs, especially at depository institutions, increasingly tighter credit standards for applications through much of the credit crunch period, and continued stringent nonrate terms on loans. These persistent credit restraints were reflected, among other things, in large increases in hold ings of government securities relative to loans at banks. Third, significant evidence points to a capital crunch as one of the major causes of credit supply problems over 1989-92. None of the earlier credit crunches were charac terized by a widespread weakening of the capital positions of banks and nonbank financial institutions. Broadly, the actual or perceived capital crunch seems to have reflected three underlying forces (in addition to the normal cyclical effects): (1) the need to correct balance sheet problems resulting from the lax lending standards that had prevailed through much of the 1980s and had left balance sheets badly exposed to asset prices and other shocks; (2) increased capital requirements induced by legislative and regulatory measures and by more intensive supervisory oversight; and (3) the weakening of capital positions reflecting declining real estate and other asset values start ing about late 1988. Fourth, market forces seem to have played a critical role in generating the latest credit crunch. To be sure, as noted above, regulatory measures and pressures contributed to the actual or perceived capital crunch but, unlike earlier credit crunches, the current episode emerged in an envi ronment of accommodative monetary policy and declining interest rates. More fundamental to the process of credit slowdown appears to have been the need to correct the debt excesses of the mid-1980s, which had become unsustainable over time. Faced with major balance sheet and other difficulties, borrowers and lenders alike responded to market forces, borrowers by lowering their credit demands and lenders by reducing credit availability. In particular, the so-called credit crumble phenomenon— the chain running from asset price declines to capital position weakness to lower capacity and willingness to lend— contributed importantly to the process of credit slowdown.12 The role of market forces was rein forced and perhaps intensified by the regulatory pressures that highlighted prudential concerns about loan quality and capital positions and argued for the need to strengthen lenders’ balance sheets. The capital crunch itself was at least partly a by-product of the correction process as weak ening capital positions and mounting loan losses called 12 See Johnson (1991) for a detailed description of this phenomenon. increasingly greater attention to the need for correction of earlier debt excesses and for additional capital. The accumulating loan losses, continuing balance sheet problems, and full realization of the debt overhang also led to more conservative lending attitudes—well beyond what could be attributed to the measurable weakness in capital positions— and to a complete reversal of the earlier lax lending standards. To a considerable extent, the pervasive ness of credit supply problems reflected the widespread nature of the correction process, with both bank and non bank creditors experiencing the need to improve loan qual ity and repair their balance sheets. Finally, the debt overhang correction process and its conjunction with a prolonged cyclical weakness in the economy made the already difficult task of distinguishing credit supply malfunctions from credit demand factors even more difficult. Both borrowers and lenders were deleverag ing and restructuring their balance sheets in response to earlier debt excesses and cyclical weakness. In the process, credit demand and credit supply narrowed simultaneously, but the drop in demand is likely to have overwhelmed the fall in supply. As a consequence, it is very difficult, if not im possible, to detect empirically the contribution of supplyside factors net of demand influences. Channels of monetary policy influence Monetary policy influences the economy through at least four important channels: the money-interest rate channel (or the “money” channel, as it is commonly known); the credit channel; the asset valuations or balance sheet chan nel; and the exchange rate channel.13 The discussion here deals with only the first three, ignoring the exchange rate channel. In the money-interest rate channel, as enshrined in the standard IS-LM model, monetary policy affects aggregate spending by raising or lowering the cost of funds through changes in the supply of money relative to the demand for money. Specifically, monetary policy actions— open market operations and so forth— induce changes in bank reserves, money, short-term interest rates and, through substitution and expectational effects, long-term interest rates. Higher (lower) interest rates, in turn, raise (lower) the cost of funds, other things equal. The credit channel, which may operate alongside the money-interest rate channel, affects aggregate demand through direct changes in the availability and terms of bank loans. A tightening of monetary policy may reduce the sup ply of bank loans through higher funding costs for banks or 13 A large number of theoretical and empirical studies on the transmission of monetary policy influence to the economy have appeared since the mid-1980s. For some recent discussions of various channels of monetary policy, see Akhtar and Harris (1987), Bennett (1990), Bernanke (1993), Bernanke and Blinder (1988, 1992), Bosworth (1989), Friedman (1989), Gertler (1988), Gertler and Gilchrist (1992), Gertler and Hubbard (1988), Mauskopf (1992), Mosser (1992), and Romer and Romer (1993). through increases in the perceived riskiness of bank loans. Since the credit channel views bank loans as imperfect substitutes for other assets in bank portfolios (government securities, corporate bonds, commercial paper and the like), monetary policy actions that reduce bank reserves and, therefore, deposits will be matched by decreases in both securities and bank loans. As a consequence, borrow ers with no access to other sources of credit will be obliged to reduce their spending, while others with nonbank sources of credit, though less affected, will not be immune to monetary policy influence as long as the alternative sources of credit are more expensive or less convenient. The asset valuations channel of monetary policy influ ence on the economy works through changes in balance sheet positions. Monetary policy actions that lower interest rates, for example, tend to increase asset values and improve liquidity for firms by lowering interest-to-cash flow ratios. These balance sheet improvements, in turn, may increase business spending by raising the availability of internal funds and improving the access to and the terms on external funds. Lower interest rates may also work to improve household balance sheet positions through debt restructuring and higher asset values, thereby increasing the availability of funds for debt retirement and additional spending. Note that the argument of this channel is that interest rate changes may affect spending by weakening (strengthening) balance sheets or wealth holdings, quite apart from their effects on the cost of funds in the moneyinterest rate channel. Effectiveness of monetary policy Factors relating to the credit crunch seem to have created significant blockages for the workings of all three channels of monetary policy. Overall, the blockages are likely to have muted the impact of monetary policy actions on economic activity. The empirical size and significance of the block ages are far from clear, however. Whether any of these blockages will turn out to have permanent consequences for the conduct of monetary policy is also not clear at this time. The credit channel of monetary policy was seriously dis rupted over 1989-92. With the decline in the willingness and capacity of banks to lend, monetary policy actions increasing bank reserves were not translated into addi tional bank lending. Specifically, easing of monetary policy apparently had very little impact on the supply of bank loans over 1989-92. This view is clearly supported by increasingly tighter credit standards, higher (or at least con tinued high) lending rates relative to funding costs, and restrictive nonrate loan terms. With nonbank credit sources also experiencing supply disruptions, frustrated bank bor rowers were not satisfied elsewhere. Much academic dis cussion of the credit channel assumes that nonbank credit alternatives are easily available to many (perhaps most) borrowers. This view clearly runs counter to the recent FRBNY Quarterly Review/Winter 1993-94 19 credit crunch experience. In fact, widespread nonbank credit supply disruptions appear to have added substan tially to the severity of the blockage in the credit channel. The money-interest rate channel of monetary policy also seems to have experienced some blockage during 198992. Policy-induced increases in bank reserves did translate into lower short-term open market rates and faster growth of narrow money, M1. But the response of long-term inter est rates and broader monetary aggregates to policy actions was very sluggish and weak throughout 1989-92. The decline in credit supply, as shown in the Hilton/Lown study, contributed importantly to slowing the growth of M2. And presumably the shift in credit supply also played some role in maintaining high long-term interest rates by putting upward pressures on rates, other things equal. As a result, monetary policy actions were less effective in lowering the cost of capital, hampering the workings of the money-interest rate channel. The process of correction for earlier debt excesses may also have weakened the asset valuations or balance sheet channel of monetary policy influence on the economy. Given the actual or perceived need to correct the large debt overhang, lower interest rates may not have induced much additional spending by businesses and households be cause the improvements in balance sheets and the under lying asset values materialized only slowly. Put differently, easier monetary policy as reflected in lower interest rates may have encouraged households and businesses to repair the perceived weakness in their balance sheets by deleveraging and debt restructuring, without increasing spending significantly. While credit supply problems during 1989-92 may have been important in reducing the effectiveness of monetary policy, it is difficult to isolate their effects from those of a broad range of other fundamental developments that are likely to have disrupted, weakened, or changed the link ages between monetary policy and economic activity. Mosser discusses a number of these other fundamental developments. Of the factors not directly related to the credit crunch, the following appear to be particularly important: • the response of long-term interest rates to short-term open market rates may have been weakened by infla tion fears or by a high level of investor uncertainty stemming from large federal budget deficits; • effects of lower interest rates may have been weak ened by very high levels of real after-tax interest costs; • looking from a longer term perspective, financial inno vation and deregulation over the last two decades are widely believed to have caused significant changes in both the size and the speed of monetary policy effects on various sectors of the economy. Economic growth in recent years has also been re Digitized for 20FRASER FRBNY Quarterly Review/Winter 1993-94 strained by factors unrelated to both the credit crunch and monetary policy transmission— relatively tight fiscal policy, a military build-down, excess capacity in the construction industry, and low levels of consumer and business confi dence. It is difficult to control for these nonmonetary influ ences in assessing the effectiveness of monetary policy. Against this background, the quantitative significance of the 1989-92 credit supply problems for the transmission channels of monetary policy is far from clear. As reported by Mosser, econometric forecasting equations, both reduced-form and structural estimates from large models, significantly overpredict real spending from 1989 to 1992. This finding is consistent with the notion that monetary pol icy actions have been less effective in recent years than in the past. Presumably the overprediction reflects both the credit crunch and other factors, however. Indeed, Mosser is unable to account for all of the overpredictions by making use of credit supply proxies. Moreover, the overpredictions are not limited to sectors that are directly sensitive to mone tary policy. Instead, they are widely spread across all sec tors, suggesting a general malaise in aggregate demand not captured by economic fundamentals. Notwithstanding these measurement difficulties, credit supply problems during 1989-92 are likely to have con tributed to reducing the effectiveness of monetary policy. Clearly, the credit crunch weakened the credit channel and caused disruptions in credit flows, producing at least some adverse consequences for economic activity. The credit supply shifts are also likely to have hampered the workings of the standard money-interest rate channel and possibly to have weakened the balance sheet-related contribution of lower interest rates to aggregate spending. The long-term implications of the credit crunch for the effectiveness of monetary policy are less clear. Recent credit supply problems may well cause durable changes in the workings of monetary policy transmission channels by altering, for example, the relationship between changes in monetary policy and bank loans, between bank loans and deposit flows, and/or between debt and income.14 But such an outcome is by no means certain. Moreover, with numer ous other potential influences on the linkages between monetary policy and economic activity, it may not be possi ble to isolate any permanent traces of the recent credit crunch on those linkages. Monetary policy guides Disruptions in the linkages between monetary policy and the economy imply adverse consequences for the useful ness of financial variables as monetary policy guides, whether viewed as intermediate targets or simply as infor- 14 If, for example, the recent experience makes banks permanently more risk averse in their lending, monetary policy effects on bank lending would be smaller than before. mation variables. The usefulness of any monetary policy guide depends primarily on two considerations: the strength and predictability of the relationship between the guiding variable(s) and the ultimate objectives of price sta bility and economic growth, and the ability of the Federal Reserve to define, interpret, and control the guiding vari able^).15 The recent credit crunch seems to have added to problems on both counts. Credit supply problems since 1989 have almost certainly contributed to reducing the usefulness of M2 and M3 as policy guides. Hilton and Lown argue that the reduced will ingness of depositories to lend was an important factor behind the weakness in deposits, although their work does not fully isolate the effect of credit supply problems from that of noncyclical credit demand factors. Specifically, the authors point out that relatively high lending rates and the pervasiveness of stringent nonrate loan terms and nonprice credit rationing reduced the supply of credit and, together with lower yields on deposits relative to alternative assets, led to weak depository flows. Controlling for cyclical effects, Hilton and Lown estimate that by the middle of 1992, the credit slowdown had lowered M2 growth by about 10 per cent. Their regression results indicate that the breakdown of M2 demand equations is at least partially attributable to the exceptional weakness in credit formation; the predictive performance of M2 demand equations improves signifi cantly when direct measures of credit or other factors cap turing cutbacks in lending are included as explanatory variables. Credit supply malfunctions have also affected the rela tionship between credit aggregates and the economy. None of the studies in this volume is able to account for develop ments in various credit measures— household, business, bank and nonbank, and so forth— over 1989-92 by using standard historical relations for macroeconomic variables. Of course, the underlying relationships of credit and mone tary aggregates to prices and economic activity have not been particularly reliable during the last decade, even before the emergence of recent credit supply problems. The usefulness of interest rates as information variables for monetary policy has also been adversely affected by the credit crunch. With the pervasiveness of nonprice credit rationing and stringent nonrate loan terms, changes in open market rates have had a smaller impact on credit con ditions and economic activity than would otherwise have been the case. Put differently, disruptions in the credit mar ket mechanisms have made past experience less pertinent as a reference point for understanding the effects of recent interest rate changes on credit conditions and the econ omy. Similarly, to the extent that credit supply problems influenced the yield curve and various interest rate spreads— such as that between lending rates and funding 15 See Friedman (1993c) for a recent perspective on the role of financial variables in guiding monetary policy. costs or that between the (riskless) Treasury bill rate and the (risky) commercial paper rate— all these variables became less useful indicators, at least over 1989-92. By reducing the information content of a broad range of financial variables, the credit crunch has compounded the problems of finding appropriate guides for steering mone tary policy. More specifically, credit supply problems in recent years have made it more difficult to use M2 or the federal funds rate (or any other financial variable for that matter) for determining appropriate money and credit con ditions relative to the needs of the economy. Even before the latest credit crunch, however, there was no significant agreement on the use of any one or two variables as mone tary policy guides. Thus, the recent experience with finan cial sector developments seems to have moved us further away from a narrow focus on one or two intermediate tar gets toward the use of a broad set of financial indicators as information variables to steer monetary policy. Some concluding observations Collectively, studies in this volume offer evidence of a sub stantial, prolonged, and broad-based contraction in credit supply over 1989-92. This finding strongly contradicts the view that the recent credit slowdown originated solely on the demand side.16 Research work reported here conclu sively demonstrates that demand influences are unable to explain a significant part of the recent credit slowdown or decline in nonfinancial borrowings from bank and nonbank sources. Moreover, the existence of credit weakness across a wide range of nonfinancial borrowings also chal lenges the notion that the recent credit slowdown was noth ing more than the bursting of a speculative bubble in com mercial real estate.17 The studies in this volume also indicate that the nature and causes of the recent credit supply problems were markedly different from those of earlier credit crunches. In particular, unlike earlier crunches, the credit supply prob lems during 1989-92 were broadly spread across both bank and nonbank sources of credit, with stringent loan terms and nonprice credit rationing persisting over a relatively long period. Also, unlike earlier episodes, the recent credit crunch was marked by a capital shortage and was driven to an important degree by market forces. Set in motion by the widespread balance sheet difficulties of both borrowers and lenders, these market forces led to the correction process for the debt overhang of the 1980s. The sharp, prolonged, and widespread decline in credit supply over 1989-92 would be expected to have had signif icant adverse consequences for the economy. It is there fore not surprising that the credit crunch has sometimes 16 See Meltzer (1991) and Klieson and Tatom (1992) for particularly strong expressions of this view. 17 See, for example, Jordan (1992). FRBNY Quarterly Review/Winter 1993-94 21 been blamed for much of the weakness in econom ic activity since 1989. Yet the studies in the volume do not support this conclusion. On the contrary, they clearly indicate that credit supply problem s were not the prim ary or dom inant cause of the w eakness in econom ic activity over 1989-92. Nevertheless, the studies do suggest, at least collectively, that credit constraints alm ost surely made some co n trib u tion to that w eakness, and probably played a significant role in slowing the econom y before the recession and in impeding the recovery process.18 The apparent inconsistency between sharply reduced credit availability and its m odest effects on econom ic activ ity is not hard to reconcile. The credit crunch has by no m eans been the only factor depressing the economy. O ther factors that contributed significantly to the 1990-91 reces sion and the subsequent w eak recovery include the Gulf War, the defense build-down, relatively tight fiscal policy throughout the period, generally high real long-term inter est rates, low levels of consum er confidence, corporate restructuring, and the com m ercial real estate depression 18 Perry and Schultz (1993) and Friedman (1993b) reach a roughly similar conclusion. that followed the great buildup of excess capacity during the 1980s. W ith so many powerful forces slowing econom ic activity in recent years, one can hardly expect the credit supply problem s to dom inate the picture. M oreover, the co n flu e n ce of w id e -ra n g in g a d ve rse in flu e nce s on e co nomic activity and the m arket-driven elem ents in the credit crunch make it difficult to isolate em pirically the effects of credit constraints on the economy. Finally, this collection of studies suggests that credit su p ply problem s over 1989-92 contributed to w eakening the influence of m onetary policy actions on the econom y and to reducing the usefulness of M2 and other financial variables as policy guides. W hether recent shifts in credit supply fa c tors will have any long-term consequences fo r the conduct o f m o n e ta ry p o lic y is fa r fro m c le a r, h o w e v e r. In th e absence of further changes in the regulatory environm ent, the long-term effect will depend to a considerable extent on the durability of recent changes in attitudes tow ard debt on the part of lenders and borrow ers— specifically, w hether le n d e rs w ill c o n tin u e to fo llo w th e re c e n t ris k -a v e rs e approach to lending and w hether the decline in the desired ratio of debt to income will turn out to be perm anent. The new co n se rva tive a ttitu d e tow ard d ebt m ay p ersist, but such an outcom e is by no m eans certain. References Akhtar, M.A., and Ethan S. Harris. “Monetary Policy Influence on the Economy: A Empirical Analysis.” Federal Reserve Bank of New York Quarterly Review, Winter 1987. Cantor, Richard, and John Wenninger. “Perspective on the Credit Slowdown.” Federal Reserve Bank of New York Quarterly Review, Spring 1993. Bennett, Paul. “The Influence of Financial Changes on Interest Rates and Monetary Policy.” Federal Reserve Bank of New York Quarterly Review, Summer 1990. Dwight, Jaffee, and Joseph Stiglitz. “Credit Rationing.” In Benjamin Friedman and Frank Hahn, eds., Handbook of Monetary Econom ics, vol. 2. New York: North Holland, 1990. Bernanke, Ben. “Credit in the Macroeconomy.” Federal Reserve Bank of New York Quarterly Review, Spring 1993. Friedman, Benjamin. “Changing Effects of Monetary Policy on Real Economic Activity.” In Monetary Policy Issues in the 1990s. Federal Reserve Bank of Kansas City, 1989. Bernanke, Ben, and Cara Lown. “The Credit Crunch." Brookings Papers on Economic Activity, 1992:2. Bernanke, Ben, and Alan Blinder. “Credit, Money, and Aggregate Demand.” American Economic Review, May 1988. __________ . “The Federal Funds Rate and the Channels of Mone tary Transmission.” American Economic Review, September 1992. Bosworth, Barry. “Institutional Change and the Efficacy of Mone tary Policy.” Brookings Papers on Economic Activity, 1989:1. Cantor, Richard, and Rebecca Demsetz. “Securitization, Loan Sales, and the Credit Slowdown.” Federal Reserve Bank of New York Quarterly Review, Summer 1993. Digitized for 22 FRASER FRBNY Quarterly R eview /W inter 19 9 3 -9 4 __________ . “The Minsky Cycle in Action: But Why?” Federal Reserve Bank of New York Quarterly Review, Spring 1993a. __________ . Comments on “Was This Recession Different? Are They All Different?” by George Perry and Charles Schultz. Brook ings Papers on Economic Activity, 1993b: 1. _____ . “Ongoing Change in the U.S. Financial Markets: Implications for the Conduct of Monetary Policy.” In Changing Cap ital Markets: Implications for Monetary Policy. Federal Reserve Bank of Kansas City, 1993c. Frydl, Edward J. “Overhangs and Hangovers: Coping with the Imbalances of the 1980s.” Federal Reserve Bank of New York Annual Report, 1991. References (Continued) Gertler, Mark. “Financial Structure and Aggregate Economic Activ ity: An Overview.” Journal of Money, Credit, and Banking, August 1988. Owens, Raymond, and Stacey Schreft. “Identifying Credit Crunches." Federal Reserve Bank of Richmond, Working Paper no. 92-1, March 1992. Gertler, Mark, and Simon Gilchrist. “The Role of Credit Market Imperfections in the Monetary Transmission Mechanism: Argu ments and Evidence." New York University, unpublished paper, May 1992. Peek, Joe, and Eric Rosengren. “Crunching the Recovery: Bank Capital and the Role of Bank Credit.” In Real Estate and the Credit Crunch. Federal Reserve Bank of Boston, 1992. Gertler, Mark, and R. Glenn Hubbard. “Financial Factors in Busi ness Fluctuations.” In Financial Market Volatility. Federal Reserve Bank of Kansas City, 1989. Greenspan, Alan. Remarks at the 55th Annual Dinner of the Tax Foundation, New York, November 18,1992. Johnson, Ronald. “The Bank Credit Crumble.” Federal Reserve Bank of New York Quarterly Review, Summer 1991. Jordan, Jerry L. ‘The Credit Crunch: A Monetarist’s Perspective.” Paper presented at the Annual Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, May 7,1992. Kaufman, Henry. “Credit Crunches: The Deregulators Were Wrong.” Wail Street Journal, October 9,1991. Kliesen, Kevin, and John Tatum. ‘The Recent Credit Crunch: The Neglected Dimensions.” Federal Reserve Bank of St. Louis Review, September-October 1992. Perry, George, and Charles Schultz. “Was This Recession Differ ent? Are They All Different?” Brookings Papers on Economic Activ ity, 1993:1. Rodrigues, Anthony. “Government Securities Investment of Com mercial Banks.” Federal Reserve Bank of New York Quarterly Review, Summer 1993. Romer, Christina, and David Romer. “Credit Channel or Credit Actions? An Interpretation of the Postwar Transmission Mecha nism.” In Changing Capital Markets: Implications for Monetary Pol icy. Federal Reserve Bank of Kansas City, 1993. Sinai, Allen. “Financial and Credit Cycles—Generic or Episodic?” Federal Reserve Bank of New York Quarterly Review, Spring 1993. Syron, Richard. “Are We Experiencing a Credit Crunch?” Federal Reserve Bank of Boston New England Economic Review, JulyAugust 1991. Jones, David M. “The Role of Credit in Economic Activity.” Federal Reserve Bank of New York Quarterly Review, Spring 1993. Syron, Richard, and Richard Randall. “The Procyclical Application of Bank Capital Requirements.” Federal Reserve Bank of Boston Annual Report, 1991. LaWare, John. Testimony before the House Committee on Bank ing, Finance and Urban Affairs, July 30,1992. Wojnilower, Albert. “The Central Role of Credit Crunches in Recent Financial History.” Brookings Papers on Economic Activity, 1980:2. Mauskopf, Eileen. “The Transmission Channels of Monetary Pol icy: How Have They Changed?” Federal Reserve Bulletin, Decem ber 1990. __________ . “Credit Crunch.” In The New Palgrave Dictionary of Money and Finance. New York: Stockton Press, 1992a. McCauley, Robert, and Rama Seth. “Foreign Bank Credit to U.S. Corporations: The Implications of Offshore Loans.” Federal Reserve Bank of New York Quarterly Review, Spring 1992. __________ . Discussion of “Crunching the Recovery: Bank Capi tal and the Role of Bank Credit,” by Joe Peek and Eric Rosengren. In Real Estate and the Credit Crunch. Federal Reserve Bank of Boston, 1992b. Meltzer, Alan. “There Is No Credit Crunch.” Wall Street Journal, February 8,1991. __________ . “Not a Blown Fuse.” Federal Reserve Bank of New York Quarterly Review, Spring 1993. Mosser, Patricia C. “Changes in Monetary Policy Effectiveness: Evidence from Large Macroeconometric Models.” Federal Reserve Bank of New York Quarterly Review, Spring 1992. Wolfson, Martin. Financial Crisis. New York: M.E. Sharpe, 1986. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 23 International Interest Rate Convergence: A Survey of the Issues and Evidence by Charles Pigott The international integration of financial markets has increased dramatically over the last two decades. Techno logical advances and the progressive elimination of official barriers to capital flows have spurred an enormous increase in cross-border financial transactions and activi ties and rapid growth in the Eurocurrency and other interna tional financial markets. As a result, linkages among national financial markets have been greatly strengthened, and financial conditions in individual countries have become increasingly sensitive to developments in the mar kets of their partners. It was widely expected that international financial inte gration would also lead to convergence of interest rates across countries, or at least to greater synchronization of interest rate movements than in the past. In fact, however, considerable international interest divergences have per sisted across a broad spectrum of assets, even very recently. Over the last two years, for example, domestic short-term rates in the United States have fallen sharply while those in Germany and other continental European countries have remained at considerably higher levels. This article examines why interest rates have apparently failed to converge internationally. We first consider in con ceptual terms what financial integration means for interest rate relations in an international context. We then examine the evidence on interest rate convergence and the circum stances under which it has or has not occurred. As we will see, the key feature distinguishing the interna tional economy from a single country is the presence of multiple currencies whose exchange rates are subject to change. Interest rate convergence has several meanings in this context. Where currency distinctions are absent, inte Digitized for 24 FRASER FRBNY Quarterly Review/Winter 1993-94 gration generally has led to interest rate convergence. But where assets differ in their currency denomination, as they typically do in comparisons of national interest rates, finan cial integration does not imply convergence unless the eco nomic conditions determining the rates become more closely aligned and exchange rates are fixed, or nearly so. In fact, the evidence strongly suggests that for countries with flexible exchange rates, national interest rates have varied nearly as freely with financial integration as earlier, although with much greater repercussions on exchange rates. There appears to be no systematic tendency for cross-country disparities among either nominal or real interest rates to decline, much less disappear — despite a dramatic reduction in barriers to international capital flows. The article examines one further concept of interest rate convergence particularly relevant to international investors. This is the extent to which national interest rate differentials tend to be systematically offset by currency movements, so that returns expressed in a common currency are equalized on average. This seemingly simple and intuitive presump tion has raised a number of somewhat complex, and to this point largely unsettled, issues. Currency risks arising from uncertainty about future exchange rates as well as system atic errors made by investors in predicting currency move ments can, and probably have, prevented full convergence in this sense. However, the evidence suggests that these considerations, at least as they are presently understood, do not seem to provide an adequate explanation for the large systematic return differentials among currencies that are observed in practice. These findings raise questions very similar to those long encountered in analyses of the behavior of stock and bond returns within a single country. The meaning of international financial integration Complete integration of an economy’s financial markets basically means that all participants have equal access to all markets. Equal access implies that interest rate and other terms faced by participants depend only on objective indicators of creditworthiness such as financial position and credit history— not on residence or nationality. Integration allows portfolio diversification across markets and instru ments; thus the tendency of investors to hold assets issued in their own locale when markets are isolated is likely to be substantially reduced, if not altogether eliminated, when markets become integrated. Financial integration within a single country, where all assets are denominated in the same currency, affects the behavior of interest rates in several important ways. First, because integration allows arbitrage across markets, returns on instruments with identical characteristics are equalized regardless of where they are issued or traded. For example, within the United States, regional interest dif ferentials among comparable assets are quite small or neg ligible in most cases. Second, and more generally, integra tion is likely to lead to much greater synchronization of interest rate movements across markets and to lower inter est differentials among similar (though not necessarily identical) assets. The basic reason is that with integration, local differences in credit conditions tend to be largely elim inated by flows of funds among markets. Thus, regional fluctuations in real income, saving, or other determinants of credit demands and supplies do not themselves lead to sig nificant interest rate divergences, as they would if markets were isolated. Instead, interest rates tend to vary with national credit conditions as determined by real growth, inflation, government fiscal positions, and other domestic macroeconomic conditions. Reinforcing this tendency is the fact that a single-currency economy sharply limits the degree to which certain key interest rate determinants, notably inflation, can differ among regions. It follows that interest differentials within a single country largely reflect differences in instrument characteristics such as maturity, liquidity, and risks that are valued, or priced, in the common national market. For example, inter est rates on ten-year corporate and U.S. government bonds move together quite closely over time, but the corpo rate rate is typically greater by an amount that largely reflects market perceptions about the risks of business defaults. Integration in the international economy While the implications of financial integration for the inter national economy are broadly similar to its implications for a single country, the specific consequences for interest rates are much less straightforward, for three reasons. First, impediments to financial flows among nations arising from overt restrictions on capital flows and from differing tax laws, regulatory policies, and other institutional arrangements typically far exceed the barriers that exist among states, provinces, or regions of a single country. Second (and substantially as a result of the first), key instrument characteristics such as available maturities, minimum denominations, and liquidity generally vary much more across countries than within any single country. Third, and most fundamentally, the international econ omy is distinguished by the existence of multiple currencies whose values are subject to change. Interest rate conver gence in such an environment has two quite distinct, if closely related, meanings. The first, the convergence of national interest rates (as they are normally expressed), involves a comparison of returns denominated in different national monies: the quoted yields on U.S. and German government bonds, for example, refer to their yields in terms of dollars and German marks, respectively. Likewise, comparisons of real interest rates across countries usually involve returns expressed in terms of national commodity bundles whose composition typically varies across coun tries.1 For investors deciding how to allocate funds among assets, however, it is the degree to which their prospective relative returns expressed in a common currency converge that matters. These relative returns are determined not only by the national interest rates themselves but also by the change in the relevant exchange rates over the investment horizon: the dollar return on, say, a three-month German mark-denominated asset depends upon the rate at which marks can be exchanged for dollars at maturity. Even with multiple currencies, linkages among markets in a financially integrated international economy are no less strong than within a single country. The connections are more indirect, however, because the national markets are linked through the markets for foreign exchange. This fact would be of little practical consequence if exchange rates were completely and irrevocably fixed. In that case, inte gration would have virtually the same effects internationally as within a single economy: national interest rates would largely converge and their movements would be closely synchronized; remaining interest differentials would be determined by disparities in market and (noncurrency) instrument characteristics rather than by macroeconomic disparities among the countries. In the actual world economy, however, exchange rates are very seldom completely fixed. The fact that national markets are linked through foreign exchange markets then has two important practical consequences. First, dispari ties in underlying determinants of national interest rates can be, and generally are, much greater than within a sin gle country. In particular, inflation rates can diverge indefi1 The U.S. real interest rate, typically defined as the nominal interest rate less some measure of anticipated domestic inflation, is effectively a return in terms of U.S. products, while German real interest rates measure returns in German goods. FRBNY Quarterly Review/Winter 1993-94 25 nitely provided that exchange rates can change to offset the differences. Second, divergences in macroeconomic forces typically will lead to cross-country differences in national interest rates when exchange rates are free to vary. In the world economy, as in a single economy, a tightening of credit that pushes up interest rates in one country’s markets tends to attract funds from abroad. This inflow, however, first places upward pressure on the home currency, raising its current value above the level expected to prevail in the future (and thus increasing the amount by which the currency is expected to fall subsequently). If the home government allows its exchange rate to float freely, this process will continue until the currency’s prospective future decline is sufficient to eliminate the incentive for funds to flow in — leaving national interest rates both at home and abroad largely unaffected. In a variable exchange rate environment, therefore, dif ferentials among national interest rates stem not only from differences in their characteristics or imperfect integration of the markets, but also from divergences in macroeco nomic and other determinants and their interactions with exchange rates. Disparities in economic conditions lead to national interest rate differentials, which in turn reflect per ceptions about the magnitude of, and (as we will see shortly) the risk associated with, future currency move ments.2 Financial integration, even if complete, need not lead to interest rate convergence nor indeed to any increased synchronization of national rate movements across countries; interest differentials are likely to vary in magnitude as their underlying determinants become more or less aligned across countries. The main, and critically important, effect of financial integration in this context is to greatly increase the sensitivity of exchange rates to national interest rate fluctuations: as explained earlier, inte gration has meant that changes in a nation’s interest rates relative to rates abroad lead to offsetting currency move ments. The result is that financial developments in one country tend to affect conditions in others through their impact on foreign exchange markets. Convergence in a common currency? Although financial integration need not lead to equalization of national interest rates, it might seem that it should result in the convergence of returns expressed in a common cur rency. This is true in a narrow sense: yields on otherwise identical instruments whose returns are guaranteed by hedging (“covering”) in forward foreign exchange markets must be equalized with complete integration. In the 2 In effect, therefore, national interest differentials (aside from characteristic differences and imperfect financial integration) can be viewed as the proximate reflection of expected future exchange rate changes and currency risks that, at least in principle, are ultimately determined by divergences in countries’ fundamental interest rate determinants. Digitized for 26 FRASER FRBNY Quarterly Review/Winter 1993-94 Eurocurrency markets, for example (where the instruments are identical except for their currency), the dollar return on a three-month German mark deposit whose proceeds at maturity are covered through forward market sale (for dol lars) is the same as that on a three-month dollar deposit. Note, however, that hedging the mark asset amounts to its redenomination in dollars (since the hedged instrument is a fixed claim to future dollars); currency distinctions among assets are effectively abolished in comparisons of their covered returns. The sources of covered interest differen tials therefore are the same as those present within a single nation — barriers to financial flows across markets and dif ferences in instrument characteristics.3 The broader and much more controversial question is whether returns that are not hedged (in other words, that are “uncovered” in the sense that they depend upon actual exchange rate movements that cannot be fully predicted) converge when expressed in a common currency. In practi cal terms, this question amounts to asking whether exchange rate movements tend on average to offset differ ences in national interest rates on otherwise similar assets. If so, investing in one currency as against another will pro duce no systematic difference in realized returns, and national interest rate differentials (apart from differences in asset characteristics) will simply reflect market expecta tions about future exchange rate movements. This principle is commonly referred to as “uncovered interest parity.” As explained further below, the degree to which uncov ered interest parity holds in a practical sense depends pri marily upon the importance of two factors. The first and, until recently, the predominant focus of debates in this area is the importance of the “currency risks” associated with investing in one currency as opposed to another. Currency risk in this context refers to the differential riskiness among assets that arises from their denomination. To understand what currency risk means, consider a U.S. investor who holds two government bonds, one denominated in dollars and the other in German marks. Both bonds are risky in that their prices, in dollars and German marks, respectively, are to some degree unpredictable; in addition, the return in dol lars of the German mark bond depends upon future exchange rate changes — which are also unpredictable. The risks of the two bonds therefore are likely to differ, 3 Complete hedging is generally available only to fairly large market participants and for fairly widely used or traded instruments. Moreover, there are well-known factors other than unanticipated exchange rate movements that may impair the liquidity or solvency of an instrument and that tend to be currency-associated, including the possible default of a government or government-guaranteed borrower on its external foreign currency obligations (“sovereign" risk) and the potential inability of private domestic entities to obtain foreign exchange to meet their external obligations because of actual or prospective capital controls (“transfer” and “political" risks). These risks are currency-associated mainly because national authorities can regulate or otherwise impede the convertibility of their national money. In this discussion, however, these factors are treated as barriers to capital mobility or as sources of differences in asset (noncurrency) characteristics. most obviously (although, as we will see later, not entirely) because of the uncertainty about exchange rates. Currency risks are reflected (as “currency risk premia”) in the uncovered returns that investors anticipate receiving in a common currency; the corresponding national interest differentials also incorporate these risks in addition to expectations about future currency changes. As with any other type of risk, the importance of currency risk depends not upon the volatility of any particular currency when viewed in isolation, but rather upon the extent to which holding an asset denominated in one money as against another contributes to the overall risk a typical investor faces; thus, uncovered interest parity is likely to hold exactly only if currency risks can be completely diversified, that is, offset by other sources of risk. From this perspec tive, the key question is not whether currency risk premia exist at all (the considerable volatility of exchange rates makes it very likely that they do) but how important they are in practice. If representative investors view these risks as comparatively large, there are likely to be significant aver age differences in dollar returns from investing in one cur rency relative to another. Even if currency risks were quite small, however, com mon currency returns could still differ considerably and sys tematically for a second reason, namely biases in market forecasts. Suppose, for example, that investors consis tently underpredicted increases in the value of the German mark versus the dollar during some period: mark-denomi nated instruments would tend to outperform their dollardenominated counterparts even though the ex ante returns anticipated by investors would be the same. Economists have normally assumed that such biases are very small or sporadic but, as we will see later, growing evidence sug gests that they may be sizable and pervasive. Evidence on the convergence of national interest rates There can be little doubt that the major financial markets of the industrial countries have become much more closely integrated over the last two decades. Official barriers to capital flows have largely been eliminated by the industrial countries and substantially reduced by many developing nations. Larger financial institutions and nonfinancial cor porations now have access to an array of international financial markets with relatively low transactions costs, as well as to major domestic markets of the larger countries; portfolio diversification, particularly by banks and, in some countries, by institutional investors, has increased markedly since the late 1970s.4 International financial integration is certainly not complete (indeed barely begun for markets 4 See Benzie (1992) for a detailed description and analysis of the remarkable development of the international bond market during the 1980s. For an excellent analysis of the international diversification by pension funds and insurance companies, see E. P. Davis (1988, 1991). catering to smaller businesses and individuals), nor is it as great as that found within the United States or most other countries, but it is still considerable in economic terms. Nonetheless, despite the obvious interdependence among financial markets resulting from integration, national interest rates, whether nominal or real, do not seem to have converged in any very meaningful sense. Indeed, the recent record is quite consistent with the conclusion of an earlier study by Kasman and Pigott (1987) that the disper sion in national interest rates fluctuates considerably over time but without any systematic tendency to decline. At pre sent, U.S. short-term interest rates are fairly close to those of Japan but substantially below those in Germany, the United Kingdom, and Canada; substantial gaps among the countries’ long-term interest rates also remain. As Chart 1 shows, divergences among short-term interest rates are now actually somewhat above their average of the last twenty years, and while the dispersion in longer term rates has declined over the last decade, it is still noticeably higher than in the early 1970s.5 Although financial integration has led to no discernible convergence of national interest rates, its effects are dra matically manifest in covered interest differentials. As explained earlier, these differentials largely reflect barriers to capital flows and instrument characteristics rather than currency distinctions and so provide a direct indicator of the progress of integration. By this standard, the major short term industrial country financial markets have become very highly integrated: as Chart 2 indicates, covered interest rate differentials among national money markets, which were at times quite large during the 1970s, have largely dis appeared, as have gaps between the domestic money mar kets and the corresponding Eurocurrency markets.6 Analo gous evidence suggests that integration has also increased in the markets for longer term instruments, although the 5 Despite this evidence, some observers have argued that integration has at least increased the synchronization of interest rate movements over the last decade. Several studies, in fact, have reported that by some measures, correlations between U.S. and foreign interest rates were somewhat greater during the 1980s as a whole than in the 1970s; see, for example, Frankel (1989) and the introduction to Bank for International Settlements (1989). But other, equally plausible measures do not show any consistent increase in this tendency (for example, see Kasman and Pigott 1988), and in many cases national interest rates appear to have been less synchronized during the latter 1980s than during much of the 1970s, when markets were presumably less integrated than now. Variations in these correlations are more likely a reflection of changing alignments among national economic conditions than a product of financial integration. 6 Numerous studies have documented the decline in short-term interest differentials resulting from the lowering of official capital controls, beginning with the major industrial nations in the 1970s and early 1980s and spreading to virtually all the industrial countries in the latter half of the decade. Among the more extensive studies are Caramazza et al. (1986) and Frankel (1988). In addition, Akhtar and Weiller (1987) and Frankel (1990) provide excellent discussions of conceptual issues concerning the definition and measurement of international capital mobility. FRBNY Quarterly Review/Winter 1993-94 27 change has been more recent and less com plete. In partic ular, as shown in Table 1, hedged (dollar) returns on go v ernm ent bonds are also now fairly closely aligned fo r at least the major currencies.7 Financial integration thus has significantly altered the re l ative importance of the factors underlying national interest rate d iffe re n tia ls m entioned earlier. Institutional barriers along with noncurrency instrum ent characteristics are now a relatively m inor source of the divergences; national inter7 Long-term instruments can be hedged through currency and interest rate swaps. The development of these facilities beginning in the mid1980s is itself a strong indication of the growing integration of major bond markets. Popper (1990) was the first to use this data to demonstrate the near-parity of hedged returns for such instruments. est diffe re n tia ls reflect, nearly entirely, disp a ritie s in the m acroeconom ic determ inants of interest rates and the cor responding exchange rate m ovem ents they induce.8 Indeed, at least the broad m ovem ents in national rate d if ferentials in recent years can be fairly plausibly explained by fluctuations in real income, inflation, m onetary and fiscal policies, and the changing alignm ent of these conditions across countries. For exam ple, the largest divergences in nominal interest rates, particularly longer term rates, have tended to occur during periods of rising and relatively high inflation such as the m id- and late 1970s and the early 8 Admittedly, heterogeneity of instrument characteristics is more important for mortgages and other assets that are less standardized than typical money market securities or government bonds. Chart 1 Cross-Country Dispersion of National Interest Rates Percent /VV's m S jV"Jitf 11 \ AXV\A j\ jM JV\ir^Uv^ / /Vh v/r \ r i' M i rji ffv-v ’ ,Ij W V No minal Rates Short-term rates „ I I .I--I—--1-1-L_J--..i..i.i i.. i—i—i—i——i—i—I — Notes: Dispersion is calculated as the average absolute deviation from the country mean of each month. Short-term rates are the call money rate for Japan and three-month money market rates for the United States, Canada, Germany, France, and the United Kingdom. Long-term rates are long-term government bond yields for the above six countries plus Italy, Belgium, and Switzerland. The real short-term rate is the nominal rate less the inflation rate over the last year; the real long-term rate is the nominal rate less the inflation rate over the last three years. 28 FRBNY Quarterly R eview /W inter 19 9 3 -9 4 1980s, largely because cross-country disparities in infla tion, the stance of m onetary policy, and business cycle positions have generally been greatest in these periods. Likewise, the decline in long-term interest rate divergences over much of the last decade can be attributed in large part to the general fall (and convergence) of national inflation rates during the same period.9 F urtherm ore, m ajor shifts in the a lignm ent of interest rates across countries have usually been associated with substantial m ovem ents in exchange rates. A dram atic illus tration is the prolonged appreciation of the dollar accom pa nying the rise in U.S. interest rates relative to rates abroad during the first half of the 1980s. The persistence of real interest rate differentials, while more surprising to many observers, is also understandable in these terms. As norm ally measured, the real interest rate on a given country’s asset is effectively its return in term s of some aggregate of com m odities produced or consum ed in that country. The com position of these com m odity aggre9 These conclusions are also broadly consistent with more direct evidence about the forces shaping domestic interest rates. This evidence suggests on the whole that while the influence of international factors has risen in some cases, traditional domestic macroeconomic factors remain the most important determinants. For example, although international factors may now have some modest influence, short-term interest rates still appear to be largely determined by variations in the domestic supply and demand for liquidity. See, for example, Radecki and Reinhart (1988). There are reasons to believe that international factors may have somewhat greater influence on long-term interest rates, but the evidence is limited. gates typically varies across countries because of the inclu sion of nontraded goods and services and differences in production and consum ption patterns. The belief that real interest rates should converge internationally is based on the presum ption that returns to capital will ultim ately be eq ualized and th a t p u rchasing pow er p a rity de te rm in es nominal exchange rates — conditions that are likely to hold, if at all, only in the very long run. O ver the medium term, Table 1 Covered Interest Bonds for Government (Foreign minus U.S. Yield to Maturity) Germany Japan Switzerland Average Standard Deviation -7 0 -4 6 18 15 42 19 Notes: Table reports the difference between the domestic (tenyear) yield to maturity on the foreign bond and the yield in the same currency of a “sw apped” U.S. ten-year Treasury bond. The differential com bines the applicable interest rate swap rate for tenyear Treasuries (that is, from ten-year fixed payments into floating rate LIBOR payments in dollars) and the currency swap rate (from floating LIBOR payments in dollars into ten-year fixed payments in the relevant foreign currency). All figures refer to averages for the period 1987-90. .....Ill# Chart 2 Covered interest Differentials Domestic Three-Month Rates Percent Notes: Data are end-of-month. The three-month commercial paper rate is used for the United States. The foreign rates are three-month interbank rates whose dollar returns are covered in the three-month forward exchange rate. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 29 real exchange rates have varied nearly as much as nominal exchange rates. Long, variable, and persistent fluctuations in real interest rates are quite consistent with this pattern, as is the corresponding tendency for domestic real interest rates to be the primary source of nominal interest rate movements over similar intervals.10 Overall, therefore, actual experience is quite consistent with the conceptual arguments presented earlier in this article. Integration has had clear and dramatic effects, most noticeably on covered interest rate differentials. Inte gration has not, however, led to any appreciable conver gence of national interest rates, because of the combina tion of variable exchange rates and continued large disparities among nations’ macroeconomic conditions that has characterized the world economy for over twenty years. Indeed, the experience of the European Monetary System, which is summarized in the box, strongly sug gests that only when exchange rates are very nearly fixed and national macroeconomic policies are largely harmo nized is integration likely to lead to any genuine conver gence of national interest rates. Uncovered interest parity? While most investors and analysts have become quite accustomed to large and persistent divergences among national interest rates, there remains a very widespread belief that these differences tend to be offset by currency movements. Investing in one currency rather than another may yield higher or lower returns at certain times, but, according to this view, the returns should be equal on aver age over longer periods. Some tendency toward this “uncovered” interest parity is evident even when markets are isolated: countries with high inflation rates tend to have relatively high interest rates but also depreciating curren cies. Moreover, as noted in the first section, currency-asso ciated risks are likely to prevent uncovered returns from being fully equalized even with complete integration. Nonetheless, it seems plausible to assume that uncov ered returns would be more closely aligned now that mar kets are substantially more integrated and investors more diversified internationally than they were in the 1960s or 1970s. As we will see shortly, however, it is far from clear that this presumption is valid. Indeed, we will see that the 10 In most empirical models of the U.S. and other economies, fluctuations in real income, inflation, and other macroeconomic determinants of credit market demands and supplies produce substantial variations in real interest rates. The corresponding international macroeconomic models — of the type first introduced by Dornbusch (1976) — view variations in real interest differentials across countries as a major, if not dominant, source of real exchange rate fluctuations. In an empirical analysis of several large industrial countries, Howe and Pigott (1992) develop evidence suggesting that long-term real interest rates vary substantially and are influenced both by persistent factors, such as aggregate debt and returns to physical capital, and, in the mediumterm, by changes in macroeconomic policies. There is some evidence (see Mishkin 1984) of long-run real interest rate convergence, however. 30 FRBNY Quarterly Review/Winter 1993-94 issues raised by empirical analyses in this area have proved to be (at least by comparison with those encoun tered in the last section) often complex and perplexing — as well as substantially unresolved. Historical evidence on uncovered interest parity The historical record of return differences across curren cies provides one very rough indication of the degree to which uncovered yields have converged under financial integration. Table 2 lists average ex post differential returns, expressed in dollars, of foreign relative to U.S. assets over five-year intervals for three types of instru ments, namely short-term (three-month) money market securities, longer term government bonds, and stocks.11 In principle, these differentials reflect the returns anticipated (ex ante) by investors as well as any errors made in fore casting future exchange rates and the assets’ prices. The differentials are often remarkably large. Indeed in certain periods they appear (even for short-term assets) to be of greater magnitude than the national interest rates them selves. The return disparities are also highly variable: in some periods, foreign assets strongly outperform their U.S. counterparts, while in other periods, they underperform them. (Partly as a consequence, average divergences over decade intervals, as well as the entire period, are generally smaller in magnitude than the five-year average.) And, of most relevance here, the differentials seem to show no ten dency to decline over time.12 While unexpected changes in currency and asset prices are undoubtedly responsible for some portion of the recorded divergences, a large and growing body of evidence strongly suggests that they cannot be the only explanation. If return differentials on comparable instruments result sim ply from random and unbiased forecast errors, they ought to vary randomly and average out to zero. Most evidence, though, indicates that the divergences are larger than is explainable by pure chance (that is, they are statistically sig nificant). Moreover, variations in return differentials appear to be systematic in the sense that they are at least partially predictable. Several studies have found, for example, that trading rules specifying when to invest or withdraw from one currency or another tend to yield significantly greater returns 11 All data are computed from monthly holding period returns. The bond return estimates are taken directly from Ibbotson and Siegal (1991) and are based on long-term interest rate figures from the International Monetary Fund’s International Financial Statistics. Note that the corresponding instruments are almost certainly not as comparable as those used for the data in Chart 1 and Table 1 (which generally are available only for a much shorter period). The stock returns are derived from aggregate stock price indexes and dividend-price ratios for the major exchanges in each country. 12 Return differentials during the 1980s as a whole are smaller than during the 1970s in slightly more than half the cases. More often than not, however, the divergences in the three-month instruments and the bonds recorded in the first half of the 1980s are greater than during either half of the preceding decade. Box: When exchange rate flexibility is limited Because interest rates do diverge considerably when cur rencies are relatively free to vary, a natural question is, what happens when exchange rate flexibility is substantially lim ited? Some light is shed on this question by the experience of the members of the Exchange Rate Mechanism (ERM) of the European Monetary System (EMS). Until last fall, about half of the members (Germany, France, Belgium, Denmark, Ireland, and the Netherlands) limited their exchange rate movements to a band of 2.25 percent around the central parity; the remainder (Italy, Spain, Portugal, Greece, and for most of its period of partic ipation, the United Kingdom) adhered to 6 percent bands.* The central parities have been changed several times since the system’s inception in the late 1970s, although with somewhat decreasing frequency up to the fall of 1992. Moreover, capital controls among the members have been removed gradually over a number of years—as early as the mid-1970s in Germany and the United Kingdom but not until the latter 1980s in several other countries. As Chart 3 shows, interest rates among the ERM coun tries have moved considerably closer, but only fairly recently. Except for the Netherlands, short-term interest rates did not achieve near-parity with Germany until about 1990. Most effective barriers to financial flows among these markets were removed some years earlier, as indicated by the fact that gaps between domestic money and Eurocur rency rates were largely closed by 1986 for France, and well before that for Belgium and the Netherlands. Moreover, it was not until 1991, at the earliest, that any genuine align ment of longer term rates occurred (again except for the Netherlands, whose long rates have followed those of Ger many for much of the 1980s). This sequence of developments suggests that it was not financial integration alone but rather the interaction of inte gration, the exchange rate regime, and the evolution of macroeconomic conditions that produced the gradual con vergence of ERM interest rates. Given the margin for exchange rate fluctuations within the system, substantial divergences in shorter term interest rates are consistent even with complete integration. For example, under the nar rower bands, three-month interest rates can differ by as much as 9 percentage points.* Even the larger divergences among European rates in the mid-1980s were well within such limits. The marked narrowing of the differentials in recent years is substantially the result of changes in mon- tary policy operating procedures: monetary authorities in France and several other countries have chosen to keep their official rates closely in line with those of Germany. This shift has been prompted by the planned European Monetary Union, but it is also reflective of the considerable conver gence in macroeconomic conditions, particularly inflation, that has occurred.5 5 For a useful recent analysis of interest rates in the ERM, see Mizrach (1993). Chart 3 Evolution of European Interest Rates Percent — Tht ee-Month Interest RahiS ♦ % United KingcJom A * % 0 * \ o \ j * » ♦% V * V ***** \ V ♦ \ F ra n c e • \ a 0 ■ • > / A /i i \ \ • w N e th e r la n d s '* A * ■**«P Germany III l l l l l l l l l l l l l l l l l l l 111 1 1 111, 1,1 i l l 11,1, 1 1 1 -......... II 1 1 1 111111 -.......... Lc>ng-Term Interest Rate 'ranee /,**. \ United Kin gdom > *• « - - - • , 9 - \ " 'l v ** V / Germany \ Netherlands * In addition, Austria, and more recently Sweden and Norway, have sought to closely tie their currencies to the German mark even though they are not formal members of the ERM. * This figure corresponds to the annualized movement of a currency across the full “width" of the permissible band. In practice, the maximum possible interest differentials depend upon a currency’s position within the band. III 1980 lllllllllllllllllll 82 84 .lllllllllllllllllll 86 88 90 1111 1111 1111 92 93 Notes: The short-term rate is the three-month interbank rate; the long-term rate is the government bond rate. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 31 B o x : W hen e x c h a n g e ra te fle x ib ility is lim ite d (C ontinued) Such macroeconomic harmonization has contributed even more to the convergence of longer term rates. In the ERM, gaps among long-term rates p rim arily reflect prospects that the central parities will be maintained over the longer term, a virtually impossible feat unless inflation rates remain equalized. Thus the near-equality of Dutch and German long-term rates for most of the 1980s essentially stemmed from the very close alignment of their inflation per formances and policies. Also understandable in these terms is the relatively late convergence of French with German long-term interest rates: not until the end of the 1980s had France’s underlying inflation rate clearly fallen into line with that of Germany. than sim ply holding a diversified portfolio of assets.13 Particularly rem arkable in this respect is an apparent te n dency, first pointed out by Fama but since supported by other studies, fo r returns on shorter term assets to rise w hen the corresponding national interest rate differential in c re a s e s .14 T hus, fo r exam ple, when G erm an national in te re st rates rise relative to U.S. rates, realized d o lla r returns on m a rk-den o m in a te d assets ty p ic a lly increase also. This pattern is clearly inconsistent with uncovered interest parity, w hich im plies that an increased Germ anU.S. interest rate gap should be fully offset (again on a ve r age) by greater mark depreciation (or less appreciation). Overall, the evidence indicates that financial integration has not led to convergence of asset returns expressed in a comm on currency. Indeed it is even unclear w hether in te gration has produced any closer alignm ent of uncovered returns. Instead we fin d a p p a re n tly siza b le sy s te m a tic uncovered differentials w hose m agnitude and sign appear to vary over time. To most observers, the most plausible explanation of these patterns is currency risk. We will see, however, that this explanation seems to be incom plete in im portant respects. A matter of risk? We noted earlier that otherw ise identical assets denom i nated in different currencies are inevitably subject to d iffe r ent risks unless their exchange rates are com pletely fixed. T ypically when any asset has an uncertain return, its inter13 Prominent examples are Dooley and Shafer (1983), Sweeney (1986), and Levich and Thomas (1993). In general, the profits found under these rules easily exceed the transactions costs incurred (by a large investor) in their implementation. 14 See Fama (1985). Even more remarkable, the results suggest that a rise in national interest rates in favor of a country is associated with an appreciation of its currency (or a diminished rate of depreciation). More generally, Fama’s findings and related results imply that variations in national interest rates predominantly reflect changing risk premia. 32 FRBNY Quarterly R eview /W inter 1993-94 The record of the ERM thus indicates that under financial integration, national interest rates probably would have con verged had a completely fixed exchange rate system, including the harmonization of policies required to sustain it, been maintained. That same record also shows, however, that even modest departures from completely fixed rates can lead to very substantial interest rate divergences of a magnitude and variability barely distinguishable from those observed under floating exchange rates. The reason is that interest rates, particularly longer term rates, are very sensi tive to prospective disparities in economic conditions and policies. Thus an option to depart from completely fixed rates, however improbable or distant its exercise, may sus tain considerable interest rate divergence. est rate must incorporate a risk prem ium as com pensation. From this perspective, system atic uncovered return diver gences are the natural result of risk fa cto rs s p e cifica lly associated with currency denom ination. C u rre n cy risk is often vie w e d as sim p ly re fle c tiv e of uncertainty about future exchange rates and in this respect q u ite d is tin c t from risks m ore n o rm a lly e n co u n te re d in dom estic markets. This view is misleading fo r at least two reasons. First, as we have seen, when exchange rates are va ria b le , the d e te rm in a n ts of in te re s t rates, and hence dom estic asset prices, are likely to be only im perfectly co r related across currencies. As our earlier exam ple of the U.S. and Germ an bonds indicated, instrum ents d eno m i nated in different currencies thus are subject to differing risks from fluctuations in their dom estic price (price risk) in a d d itio n to th e ris k s a ris in g d ire c tly fro m u n e x p e c te d exchange rate movem ents. Second, the factors underlying the risks associated with foreign cu rre n cy a sse ts are not fu n d a m e n ta lly d iffe re n t from those determ ining risks on dom estic instrum ents. Any investor holding U.S. bonds or Japanese bonds, for exam ple, has to consider the outlook for inflation, real growth, and other factors in those countries that contribute to flu ctu ations in the bond’s dom estic currency price. M oreover, exchange rate m ovem ents, at least in principle, are de te r mined by differences across countries in very much the same set of underlying conditions. From this perspective, the overall size of currency risk prem ia largely reflects the extent to which the im portance of these standard d eterm i nants differs am ong currencies — whether, for exam ple, uncertainties about U.S. inflation are more or less im por tant to investors than uncertainties about inflation in other countries. Likewise, the risk prem ia are likely to change o ve r tim e if and w hen the d e te rm in a n ts change. Thus, assessing relative currency risks involves considerations fairly sim ilar to those that have traditionally guided assess ments of dom estic instrum ents. Risk prem ia generally should decline with international fin a n c ia l in te g ra tio n b e ca u se in te g ra tio n a llo w s m uch greater risk diversification than is norm ally available from holding dom estic assets only. The scope for such d iversifi cation is greatest when exchange rates vary sim ply to o ff set differences in national inflation rates. In that case the relative risks of assets denom inated in different currencies w ould be the sam e fo r all inve sto rs regardless of th e ir nationality (that is, w hether returns are calculated in term s of U.S. or foreign consum ption goods), and their portfolios w ould be very sim ilar in com position. In reality, purchasing power parity does not hold, except perhaps in the very long run, and the variability of real exchange rates does reduce the p o s s ib ilitie s fo r w o rth w h ile d iv e rs ific a tio n by g iving d o m e stic in v e s to rs an e ffe c tiv e h a b ita t p re fe re n c e fo r assets denom inated in th e ir own currency. T hat is, to a German investor (one who assesses returns in term s of German goods), dollar instrum ents appear to be substan tially more risky than a German mark asset, while the oppo site is the case fo r a U.S. investor. N onetheless, even though real exchange rates have often been quite volatile, much evidence suggests that investors can sig n ifica n tly improve their tradeoff between risk and return by devoting a significant portion of their holdings to foreign a sse ts.15 Most standard fram ew orks for assessing risk also sug gest that currency-associated risk prem ia are likely to be fairly modest. In the most w idely used approach, the risk prem ium of any asset is proportional to its contribution to the fluctuations in the value of the m arket portfolio as a w h o le .16 From th is p e rs p e c tiv e , c u rre n c y flu c tu a tio n s account for only a small fraction of the total risk facing a typical investor; unforeseen fluctuations in dom estic asset prices, for exam ple, generally are a much more im portant 15 Recent studies include Levich and Thomas (1993) and Tesar and Warner (1992). Real exchange rate variability is probably one important reason why the portfolios of even the most internationalized financial institutions are far from fully diversified. 16 The framework is known as the “capital asset pricing model," first developed by Sharpe (1964) and Lintner (1965). An individual asset’s risk premium in this framework is proximately determined not only by the asset’s own return volatility but also by its correlation with fluctuations in the other asset prices. Both are determined by the fundamental economic conditions prevailing during a given period and are subject to change over time. Many extensions of this approach have been developed, the most common of which bases asset risk premia on their contribution to the variability of consumption rather than the market portfolio's value. Table 2 Foreign-U .S. Return Differentials in Dollars (Annual Average Percentage Rates) 86-90 70-80 81-9( -2.4 -7 .0 -9 .6 -11.7 17.8 -4 .9 7.2 6.9 8.7 10.3 -5 .0 4.6 -1 .0 3.6 2.5 2.1 — -0 .2 2.4 -0 .2 -0 .7 -1.1 5.7 -0 .2 6.0 — 0 9.8 -7 .3 5.4 -8.1 6.6 -1 .5 -1 .2 -1.1 4.6 -5 .0 1.9 9.0 5.8 -1 .2 6.7 3.2 -6 .6 2.8 3.8 -10.9 10.5 -6 .6 4.2 3.6 2.2 -1 .2 5.7 1.0 11.0 2.3 17.4 -0 .8 14.9 -3 .7 -9.7 -10.0 -10.3 -6 .8 -4 .8 1.2 3.2 7.9 4.6 12.6 0.6 -0 .4 10.6 4.0 5.6 -2.1 9.1 -1 .0 -3.3 -1.1 -2 .9 2.9 -2.1 71-75 76-80 Short rates Canada Germany France United Kingdom Italy Japan 0.4 7.2 5.4 -1 .6 — 4.4 -2 .4 0.2 -0.1 5.8 11.3 6.2 Equity1 Canada Germany France United Kingdom Italy Japan 0.5 — 4.6 -1 .2 -11.8 12.6 Bonds* Canada Germany France United Kingdom Italy Japan -1 .8 10.1 5.7 -6 .2 -3.4 3.3 81-85 Note: Reported values represent the difference between foreign and U.S. average monthly returns, including reinvested earnings, expressed at an annual rate. t The 1970s periods are 1970-75, 1976-80, and 1970-80. 4: Figures are taken from Ibbotson and Siegal (1990). FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 33 source. This point is illustrated in Table 3, which lists e sti mates of the average (ex ante) differential between foreign currency and U.S. dollar-denom inated bonds predicted for the period 1978-91 on this basis.17 The differential returns seem relatively m odest in m agnitude — between 1/4 and slightly more than 1/2 of 1 percentage p oint.18 Limitations of the risk explanation These estim ates suggest that currency-associated risk pre mia based on econom ic fundam entals provide a plausible explanation of w hy system atic return differentials exist and w hy they m ight vary over time. At the same tim e, however, em pirical analyses based on risk considerations have not accounted satisfactorily fo r key aspects of observed return differentials. The main problem is that even after the influ ence of random fo re c a s t e rro rs is ta ke n in to a cc o u n t, 17 The model for these estimates extends the standard capital asset pricing model to an international context and allows for the effects of real exchange rate variability and differing investor consumption preferences; see Lewis (1988). The estimates are derived from the variances and correlations of (real) bond returns and exchange rates for the period. Figures for different intervals will generally differ from those in the table because of the differences in the distribution of the asset returns. The framework used here is essentially the same as that used in Hung, Pigott, and Rodrigues (1989) to estimate the potential effects of the accumulation of U.S. debt to foreign countries. 18 By comparison, since the 1920s, the annual returns on U.S. common stocks have exceeded the yield on U.S. Treasury bills by an average of 6 percentage points, while government bond yields have averaged about 1 percentage point over the bill return (see Ibbotson 1992). Nevertheless, differential returns are highly variable, even across decades. The return differentials for short-term assets implied by this analysis are even smaller than those shown in Table 3 since short-term assets are largely free of price risk. Table 3 Hypothetical Differential Currency Risk Premia for Bonds observed ex post return differentials (such as those shown in Table 1) seem to be too large as well as too variable to be explainable sim ply in term s of risk factors — at least as they are understood by standard risk assessm ent fram e works of the type used fo r the figures in Table 3 .19 Further more, em pirical studies generally have had little success in explaining observed uncovered return differences in term s of the fundam ental econom ic factors thought to determ ine asset risks.20 The shortcom ings of such approaches have led a num ber of analysts to consider an alternative possibility, m en tioned earlier: ex post return differentials am ong currencies m a y re fle c t s y s te m a tic e rro rs in m a rk e t fo re c a s ts of exchange rates and dom estic asset prices, and not sim ply (or even prim arily) risk. Such errors could lead to system atic d ive rg e n ce s in ex post returns even if the ex ante returns expected by investors were equalized (that is, risk prem ia were negligible). A lthough usually ruled out in fo r m al e c o n o m ic a n a ly s e s , w h ic h ty p ic a lly a s s u m e th a t expectations are rational and therefore unbiased, the view that expectations are biased is not im plausible. Studies of survey data on the forecasts of m arket participants and analysts indicate that forecasts are generally biased, often substantially so.21 Market survey data do not, however, support the notion that expectations biases are the main reason fo r the large system atic return differentials observed across countries. If such biases were the reason, we w ould expect that antici pated (ex ante) returns on com parable assets calculated using survey data as a m easure of expected exchange rate changes w ould be fairly small. In fact, as illustrated in Chart 3, this does not seem to be the case. The chart show s the expected return differential, expressed in dollars, between U.S. and foreign three-m onth E urocurrency deposits. The d iffe re n tia ls are ca lcu la te d by su b tra ctin g th e expected change in the relevant exchange rate, taken from a prom i nent su rve y of m a rke t fo re c a s ts , from the U .S .-foreign (Ex ante Return Differential for Foreign Relative to U.S. Government Bonds) Basis Points Canada Germany France United Kingdom Japan -2 4 -6 0 -2 4 -2 4 24 Notes: Figures refer to the annualized differential ex ante yield of a representative foreign government bond over a U.S. counterpart. The estimates are averages for 1986-91 calculated from monthly realized returns on a portfolio of bonds from seven industrial coun tries (the above plus Belgium). The estimates are calibrated so that the ex ante return on the aggregate (world) bond portfolio cor responding to these figures is about 150 points above the U.S. Treasury bill yield. For details of the model used for these calcula tions, see Lewis (1988). 34 FRBNY Quarterly R eview /W inter 19 9 3 -9 4 19 Indeed, the Fama evidence cited earlier implies that risk premia, if viewed as the sole source of observed uncovered return differentials, are the dominant contributor to fluctuations in national short-term interest rates. This implication is both remarkable and implausible; it is hard to see why the normal determinants of domestic interest rates should be so strongly associated with risk. 20 Generally, empirical applications of capital asset pricing models (including consumption-based versions) have not been able to explain observed return differentials either domestically or internationally, and their underlying assumptions are quite often statistically rejected. See, for example, Engle and Rodrigues (1989) and Lewis (1990). Moreover, research to identify the underlying economic determinants of asset price volatility, asset risks, and risk premia has barely begun. 21 Frankel and Froot (1989, 1990) and numerous subsequent papers have demonstrated considerable biases in market forecasts of exchange rates as measured by surveys. Forecasts over near-term horizons tend to draw heavily on recent experience. Earlier studies have shown a similar pattern in surveys of expected inflation. interest rate d iffe re n tia l.22 The return differences, w hich can be viewed as the risk prem ium between the dollar and foreign currency assets that m arket investors expect to receive, appear to be quite substantial, indeed com parable in m agnitude and variability to the historical return d iffe re n tials shown in Table 1. In short, the survey data (assum ing they reasonably represent expectations) seem to confirm the im pression from the ex post return data that investors believe that substantial currency-associated risk prem ia exist. But the question raised e a rlie r rem ains: W hy are these apparent risk prem ia so large compared with those predicted by standard theoretical fram ew orks?23 O verall, th e re fo re , u n c e rta in tie s rem ain a b o u t d iffe r ences in uncovered returns among assets denom inated in alternative currencies as well as the effects that financial integration has had on these differences. Significant and variable com m on-currency return divergences apparently have persisted, but we cannot say to what degree currency risk factors or m arket expectations are responsible, individ ually or collectively, much less w hat the basic econom ic determ inants of the divergences are. Before closing, however, we note that these uncertainties are not peculia r to intern a tio n a l com parisons or foreign exchange markets. Systematic divergences among returns on bonds, stocks, and indeed a wide range of assets have long been observed in domestic markets in the United States as well as abroad.24 Attempts to attribute these divergences to risk or other factors have likewise met with only limited success. As here, these divergences have suggested to m any analysts tha t the determ ination of asset risks and expectations may be much more com plex, and financial markets much less “efficient,” than was previously thought. Quite possibly, com plexities of this sort may be more im por tant in international financial markets, given their shorter his tory and more limited experience relative to domestic finan cial markets, but they probably are not unique. Conclusions There can be little doubt that financial m arkets across the Chart 4 Ex Ante Return Differentials in Dollars Implied by Surveys of Market Expectations Percent United States versus Germany j \ A A \ J / y l \ I / U W \ / \ . i l l i ! \ J v on L m 1111111 i i 11 1 1 1 1 1 1 1 1 1 10 United States versus Japan 22 The premia shown are calculated as the difference between the threemonth U.S. and foreign interest rates for the date of the survey, less the (consensus) expected dollar depreciation over the next three months. The survey data are from Consensus Forecasts, various issues. 23 An alternative possibility is that deviations from uncovered interest parity reflect market expectations about discrete events, such as major policy shifts, that occur only infrequently but have large impacts on asset prices if they materialize. (See, for example, Evans and Lewis 1992.) The situation of the Mexican peso during the 1980s is often cited as an example. Mexican rates were substantially above those for some time in large part because of market perceptions that a devaluation was inevitable. Thus, for a substantial interval before the actual devaluation, dollar returns on peso-denominated instruments were consistently higher than the returns on com parable U.S. alternatives. Deviations from uncovered interest parity seem so pervasive, however, that such factors could only be responsible in fairly isolated instances. 24 A provocative analysis by Cutler, Poterba, and Summers (1990) reveals several stylized facts common to a wide range of asset markets, including those for foreign exchange and those for art and other collectibles. These facts are 1) systematic persistence of excess returns over the near term, 2) some tendency for those returns to be reversed (“mean reversion”) over longer periods, and 3) a tendency for actual asset prices and returns to converge over the long run with the values predicted by economic fundamentals (according to some model). The latter two tendencies, however, appear to be considerably weaker than the first. *5 -15 -?5 L L I I I I I I I I I I _ u 1991 111J1.1L11 1992 i i I, i i J 1993 Notes: The ex ante dollar return difference is the U.S.-foreign interest differential less the survey's consensus forecast of the rate of dollar depreciation over the three months to maturity, expressed at an annual rate. Interest rates are returns on threemonth Eurocurrency deposits. Market forecasts of currency movements are from Consensus Forecasts. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 35 w orld have becom e highly interdependent. News a bout conditions in one country’s markets typically has repercus sions in foreign exchange m arkets and nearly as often in the dom estic money, bond, and equity markets of the c o u n try’s partners. So rapidly do these reactions am ong m ar kets occur that an observer of their daily m ovem ents might easily conclude that dom estic and foreign interest rates are directly and very closely linked. W e have seen that financial integration has indeed had im portant and tangible effects on international interest rate relations. Most obviously, integration has nearly elim inated covered interest differentials among the m ajor m arkets of the industrial countries. But we have also seen that, largely because of the e xis tence of multiple currencies with changeable relative v a l ues, the effects of integration on the international econom y are much less straightforw ard than they are within any sin gle country. In the international environm ent, there are sev eral distinct relations among interest rates that are jointly determ ined by the currency regim e, m arket perceptions about currency fluctuations, and countries’ m acroeconom ic conditions. Localized fluctuations in credit dem ands or su p plies th a t w ould be tra n sm itte d d ire ctly a cro ss m arkets w ithin a single country are, in the international economy, m ore o fte n th a n not s u b s ta n tia lly a b so rb e d in fo re ig n exchange markets. Thus in principle— and as the evidence review ed here stro n g ly suggests, in p ra c tic e — fin a n cial integration need have little if any im pact on divergences am ong natio n a l in te re s t rates, e xce p t w here e xchang e rates are fixed or very nearly so. Financial integration has also led to considerable interna tional d ive rsifica tio n of financial holdings. It thus seem s plausible to expect that national interest rate differentials would tend to be offset by exchange rate changes, so that average returns on com parable assets w ould be substan tially if not com pletely equalized when expressed in a com m on c u rre n c y . In fa c t, h o w e v e r, re tu rn d iffe r e n tia ls recorded over the last two decades appear to have been sizable and system atic. Little is yet known about the exact nature of these differentials or how they are determ ined: in particular, they seem to be too large and variable to be explainable purely in term s of risk considerations — at least as th e y are p re se n tly understood. T hese fin d in g s raise questions about the form ation of in ve sto rs’ expectations and the assessm ent of risk quite sim ilar to those encoun tered in analyses of the term structure of interest rates or the pricing of equities. Thus the issues posed by the inte r national integration of financial markets, while new in cer tain respects, are in others quite fam iliar. References Akhtar, M.A., and Kenneth Weiller. “Developments in Inter national Capital Mobility: A Perspective on the Underlying Forces and Empirical Literature.” In Research Papers on International Integration of Financial Markets and U.S. Mon etary Policy. Federal Reserve Bank of New York, December 1987. Bank for International Settlements. International Interest Rate Linkages and Monetary Policy. Spring Economists’ Meeting, March 1989. Benzie, Richard. “The Development of the International Bond Market.” BIS Economic Papers, no. 32, January 1992. Caramazza, Francesco, Kevin Clinton, Agathe Cot6, and David Longworth. International Capital Mobility and Asset Substitutability: Some Theory and Evidence on Recent Structural Changes. Bank of Canada, Technical Report no. 44,1986. Cutler, David, James Poterba, and Lawrence Summers. “Speculative Dynamics.” National Bureau of Economic Research, Working Paper no. 3242, January 1990. 36FRASER FRBNY Quarterly R eview /W inter 19 9 3 -9 4 Digitized for Davis, E. P. “Financial Market Activity of Life Insurance Companies and Pension Funds.” BIS Economic Papers, no. 21,1988. __________ . “International Diversification of Institutional Investors.” Bank of England, Discussion Paper no. 44, Sep tember 1991. Dooley, Michael, and Jeffrey Shafer. “Analysis of Short-Run Exchange Rate Behavior: March 1973-November 1981.” In D. Bigman and T. Taya, eds., Exchange Rate and Trade Instability. Ballinger, 1983. Dornbusch, Rudiger. “Expectations and Exchange Rate Dynamics.” Journal of Political Economy, August 1976. Engle, Charles, and Anthony Rodrigues. “Tests of Interna tional CAPM with Time-varying Covariances.” Journal of Applied Econometrics, vol. 4 (1989). Evans, Martin D., and Karen Lewis. “Peso Problems and Heterogenous Trading: Evidence from Excess Returns in Foreign and Euromarkets.” National Bureau of Economic Research, Working Paper no. 4003,1992. Fama, Eugene. “Forward and Spot Exchange Rates.” Jour nal of Monetary Economics, vol. 14(1984). Frankel, Jeffrey, and Kenneth Froot. “Forward Discount Bias: Is It An Exchange Risk Premium?” Quarterly Journal of Economics vol 104(1989 __________ . “Exchange Rate Forecasting Techniques, Survey Data, and Implications for the Foreign Exchange Market.” International Monetary Fund, Working Paper no. 90/43, May 1990. Frankel, Jeffrey. “Quantifying International Capital Mobility in the 1980s.” National Bureau of Economic Research, Working Paper no. 2956,1990. __________. “International Capital Flows and Domestic Economic Policies.” In Martin Feldstein, ed., The United States in the World Economy. University of Chicago Press, 1988. __________ . “International Financial Integration, Relations among Interest Rates, and Exchange Rates and Monetary Indicators.” In Charles Pigott, ed., International Financial Integration and U.S. Monetary Policy. Proceedings of a Fed eral Reserve Bank of New York colloquium, October 1989. Howe, Howard, and Charles Pigott. “Determinants of LongTerm Interest Rates: An Empirical Study of Several Indus trial Countries.” Federal Reserve Bank of New York Quar terly Review, Winter 1991-92. Kasman, Bruce, and Charles Pigott. “Interest Rate Diver gences among the Major Industrial Nations.” Federal Reserve Bank of New York Quarterly Review, Fall 1988. Levich, Richard, and Lee Thomas. “Internationally Diversi fied Bond Portfolios: The Merits of Active Currency Risk Management.” National Bureau of Economic Research, Working Paper no. 4340, April 1993. Lewis, Karen. “Inflation Risk and Asset Market Distur bances: The Mean-Variance Model Revisited.” Journal of International Money and Finance, September 1988. Lintner, John. “Security Prices, Risk, and Maximal Gains from Diversification.” Journal of Finance, vol. 20 (1964). Mishkin, Frederic. “Are Real Interest Rates Equal Across Countries: An Empirical Investigation of International Parity Conditions.” Journal of Finance, vol. 39 (1984). Mizrach, Bruce. “The ERM since Basle-Nyborg.” Federal Reserve Bank of New York, unpublished paper, June 1993. Popper, Helen. “International Capital Mobility: Direct Evi dence from Long-Term Currency Swaps.” Board of Gover nors of the Federal Reserve System, International Finance Discussion Papers, no. 382, June 1990. Radecki, Larry, and Vincent Reinhart. “The Globalization of Financial Markets and the Effectiveness of Monetary Policy Instruments.” Federal Reserve Bank of New York Quarterly Review, Fall 1988. Hung, Juann, Charles Pigott, and Anthony Rodrigues. “Financial Implications of the U.S. External Deficit.” Federal Reserve Bank of New York Quarterly Review, Winter-Spring 1989. Sharpe, William F., “Capital Asset Prices: A Theory of Mar ket Equilibrium under Conditions of Risk.” Journal of Finance, vol. 19 (1964). Ibbotson Associates. Stocks, Bonds, Bills and Inflation: Market Results for 1926-91. Yearbook, 1992. Sweeney, Richard J. “Beating the Foreign Exchange Mar ket.” Journal of Finance, March 1986. Ibbotson, Roger G., and Laurence B. Siegal. "The World Bond Market: Market Values, Yields and Returns.” Journal of Fixed Income Research, vol. 1 (1991). Tesar, Linda, and Ingrid Warner. “Home Bias and the Glob alization of Securities Markets.” National Bureau of Eco nomic Research, Working Paper no. 4218, November 1992. FRBNY Quarterly R eview / W inter 1 9 9 3 -9 4 37 Debt Reduction and Market Reentry under the Brady Plan by John Clark In March 1989, U.S. Treasury Secretary Brady proposed a new approach to resolving the developing country debt problem and restoring the creditworthiness of restructuring countries. From the outbreak of the debt crisis in mid-1982, financial packages for restructuring countries had empha sized new lending to give countries time to grow out of their debt-servicing difficulties. However, seven years later, few countries appeared close to returning to normal debt ser vicing and financing was becoming progressively harder to arrange. Drawing on banks’ and countries’ widening experi ence with agreements to convert and reduce debt, Secre tary Brady urged a shift in emphasis toward permanent relief through market-based debt and debt service reduc tion for countries adopting strong economic reform pro grams. This article examines the impact of this new approach on participating countries and their creditors.1 The agreements that followed the new approach pro vided for long-term net cash flows broadly comparable to the net flows previously achieved on a temporary basis through new money packages. Thus, countries were encouraged to embark on reform efforts by a new confi dence that needed financial support would be available over time. Moreover, by marshaling this support through 1 The analysis focuses on the experiences of eight middle-income countries— Argentina, Brazil, Costa Rica, Mexico, Nigeria, the Philippines, Uruguay, and Venezuela— that had obtained agreement to reduce their bank debts by end-1992. Some comparisons are also made to Chile, which significantly reduced its debt through market-based debt conversions. Bank claims have been substantially reduced in several other cases, but the affected claims accounted for a small portion of these countries’ total indebtedness. For example, since 1988 five lowincome countries— Bolivia, Guyana, Mozambique, Niger, and Uganda— have completed buybacks of their debts at steep discounts. These latter operations were largely financed out of grants and concessional loans from official creditors. FRBNY Quarterly Review/Winter 1993-94 Digitized38 for FRASER market-based debt reduction, the new approach contained the growth in debt and fostered cooperation between debtors and creditors. Nonetheless, the immediate benefits to countries should not be exaggerated. The need to con tinue reform efforts was underscored by countries’ ongoing debt burdens, which remained heavy notwithstanding the reductions in claims, and the persistence of deep discounts on the countries’ external obligations immediately after the restructurings. Indeed, for countries that in recent years had unilaterally curtailed interest payments, such as Argentina and Brazil, restoring normal relations with credi tors through Brady restructurings required significant increases in debt service payments. The ultimate results of the change in approach have been impressive. In particular, several countries that mounted sustained reform efforts and reduced their debts have bene fited from growing market access on improving terms. Although stronger economic performance by debtors has undoubtedly been the key to reopening market access, the Brady operations catalyzed and accelerated this process. Because the Brady agreements provided cash flow relief over a longer time horizon than conventional restructuring packages and insulated countries from possible future inter est rate increases, the operations improved prospects for breaking the cycle of continual renegotiation that impeded capital flows under the previous approach. In the event, lower global interest rates have made the Brady operations more effective by giving countries additional cash flow relief and encouraging investors seeking alternatives to low-yield ing industrial country investments to reevaluate restructur ing countries’ payment prospects. The change in approach has also contributed to the recovery in the secondary market value of creditor claims, enhanced the claims’ liquidity, and helped create expanded income opportunities in the secondary market trading of restructured bank claims and the underwriting of securities flows to restructuring countries. From mid-1988 until Febru ary 1989, as the market’s confidence in the existing new money approach waned, the price of claims in what remained a fairly thin secondary market declined sharply. In fact, the amount of debt reduced in relation to cash out lays in the early Brady deals was broadly consistent with what could have been achieved through a cash purchase at these lower market prices. The subsequent substantial appreciation of prices, which came with a lag, reflected the market’s reassessment of the reinforced strategy’s overall prospects for success in an environment of improved macroeconomic performance by several countries as well as lower global interest rates. The Brady Plan and the evolving debt strategy While reaffirming the basic tenets of the existing debt strat egy—a case-by-case approach stressing reform by debtor countries and financial support from private and official creditors— the Brady Plan introduced important innova tions. Tactically, the new approach emphasized using financial incentives such as collateralized partial guaran tees to encourage banks to provide financial relief. At the strategic level, the new initiative completed an evolution toward longer term horizons in bank debt restructuring packages by emphasizing permanent relief through princi pal write-downs and interest reductions. The pre-Brady new money approach When the debt crisis erupted, the international commu nity— debtors, creditors, governments, central banks, and international financial institutions— moved swiftly to avert a systemic disruption of international trade and finance.2 The strategy emphasized cooperation among debtors and cred itors and timely financing to allow countries to reorient their economies while remaining current on interest payments. Banks rescheduled amortization payments falling due and in arrears, maintained short-term credit lines, and in effect partially refinanced interest obligations by extending new loans. Multilateral creditors— initially the International Mon etary Fund (IMF) and later the World Bank— increased their lending. Countries tightened their belts by cutting public investment and noninterest current expenditures and by 2 Worries about the international financial system grew out of the risks to the international banking system posed by the high exposure to developing country debt. For example, at the end of 1982, exposure to restructuring developing-country borrowers equaled 215 percent of the capital and 260 percent of the equity of the U.S. money center banks. Many of the large regional banks also were heavily exposed, as were leading banks of other industrial countries. For example, at the end of 1984 the less developed country exposures of the major banks of the United Kingdom and Canada were about 275 percent and 195 percent of equity, respectively; see David Mengle, “Update: Banks and LDC Debt,” Morgan Guarantee Trust Co., Economic Research Note, May 1992. devaluing their currencies and slashing imports.3 As restoring creditworthiness proved a time-consuming process, the strategy adopted a progressively longer hori zon. Debt packages became more comprehensive, often restructuring the entire stock of medium-term bank debt rather than just the obligations falling due in a one- to twoyear period. Repayment periods lengthened from around eight years out to as much as twenty years, and interest rate spreads narrowed to 13/16 of a percent over bank funding costs.4 On the policy side, the emphasis broadened under the “Baker Plan” to include structural reforms, such as trade liberalization and tax reform, that were designed to enhance countries’ longer term growth prospects.5 To sup port faster growth, the Baker initiative also called for increased official and commercial bank lending. By 1989, this basic case-by-case approach had achieved some measure of success. It had afforded banks the time to increase their capital, thereby containing systemic threats to the international financial system.6 In addition, after peaking at mid-decade, most restructuring countries’ debt and debt service indicators had begun to decline (Chart 1). Nonetheless, important strains had emerged, leading to deepening fatigue and frustration for both debtors and creditors. The net cash drain on debtors remained burden some and the goal of countries’ servicing their obligations without further extraordinary financing arrangements remained distant. While principal deferrals were longer, relief from interest payments continued to be of short dura tion because new loans covered a fraction of the interest falling due only during a tw o-year period. Debtor economies had grown disappointingly slowly and in many cases policy reform had not been adequate. Rates of capi tal formation had failed to recover from their sharp declines at the onset of the crisis, and domestic investors continued to express their lack of confidence by hoarding financial 3 Nonetheless, as a result of transfers of external debt obligations from the private to the public sector and the public sector’s greater reliance on more expensive internal financing following the cutoff of international bank lending, overall deficits declined by less than the improvements in the noninterest balances of the central governments. For a review of fiscal adjustment by several major debtors, see William Easterly, “Fiscal Adjustment and Deficit Financing during the Debt Crisis," in Ishrat Husain and Ishac Diwan, eds., Dealing with the Debt Crisis (Washington D.C.: World Bank, 1989). 4 Reschedulings at the onset of the debt crisis typically entailed spreads over LIBOR ranging between VA and 2 / percent. 5 This adaptation of the official strategy, emphasizing growth-oriented reform and new lending, was adopted along lines suggested by United States Treasury Secretary Baker during presentations at the October 1985 annual meetings of the World Bank and IMF, and hence was named the “Baker Plan.” 6 By end-1988, U.S. money center banks’ exposure to restructuring countries in relation to capital had been cut by more than half, to about 95 percent. FRBNY Quarterly Review/Winter 1993-94 39 assets abroad. The growing liabilities of international banks to depositors from restructuring countries partly indicate the extent of this capital flight (C hart 2). A t the sam e tim e, on the fin a n cin g side, new m oney packages becam e increasingly difficult to arrange. The new money approach relied on banks to act in their collective interest even though, individually, banks m ight have pre ferred to “free ride”— that is, to benefit from the financial packages by receiving interest payments w ithout providing new money. The approach was successful so long as the contradiction betw een co lle ctive and individual interests was not too severe. However, as the m arket’s confidence in the prevailing strategy tum bled— as revealed by secondary m arket discounts that w idened from one-third at end-1986 to an average of tw o-thirds by early 1989— disbursing cash in return fo r uncertain loan claim s appeared ever more unattractive.7 The slide in secondary market prices in part reflected co u n trie s’ uneven econom ic perform ances and the souring of the general atm osphere that followed some c o u n trie s’ im positions of unilateral paym ents m oratoria. M o re o ve r, th e s tre n g th e n in g of bank b a la n c e s h e e ts , including increased loa n -lo ss provisioning by the m ajor banks, reduced the adverse consequences of tem porary paym ent interruptions and allowed banks to take a harder line in negotiations.8 This tension between stronger coun tervailing individual interests and weakened perceptions of collective interest produced a growing num ber of free riders and in cre a sin g ly con stra in e d the fe a sib le fin a n cing that could be raised through new money. W hile ever more banks resisted new lending, at least some banks were willing to sell their claim s. By 1987 most m ajor debtors had instituted conversion schem es under w hich foreign debt could be exchanged for local currency to make direct or portfolio investm ents. Some banks directly transform ed th e ir loan claim s into equity stakes in local businesses; others sold their claim s at a discount for cash to foreign or local investors who in turn undertook the co n version. Debt retirem ents under ongoing official debt co n version schem es rose from a total of $3.7 billion in 1984-86 8 When exposure was high in relation to banks’ capital, banks had stronger incentives to cooperate with the debtor to prevent the loan from lapsing into nonperforming status. Analyses of the rationale for and drawbacks of the new money process can be found in William Cline, International D ebt and the Stability of the World Economy (Washington, D.C.: Institute for International Economics, 1983); and Paul Krugman, “Private Capital Flows to Problem Debtors," in Jeffrey Sachs, ed., Developing Country Debt and Economic Performance (Chicago: University of Chicago Press, 1989). 7 This concept of cost, based on the difference between the amount of new money disbursed and the expected future receipts associated with the new claim, did not necessarily accord with the “accounting cost” of providing new money. The regulatory authorities of some creditor countries required banks to establish reserves against their new money loans. Even where such provisioning requirements did not exist, however, new money could be perceived as lowering shareholder wealth if free riding presented a viable alternative. Chart 2 Cross-Border Liabilities of International Banks to Nonbank Depositors from Selected Restructuring Countries Billions of U.S. dollars 80----------- —-----------------------------------------------------Volume at End of Period Chart 1 Evolution of the External Debt Burden of Selected Developing Countries, 1981-89 70 60 Percent 400 ............................................................................ .... .. External Debt as a Percentage of Exports of Goods and Services _ 50 300 40 Group of fifteen heavily indebted countries* 30 200 20 ...................... 100 ................ Countries without recent debt-servicing difficulties I ° I 1981 I 82 I 83 I 84 10 I 85 I 86 87 1..1 J 88 89 Source: International Monetary Fund, World Economic Outlook. * Argentina, Brazil, Bolivia, Chile, Colombia, Cote d'Ivoire, Ecuador, Mexico, Morocco, Nigeria, Peru, the Philippines, Uruguay, Venezuela, and Yugoslavia. FRBNY Quarterly R eview /W inter 1 993-94 Digitized 40 for FRASER 0 1981 82 83 84 85 86 87 88 89 90 91 92 Source: International Monetary Fund, International Financial Statistics. Note: The selected restructuring countries are Argentina, Brazil, Chile, Mexico, and Venezuela. to $4.7 billion in 1987 and $8.8 billion in 1988.9 In one important transaction in 1988, Mexico used its reserves to finance a debt-for-debt exchange in which $3.7 billion of bank loans were swapped for $2.6 billion of partially collat eralized twenty-year bonds.10 In addition, large amounts of cross-border debt were extinguished through unofficial conversions, particularly in Mexico and Brazil.11 These lat ter transactions usually involved direct negotiations between corporations and their foreign bank creditors. Nonetheless, despite the demonstrated increased willing ness of banks to sell, debtor countries became increasingly disenchanted.12 The burden of essentially prepaying exter nal debt at a discount proved difficult for fiscal and mone tary authorities to manage. Concerns about possible adverse inflationary or balance-of-payments impacts led many countries to suspend or curtail their official programs by early 1989. The Brady Plan The need for a new, more comprehensive, and longer last ing approach was widely appreciated.13 Against this back ground, Secretary Brady proposed a shift in emphasis toward permanent relief through market-based debt and debt service reduction. Instead of providing new money, 9 Charles Collyns and others, Private Market Financing for Developing Countries (Washington D.C.: International Monetary Fund, December 1992). 10 In some other notable experiments in 1988, Venezuela’s bank creditors disbursed $100 million in cash and swapped $400 million of loans for $500 million of new securities, and banks exchanged $1.1 billion of loans for an equivalent amount of uncollateralized “exit" bonds carrying a 6 percent fixed interest rate as part of the financial package for Brazil. Most of the experiments of 1987-88 were less successful than countries had hoped. However, the lessons of these experiments were later applied to the Brady restructurings. 11 Eli Remolona and Paul DiLeo estimate that $11.4 billion of Brazilian and Mexican debts were canceled through informal conversions in 1987-88 (“Voluntary Conversions of LDC Debt,” in Kate Phylaktis and Mahmood Pradhan, eds., International Finance and the Less Developed Countries [London: MacMillan, 1989], p. 75). 12 Major U.S. banks initially remained on the sidelines. However, following the increases in reserves against developing country debt by money center banks in 1987, several U.S. money center banks became more active. For example, Citibank reported in 1989 that it had reduced cross-border exposure to developing countries by some $2.4 billion at an average discount of about one-third. Regulatory changes that allowed U.S. banks to take larger equity shares in companies as a result of debt conversions also facilitated greater participation. See Mark Allen and others, International Capital Markets: Developments and Prospects (Washington D.C.: International Monetary Fund, April 1989). 13 By the eve of Secretary Brady’s speech, many prominent individuals and politicians had made public proposals for new approaches involving elements of debt reduction. Included in this group were French President Mitterand and Japanese Finance Minister Miyazawa. Moreover, in 1988 the U.S. Congress had directed the Treasury to study the feasibility of creating an international debt management authority to purchase the bank debts of developing countries in the secondary market and to pass the discount along to the debtors. banks would voluntarily reduce their claims on the debtor countries in return for credit enhancements on their remain ing exposure, such as collateral accounts to guarantee the principal and/or interest in a bond exchange, or cash pay ments in the context of buybacks. To support countries’ economic reform programs and help debtors make the required up-front cash outlays for the debt operations, official creditors would provide finan cial assistance. Under the strategy, reforming countries would continue to benefit from loans from the IMF and World Bank, reschedulings from Paris Club creditors, and loans and loan guarantees from government agencies. However, a portion of the loans from the Fund and Bank would be set aside specifically to finance operations involv ing debt reduction. Additional Fund and Bank financing could also be made available to fund interest guarantees.14 To receive such support, countries would need to adopt strong policies to ensure that they would be able to service their reduced debt burdens. Measures to promote domestic savings and the repatriation of flight capital, such as remov ing interest rate controls, received particular emphasis. In addition, countries would be encouraged to maintain ongo ing debt conversion schemes to provide additional relief. By offering individual banks direct financial incentives, such as collateralized guarantees, to provide the targeted levels of financial relief, the new approach addressed the contradiction between individual and collective interests that had increasingly troubled its predecessor. Whereas high discounts and increased capital levels had worked against the new money strategy, they actually supported the new approach. High discounts allowed limited amounts of public moneys to “buy” a higher targeted level of cash flow relief, while strengthened capital and reserves allowed banks to take the hit on their balance sheets. The new plan in action: the menu approach In implementing the Brady approach, countries and bank steering committees negotiated comprehensive packages that offered “menus” of debt and debt service reduction options. These menus, which differed in their details from case to case, gave banks a range of choices that varied from as few as two to as many as six. From the debtors’ perspective, these packages were equivalent in impact to a combination of a partial debt buy back at market prices and a restructuring of the remain der.15 The restructurings usually securitized the claims— that is, converted the form of the claims from loans to 14 The distinctions between the uses of “set-asides" and “additional financing" were relaxed in January 1994. 15 The first debt package for the Philippines, completed in 1990 and involving a buyback and new money, differed from other Brady packages in that it deferred the handling of the remaining exposure to a subsequent operation. The second stage was completed three years later. FRBNY Quarterly Review/Winter 1993-94 41 bonds—and lengthened the repayment periods, sometimes to as much as thirty years. Much of the remaining exposure (about half in total) was converted from floating to fixed rate obligations. For the banks, the Brady operations offered complex ranges of options designed to accommodate banks’ diverse needs and expectations (Chart 3 ).16 At one extreme, some menus included buyback options—that is, outright sales of bank claims at a discount for cash— that enabled well-provisioned, pessimistic, or risk-averse banks to exit completely. At the other extreme, new money/debt conversion options permitted banks to exempt their existing exposure from debt and debt service reduction and usually to convert the exposure into a security, provided that they disbursed fresh money. Such financing in turn helped coun tries replenish reserves used to finance the up-front costs of debt reduction options chosen by other banks. Although the disbursement of new money for risky bonds was costly, some optimistic banks were attracted to the possibility of capital gains on their base exposure. Such gains might be anticipated because of securitization or because the debt reduction agreed to by others decreased competing claims. Banks valued securitization because it imparted greater liq uidity to their claims. In addition, since securitized claims would be more widely held, a future restructuring would be more difficult to organize and hence less likely.17 Discount and par exchanges, which combined elements of both a buyback and a restructuring, proved the most popular options (Chart 3). Creditors swapped existing loans for new bonds with a lower principal amount (discount exchange) or with the same principal but submarket, fixed interest rates (par exchange). Instead of receiving cash as in a buyback, creditors benefited from the attachment of irrevocable collateral accounts to the securities. Most com monly, the principal would be fully secured by zero coupon U.S. Treasury bonds, and the next twelve to eighteen months of interest payments would be backed by escrowed high-grade short-term securities.18 If the country remained current on interest, the interest guarantee would roll for ward, covering the next twelve- to eighteen-month period, but usually the interest earnings from the escrow account would return to the debtor.19 Altogether, par and discount exchanges reduced banks’ economic exposure by the pre sent value of the outright interest or principal reduction plus the present value of any principal or interest guarantees. For the debtor, the collateral accounts also effectively reduced the burden of the debt because expected rebates of interest and later principal from the accounts would eventually cover the cost of funding the collateral accounts.20 By contrast, in the case of a simple buyback, there would be no prospect of future rebates; a country’s debt would merely decline by the amount of debt purchased and increase by the borrowings to finance the operation. Par and discount exchanges generally entailed lower cash outlays in relation to exposure reduction than did buy backs or secondary market sales.21 Many banks were nonetheless attracted to bond exchanges rather than out right sales because of the upside potential on the remaining exposure— again owing to securitization and the reduction in claims. Relative to new money/debt conversion options, bond exchanges held the additional attraction for banks of concentrating remaining unsecured exposure into interest claims, which were less commonly rescheduled than amor tization obligations. Against this, par and discount ex changes required a longer maturity for the remaining expo sure than debt conversion options and usually involved registered rather than bearer securities. 16 The range of options has varied across packages from just two for Argentina (par and/or discount exchanges) and Costa Rica (buyback and/or par exchange) to as many as six (Brazil). All packages have included at least one bond exchange option. Buybacks were included in all packages except those for Argentina, Brazil, and Mexico. New money options were omitted in the Argentina, Costa Rica, Dominican Republic, and Jordan agreements. More information on the structure of individual agreements and bank choices may be found in World Bank, World Debt Tables, various issues, and Collyns and others, Private Market Financing for Developing Countries, 1992 and 1993. 20 Usually with collateralized guarantees, as the country serviced its debt it would receive rebates of interest earned by the interest collateral account; at maturity the country would also receive the principal and accrued interest in the principal collateral account and the collateral deposited in the interest account. The rebates of course would be expected to equal in present value the money originally borrowed or drawn from reserves and deposited in the accounts. Hence, the gross debt reduction achieved through, say, a discount exchange would typically be equal to the discount times the exchanged debt plus the present value of the expected rebates, while the net debt reduction (which takes into account financing costs) would be equal to just the discount. For a further discussion of guarantee structures and the concepts of gross and net debt reduction, see John Clark, “Evaluation of Debt Exchanges," IMF Working Paper 90/9, 1990. 17 Against this view, it could be argued that the difficulty of rescheduling widely held bonds would make debtors' future cash flow problems more difficult to resolve and would increase the likelihood of default should difficulties arise. Some advocates for securitization pointed to restructuring countries’ record of regularly servicing their bonds throughout the 1980s as evidence that the new securities would be serviced better than the previous loans. However, this argument ignores the likelihood that the privileged servicing record of bonds has owed more to their small share of total debt than to their actual form. 42 FRASER FRBNY Quarterly Review/Winter 1993-94 Digitized for 18 The amounts deposited in the interest guarantee accounts varied from 7 to 13 percent of the expected present value of the interest streams on the bonds, depending on the number of months covered and the interest rates involved. 19 Temporary interest reduction bonds differed in that the interest would cumulate in the interest collateral account. When the temporary interest reduction expired after about six years, the collateral and accrued interest would be returned to the debtor. 21 As shown in Collyns and others, Private Market Financing for Developing Countries, pp. 12-13, the ratio of collateral costs to exposure reduction was generally slightly lower for par and discount exchanges than prevailing secondary market prices. In contrast, buybacks took place at the prevailing price. U niform ity and diversity in terms Brady packages have shown tendencies toward both uni form ity in the design of som e aspects of individual options and a tailoring to countries’ individual needs. On the one hand, the discount and par exchanges, the prim ary debt re d uction ve h icle s fo r the pa cka g e s of the fo u r largest debtors, generally kept the extent of principal or interest rate reductions at about o n e -th ird because banks were u n w illin g to g ra n t te rm s m o re fa v o r a b le th a n th o s e accorded M exico.22 The lone exception was the discount exchange for Bulgaria agreed upon in principle in N ovem ber 1993; the agreed term s in this case specified a discount of one half. On the other hand, through differing degrees of 22 The par exchange for Mexico, whose terms were agreed to in July 1989 while LIBOR stood at 8.81 percent and thirty-year Treasury bonds were yielding 8.14 percent, specified a fixed interest rate of 6.25 percent. Reflecting subsequent movements in the yield curve, some later agreements have specified initial coupon rates as low as 4 percent, which gradually rise to levels similar to those negotiated with Mexico. Chart 3 Principal Restructuring Options in a Brady Menu Original Payment Obligation 9% Menu Option Immediate Payment or Enhancement Residual Payment Obligation Buyback Cash payment None Par exchange Prepayment of principal and 9 to 12% of remaining interest; securitization of remaining obligations Fixed interest stream, usually at a rate of about 6.25%, less rebate received by debtor of earnings on interest collateral account * Discount exchange Prepayment of principal and 7 to 13% of remaining interest; securitization of remaining obligations Floating interest stream at a rate of LIBOR + 13/16 on a reduced (by 30-35%) principal amount, less rebate received by debtor of earnings on interest collateral account Temporary interest reduction exchange Prepayment of about 10% of remaining interest; securitization of remaining obligations Rising submarket fixed interest stream, switching to LIBOR + 13/16 after 5-6 years; amortization of principal t Debt conversion/ new money Securitization of remaining obligations (new loans equal to about one-fifth of base exposure)** Interest of LIBOR + 7/8 + amortization of principal 46% Source: Federal Resen/e Bank of New York staff estimates. Notes: Most menus did not include the full range of options. Several packages provided for the refinancing of outstanding overdue interest at market rates following an initial cash down payment. Percentages show proportion of aggregate principal allocated to each menu option for agreements concluded by mid-1993. Countries also achieved debt relief through debt conversions. These conversions took place before and after the Brady operations but were not part of the menu in a Brady exchange. * Initial interest rates were sometimes lower (for example, 4 percent for Argentina), reflecting the shape of the yield curve at the time of agreement in principle. t Rates reflected term structure at the time of agreement in principle. * * The Mexican new money option did not entail securitization of the base. The Brazilian agreement provides for an interest capitalization option in addition to a debt conversion/new money option. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 43 Chart 4 Buyback Equivalent Prices of Brady Packages Cents per dollar of contractual claim 6 0 ------------------------------------------------------------------------------------- 10 20 30 40 50 Average secondary market price prevailing from March 1989 until agreement in principle Sources: Salomon Brothers and Federal Reserve Bank of New York staff estimates. Notes: The buyback equivalent price is the price at which the same amount of cash could have purchased an equivalent amount of debt reduction through a buyback. It is calculated as the ratio of the actual cost of the package to the amount of gross debt reduction achieved. The gross debt reduction comprises the outright principal reduction through buybacks and discount exchanges, the present value of interest reduction on par exchanges, and effective prepayments of principal and interest through collateral accounts. The solid line plots the results of a cross-sectional regression of the buyback equivalent prices (BEP) on the price of the Mexican agreement (Mexican BEP) and the secondary market price for each country's debt during the period of negotiations (Avg Price). T-statistics are shown in parentheses: BEP= 0.77 (Avg Price) + 0.29 (Mexican BEP) + u. (6.06) (2.07) R2 = 0.78 This regression suggests that the price of a Brady deal can be expected to reflect a weighted average of the secondary market price prevailing during negotiations (3/4 weight) and the price established for Mexico (1/4 weight). * The buyback equivalent prices for Argentina and Costa Rica do not reflect down payments made at closing against interest arrears. Inclusion of these costs would raise the Argentina price by around 5 cents and the Costa Rica price by about 2 cents. t The estimated price for Brazil reflects bank choices among options and interest rates prevailing in July 1992, when the terms of the package were agreed upon in principle. The actual cost of the package may be higher because long-term interest rates have subsequently declined, raising the cost of thirty-year zero coupon bonds. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 Digitized 44 for FRASER 60 collateralization, effective pricing varied in a m anner corre lated with the differing secondary m arket discounts prevail ing before the deals (C hart 4). For example, Argentina and Brazil collateralized only tw elve months of interest w hile M exico collateralized eighteen; likewise, C osta Rica did not guarantee the principal on its par bonds. In addition, the range of options included varied from package to package, reflecting the circum stances of particular cases. In particu lar, sm aller debtors were able to achieve higher percentage reductions in claim s payable to banks by securing a greater role for buybacks.23 In many cases, countries were slow to im plem ent and sustain policy changes th a t w ould provide the basis for needed o fficia l fin a n cia l support. In a d d ition , the richer menu of options made negotiations more com plex, particu larly when precedents did not yet exist or countries tried to vary from the precedents. In cases where significant inte r est arrears had accum ulated, agreem ent on the level of paym ents to banks ahead of com pletion of the debt pack age often presented a key hurdle. Moreover, reconciling o verdue in te re st claim s proved arduous. As a result of these fa cto rs, im plem enting the new stra te g y has been tim e consum ing. Still, progress has been steady and com prehensive packages have been com pleted for eight coun tries: Mexico, C osta Rica, and V enezuela in 1990; Uruguay in 1991; N igeria and the P hilippines in 1992; A rgentina in April 1993; and Jordan in Decem ber 1993. Also in 1993 banks form ally com m itted to participate in the package for Brazil and agreem ents in principle w ere reached fo r the Dominican Republic and Bulgaria. Altogether, the first ten of these m iddle-incom e countries account fo r about fourfifths of all bank claim s on countries that had encountered debt-servicing d ifficulties at the start of the last decade. D iscussions are in progress in a num ber of other cases, including Poland, Peru, Ecuador, and Panam a.24 A dvantages a nd disadvantages o f the m enu approach The menu approach encouraged nearly universal participa tio n and helped co u n trie s m axim ize the d ebt reduction achieved with a given am ount of collateral resources by allowing banks to choose options that best fit their particu lar tax, regulatory, and accounting situations, as well as 23 Countries could encourage more banks to choose the buyback by offering a relatively attractive price. Costa Rica’s Brady agreement was contingent on banks' offering at least 60 percent of their aggregate exposure to the buyback option. To encourage individual banks to tender at least 60 percent of their claims to the buyback option, Costa Rica offered more attractive terms, in the form of guarantees and shorter maturities, on the remaining exposure of banks that met that threshold. 24 Not all countries with recent bank debt-servicing difficulties have sought Brady-type restructurings. Among the countries targeted for special attention under the Baker initiative, Chile has achieved a more manageable debt profile through debt conversions that canceled much of the country's medium-term debt, while Colombia and Morocco have refinanced principal without reducing debt. their views on interest rates and the countries’ prospects. N onetheless, for countries, the approach introduced uncer tainty as to the overall cost and impact of the packages. For example, if banks allocated too much exposure to the debt reduction options, the cost m ight exceed available fin a n c ing, whereas if banks opted excessively for new money, the country might not achieve its debt reduction objectives. In this context, par and discount exchanges were often a ttra c tive to countries because they em bodied in one option an ou tcom e close to the ove ra ll d e sire d mix, and th e re b y reduced uncertainty surrounding the overall im pact of the package. In addition to the above uncertainties, allowing banks to choose among options proved costly to countries when the external environm ent changed between the tim e of agree ment on a menu and the actual selection of options. This com plication reflected the convention, still observed, of fix ing the interest rates and guarantees at the tim e of agree ment in principle rather than indexing them to m ovem ents in m arket rates before the com pletion of the deal. As a result, m ovem ents in rates could shift the overall pricing and also favor some options over others. This problem did not arise with the early bank packages but em erged as an im portant issue for A rgentina and Brazil, which saw a fall in long-term interest rates follow ing agreem ent in principle with banks on a restructuring menu. These declines, to the extent they were unhedged, increased the cost of the T rea sury zero coupon bonds used to secure the Brady bonds’ principal, raising collateral costs for both the par and d is count bonds. The cost increase was more pronounced for the par bonds because they had a larger principal am ount to be secured. In addition, when the gap narrowed between m arket rates and the agreed fixed rates for the par bonds, banks stron gly preferred the par option, w hich becam e more costly for the debtors. In both cases the countries sought a “ rebalancing” or reallocation of choices away from the unexpectedly less concessional par exchange. Financing The debt operations entailed large up-front cash outlays for buybacks, collateral purchases fo r the bond exchanges, and in som e cases dow n paym ents on arrears. O fficial sources provided the bulk of the financing of these costs for the seven operations that have been com pleted.25 In partic ular, three-fifths of the overall financing came from official sources, although in every case the debtor also made a sig nificant contribution (Table 1). However, fo r M exico, the P h ilip p in e s, and V e n ezue la , new m oney c o m m itte d by b anks e ffe c tiv e ly co ve re d a s u b s ta n tia l p o rtio n of the debtors’ share of the financing burden. Interpretation of the fin a n c in g of the A rg e n tin e p a ck a g e is m ore co m p le x . Although A rgentina did not receive new money, banks refi nanced accum ulated interest arrears. H ence, to a large extent the resources th a t A rgentina is expected to con tribute are the coun te rp a rt of e a rlie r unpaid interest. To date, only the financing for N igeria’s debt operation breaks with the prevailing pattern. N igeria received neither new m oney nor direct official financial support.26 D ebt conversions As noted above, the revised official strategy encouraged the m aintenance of debt conversion schem es.27 In n e g oti ating th e ir debt packages, m ost countries agreed to m ain tain or reestablish debt conversion schem es and to carry out an agreed m inim um level of conversions. In contrast with the Brady packages, w hich w ere concerted o p e ra tions that dealt w ith all the debt at one go on preset term s, these co n ve rsio n s w ere u su a lly sm a lle r scale, ongoing operations that involved auction m echanism s. O verall, the pace of conversions did accelerate after 1989, with some $28 b illio n in cla im s c o n ve rte d u n d e r o ffic ia l schem es from 1989 thro u g h 1992. For m ost countries, how ever, these debt co nversions played a sm aller, com plem entary role to the B rady p ackages in reducing c o u n trie s ’ debt 26 Under the Nigeria agreement, since all debt service arrears were to be eliminated before the closing, no effective financing was achieved through arrears. 27 The IMF and World Bank guidelines on support for debt and debt service reduction explicitly endorsed the existence of debt equity swap programs as a useful step in encouraging investment. Banks pressed strongly for debt conversion schemes, reflecting beliefs that such programs enhanced the value and liquidity of their claims. Table 1 Financing for Debt Reduction Packages Billions of Dollars Own Total Cost Official of Operation Support1 Reserves IVICI1lUl dl iUUIIl. New Money from Commercial Banks 7.12 0.22 2.38 0.46 1.70 1.80 3.64 5.33 0.18 1.46 0.06 0.00 0.88 2.53 1.79 0.04 0.92 0.40 1.70 0.92 1.12 1.20 0.09 — 0.85 — Total 17.32 (as a percentage of total cost) (100.0) 10.44 6.89 3.23 (60.3) (39.8) (18.7) Mexico Costa Rica* Venezuela Uruguay Nigeria Philippines Argentina* 1.09 — in c lu d e s disbursements of parallel financing from Japan Eximbank. Although not directly tied to debt reduction, this financ ing supported the programs of several countries that com pleted debt packages. ♦Includes down payments made at closing against interest 25 Most of the operations involving middle-income countries were directly or indirectly financed with loans or reserves; grants have more commonly been used to finance operations for low-income countries. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 45 service burdens (Table 2). The principal exceptions to this rule have been Chile and Argentina, w hich account for about three-fifths of the debt converted under official schem es since 1989. For Chile, debt-equity conversions constituted the prim ary m eans of reducing debt ow ed to banks, although som e debt was re d u c e d th ro u g h b u y b a c k s in 1 9 8 8 a n d 1 9 8 9 . F o r Argentina, debt conversions ahead of its Brady operation were an integral part of the country’s overall debt reduction strategy. From 1990 onward these conversions consisted entirely of exchanges of debt for equity in privatized firms. T he re d u ctio n in b ank cla im s th ro u g h such o p e ra tio n s exceeded th a t achieved th rough the d ebt package and more than offset the $8 billion accum ulation of bank debt from 1988-92 stem m ing from interest arrears. Impact on countries’ debt and debt service burdens T he B rady o p e ra tio n s gave c o u n trie s a leg up in th e ir efforts to surm ount the ir debt-servicing difficulties. E ssen tially the operations provided perm anent cash flow relief on a sca le c o m p a ra b le to th e te m p o ra ry re lie f p re v io u s ly achieved through new money packages. N onetheless, sig nificant debt service obligations remained, to other cre d i tors as well as to banks, so that debtors had to continue to pursue sound econom ic policies to service the rem aining debt and maintain growth. Reduction in debt service obligations The seven Brady Plan operations com pleted to date are e x p e c te d to c a n c e l d e b t s e rv ic e o b lig a tio n s w ith an expected present value of roughly $50 billion, or about onethird of the eligible bank debt (Table 2). The expected per centage reductions in the present value of gross claim s payable to banks have differed across cases, from a low of about three-tenths for Venezuela and A rgentina to about four-fifths for C osta Rica and N igeria.28 W ith the com pletion of the Brazil package, the present value of obligations ca n celed is expected to rise to about $65 billion. Com parison o f changes in debt stock a nd d ebt service obligations The stock of debt to banks, however, will decrease by a much sm aller am ount. R oughly th re e -fifth s of the gross reduction in debt service burdens is expected through in ter est rate reductions on par exchanges and effective prepay28 The gross reduction in claims payable to banks might alternatively be called the gross reduction in bank exposure. It measures the partial effect of those features of the packages that reduce debt and debt service; it does not include the increases in debt to banks through new money. It is the sum of the reduction in principal through discount exchanges and buybacks, the present value of debt service reduction on the reduced interest par bonds, and the prepayment of principal and interest through collateral accounts. The present value of the interest reduction is an ex ante calculation based on the long-term interest rates prevailing when agreement in principle was reached. U B illiliiil Table 2 ll& tH i . I Debt Reduction through Concerted Bank Packages Gross Reduction in Claims Payable to Banks1 Net Debt Reduction* as a Percentage of: Memorandum: Debt Retired under Official Debt Conversion Schemes5 (Billions of Dollars) Billions of Dollars Percent of Eligible Bank Debt11 GDP (1991) Exports (1991) Total External Debt (1989) 1984-92 1989-9; Mexico Costa Rica Venezuela Uruguay Nigeria Philippines Argentina Braziltt Chile 21.1 1.2 6.4 0.9 4.3 3.7 10.5 16.0 N.A. 43.5 75.0 32.1 55.3 79.6 57.8 36.7 27.1 N.A. 5.1 13.5 7.5 4.7 8.2 4.2 5.3 2.8 N.A. 30.8 43.0 21.6 18.1 19.4 13.0 46.3 32.1 N.A. 14.6 25.2 12.2 10.3 8.1 6.8 10.5 10.1 N.A. 7.3 0.3 1.7 0.2 0.8 3.2 12.9 5.2 11.4 3.1 0.2 1.6 0.1 0.8 2.6 11.3 1.3 5.2 Total/(average)n 64.0 (38.2) (4.8) (30.6) (11.8) 43.0 26.1 Sources: International Monetary Fund; World Bank; Federal Reserve Bank of New York staff estimates. f Principal reduction through discount exchanges and buybacks, present value of reduction in interest rates on interest reduction bonds, and prepay ments of principal and interest through collateral accounts. *Gross reduction in claims payable to banks less the cost of financing the operation. in c lu d e s the 1988 Mexican collateralized bond exchange but excludes estimates of unofficial debt conversions. flAs a percentage of public sector medium-term bank debt, including interest arrears, at the time of the operation. ™Author's estimate, based on banks’ latest allocation among eligible options **Averages are weighted by shares in total bank debt as of en d-1986. 46FRASER FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 Digitized for m ents of principal and interest through collateral accounts, that is, operations that reduce the present value of debt service but not the stock of debt (Chart 5). The impact on total debt stocks is even sm aller than the direct reduction in bank debt because of new borrowing to finance the o pera tions.29 For exam ple, because they involve loans to finance the collateral accounts, par exchanges a ctually increase the stock of debt even though they fix interest rates below prevailing m arket rates. Cash flow im pacts A lth o u g h the B rady o p e ra tio n s c a n c e le d a s ig n ific a n t amount of claims, they did not necessarily directly provide countries with more cash to finance growth and investm ent than did the previous approach. Recall that under the new money approach, banks often effectively refinanced a por tion of the interest due by extending new loans. Moreover, some countries forced an even greater degree of cash flow relief by instituting unilateral partial or com plete m oratoria on interest payments. In fact, the Brady operations on average tended to leave net transfers largely unchanged. This observation is borne out by a com parison of the absolute levels of net financial flow s (debt service actually paid less disbursem ents from banks) during the Baker plan period and before and after the Brady operations (Table 3 ).30 These calculations also re fle ct the co m p le m e n ta ry im pact of debt c a n c e lla tio n s under debt conversion schem es as well as the level and structure of interest rates at the tim e of each agreem ent. Net cash flow im pacts have varied, however, fo r the d if ferent countries. C ountries that were paying full interest before their Brady deals, that is, countries not benefiting from new m oney loans nor incurring arrears, achieved the largest expected savings in cash outflows. In contrast, for Argentina and Brazil, restoring normal relations with credi to rs th ro u g h B ra d y re s tru c tu rin g s re q u ire d s ig n ific a n t increases in debt service paym ents. The agreem ents pro vided for net paym ents that were expected to rise over time to levels com parable to those of the e a rlie r B aker plan period.31 For Mexico, which had benefited from large new m oney packages in 1983, 1984, and 1987, projected debt service paym ents after the Brady operation were slightly higher than the average net paym ents made in 1986-88 but lower than those made in the years im m ediately following the onset of the crisis. The Brady operations departed more strikingly from the previous new m oney approach by greatly extending the tim e horizon of contractual relief. A com parison of M exico’s net debt service obligations on restructured principal result ing from the financial packages of 1983, 1986, and 1989 shows the lowering and flattening of contractual obligations (Chart 6). C om pared with the earlier agreem ents, the Brady packages su b sta n tia lly reduced the likelihood of fu rth er rescheduling or new m oney requests. Thus they enhanced countries’ access to the international capital markets, fu r ther improving their net cash flow. Implications of declining U.S. interest rates The central goal of the Brady operations w as to reduce Footnote 31 continued accruals. In addition, full interest has been paid on 1989-90 interest arrears refinanced in 1992. Chart 5 Reduction in Claims Payable to Banks through Concerted Bank Packages, by Modality of Debt Service Reduction Interest reduction through par exchanges 32.8% Principal reduction through discount exchanges 21.0% 29 On average, countries achieved net reductions in total debt service obligations of about one-eighth. The net debt reduction is defined as the gross reduction in claims payable to banks less the cost of financing the operation. The low reduction in net debt reflects the fact that the Brady restructurings to date have dealt only with medium- and long-term public sector debt to banks; these debts have generally accounted for between one-half and one-quarter of the total debt of the participating countries. 30 For alternative calculations of debt service savings for packages com pleted through mid-1991 and a discussion of alternative counterfactual scenarios, see Eduardo Fernandez, “Cost and Benefits of Debt and Debt Service Reduction,’’ World Bank Working Paper no. 1169, August 1993. 31 Brazil has already begun to step up its payments. After interest payments were suspended in 1989, a $2 billion down payment on overdue interest was made in 1991 along with 30 percent of the current interest accruals on principal. The partial payment rate was stepped up to 50 percent of the accrued interest in 1993, with retroactive payments made on 1992 Source: Federal Reserve Bank of New York staff estimates. Notes: Chart does not show claim reductions resulting from packages agreed upon but not yet finalized (as in the case of Brazil). Separated portion of pie represents outright reduction of principal (39.9%). FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 47 co untrie s’ debt service outflow s to m anageable levels on a perm anent basis. To achieve this, countries prepaid a por tio n of th e ir debt se rvice o b lig a tio n s at a d is c o u n t and locked in the interest rates on a significant portion of the rem ainder.32 These steps helped insulate them from future interest rate changes, up or down. T his locking in of in te re st rates may have resulted in additional ex ante costs w hich are not reflected in the c a l cu la tio ns above if lon g -te rm rates g e n e ra lly exceed an average of relevant short-term rates. The calculations of the present value of countries’ expected debt service sav ings are based on long-term interest rates at the tim e of agreem ent in principle for each of the packages. S pecifi cally, the interest rates on the par bonds are com pared with the hypothetical fixed rate that w ould result from swapping a LIBOR plus 13/16 paym ent stream into a fixed paym ent stream .33 In this way, the ex post costs or benefits from 32 The transformation of debt obligations from floating to fixed rates may be considered an extra benefit of the restructurings to the extent that the debtor country prefers fixing the rates on a portion of its liabilities but is prevented by its credit standing from achieving such rates through the swap market. 33 The swap rate is taken as the market’s expectation of the average level of future short-term rates. unanticipated interest rate changes are separated from the ex ante relief negotiated with creditors. H owever, to the extent that long-term interest rates have an upward bias in predicting short-term rates, this m easure overstates the expected savings resulting from the par exchanges.34 In the e vent, d e v e lo p m e n ts in d o lla r m oney m arkets since the launching of the B rady initiative have thus fa r turned out rem arkably well fo r debtor countries. The LIBOR rate fo r U.S. dollar deposits, to which m ost loan contracts 34 The literature on the predictive power of the term structure has cast strong doubts on the accuracy of the pure expectations theory of the term structure, particularly as the theory relates to the ability of short term rates to predict movements in shorter maturities. However, some research suggests that at longer time intervals, medium- and long-term rates do tend to be useful predictors of medium-term movements in short rates: see, for example, Eugene Fama and Robert Bliss, “The Information in Long-Maturity Forward Rates," American Economic Review, vol. 77 (1987); Kenneth Froot, “New Hope for the Expectations Hypothesis of the Term Structure of Interest Rates,” Journal of Finance, vol. 44 (1989); and John Campbell and Robert Shiller, “Yield Spreads and Interest Rate Movements: A Bird’s Eye View,” Review o f Economic Studies, vol. 58 (1991). Studies have also found that when the yield curve slopes upward, the yields on longer bonds subsequently tend to decline, while short-term interest rates tend to rise. Although many theoretical models have been developed to explain the existence of a possible term premium, no consensus has emerged on the degree to which long-term rates overpredict future short-term rates. Table 3 Annual Net Transfers to Banks before and after Completion of Debt Reduction Packages Billions of Dollars Memorandum: Before Conclusion of Bank Package1 Mexico Costa Rica yenezuela Uruguay Nigeria Philippines Argentina Brazil Total 3.24 0.04 2.02 0.29 0.64 1.04 § 0.59 2.20 t t 10.05 After Conclusion of Bank Package* Short-Run 3.59 0.05 1.53 0.10 0.22 0.28 1.19 2.45 9.42 Cumulative New Money Disbursements Net Transfers Long-Run 3.59 0.05 1.69 0.11 0.28 0.49 1 2.09 4.44 12.73 1983-85 1986-88 1983-88 3.95 0.22 1.12 0.06 1.05 0.00 0.68 0.74 3.22 0.06 2.21 0.17 0.48 0.71 1.33 3.70 14.27 0.28 0.00 0.24 0.00 0.93 6.50 14.90 7.82 11.88 37.11 Sources: World Bank, World D ebt Tables; author’s estimates. Notes: Net transfers before debt reduction are defined as cash debt service payments less disbursements from banks. Transfers after debt reduc tion are defined as net interest payments due on new debt instruments issued plus interest on funds used to finance the transaction, including use of reserves and new money from commercial banks. Floating rate interest obligations are projected on the basis of swap rates prevailing at the time of agreement in principle. The calculations do not reflect additional expected savings due to downward shifts in the yield curve following the initial agreements. t Average net transfer in the three years preceding the completion of the bank package. *The difference between short- and long-run projected net transfers reflects temporary interest reduction on par bonds and the expected path of floating rate interest rates based on the term structure of interest rates at the time of agreement in principle. For cases with rising interest payments, the long-run level of interest payments is generally expected to be reached in five to seven years. §Average net transfers during 1987-89. in c lu d e s interest but not principal on bank debt not eligible for debt reduction. ^E stim ated average during 1991-93. Includes 1991 down payment against interest arrears and interest on refinanced interest arrears (so-called interest due and unpaid bonds). FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 w ere indexed before the Brady operations, d eclined by about 500 basis points between Novem ber 1990 and Feb ruary 1993. For most debtors the interest savings implied by this decline are significantly larger than those resulting from the debt restructurings. For countries that had already agreed on debt packages (Mexico, Venezuela, C osta Rica), the declines did produce additional savings on the portion of their rem aining medium -term bank debt that was left at flo a tin g ra te s (fo r e xa m p le , d is c o u n t and new m o n e y bonds). Moreover, savings accrued on other floating rate debts, such as short-term debt and debt to international Chart 6 Mexico: Debt Service Due on Restructured Principal under Selected Restructuring Agreements Principal and interest due, in percent of restructured principal 4 0 --------1983 Restructuring Agreement 30 20 Buying back into the market 10 0 institutions. Nonetheless, fo r Mexico and Venezuela, coun tries that restructured early, near-term interest obligations on their medium -term debt to banks are currently about the same as if the restructurings had not taken place (Chart 7). Of course, the sharp upward slope of the yield curve indi ca te s th a t the m a rke t e xp e cts an e ve n tu a l re co ve ry in short-term rates. Should this occur, these countries’ inter est obligations will rise, but by much less than if the debts had not been restructured. C ountries that restructured later, particularly A rgentina and Brazil, benefited more from the decline in rates. The coupon profiles on their par bonds m imicked the slope of the U.S. yield curve at the tim e of their agreem ents.35 Most of the countries that had not yet reached agreem ents by early 1991 were making at best only partial paym ents on accruing interest obligations. For these countries, a Brady package required a significant increase in cash outflows. The decline in short-term rates made for a more gradual step-up in paym ents, giving these countries tim e to grow into their long-term debt-servicing capacity (Chart 7). i, I I Li I I I I I I IU I IIM I 30 1986-87 Multiyear Rescheduling Agreem ent* 20 10 ........ ............... 1989 Par Bond Exch ange Agreement 20 As argued above, the Brady operations did more to lock in a longer horizon of debt service relief than to change immedi ate net debt service outflow s from their levels during the B aker Plan p eriod. T h is locking in im p ro ve d c o u n trie s ’ prospects for breaking out of the cycle of continuous rene gotiation that characterized the previous approach (Table 4). Of course, it was crucial that countries implem ent and sustain the policy reforms that would allow the servicing of the rem aining reduced claim s as well as any new borrow ings. In this way, the Brady operations in concert with sound econom ic policies helped countries return to the interna tional capital m arkets and played an indirect but catalytic role in helping countries achieve a positive net cash flow .36 Breaking the cycle of continuous renegotiation 10 1111111111 inniiiiii lllllilillai i 10 15 20 Years from start of agreement 25 30 Source: Federal Reserve Bank of New York staff estimates, based on the terms of the respective agreements. Notes: The 1983 and 1986 financial packages also provided for new loans that effectively covered a portion of the interest due in the initial years. In contrast, the 1989 agreement required outlays for principal and interest guarantees. Floating rate interest obligations on the 1983 and 1986-87 agreements are projected on the basis of long-term U.S. Treasury yields at the time the packages were finalized. Payments on par bonds exclude principal and interest payments prepaid through collateral accounts. * The terms of the rescheduling agreement were agreed upon in principle in September 1986, and the package was finalized in April 1987. Under the new m oney approach that predated the Brady initiative, bank packages provided “front-loaded” cash flow 35 For example, the Argentina and Brazil par exchange agreements provided for interest rates that rose gradually from 4 to 6 percent over a six-year period. 36 Restoration of market access has always been a central goal of the debt strategy, and the shift toward debt reduction was presented as an important means toward this end. Secretary Brady, in his March 10, 1989, address to the Bretton Woods Committee, argued that “the path towards greater creditworthiness and a return to the markets needs to involve debt reduction” (reprinted in Edward Fried and Philip Trezise, eds., Third World Debt: The Next Phase [Washington D.C.: Brookings Institution, 1989]). The IMF guidelines on Fund support for debt and debt service reduction, approved in May 1989, stated that in consid ering requests for support, particular reference would be made to the strength of economic policies, “the scope for voluntary market-based debt operations that would help the country regain access to credit markets and attain external viability with growth,” and the efficiency of resource use. See International Monetary Fund, Selected Decisions and Selected Documents, no. 16, 1991. FRBNY Quarterly R eview /W inter 1 9 9 3 -9 4 49 Chart 7 Mexico's and Argentina's Interest Savings on U.S. Dollar Brady Bonds Interest due per period in percent of restructured principal Interest due per period in percent of restructured principal 12 -------------------------------------------------------------------------- 12 Argentina: Ex Ante Outlook Mexico: Ex Ante Outlook No-debt-reduction scenario No-debt-reduction scenario (based on July 1989 yield curve) i--------- (based on April 1992 yield curve) — j --------------------------------■Ex ante interest savings Ex ante interest savings Expected interest obligations on Brady bonds Expected interest obligations on Brady bonds (based on yield curve at the time of agreement in principle) Mexico: Ex Post Interest Savings and Current Outlook . (based on yield curve at the time of agreement in principle) Argentina: Ex Post Interest Savings and Current Outlook No-debt-reduction scenario No-debt-reduction scenario (reflects actual LIBOR rates through 1993; projections based on October 1993 yield curve) (reflects actual LIBOR in 1993; projections based on October 1993 yield curve) E x p o s t in te re s t s a v in g s Actual and projected interest obligations on Brady bonds ■ Actual and projected interest. obligations on Brady bonds (projections based on October 1993 yield curve) 1990 1993 1995 2000 (projections based on October 1993 yield curve) 1993 1995 2010 Notes: Mexico's and Argentina's annual interest savings on their par and discount bonds are shown as the shaded distances between the "no-debtreduction scenario" lines and the lines showing the interest obligations on their Brady bonds. The no-debt-reduction line presents a counterfactual scenario under which interest accrues at a rate of LIBOR plus 13/16, and amortization is continually deferred. The Brady bond line shows the interest coupons expected on par and discount bonds weighted by their shares of restructured principal. The coupons are expressed as a percentage of the base exposure; hence, the market interest rates for discount bonds are reduced by the size of the discount. In the ex ante panels, interest coupons (including those for the floating rate Brady discount bonds) are projected from forward interest rates implied by the yield curve prevailing when the Brady operation was agreed upon in principle (July 1989 for Mexico and April 1992 for Argentina). In the case of Argentina, the rising projected interest obligations on the Brady bonds reflect the step-up in coupon rates on the fixed rate par bonds as well as the expected rise in interest rates on the discount bonds. The ex post panels reflect the actual path of short-term rates through 1993 and the future path of short-term rates implied by the October 1993 yield curve. 50 FRASER FRBNY Quarterly R eview / W inter 1 9 93-94 Digitized for I relief. Since new money loans would generally cover a fraction of the interest due only over the next year or two, further packages would be necessary unless debt-servicing prospects improved sharply. On the face of it, this approach served the interests of banks by maintaining a contingent claim on future improve ments in debt-servicing capacity. If a country’s debt-servicing prospects improved in the years following a new money package, then continuing with the current agreement (which usually provided for rising amortization payments and no fur ther new money) would ensure that banks would benefit.37 However, as indicated earlier, this approach was becom ing harder to implement over time. Renegotiation was time consuming and distracting for decision makers. Moreover, such arrangements increased the already high stock of debt and blunted the perceived incentives for restructuring countries to improve their debt-servicing capacity. Finally, this approach impeded the resumption of voluntary lending. Potential new creditors were wary of being caught up in this cycle of continuous renegotiation. Given the unclear rules of the game, in which contracts were continually reopened, they feared that cash flow relief might be required from 37 A formal model exploring how bank packages would be expected to address only the debtor’s near-term need for debt service relief is presented in Jeremy Bulow and Kenneth Rogoff, “A Constant Recontracting Model of Sovereign Debt," Journal of Political Economy, vol. 97 (1989). To encourage creditors to agree to contractual relief over a longer time horizon, a number of countries agreed to “recapture clauses” in their Brady packages that provided for increased debt Footnote 37 continued service payments in the event of certain largely exogenous improvements in debt-servicing capacity; these clauses were usually linked to higher export proceeds (oil for Mexico, Venezuela, and Nigeria and agricultural commodities for Uruguay), although the Costa Rica clause was linked to GDP growth. The more recent agreements for Argentina, Brazil, and the Philippines have not included such clauses. Table 4 Chronology of Restructuring Agreements, 1983-92 Country Brady Countries Mexico Costa Rica Venezuela Uruguay Nigeria Philippines Argentina Brazil Dominican Republic Jordan Bulgaria 1983 1984 1985 N N N R N N R N N R P N 1986 P R R N N R R 1987 1988 1989 N P P B Bp Np R N P Selected middle-income countries currently negotiating debt reduction packages Ecuador N N Np N m N Panama m N m m Peru Rp N N N Poland Other selected middle-income countries with recent debt servicing difficulties N N R Chile R Colombia Rf m N R Cote d’Ivoire Morocco R R Yugoslavia1' N R R m N N B m m m R m m m R P m m R Rf Nm N 1990 1991 1992 B B P Bp P m m m d P P m P d B Bp Bp m P P P m m m P m m P m m m m Rf m m R m m R m Sources; International Monetary Fund, International Capital Markets: Developments and Prospects; World Bank, World Debt Tables; Federal Reserve Bank of New York. Notes; Brady countries are those that have completed or reached agreement in principle on operations to substantially reduce their commercial bank debt. Brady countries are ordered by the date of agreement in principle. N: Agreement includes provisions for new financing. R: Agreement provides for principal rescheduling only. B: Agreement in principle on a Brady Plan debt restructuring, d: Rolling agreement to defer all payments of principal and interest. Rf: Principal refinancing agreement. m: Indicates that at year-end country had suspended interest payments to banks. Excludes moratoria of less than twelve months, p; Indicates that the country was making partial interest payments to banks; bold indicates that the level of payments was consistent with an agree ment in principle with the bank steering committee. fSerbia and Montenegro in 1992. FRBNY Quarterly Review/Winter 1993-94 51 them as well or that new inflows from them would be used to justify a cut in debt relief by existing creditors.38 Then, instead of improving debtor countries’ capacity for growth, the new creditors would in effect be buying out the old cred itors’ heavily discounted debt at par. In contrast, debt reduction and longer maturities capped existing creditors’ claims on current cash flows, allowing new flows to finance new growth and investment. Lowering the profile of contractual obligations would not, by itself, be expected to lead to renewed market access. To be successful on its own, the Brady operation would need to convince the market that enough reduction had taken place so that the debtor, without further changes, would have sufficient capacity to service the remaining claims. Here market-based debt reduction faced an inherent limita tion: to the extent that the reduction in the stock of claims was expected to raise the probability that the remainder would be more fully serviced, creditors would only be willing to sell at higher prices. With prices being bid up and financ ing limited, less debt reduction would be achieved and the overhang would persist, deterring new flows.39 In the event, despite the reductions in claims owed to banks, discounts on unsecured restructured obligations generally remained steep, albeit somewhat lower, immediately after the Brady operations. In part, this discount reflected the longer matu rities arranged under the restructurings. Hence, in order to gain significant market access, countries have had to show evidence of improved debt-servicing capacity as reflected in declining yield spreads. 38 Pari passu clauses in the contracts on the existing debt specified that the old debt would be treated equally with all other debts of the borrower. Charging a higher interest rate on new credits to cover the possibility of a debt consolidation could lead to an explosion of debt; moreover, new creditors might not be able to maintain their interest premia if their claims were consolidated under a restructuring exercise. The debt overhang deterred nonbank flows as well as new bank loans. For example, while most rescheduling countries exempted their external bonds from refinancing during the 1980s—a move reflecting both the bonds' small share of total indebtedness and difficulties in organizing debt relief—discounts on bonds still tended to be high. These discounts reflected fears that de facto seniority was not absolute, and access to new flows appeared to be ruled out. Portfolio and direct investment equity flows could also be reduced because of concerns that debtservicing difficulties could lead to restrictions on repatriations of profits and capital and/or costly confrontations with creditors that could adversely affect the return on capital. Similar fears of cost shifting could induce domestic investors to engage in capital flight. 39 Countries did try to circumvent the problem of capital gains for nonexiting banks through a combination of novation (converting the remaining claims of participating creditors into bonds that might be treated more favorably than the claims of free-riding creditors) and requirements that nonexiting banks provide new money. In theory these efforts could have led to a complete elimination of the discount at little cost; however, in practice the Brady operations generally entailed up front resource costs (that is, buyback equivalent prices) broadly similar to those prevailing in the period of negotiations. Discussions of the limitations of market-based debt reduction can be found in Stijn Claessens, Ishac Diwan, Kenneth Froot, and Paul Krugman, “MarketBased Debt Reduction for Developing Countries: Principles and Prospects,” World Bank Policy and Research Series no. 16 (1990); and Jeremy Bulow and Kenneth Rogoff, “The Buyback Boondoggle,” Brookings Papers on Economic Activity, 1988:2, pp. 675-703. http://fraser.stlouisfed.org/ 52 FRBNY Quarterly Review/Winter 1993-94 Federal Reserve Bank of St. Louis Mexico as prototype The experience of Mexico, the first and most successful of the Brady countries to return to the international capital markets, illustrates the interactions between debt reduc tion, improved debt-servicing capacity, and market reentry (Chart 8). Mexico’s implementation of a broad-based macroeconomic stabilization and structural reform program was already well advanced before the Brady initiative was announced.40 Even as the Brady operation was being negotiated, Mexican borrowers began returning to the inter national bond markets. However, initial yield spreads were very steep (although lower than those on the old bank debt), overall volumes were not high compared with later levels, and most of the initial placements were enhanced by the attachment of receivables accounts or favorable equity conversion rights.41 Later, as perceptions of improving eco nomic performance and rising payments capacity led to lower yield spreads on the restructured long-term (Brady) debt, yield spreads on new unenhanced issues decreased, the volume rose significantly, and the composition shifted toward unenhanced issues. Maturities were initially short because creditors were not sure that the improvements in debt-servicing capacity would last. By lending over the short term, creditors could monitor whether the improve ment in capacity was being sustained; if not, they could then try to reduce their exposure as it matured. More recently, most Mexican Eurobond issues, particularly by public sector borrowers, have carried maturities of at least five years, and the maturity of several issues has exceeded ten years. In fact, in November 1993, Pemex, the state oil company, was able to place a thirty-year issue. Market reentry: broad based but not universal Certainly one of the more remarkable recent developments in the international financial arena has been the explosion of private capital flows to borrowers, especially Brady coun tries, that were once credit constrained. Most of the new flows have been in the form of direct and portfolio invest ment, both through equity and securities markets, and repatriation of flight capital. Syndicated lending from com mercial banks has not resumed on a significant scale. Not all Brady countries, however, have been able to 40 See Claudio Loser and Eliot Kalter, eds., Mexico: The Strategy to Achieve Sustained Economic Growth, International Monetary Fund, Occasional Paper no. 99, September 1992. 41 The fact that new issues carried lower spreads than the Brady bonds may have reflected perceived de facto seniority owing to the new issues’ shorter maturity and small share of total indebtedness. Receivablesbacked borrowings eliminated convertibility risk by directing an entity outside of Mexico to pay funds owed to the Mexican borrower into a special purpose vehicle (a specially created trust, partnership, or corporation) that would then issue securities on behalf of the Mexican borrower. For example, Telmex, the Mexican telephone company, directed AT&T to deposit long-distance payments owed to Telmex into a trust located in the United States. See Andrew Quale, “Securing the Future," LatinFinance, May 1991. regain access to the international capital markets (Table 5, Chart 9).42 As reflected in secondary market prices, the m arket did not perceive an im provem ent in N ige ria ’s prospects for growth and reform following the completion of its package, and the country has not returned to the inter national capital markets. In addition, the Philippines, which experienced a decline in secondary market prices following its 1990 buyback, was largely absent from the capital 42 Observed credit flows are of course only partial indicators of credit availability. Some countries, such as Chile, have taken active measures to limit the extent of capital inflows, while other countries have not been willing to borrow unless the terms were sufficiently attractive. Chart 8 Mexico: Volume and Yield Spreads on New International Bond Issues 1989 to 1993, First Half Basis points * Millions of dollars 2000 1750 markets until the completion of the second stage of its Brady restructuring further reduced its debt and lengthened the maturity of the remaining exposure. Argentina and Brazil The success of Argentina and Brazil in regaining access to the international capital market when their debt packages were not yet completed and they were still incurring interest arrears raises questions about the relative importance of the debt operations. This is particularly the case for Brazil, where high levels of inflation persist and uncertainties con tinue regarding when the package will be completed. Support for the view that the debt operations were a cata lyst for reentry can be found in the timing of the countries’ entries into the market in the third quarter of 1991. For both countries, market reentry followed developments indicating that the probability of a “Brady package” in the near future was increasing sharply. Brazil had just recently reached a preliminary agreement with the banks on the treatment of accumulated interest arrears that cleared the way for nego tiations on a debt reduction package.43 While Argentina’s negotiations were not as advanced as Brazil’s (although partial payments of interest had resumed sooner), the 1500 43 Brazil initiated partial interest payments on its medium- and long-term public debt in early 1991 after an eighteen-month moratorium. Agreement with the banks on the treatment of interest arrears accumulated during 1989-90 was reached in April 1991. Moreover, earlier in the year, in an effort to restore market access for Brazilian corporations, all private sector borrowers as well as several leading publicly owned corporations were given permission to negotiate directly with their bank creditors. Concerted interbank and short-term trade facilities were allowed to expire in April 1991 and were replaced with voluntary facilities. 1250 1000 750 500 250 Table 5 Net Capital Inflows to Restructuring Countries 0 Billions of Dollars 1989 1990 1991 1992 1993 Sources: International Financing Review, Euroweek, Financial Times, Salomon Brothers, J.P. Morgan, and Federal Reserve Bank of New York staff estimates. * Spread over comparable maturity U.S. Treasuries. t Unenhanced bonds are new issues that do not carry equity conversion rights and are not backed by collateral or receivables accounts. * * Volume of bonds enhanced by attachment of collateral or receivables accounts or equity conversion rights. t t Yield spread on unguaranteed portion of par bonds, that is, stripped yield spread. Before the issuance of Brady bonds, spreads are implied yield spreads on Brady-eligible bank loans. Implied yields on bank loans are constructed by dividing the long-run average expected interest rate by the price of the loan. 1989 1990 1991 1992 1993e Mexico Costa Rica Venezuela Uruguay Nigeria Philippines Argentina Brazil Chile 4.5 0.6 -1.1 -0.1 0.1 1.9 -0.5 -0.1 1.3 10.4 0.3 -3.4 -0.1 -2.5 2.3 1.2 4.3 2.7 21.9 0.5 0.7 0.1 -0.6 3.0 4.9 1.0 0.9 24.0 0.5 2.2 0.2 -6.0 2.7 12.8 8.4 2.9 23.9 0.7 1.3 0.3 1.3 3.7 12.7 8.7 2.8 Total 6.5 32.5 47.7 55.4 . ss«iisssi 15.2 | JiplP I £ Sources: International Monetary Fund, International Financial Statistics; Federal Reserve Bank of New York staff estimates. Notes: Net capital inflows are defined as the current account deficit plus the increase in gross reserves. The inflows include errors and omissions and exceptional financing FRBNY Quarterly Review/Winter 1993-94 53 country was making important progress in controlling infla tion and restructuring the economy. As a result, Argentina’s prospects for receiving official financial support for a future Brady restructuring appeared to be on the rise.44 In addi tion, the precedents established through the debt opera tions with Mexico, Venezuela, and other countries tended to make the timing of an agreement less crucial because potential new creditors were able to project reasonably well how existing bank claims would be treated under a debt package. The likely future structures were further clarified once the April 1991 Brazilian arrears agreement estab lished a pattern for the treatment of such claims. The notion that term s for the treatm ent of old debt were already broadly defined apparently contributed to a presumption, reinforced by the countries’ policies, that pending a restruc turing of the old debt, new obligations would be given a de facto senior status. Hence the new flows were priced more on the basis of expected post-deal creditworthiness. 44 Since mid-1989 Argentina had been implementing sweeping measures to encourage competitiveness, including liberalizing the trade regime and privatizing several major public enterprises. Argentina’s reentry into the international capital markets in August 1991 followed the adoption of a new stabilization program in March 1991. The program, which involved a tightening of public finances, further structural measures, and a fixed exchange rate, was showing success in sharply curbing inflation and formed the basis for a stand-by arrangement with the IMF approved that same month. For both Argentina and Brazil, access to new capital flows followed changes in the market’s perception of their capacity to service existing debts. In both cases, the yield spreads on their long-term debt sharply improved ahead of their reemergence in the international bond markets. In the case of Brazil, it is notable that bond issues peaked in the first half of 1992; this development coincided with a low point in yield spreads on the long-term debt as the country approached an agreement in principle, announced in July 1992, on a debt reduction package (Chart 10). Afterwards, in the face of political uncertainties culminating in the resig nation of President Collor and continued high inflation, prospects for a deal dimmed, the yield spread on the long term debt widened, and the flow of new issues slowed m a rk e d ly .45 A rg e n tin a made an in itia l fo ra y into the Eurobond market in the third quarter of 1991, but it was not until after agreement on a term sheet for the bank operation in June 1992 that further significant bond issues took place (Chart 11). Moreover, the completion of the par and dis count exchanges in April 1993 and the deepening success of the country’s stabilization and reform effort, reflected in further declines in yield spreads in 1993, led to an explo sion of new issues in the second and third quarters of 1993. 45 Net foreign purchases of Brazilian equities followed a similar pattern: they fell to $0.3 billion in the second half of 1992 after rising to $1.4 billion in the first half of 1992 from $0.6 billion in all of 1991. Chart 9 Intern atio nal Capital M arket Financing Received by R estructuring C ountries, 1990-93 Billions of dollars 20 Gross international bond issues International equity issues 15 Net Euro-CD issues Net Euro-commercial paper issues 10 5 0 Mexico Costa Rica Venezuela Uruguay Nigeria Philippines Argentina Brazil Chile Sources: For bonds and equities, Financial Times, International Financing Review, and Euroweek\ for Euro-CDs, Euroclear; for Euro-commercial paper, Bank for International Settlements, International Banking and Financial Market Developments. Note: Data include issues through September 1993, except for Euro-commercial paper issues, which are through June 1993. http://fraser.stlouisfed.org/ 54 FRBNY Quarterly Review/Winter 1993-94 Federal Reserve Bank of St. Louis Brazil’s access to capital market inflows still appears somewhat tentative in comparison with that achieved by Mexico. Argentina, which in 1993 saw growing interest from institutional investors, occupies a somewhat intermediate position . Through 1992, alm ost all of the unsecured Eurobond issues by Brazilian borrowers carried maturities of two to three years, whereas most recent Mexican issues have had maturities of five years or more. Argentina has been relatively more successful in placing longer term bor rowings; notably, all of the five-year issues came after agreement on a term sheet for the debt exchange (Chart 12). While institutional and retail investors from developed countries are reportedly showing substantial interest in Mexico, Brazil has not captured institutional investors’ interest to nearly the same degree. Market participants reported in mid-1993 that flight capital still accounted for the bulk of the demand for Brazilian Eurobond issues, par ticularly for private sector borrowers; in contrast, industrial country investors, particularly from the United States, accounted for most of the demand for recent bond issues by Mexican corporations.46 These differences suggest that if Brazil is to emulate some other countries’ success in broadening the investor base, achieving a longer maturity structure, and narrowing yield spreads, investor concerns a bo ut c ro s s -b o rd e r risk m ust be add re ssed th rough improvements in debt-servicing capacity and completion of the debt package. Overall, the pattern of sharply increased capital inflows received by many Brady restructuring countries since 1990 confirms that the debt operations, when accompanied by 46 Inform ation on final h old e rs of E u rob o n ds is s ke tch y at best. H ow ever, e v id e n c e of the intere st of d e v e lo p e d c o u n try investors in M e xico can be found in the stro n g g row th of fo re ig n h o ld in g s of d o m e s tic a lly issued M exican g ove rn m e n t b on d s, w hich in cre a se d b y a b o u t $20 billion b etw een e nd -1 9 90 and m id-1993. Chart 11 Argentina: Volume and Yield Spreads on New International Bond Issues Chart 10 Brazil: Volume and Yield Spreads on New International Bond Issues 1990 to 1993, Third Quarter Basis points* 4000 Millions of dollars ---------------- 2000 1990 to 1993, Third Quarter Basis points* Millions of dollars 8000--------------------------------------------------------------------------------- 2000 7000 Implicit yield spread on B ra d y -_______ eligible debt -|Volum e of bond issues - Scale Scale 3500 3000 Yield spread on ■Brady-eligible debt t ---- Scale 1750 1250 5000 Volume of Eurobond issues Scale------- ► 2500 4000 2000 3000 1500 Average yield spread on new - Eurobond issues • * ------Scale 2000 1000 1000 Yield spread Average yield spread on new 500 — Eurobond issues. • * ------- Scale on 89 bonex ------ Scale 1991 1990 1991 1992 1993 0 1993 Sources: International Financing Review, Euroweek, Financial Times, LatinFinance, Salomon Brothers, J.P. Morgan, and Federal Reserve Bank of New York staff estimates. * Spread over comparable maturity U.S. Treasuries. t Implied yields on medium-term bank loans are constructed by dividing the long-term expected interest rate by the price of the debt. ** T h e 1988 new money bonds were issued as part of Brazil's 1988 financing package. They are amortizing bonds with a final maturity in 1999. 1992 Sources: International Financing Review, Euroweek, Financial Times, LatinFinance, Salomon Brothers, J.P. Morgan, and Federal Reserve Bank of New York staff estimates. * Spread over comparable maturity U.S. Treasuries. t Implied yields on medium-term bank loans are constructed by dividing the long-term expected interest rate by the price of the debt. From April 1993, yield spreads are based on the stripped yields on par bonds. Stripped yields measure the yield to maturity on the uncollateralized or risky portion of a Brady bond. * * Bonex 89 are sovereign bonds issued in 1989 that fully mature in 1999. FRBNY Quarterly Review/Winter 1993-94 55 improved policy performance, have played a catalytic role. Countries that have boosted their debt-servicing capacity and reduced their debts have been rewarded with growing market access on improving terms. However, the pattern of inflows suggests that other factors are at work as well. Most im portant, lower global interest rates, p articularly the medium- and long-term declines in 1993, have encouraged yield-sensitive investors to reconsider the prospects of restructuring countries. The generally more favorable envi ronment for capital flows helps account for the magnitude of net capital flows to Mexico, which greatly exceed the debt reduction achieved through the Brady operation, and the 1993 rebound in flows to Brazil despite uncertain funda mentals. In the current environment, some investors seem more willing to lend on the promise of reform, provided the contractual yield is sufficiently attractive. Impact on investment performance Many advocates for debt reduction argued that lowering countries’ debt and debt service burdens would lead to Chart 12 Maturity Structure of Unsecured International Bond Issues, 1990-92 Argentina Brazil Mexico I_____ I_Maturity of two years or less I Maturity of between two and three years ] j Maturity of between three and five years | | Maturity of five years I Maturity of more than five years Sources: International Financing Review, Euroweek, and Financial Times. Note: In cases where put options are incorporated, time to put is used. 56 FRBNY Quarterly Review/Winter 1993-94 higher rates of capital formation. In fact, for most Brady countries, investment rates have been increasing in recent years (Table 6). Nonetheless, although in some cases cap ital inflows now rival those observed before the debt crisis and secondary market discounts have narrowed, in most Brady countries investment still accounts for a substantially smaller share of GDP than in the pre-debt crisis period. The “debt overhang” hypothesis, advanced by a number of analysts of the developing country debt crisis, had sug gested the possibility of a stronger investment response, at least in some cases. According to this hypothesis, elimina tion through debt reduction of the substantial discounts on countries’ external debts would encourage investment, thereby producing important efficiency gains.47 The “over hang,” or excess of what debtors owed over what they could pay (as indicated by the market value of the debt), was thought to dissuade countries from improving their debt-servicing capacity: any improvements were expected to be largely “taxed away” through reduced debt relief in the future. This disincentive was seen to act both at the level of governments reluctant to adopt unpopular austerity mea sures and on the microeconomic level of investors who feared confiscatory tax policies. A variant of the overhang hypothesis argued that the discounts constrained invest ment by restricting the availability of financing. Absent credible seniority for new flows, potential creditors refused 47 Amongst the most widely cited expositions of this view are the theoretical arguments of Jeffrey Sachs, “The Debt Overhang of Developing Countries,” in Jorge de Macedo and Ronald Findlay, eds., Developing Country Debt and the World Economy (Helsinki: WIDER Institute, 1988), and Paul Krugman, “Market Based Debt Reduction Schemes," in Jacob Frankel, Michael Dooley, and Peter Wickham, eds., Analytical Issues in Debt (Washington, D.C.: International Monetary Fund, 1989). The overhang hypothesis was by no means universally endorsed. For example, Jonathan Eaton in “Debt Relief and the International Enforcement of Loan Contracts,” Journal of Economic Perspectives, vol. 4 (1990), and Jerem y Bulow and Kenneth Rogoff in “Cleaning up Third World Debt Without Getting Taken to the Cleaners,” Journal of Economic Perspectives, vol. 4 (1990), strongly questioned the empirical significance of the overhang effect. to finance new investments for fear that their loans, like the old loans, would not be fully serviced.48 To be fair, overhang proponents were skeptical about market-based debt reduction, as opposed to mandatory writedowns of excess claims, arguing that the former would not make much of a dent in the prevailing discounts. Indeed, discounts often remained high immediately follow ing the implementation of the Brady packages. Nonethe less, even in those countries experiencing the largest ex post reductions in discounts—for example, Mexico— or the greatest restoration of capital flows, the improvements in investment rates have not generally been striking compared with the deterioration at the outset of the debt crisis.49 Impact on banks The secondary market value of claims on restructuring countries has recovered significantly in the period following the change in strategy. Some banks have also gained from 48 Ishac Diwan and Dani Rodrik developed an argument broadly along these lines in "Debt Reduction, Adjustment Lending, and Burden Sharing," World Bank, mimeo, September 1991. Eduardo Borenzstein presented numerical simulations suggesting that credit rationing associated with excess indebtedness may be more important in restraining investment than negative incentive effects; see “Debt Overhang, Credit Rationing and Investment,” International Monetary Fund, Working Paper no. 89/74, 1989. Daniel Cohen presented empirical evidence of a negative linkage between net debt service outflows and investment in “Low Investment and Large LDC Debt in the Eighties,” CEPREMAP Working Paper no. 9002, 1989. Cohen’s results implied that a restoration of capital inflows should lead to increased investment. 49 Adherents of the debt overhang hypothesis did not specify how rapidly investment would recover. However, the comparisons that some made with the collapse in investment at the start of the debt crisis appeared to imply that a rapid rebound would be possible. The weak association observed to date between debt reduction and investment may reflect in part countries’ monetary and fiscal policies. In the aggregate, about half of the increased capital inflows in recent years have been channeled into increased holdings of official reserve assets. Public sector investment has declined relative to the period immediately preceding the debt crisis, a change that reflects both public sector austerity and reductions in the size of the state sector through privatization. Investment Performance in Restructuring Countries Nominal Gross Fixed Capital Formation, Percent of GDP ____________________ 1978-82________1983-89________ 1990-92___________________ 1989___________1990___________1991___________ 1992 24.4 23.5 29.6 15.9 22.7 26.4 23.7 22.7 19.1 Mexico Costa Rica Venezuela Uruguay Nigeria Philippines Argentina Brazil Chile 19.3 19.3 19.4 11.3 8.7 206 18.5 20.2 18.0 19.9 18.2 21.1 20.5 17.6 17.2 11.4 11.6 13.1 8.2 22.3 20.9 15.1 15.5 19.9 24.8 23.3 23.1 '"jlH 1 H$§§I Sources: International Monetary Fund, International Financial Statistics; Federal Reserve Bank of New .v; mm : IB # illlii "p : r-g ; ■: V'-' .C-,. 19.4 19.7 18.2 11.3 12.7 20.6 14.6 19.0 21.7 *t _ ;■ York staff estimates. 18.6 22.4 14.1 10.8 11.9 24.1 14.0 21.6 24.6 •* . 7 r ; 21.6 21.2 20.6 12.1 14.6 22.3 16.7 19.1 23.7 -wmjm ,i® FRBNY Quarterly Review/Winter 1993-94 57 expanding income opportunities in the secondary market trading of restructured debts and the underwriting of new securities flows to restructuring countries. Although in the early cases the prices paid to banks in the form of collateral and cash for their forgone claims were close to the histori cal lows that had prevailed in the secondary market, banks have regained ground through subsequent price apprecia tions on their remaining exposure. This price rebound, which came with a lag, reflects growing optimism about the effectiveness of the new strategy. Moreover, the dramatic increase in secondary market liquidity, thanks in large mea sure to the securitization of claims through the Brady restructurings, has given banks new flexibility in managing their developing country exposure. Sorting through the aftermath of the debt crisis has been a painful and costly process for the banks. From 1987 to 1992, the leading U.S. money center and regional banks charged off more than $25 billion of their loans to restruc turing country borrowers, or about one-third of their aggre gate exposure at the end of 1987.50 For the largest banks, these losses equaled two-thirds of these banks’ capital at the start of the debt crisis.51 It is difficult to determine the extent to which the change in the debt strategy caused or contained these losses. Dif ferent views reflect largely unconfirmable hypotheses about what would have happened had another course been followed. In one view, shifting of the rules of the game to recognize that the loans were no longer fully collectible weakened the position of banks and created losses. This perspective imputes a strong role to the official community in arbitrating between countries and their creditors. By con trast, others maintain that banks were bound to incur losses anyway; providing official financing to help countries buy back their debts benefited banks by driving up prices and shifting risk to the official sector.52 One could also argue that the strategy helped all parties by encouraging greater economic efficiency. A Brady bounce or a Brady dip? One kind of evidence that bears on this problem is the reaction of secondary market debt prices to the Brady 50 Some charge-offs and provisions were made before 1987, but these were relatively insignificant compared with the post-1986 actions. Although some of these charge-offs are potentially recoverable because the banks have retained their legal claims, many are not because they reflect losses through swaps and sales. 51 In contrast, leading U.S. banks’ net earnings on foreign operations, which include many activities unrelated to developing country lending, were on the order of $1.1 billion per year in 1981-82. 52 See, for example, Bulow and Rogoff, “Cleaning up Third World Debt." Why creditors might or might not be better off is also discussed in W. Max Corden, “An International Debt Facility?” in Analytical Issues in Debt; and Michael Dooley, “Buy-Backs, Debt-Equity Swaps, Asset Exchanges, and Market Prices of External Debt,” in Analytical Issues in Debt. 58 FRBNY Quarterly Review/Winter 1993-94 initiative.53 Secondary market prices have generally been on an upward trend since the launching of the Brady initia tive; in particular, prices rose just after the proposal was announced (Chart 13). Some critics have pointed to such price behavior even against a background of continued steep discounts to suggest that the new plan was beneficial for banks, in some arguments to the exclusion of other par ties.54 However, any focus on the short-term movement immediately after the announcement needs to be tempered by awareness that the market anticipated the possibility of a tilt toward debt reduction well before Secretary Brady’s March 1989 speech. Once this is taken into account, the initial reaction of the market to the change in approach appears on balance to be unfavorable. In particular, the free fall in secondary market prices from mid-1988 to the eve of Secretary Brady’s speech must be regarded as at least partly reflecting fears that a change in strategy would adversely affect banks.55 In the debate leading to the change in strategy, banks expressed concerns that any new approach be voluntary and that the Baker Plan’s emphasis on policy reform continue. Secondary market prices did rise in the two months following Secretary Brady’s speech as the official community worked out the details of the new approach, but this rebound offset the declines that had taken place only since December 1988, when President-elect Bush announced that the debt strat egy was under review, and was still much smaller than the fall from mid-1988. Buyback equivalent prices and post-deal price improvements Brady deals can also be examined on a country-by-country basis to determine the effects on banks. The financial impact of Brady restructurings can be separated into two aspects: (1) the effective purchase price in the form of collateral and 53 Because of the thinness of the secondary market before the launching of the Brady initiative, prices from this period might be regarded as unreliable. However, even after secondary market volumes increased in 1989 and 1990, trading stayed broadly in the ranges reached at end1988, with prices rising somewhat in cases where economic performance was improving and falling where it did not. 54 Conclusions about the implications of price movements for other parties, such as the borrowing countries, are generally based on strong assumptions that usually rule out the very efficiency gains that the reinforced strategy was seeking. 55 In March 1989, the U.S. Treasury cited “heightened publicity on establishing debt facilities in the latter part of 1988" as one of the factors contributing to downward pressure on secondary market prices in the second half of 1988 (Department of the Treasury, “Interim Report to the Congress Concerning International Discussions on an International Debt Management Authority,” in Third World Debt—Reports and the Brady Plan, Hearings before the Subcommittee on International Development, Finance, Trade and Monetary Policy of the House Committee on Banking, Finance and Urban Affairs, 101st Cong., 1st sess. [Washington, D.C.: GPO, 1989], p. 64). Clearly the decline reflected pessimism about the pre-Brady strategy. However, neither anticipation of a new approach nor clarification of how the strategy would change fully reversed the downward adjustment. cash paid to compensate banks for forgone claims and (2) the returns on remaining exposure. Banks realize a benefit on market accounting when the effective purchase price is high compared with the prevailing secondary market price. However, even when it is low, they may be better off because of an induced capital gain on their remaining expo sure due to a reduction in the amount of debt outstanding.56 The effective pricing of the early deals was consistent with the low levels to which prices had fallen (Table 7). In the case of Mexico, the buyback equivalent price—that is, the price at which the same amount of cash could have pur chased an equivalent amount of debt reduction through a buyback—was below the secondary market prices prevail ing during the period of negotiations.57 Despite a rise in 58 In fact, some have argued that banks could have been paid less in anticipation of a post-deal price rise. However, this would give rise to free rider problems because any single bank selling off its exposure would be worse off than those that did not. 57 The buyback equivalent price for a Brady package is the ratio of total up-front cash outlays for buybacks and collateral purchases to the present value of the exposure reduction by exiting banks. For a further discussion of buyback equivalent prices, see John Clark, “Evaluation of Debt Exchanges,” International Monetary Fund, Working Paper no. 90/9, February 1990. prices just before the agreem ent with Venezuela was reached, the buyback equivalent price was in line with the average price prevailing during the negotiation period. Still, despite the relatively low compensation received by creditors for their reductions in nominal claims, they have benefited as their remaining exposure has appreciated in value. This recovery in prices was not immediate and it reflects a variety of factors, including some unrelated to the change in strategy. Undoubtedly the most important influence has been the increase in debt-servicing capacity. Since 1988, most Brady countries have increased their exports, lowered their fiscal deficits, curtailed inflation, and strengthened their balance of payments positions (Table 8).58 The shift in strategy may well have encouraged such changes by making needed policy reforms more politically a ccep tab le. The reduced debt burden m agnified the effects of im provem ents in debt-servicing capacity on perceived creditworthiness, helping speed the return to market access. In addition, the securitization of remaining 58 The country that made the least progress in some of these areas, Brazil, is also the country that showed the weakest price performance. Chart 13 Average Secondary Market Prices for Medium-Term Bank Debt Price in percent of face value Sources: Salomon Brothers; Federal Reserve Bank of New York staff estimates. Notes: Index weights countries' debt prices by their share in total debt at the start of the period. The countries included are Argentina, Bolivia, Brazil, Chile, Colombia, Cote d’Ivoire, Ecuador, Mexico, Morocco, Nigeria, Peru, the Philippines, Uruguay, Venezuela, and Yugoslavia. Index reflects stripped prices (that is, prices adjusted to remove the effects of partial collateralization and interest reduction) following Brady restructurings. FRBNY Quarterly Review/Winter 1993-94 59 claims greatly expanded liquidity (some Brady bonds are among the most actively traded instruments in the Euro markets) and facilitated the entry of new investors into the market. An important factor in the price rises not linked to the change in strategy has been the recent decline in global interest rates, particularly at medium- and long-term matu rities. Lower rates raise the present value of future expected Table 7 Evolution of Secondary Market Prices of Bank Claims on Selected Restructuring Countries Cents per Dollar of Contractual Claim Secondary Market Prices Mid1988 Country February 1989 Average during Negotiationsf Following Agreement in Principle December 1992* Memorandum Item: Buyback Equivalent Price of Brady Package Weighted Average§ (excluding Chile) 48.3 Mexico Costa Rica11 Venezuela Uruguay Nigeria Philippines (1990) Philippines (1992) Argentina Brazil Chile 31.9 34.3/38.3 41.8 46.8 33.6 51.3 14.7 55.6 61.3 29.5 54.8 36.5 13.5 35.0 60.5 22.0 42.8 41.1 13.0 37.8 53.4 29.1/36.1 47.4 59.4 34.1 53.9 65.4 39.5 54.7 26.5 52.1 61.6 19.0 30.0 59.5 21.1/39.3 29.5/34.2 N.A. 44.5 16.0 42.8 56.3 44.6 51.5 53.3 44.8 35.9 N.A. 33.7 17.4 * 37.5 51.0 39.0 50.0 46.2 29.9 * 28.5 n N.A. 48.0 29.5 93.0 Sources: Salomon Brothers; Federal Reserve Bank of New York staff estimates. f Average price from March 1989 until agreement in principle. In cases where formal negotiations did not begin in 1989, the second price is the aver age during the period of formal negotiations. ♦Reflects weighted average price of new instruments issued under debt exchanges. In cases where buybacks were included, the price includes the buyback price weighted by the share of debt allocated to the buyback option. §Prices are weighted by shares in total debt to commercial banks as of end-1986. 'The buyback equivalent prices for Argentina and Costa Rica do not reflect down payments made at closing against interest arrears. Inclusion of these costs would raise the Argentina price by around 5 cents and the Costa Rica price by about 2 cents. n The estimated price for Brazil reflects bank choices among options and interest rates prevailing in July 1992, when the terms of the package were agreed upon in principle. The actual cost of the package may be higher because of subsequent declines in long-term interest rates that have raised the cost of thirty-year zero coupon bonds. Table 8 Selected Indicators of Economic Policy Performance Primary Fiscal Balance1 (Percent of GDP) Inflation (Annual Percent Change in Consumer Price Index) 1988 Mexico Costa Rica Venezuela Uruguay Nigeria Philippines Argentina Brazil Chile 51.7 25.3 35.5 69.0 64.7 9.0 387.5 1,006.5 12.2 1989-91 22.8 20.8 49.5 99.9 23.8 15.0 2,119.0 1,302.9 22.5 Cumulative Export Growth* 1992 1988 1989-91 1992 (1988-92) 11.9 17.0 31.9 58.9 48.8 8.1 17.7 1,156.4 12.8 8.0 2.5 -6.1 -0.4 -2.9 2.7 -0.6 -0.4 5.0 7.1 1.5 3.5 1.3 6.5 3.6 0.3 1.5 5.5 5.6 4.3 -1.2 3.4 1.8 4.6 1.4 2.0 4.6 47.5 60.6 33.2 36.5 78.0 62.4 33.0 11.7 53.4 Sources: International Monetary Fund, International Financial Statistics; Federal Reserve Bank of New York staff estimates. Excludes privatization receipts. ♦Exports of goods and services. 60 FRBNY Quarterly Review/Winter 1993-94 net debt service payments.59 In addition, lower global rates raise the prospects for productive new inflows as yield-sen sitive investors seek alternatives to industrial country investments. These new inflows in turn can raise debt-servicing capacity, increasing the value of existing debt.60 The emerging markets fixed income business Finally, the Brady operations have helped create a new industry focused on investments in countries that have restructured their debts. The operations catalyzed a restoration of market access and encouraged the emer gence of a vibrant secondary market for restructured claims. Secondary market trading rose more than seven fold from 1988 to 1992, reaching about $0.7 trillion during the latter year.61 On the fixed income side, the main lines of business include investing and trading in Brady bonds and Brady-eligible medium-term bank claims, and underwriting and investing in new international bond, commercial paper, and certificate of deposit issues.62 In addition, derivatives underwriting has expanded in recent years. In some cases, the increased investor interest has spilled over into domes tically issued debt instruments as well, principally those of Mexico and Argentina. Most of the income earned in these markets accrues directly in the form of yield spreads and capital gains to investors willing to put their capital at risk. However, ancil lary noninterest income opportunities have arisen as well from market making and underwriting. Although banks have captured a large, although shrinking, share of the noninterest income, they generally have been reluctant to expand their exposure significantly. Most of the growth in claims has been taken up by an expanding pool of nonbank investors. reformers, particularly Mexico and Chile, demonstrated how policy reform and debt reduction could lead to restored market access and sustainable growth, and as the interest rate outlook improved, a reassessment took place. For late-restructuring countries, including those yet to negotiate a Brady deal, this reappraisal represents a mixed blessing. These countries have been helped by the en hanced credibility given to needed structural reforms, which makes their adoption more likely, and by the acceler ation of their return to the market. Hence, their debt-servic ing capacity has improved. However, this reappraisal has also tended to push up the price at which banks are willing to reduce their claims.63 For example, market participants cited this demonstration effect to explain market bullish ness in third-quarter 1991 for claims on countries in the earlier stages of policy reform and debt restructuring. Late restructuring countries also face more of an uphill debt-servicing path because of their past interest arrears. Banks have taken a harder line on the treatment of interest arrears relative to principal, as part of their strategy of dis couraging forced relief through such arrears. For example, in the bank packages for Brazil and Argentina, refinanced interest arrears were excluded from principal or interest reduction and carried maturities of ten to twelve years in contrast to maturities of as long as thirty years for restruc tured principal. As a consequence of not granting as flat a repayment profile as that accorded Mexico, the banks have effectively maintained a claim on these countries’ expected increases in market access over the medium term.64 Conclusion Since the emergence of the developing country debt crisis in 1982, policymakers have sought to avert systemic threats to the international financial system, to gain time for debtor The experience of late-restructuring countries The early Brady deals (Mexico, Venezuela, Costa Rica) were priced at levels that reflected skepticism about the effectiveness of the new approach. However, as the early 59 The value of a country’s external debt may be viewed as a function of the portion of export receipts or national income that the country is presumably willing to devote to the debt’s servicing in the future. When the discount rate is lowered, the present value of any future path of service payments rises. 60 Of course, this outcome requires that the inflows be channeled into activities with appropriate returns. 81 For estimates of the growth in trading volumes based on periodic surveys of market participants, see Richard Voorhees, “A Trillion Dollar Market," LatinFinance, no. 45, pp. 49-62. The 1992 estimate is taken from the Emerging Markets Traders Association’s (EMTA) survey of market participants. The EMTA estimate does not adjust for double counting; see “EMTA Volume Study: Brazil, Mexico Grab Top Spots in $734 Billion Debt Market,” LDC Debt Report, October 4, 1993, p. 7. 62 For a discussion of recent developments in equity flows, see John Mullin, “Emerging Equity Markets in the Global Economy," Federal Reserve Bank of New York Quarterly Review, Summer 1993. 63 In many cases, before recovering, the prices of claims on the laterestructuring countries fell well below the buyback equivalent price of the Mexican Brady package. Note that fears that countries might drive down the price of their debt so as to purchase it subsequently on the cheap do not appear to have been borne out. Banks have used arguments of precedence to resist offering more generous terms to laterestructuring countries; the power of precedent has pushed buyback equivalent prices of debt exchanges to conform more closely to those offered Mexico. Countries that did not adopt strong adjustment programs generally lacked the resources to complete comprehensive restructuring operations because direct support from official creditors was not available. When countries that had been incurring interest arrears showed signs of moving toward a debt operation, debt prices tended to recover sharply. As a result, the effective prices of the bank packages reflected precedent, expected future debt-servicing capacity, and up-front enhancements rather than past debt-servicing history. 64 Although capturing the benefits of the improved outlook for capital inflows might not have directly informed banks’ negotiating positions on the treatment of interest arrears, countries surely considered the outlook for future flows in deciding whether to agree to the banks' terms. Moreover, it seems reasonable to expect that banks as negotiators attempted to anticipate countries’ positions. Hence, in this way, the outcomes on the arrears restructurings reflected the more optimistic outlook. FRBNY Quarterly Review/Winter 1993-94 61 countries to build up their debt-servicing capacity and get back on a sustainable growth path, and to restore coun tries’ access to the international capital markets. Advances toward these goals were uneven under the new money strategy, and the process proved less and less workable over time. Designed to address these shortcomings, the Brady approach has achieved impressive results. The Brady restructurings did not achieve significantly more near-term cash flow relief for debtors than the previous approach. But they did provide a more stable long-run 62 FRBNY Quarterly Review/Winter 1993-94 financial framework that, in combination with structural reforms by debtors and a favorable environment of lower global interest rates, helped to restore market access. Although there has been a remarkable turnaround in the market’s assessment of restructuring countries, significant risks remain. Debt service obligations remain heavy for the Brady countries. While the restoration of market access is helpful, the key to sustained growth and creditworthiness continues to be sound macroeconomic policies comple mented where needed with further structural reforms. Index Amortizing Rate Swaps by Lisa N. Galaif As short-term interest rates have declined over the past sev eral years, investors have increasingly sought higher yield ing investment vehicles. The index amortizing rate (IAR) swap is one of several new instruments that have been developed in response to this investor demand for yield enhancement. An IAR swap is an interest rate swap based on a notional principal amount that may decrease over time in accordance with the path of future interest rates.1 The IAR swap market has grown rapidly since its incep tion in 1990, achieving a market size in late 1993 estimated at $100 billion to $150 billion notional principal. IAR swaps should continue to be popular because they can be an attractive investment under certain interest rate scenarios and a good hedging vehicle for dealers’ written options exposures. This article explains the structure and pricing of IAR swaps, the risks associated with the product, and the uses as well as the growth prospects for the market. We find that while the product has advantages for dealers and investors, its complexity may be a drawback. To price and hedge IAR swaps, dealers must use highly technical models with para meters whose values are difficult to forecast. Investors may have trouble comparing the risk-return tradeoffs of an IAR swap with those of more liquid and traditional instruments. The structure of IAR swaps An IAR swap is an over-the-counter contract between two parties to exchange interest payments—one based on a fixed rate and the other on a floating rate—on an amortizing notional principal amount. Like the so-called plain vanilla 1 The IAR swap is also known as an index principal swap (IPS) or an index amortizing swap (IAS). interest rate swap, the IAR swap involves no exchange of principal. But unlike the plain vanilla swap, whose net inter est payments are made on a fixed notional amount, the IAR swap calls for net interest payments made on a notional prin cipal balance that may decrease over the life of the swap. The rate at which the notional principal amount decreases will vary with a specified short-term interest rate according to a schedule predetermined by the two parties. In general, however, notional principal amortizes more quickly when short rates fall and more slowly when short rates rise.2 In a typical IAR swap, an end-user3 (or fixed rate receiver) receives interest payments based on the fixed rate while paying the dealer (or fixed rate payer) floating interest indexed to three-month LIBOR. The amortizing notional amount on which both interest payments are based is typically $100 million at origination. Net interest payments are most often made quarterly throughout the life of the swap, just as they are in a plain vanilla swap. The standard contractual maturity for an IAR swap is five years with a two-year “lockout” period, meaning that the swap does not start amortizing until the beginning of the third year. The amortization schedule is usually designed so that if short-term interest rates remain unchanged, the IAR swap will have a life of about three years. However, if the floating rate index falls sufficiently, the swap could fully amortize at the end of the lockout period. Alternatively, if rates rise, the swap would amortize at a slower rate and 2 Despite the use of the term “amortization” by market participants, the amortization of notional principal does not imply payment of principal; it refers to the declining notional principal amount on which interest payments are based. 3 An end-user or customer is typically an institutional investor such as an insurance company, bank, or mutual fund. FRBNY Quarterly Review/Winter 1993-94 63 have a longer than expected maturity, perhaps reaching its five-year maximum life. The variable maturity of an IAR swap is another feature distinguishing it from a plain vanilla swap, which has a fixed maturity date. Table 1 presents a typical IAR swap amortization sched ule. If LIBOR remains at 4.50 percent, the swap amortizes by 80 percent per year after the lockout period; if LIBOR rises to 5.50 percent, the swap amortizes at 30 percent per year. Alternatively, if LIBOR drops to 3.50 percent, the swap amortizes at 100 percent in year 3. This particular schedule assumes yearly amortization, although quarterly amortization is also common in IAR swap schedules. Changes in future short-term interest rates affect the Table 1 Amortization Schedule of Typical IAR Swap LIBOR* (Percent) Change in Basis Points Amortization Rate (Percent) 3.50 4.50 5.50 6.50 -100 0 + 100 +200 100 80 30 10 Notes: The amortization rate in the table is based on annual changes in LIBOR. The terms and conditions of the IAR swap illustrated here are as follows: Notional amount: Fixed rate: Lockout period: Final maturity: Payment frequency: Amortization: $1,000 4.745 percent 2 years 5 years Annual After the lockout period, yearly amortization of remaining notional principal balance based on changes of yearly LIBOR. f The initial spot rate is 4.50 percent. swap in three ways: they 1) directly affect future net interest payments, 2) indirectly affect future net interest payments by changing the principal amount on which interest calcula tions are based, and 3) alter the maturity of the swap. The interest rate scenarios presented in Table 2 illustrate how the notional principal of an IAR swap amortizes given the schedule set forth in Table 1. If future interest rates fol low LIBOR path 2 (case 2), then, in year 3, $800 of the notional princip al a m o rtize s, reducing the rem aining notional principal to $200.4 An IAR swap’s maturity is usually described in terms of a weighted average life because the instrument’s maturity and notional principal may vary. First, the date of the swap’s last payment will vary with the path followed by short-term interest rates. Second, the date of the last pay ment can be a misleading representation of the swap’s maturity because the remaining notional principal is also variable. Consider, for example, two IAR swaps that origi nate with the same notional principal of $100. While both may end after three years, one may end with a notional principal amount of $60 while the other may end with a notional amount of $30. The weighted average life of an IAR swap is calculated by summing the percentage of the remaining notional principal amounts over each interest rate path. These amounts are then averaged across the possible paths. Note that the weighted average life is sim ply used to describe the instrument’s maturity. It is not used for pricing and hedging because it does not describe the actual cash flows with sufficient precision. 4 Given an interest rate of 4.50 in year 3 (case 2), the amortization schedule specifies that $800 of the notional principal will amortize. This amortization leaves $200 in remaining notional principal at the end of year 3. In this example, the amortization rate applies to the current outstanding notional principal. Table 2 IAR Swap Notional Principal Balance Notional Principal Given Various LIBOR Paths 0-1* 1-2* 2-3 3-4 4-5 Case 1: declining rates LIBOR Notional principal 4.50 1,000 4.00 1,000 3.50 0 3.25 0 3.00 0 Case 2: stable rates LIBOR Notional principal 4.50 1,000 4.50 1,000 4.50 200 4.50 40 4.50 8 Case 3: rising rates LIBOR Notional principal 4.50 1,000 5.01 1,000 5.53 705 5.82 539 6.13 445 Paths Year: Notes: Amortization is applied to the remaining notional principal balance of the previous period and is based on the schedule in Table 1 Amortization for rates not given in Table 1 is computed through linear interpolation. 1 No notional principal amortization during two-year lockout period. 64 FRBNY Quarterly Review/Winter 1993-94 Optionality of an IAR swap The amortizing feature of an IAR swap is an implicit call option that essentially gives the fixed rate payer the right to “call” or cancel a portion of the swap (according to the pre determined schedule) if interest rates decline substantially. The fixed rate payer in an IAR swap thus owns an implicit option analogous (but not identical) to the prepayment option in a callable bond or mortgage security. For this right, the fixed rate payer pays a yield premium for the im plicit option. However, in contrast to the embedded options on long-term rates in callable bonds and mortgage securities, the implicit options in an IAR swap are usually options on short-term interest rates.5 Because an IAR swap’s behavior is dependent on the path of interest rates, the exact set of interest rate options embedded in an IAR swap are difficult to determine directly from the amortization schedule. Instead, these im plicit options must be determined indirectly from interest rate models that estimate the IAR swap’s exposure profile in dif ferent interest rate scenarios. For example, in Table 2, the amount of notional principal remaining in case 2, year 4, depends not only on the short-term rate that will prevail in year 4, but also on the rate that will prevail in year 3. Hence, it is not always possible to purchase the correct number of options or futures contracts in year 1 to hedge the cash flow risk in year 4, since the exposure in year 4 depends on the interm ediate path of future interest rates. Specifically, dynam ic hedging is required as the exposures to be hedged change with each period. However, for large parallel shifts in the yield curve, the IAR swap will provide a lower return than the plain vanilla swap. If both short and long rates fall, the IAR swap will amortize rapidly after the lockout period, subjecting the IAR swap’s fixed receiver to reinvestment losses at the lower rates. If both short and long rates rise, the amortization rate will slow, lengthening the maturity. In this scenario the fixed rate receiver is paid a below-market fixed rate for a longer period than would be the case in the plain vanilla swap. As the chart shows, if the net present value of the plain vanilla swap is subtracted from the net present value of the IAR swap, the difference is similar, but not identical, to the exposure profile of a short straddle.6 In other words, an IAR swap can be thought of as a plain vanilla swap (of the same maturity as the expected maturity of the IAR swap) com bined with a collection of interest rate options written by the fixed rate receiver that replicate the “straddle-like” exposure in the chart. For the fixed rate receiver, the option premium 6 A short straddle is a collection of written interest rate options. Some pay off when rates rise, while others pay off when rates fall. Our chart is modeled loosely on a chart that appeared in Derivatives Week, vol. 2, no. 3 (January 25, 1993). Net Difference between an Index Amortizing Rate Swap and an interest Rate Swap from the Perspective of a Fixed Rate Receiver Dollar net difference in present value 10000 ----------------------------------------------------------- Behavior of IAR swaps when interest rates change Like a plain vanilla interest rate swap, an IAR swap has a present value for the fixed rate receiver that will fall when interest rates rise and increase when interest rates fall. However, the magnitude of these changes for an IAR swap and a plain vanilla swap differs because of the option-like behavior of the IAR swap. Specifically, when rates fall, the gain in an IAR swap’s value is smaller than the gain in a plain vanilla swap’s value; when rates rise, the loss in value of an IAR swap exceeds that of a plain vanilla swap. The chart illustrates the performance difference between an IAR swap and a plain vanilla interest rate swap (of the same maturity as the expected maturity of the IAR swap) from the perspective of the fixed rate receiver. When long and short rates move together (producing parallel shifts of the yield curve), the IAR swap outperforms the plain vanilla interest rate swap in a stable interest rate environment and underperforms it in a volatile environment. In other words, if interest rates do not change by a large amount, an IAR swap offers the investor a more favorable fixed rate of return than the plain vanilla swap because of the option premium embedded in the IAR swap’s fixed rate. 5 Thus, IAR swaps are not ideal hedges for mortgage securities unless perfect correlation exists between long-term and short-term rates. 5000 0 -5000 -10000 -15000 -200 -150 -100 -50 0 50 100 Interest rate change in basis points 150 200 Notes: The net difference equals the present value of the cash flows of the IAR swap along the given interest rate path minus the present value of the cash flows of the interest rate swap along the same interest rate path. The interest rate changes, are based on parallel shifts in the yield curve. The weighted average life for the IAR swap is three years, with a contractual maturity of five years and a two-year lockout period. The maturity of the interest rate swap is three years. The original notional principal for both the IAR swap and the interest rate swap is $1,000,000. The fixed rate on the IAR swap is 4.745 percent and the fixed rate on the interest rate swap is 4.50 percent. FRBNY Quarterly Review/Winter 1993-94 65 embedded in the fixed rate of the IAR swap causes the IAR swap returns to exceed the plain vanilla swap returns when interest rates stay within a narrow range (because the option is not exercised). But when rates either fall or rise by a large amount, some of the embedded options will be exercised against the fixed rate receiver, thus causing the returns from the IAR swap to fall short of the returns from the plain vanilla swap. Nonparallel shifts in the yield curve The embedded options in an IAR swap have complex fea tures that become apparent as soon as nonparallel yield curve changes are considered. If long rates rise and short rates fall, an IAR swap outperforms a plain vanilla swap from the perspective of the fixed rate receiver.7 As short rates decline, an IAR swap amortizes faster, allowing the fixed rate receiver to enter into another swap at a higher long-term fixed rate, whereas the owner of a plain vanilla swap will continue to hold an instrument that now pays a below-market fixed rate. Similarly, if long rates fall and short rates rise, the IAR swap will also outperform the plain vanilla swap for the fixed rate receiver. As short rates rise, the IAR swap amor tizes at a slower pace, enabling the fixed rate receiver to continue receiving an above-market fixed rate for a longer period. In contrast, the owner of a plain vanilla swap experi ences reinvestment losses at the now lower long-term fixed rate when the plain vanilla swap matures. Pricing of IAR swaps In principle, the fixed rate of an IAR swap is set at the level that gives the swap an expected net present value of zero at origination. That is, the IAR swap is priced by taking the swap’s net cash flows over each of the possible paths of LIBOR rates (in Table 3, three equally likely paths) and solving for the fixed rate that makes the average present value of the net cash flows equal to zero. In practice, all pricing models apply weights to the possible paths. To maintain the internal consistency of the pricing model, these paths and their weights are chosen so that arbitrage possibilities are eliminated. Table 3 illustrates the difference in pricing between an IAR swap and a plain vanilla swap. Consider an IAR swap with a $1,000 initial notional principal and the amortization schedule presented in Table 1. The cash flows calculated in the example are from the perspective of the fixed rate receiver. For simplicity, assume that the possible future paths of LIBOR rates are the three paths indicated by cases 1,2, and 3. Case 2 is the path of LIBOR rates implied by forward rates derived from the initial yield curve, and the other two paths are possible alternative interest rate paths. 7 In reality, medium-term rates of under five years are relevant for IAR swaps because the contractual maturity in most IAR swaps is five years or less. Digitized66 for FRASER FRBNY Quarterly Review/Winter 1993-94 The price (or the fixed rate) of the plain vanilla swap is the fixed rate that causes the present value of the fixed pay ments to equal the present value of floating payments as forecast by the initial forward rates.8 The fixed rate of the IAR swap is 4.745 percent, while the fixed rate of the plain vanilla interest rate swap is 4.50 percent. In effect, the 24.5 basis point difference between the two rates represents the value of the implicit options in the IAR swap. The complexity of the IAR swap’s valuation process is itself a source of uncertainty. Market participants will use different assumptions about volatilities, future interest rate paths, and the correlations between long and short rates in their IAR swap interest rate models. These different assumptions can create larger price variations between dif ferent market participants’ pricing models for IAR swaps than is the case with plain vanilla interest rate instruments, which are priced using the observable yield curve. Risk issues Price risk The greatest risk for an investor (that is, fixed rate receiver) in an IAR swap is the opportunity cost of holding an IAR swap in the event of a significant interest rate move up or down. If short rates rise sufficiently, the net payout for the fixed rate receiver (end-user) can become negative if the amount of the floating rate payment exceeds the amount of the fixed rate receipt. This interest rate risk is amplified in an IAR swap because as rates rise, the swap’s amortiza tion slows and the fixed rate receiver may have a negative cash flow for a longer period. Since the birth of the IAR swap market in 1990, short term rates have declined. Thus, most IAR swaps initiated to date have ended immediately after the lockout period, and the behavior of IAR swaps in a rising rate environment has not yet been tested.9 Many end-users may find it difficult to determine pre cisely the risk-return tradeoff provided by IAR swaps. The exact set of interest rate options embedded in an IAR swap is not easily identified because of the IAR swap’s pathdependent nature. Hence, buyers cannot go to an exchange and price a specific set of options equivalent to those embedded in the IAR swap. As a result, fixed rate receivers will have a difficult time judging whether or not they have received the appropriate premium for the implicit options • Alternatively, the plain vanilla swap can be priced over the same set of possible interest rate paths used in pricing the IAR swap. If these interest rate paths satisfy a consistency condition known as the “arbitrage-free” condition—a requirement that profitable, riskless strategies be ruled out—then the two pricing methods for the plain vanilla swap will produce the same price. 9 Recently, barrier-type options called “knock-outs” have been offered on some IAR swap contracts. A knock-out clause typically states that if interest rates rise above a certain level (the knock-out rate), the swap will terminate automatically. This feature effectively eliminates the extension risk for the end-user. However, these contracts are expensive and thus tend to defeat the yield-enhancement feature of the IAR swap. as mentioned previously, the exact structure of the interest rate options embedded in an IAR swap cannot be easily determined from the swap’s amortization schedule. The path-dependent nature of the IAR swap requires dealers to use interest rate models to “reveal” and then dynamically hedge the swap’s embedded options because the pathdependency of these options cannot be replicated by any simple buy-and-hold options portfolio. Moreover, dealers must use sensitivity analysis or simulations of both the IAR swap and the rest of th eir portfolios to determ ine the degree to which the IAR swaps and other exposures in the portfolio offset each other. Hence, hedging the IAR swap’s exposures depends on the reliability of the interest rate model used in the simulations. they have sold, because they lack readily apparent and equivalent market prices for the set of options embedded in an IAR swap. Hedging risk To hedge IAR swaps, dealers use interest rate term struc ture models that incorporate several assumptions about the volatility of rates and the correlation of movements in short and long rates. As a first step, the dealers estimate the IAR swap’s exposures with an interest rate model.10 Next, they take into account the offsetting exposures already in their portfolios to determine a residual exposure. These residual exposures (both to changes in interest rate levels and changes in interest rate volatilities) are then hedged, usu ally using Eurodollar futures and interest rate options. An interest rate model is required for hedging because, Model risk Estimating the true profitability over time of an IAR swap can be difficult. Because of the IAR swap’s path-dependent behavior, the instrument cannot be easily broken down into 10 See Julia Fernald, “The Pricing and Hedging of Index Amortizing Rate Swaps," in this issue of the Quarterly Review. Table 3 Comparison of the Pricing of an IAR Swap and a Plain Vanilla Swap Year Forward Rate (Percent) Notional Principal Fixed Payment1" Floating Payment* Net§ Present Value of Net 4.50 4.00 3.50 3.25 3.00 1,000 1,000 0 0 0 47.45 47.45 0.00 0.00 0.00 45.00 40.00 0,00 0.00 0.00 2.45 7.45 0.00 0.00 0.00 2.35 6.85 0.00 0.00 0.00 IAR swap pricing Case 1 0-1 1-2 2-3 3-4 4-5 Sum Case 2 0-1 1-2 2-3 3-4 4-5 1,000 1,000 200 40 8 4.500 4.500 4.500 4.500 4.500 47.45 47.45 9.49 1.90 0.38 45.00 45.00 9.00 1.80 0.36 2.45 2.45 0.49 0.10 0.02 2.35 2.24 0.43 0.08 0.02 Sum Case 3 0-1 1-2 2-3 3-4 4-5 4.50 5.01 5.53 5.82 6.13 1,000 1,000 705 539 445 47.45 47.45 33.45 25.58 21.12 45.00 50.10 38.99 31.37 27.28 2.45 -2.65 -5.54 -5.79 -6.16 (9.20 Plain vanilla swap pricing11 1 4.50 2 4.50 3 4.50 + 5.12 1,000 1,000 1,000 + 45.00 45.00 45.00 -14.32) 45.00 45.00 45.00 -h3 5.12 2.35 -2.41 -4.79 -4.73 -4.74 Sum Average11 9.20 -14.32 0.00 0.00 0.00 0.00 0.00 0.00 0.00 f Fixed payments are calculated by multiplying notional principal by 4.745 percent. * Floating payments are calculated by multiplying notional principal by LIBOR. ® Net is the difference between the fixed and floating payments. 1 The average is calculated under the assumption that the three possible LIBOR paths are equally likely. ft Since the average life of this IAR swap is approximately three years, the comparable swap is the three-year plain vanilla swap. FRBNY Quarterly Review/Winter 1993-94 67 pieces that look exactly like other instruments whose prices are known. Hence, the product’s valuation depends criti cally on interest rate models. This dependence on interest rate models and the possibility of mispricing is known as “model risk.” The set of possible interest rate paths over which an IAR swap is priced and valued is usually generated using one or two factor interest rate models. One factor interest rate models implicitly assume perfect correlation between changes in short and long rates. Two factor interest rate models, by contrast, can simulate imperfectly correlated short- and long-term rates. In this respect, two factor models would appear to provide better representations of the term structure than one factor models. Two factor models, how ever, require their users to make explicit assumptions about the correlation between separately varying short- and long term rates. If inappropriate assumptions are made, then a two factor model’s results can be less accurate. The pricing models must also rely on assumptions about the volatility of short- and long-term rates. Assumptions about volatility, like those concerning the correlation of short and long rates, make IAR swaps difficult to “mark to market” and to hedge. The correlation of rates, however, is an especially difficult parameter to forecast, and problems can arise because pricing model results are particularly sensitive to the assumed magnitude of the correlation. For example, the assumptions about correlations can have a substantial impact on the level of the fixed rate determined by the model. Closely related to model risk is “personnel risk.” When the IAR swap market was first formed, finding personnel familiar with the instrument’s pricing and hedging demands was difficult. In some cases, only one trader at an institu tion may have been familiar with IAR swap pricing models. If that trader left the firm, a knowledge gap could arise, making the risk management of outstanding IAR swap positions more difficult. Fortunately, personnel risk tends to diminish as a product matures and market participants become more familiar with the instrument’s behavior in a variety of market conditions. Liquidity risk For end-users, significant illiquidity exists in the IAR swap market because of the difficulties of hedging and the cus tomized nature of the instrument. Because only dealers with sizable interest rate option exposures can successfully compete in the IAR swap market, only a handful actively trade this product. Smaller dealers, who generally lack siz able interest rate options positions, find it more difficult to hedge IAR swaps in a cost-effective way and typically exe cute these swap deals only if they can earn a substantial margin up front. Without a sizable interest rate options book, small dealers would have to sell options in the market to offset their IAR swap positions. 68 FRBNY Quarterly Review/Winter 1993-94 Dealers have expressed their willingness to make a sec ondary market in this product for customers, but as of yet an active secondary market has not developed.11 Normal industry practice is for the initiating dealer to make a bid to the customer who wants to liquidate an existing contract. But if the dealer chooses not to buy back the swap from an end-user and the end-user is unable to find another dealer to assume the swap, the end-user cannot easily liquidate or offset the position. Hedging, instead of unwinding, would be difficult for most end-users because the precise nature of the exposure to be hedged can be discovered only with an interest rate model, which IAR swap end-users normally do not possess. Credit risk Principal risk is not present in an IAR swap because there is no principal investment (as there is in mortgage securi ties). Hence, potential credit losses are limited to the net exchange of interest payments over the remaining life of the swap. Like plain vanilla interest rate swaps, IAR swaps are priced with a zero net present value at inception. As short-term interest rates change, the net interest payments will acquire a net positive or negative present value. This present value is the credit exposure between the two coun terparties and is usually only a small fraction of the notional principal. Thus, IAR swaps pose no additional or funda mentally different credit or settlement risks than those already present in the plain vanilla interest rate swap. The market for IAR swaps The number of dealers currently active in the IAR swap market is small but growing. While major U.S. securities firms dominate the market, U.S. money center banks and foreign bank subsidiaries also participate in the market. New York is the market center for IAR swaps, and most IAR swaps are denominated in U.S. dollars. The low short-term interest rate environment in the United States has no doubt been more conducive to the development of the IAR swap market than have other countries’ interest rate environ ments. If the yield curves of other countries begin to steepen, however, investors may begin to use IAR swaps pegged to non-U.S. rates. Initially, regional banks were the primary end-users of IAR swaps. Much of the recent growth in demand, however, has come from mutual funds, insurance companies, and other institutional investors. Uses of IAR swaps For dealers with sophisticated risk management systems, IAR swaps provide offsets to the exposures arising from 11 Secondary market liquidity has yet to be tested in the swaps initiated before or during 1991 because these swaps ended immediately after the lockout period owing to a dramatic drop in rates over the past two years. their over-the-counter interest rate options business. As fixed rate payers, the dealers own the options embedded in the IAR swap. Hence they can use these options to hedge their written interest rate option positions as well as other exposures in their interest rate swap book. From the viewpoint of investors such as mutual funds, insurance companies, and regional banks, IAR swaps pro vide enhanced yields in a low interest rate environment. These investors, as writers of the options embedded in IAR swaps, are essentially speculating that interest rate changes will be less volatile than buyers of the embedded options expect. In other words, these investors are betting that short- and medium-term rates will remain unchanged or will rise more slowly than predicted by the forward curve. If this scenario does in fact occur, investors will receive an above-market fixed return over the life of the swap from the premiums on the unexercised implicit options that they sold in the swap. Investors also find IAR swaps to be a useful substitute for mortgage-related securities such as collateralized mort gage obligations (CMOs) and pass-throughs. IAR swaps offer mortgage-bond-type yields and a similar risk profile, but remove the idiosyncratic portion of prepayment risk associated with mortgage securities. Idiosyncratic prepay ment risk refers to risk not directly related to changes in interest rates. For example, the need to relocate or a death in the family may prompt a homeowner to prepay a mort gage in what would otherwise seem to be an unfavorable interest rate environment. IAR swaps eliminate risks of this kind, leaving only the interest-rate-sensitive portion of pre payment risk. For many end-users, the IAR swap combined with a posi tion in Treasury securities provides additional advantages over owning CMOs and other types of cash mortgage instruments. IAR swaps offer a less uncertain absolute final maturity than do CMOs, and as a result, they have a more predictable weighted-average-life profile than CMOs and other mortgage assets. IAR swaps also have fewer opera tional complexities than mortgage securities. For example, the IAR swaps’ typical quarterly pay structure is easier to track than the pay structure of mortgage-backed securities, whose principal and interest payments must be recalcu lated monthly as prepayment rates change. By entering into an IAR swap while holding Treasury securities, a regional bank can increase its liquidity while receiving yields similar to those of a CMO and maintaining an interest rate exposure comparable to a mortgage pro duct’s. Dealers’ marketing materials for IAR swaps also emphasize “capital efficiency,” suggesting that some regional bank end-users use IAR swaps to reduce capital requirements. A position combining government securities and an IAR swap has low capital requirements that can offer advantages over the purchase of similar short-dated CMO securities. Note, however, that this difference in capi tal requirements is justified by the lack of any principal risk in the IAR swaps. Size and growth prospects The IAR swap market has been expanding rapidly for the past two years, showing particularly fast growth through the first half of 1993. An estimated $100 billion to $150 billion in notional principal has been originated since 1990. It is unlikely that this expansion will slow markedly unless the yield curve flattens dramatically. The market for IAR swaps to date is almost completely one-way in nature. Dealers are almost exclusively the fixed rate payers (buyers of the embedded options), and end-users are almost exclusively the fixed rate receivers (writers of the embedded options). Recently, however, a small interdealer market has developed and a modest number of transactions have been completed through interdealer brokers. Although the market seems to be expanding and matur ing, growth could ultimately be limited by dealers’ inability to sell the embedded options they have purchased by pay ing the fixed rate. Dealers must manage their options risk and thus do not want a large net long or net short options position. Dealers may be forced to cease writing IAR swaps if they cannot use the purchased options to hedge other written option risk or if they cannot resell the long options exposures. The cost of hedging the residual exposures cre ated by unmatched positions can become prohibitive, especially as the IAR swap market becomes more competi tive and the cost of the embedded options begins to increase.12 In fact, some dealers have shown reluctance to originate new transactions because the difficulties of hedg ing and evaluating the prospective profitability of these instruments become more critical as spreads narrow.13 Conclusions IAR swaps have proved useful to both investors and deal ers. Investors in this instrument can acquire a position that pays off if rates rise more slowly than predicted by the for ward curve. Investors in the swaps have also earned enhanced yields comparable to those on mortgage bond securities while remaining exempt from the idiosyncratic portion of the prepayment risk embedded in mortgage securities. Through IAR swaps, investors have been able to earn short-dated mortgage-type yields for at least two 12 If dealers were able to sell all of the IAR swaps’ embedded options, they would not be forced to go to the Eurodollar futures market to hedge residual risk not offset within their portfolio of other options. Alternatively, if a two-way market for IAR swaps existed, dealers would be able to receive the fixed rate and create a natural hedge for those existing IAR swap positions where they are the fixed rate payer. 13 The rating agencies have prohibited dealers from placing IAR swaps in their special-purpose AAA-rated swap subsidiaries. The agencies cite concerns that the one-way nature of the IAR swap market would make it more difficult to unwind such a swap book in a timely manner. FRBNY Quarterly Review/Winter 1993-94 69 years, while many cash mortgage securities have prepaid. Dealers with large interest rate options books have found IAR swaps attractive as an alternative instrument for hedg ing the exposures arising from their over-the-counter options business. In other words, IAR swaps have created a natural offset for most dealers’ net short positions in options, thereby helping dealers to meet the m arket’s demand for interest rate options. Most of the risks associated with IAR swaps are similar to those of other instruments. The IAR swap poses the same threat of negative cash flows as plain vanilla interest rate swaps or equity-index swaps, along with prepayment and reinvestment risks similar to those of mortgage securities. Nevertheless, while IAR swaps pose few unique risks for most market participants, significant problems may materi alize in a portfolio with a high concentration of IAR swaps. Certainly, model risk figures more prominently in IAR swaps than in other kinds of instruments. Pricing and hedg ing IAR swaps require highly technical interest rate models, and the absence of benchm ark market prices and the instrument’s relatively long life mean that pricing model inaccuracies may not become immediately apparent. A dealer who enters the market w ithout strong technical expertise may encounter problems arising from mispricing and mishedging. Risk management systems in place for plain vanilla interest rate swaps and options may not be sufficient to handle the complexity of IAR swaps. A firm ’s internal risk control unit must be capable of accurately monitoring the trading desk’s pricing and hedging models for IAR swaps. In sum, dealers who are active in the IAR swap market need considerable technical knowledge as well as strong risk management systems. The variable maturity feature of IAR swaps requires that an in s titu tio n ’s risk m anagem ent system take proper account of longer term exposures embodied in these instru ments. For example, excessive emphasis by management on short-term trading results may create incentives to enter into IAR swaps strictly for short-term yield enhancement or trading gains, without consideration of the long-term perfor mance results of the instrument. Note, however, that this problem exists for all instruments with medium- to long term option-like exposure, not only IAR swaps. This problem highlights potential weaknesses in current methods of recognizing trading gains in accounting sys tems. For example, the fixed rate return of an IAR swap contains an option premium for future option-like liabilities or exposures. This feature leads one to ask how much of an IAR swap’s yield premium should be incorporated in cur rent income. From a broader perspective, the proliferation of IAR swaps and similarly complex financial transactions underscores the need for accounting and disclosure prac tices suited to such instruments. Appendix: Reverse Index Amortizing Rate Swaps Instrument structure Anticipating a possible rise in short-term interest rates, investors are seeking to limit potential losses on their float ing rate exposures. In response to this demand, dealers are currently marketing a variation of the IAR swap called the reverse index amortizing rate swap or RIAR swap. Like an IAR swap, an RIAR swap is an interest rate swap whose notional principal amortizes at a rate that varies with the level of market interest rates according to a predetermined schedule. In a typical RIAR swap, as in an IAR swap, an end-user receives the fixed rate while paying the dealer a floating rate. An RIAR swap’s amortization schedule differs from that of an IAR swap, however, in calling for the notional principal to amortize more quickly as market interest rates rise. For example, if the floating rate index rises sufficiently, the swap could fully amortize at the end of the lockout period. Alternatively, if rates decrease, the predetermined structure of the RIAR swap could cause the swap to amor tize more slowly or, in some cases, not at all. The amortizing feature of an RIAR swap can be viewed as an implicit put option, giving the floating rate payer the right to “put” or reduce a floating rate liability if rates increase. For this right, the floating rate payer receives a 70 FRBNY Quarterly Review/Winter 1993-94 somewhat lower fixed rate than would be paid on a plain vanilla interest rate swap. At the present time a small number of U.S. securities firms and money center banks are developing this product. Only a handful of trades are believed to have taken place in the market to date. RIAR swaps are being marketed to corporate end-users, banks, mutual funds, insurance companies, and other insti tutional investors. Risks The RIAR market is presently one-sided. To date, only deal ers have written the embedded put option in the RIAR swap, and in their normal course of business, they are typically net sellers (writers) of options. Thus, for dealers with net short option positions, writing put options embedded in RIAR swaps may increase their overall portfolio’s residual expo sure and raise hedging costs. Like IAR swaps, RIAR swaps involve no principal risk. The greatest risk to an investor would be the opportunity cost of holding an instrument paying a below-market rate of interest if rates were to remain stable. The Pricing and Hedging of Index Amortizing Rate Swaps by Julia D. Fernald Index amortizing rate (IAR) swaps have been popular yield enhancement instruments over the past few years.1 The enhanced yields associated with these instruments result from premiums earned on options embedded in the swaps. Because these options depend on the path of interest rates, the pricing of IAR swaps requires a model of interest rate movements.2 This article presents a simple example of an interest rate model, outlines IAR swap pricing derived from the model, and develops a hedging strategy to offset the uncertain cash flows from the swap. Finally, the article discusses the complications that arise in more realistic pricing and hedg ing situations. rising or falling equal one-half.4 Description of the swap Although the interest rate tree has only two periods of uncertainty, the IAR swap in our example has three cash flow payments. If we assume an IAR swap with a one-year lockout period, the first cash flow at time 0 is based on an original notional amount of $100 and the current one-year rate. The two subsequent payments depend on the realiza tion of the one-year rates at time 1 and time 2 and on the amortization schedule in Table 1. 4 The price of a two-period zero coupon bond with an interest rate of 9.995 Interest rate model In this example, we assume that one-year interest rates are well represented by a model with the binomial tree structure illustrated in the figure.3 The tree is consistent with initial two- and three-year interest rates of 9.995 percent and 9.988 percent, respectively, if the probabilities of rates 1 See Lisa Galaif, “Index Amortizing Rate Swaps,” in this issue of the Quarterly Review. p e rc e n t e qu a ls the p ric e of a tw o -ye a r zero co up o n bon d d erive d from the tree: iTo * I * ( 4 + i“Tt) = ' 827' In the pricing and hedging of IAR swaps, the relevant probabilities are those that make the binomial tree consistent with the current term structure of interest rates. Figure: Binomial Distribution of One-Year Interest Rates One-Year Interest Rates 2 Models used to value path-dependent interest rate options must be free from arbitrage in the sense that they price fixed-income instruments consistently with the current term structure of interest rates. The models can be represented by interest rate trees or lattices that give possible outcomes of future short-term interest rates. These representations are used to calculate both the initial price of the IAR swap and the dynamic hedges that swap dealers would enter over time. Time 0 Time 1 Time 2 Path 3 Our example assumes that future short-term rates are determined by one factor. The example is consistent with one-year rates that are normally distributed with a constant annual volatility of 1.0 percentage point. FRBNY Quarterly Review/ Winter 1993-94 71 Because the swap’s notional principal amortizes on the basis of the short rate, the swap cash flows at each period depend not only on the rate that period but also on the path of previous rates. Table 2 shows the four possible cash flow paths (from the perspective of the fixed rate payer) that arise from our interest rate model. In this example, F is the fixed rate paid on the IAR swap. Pricing As with any swap, the fixed rate on a IAR swap is deter mined such that the initial present value of the swap’s cash flows is zero. The present value of the cash flows from an IAR swap is more difficult to calculate than the correspond ing value for a plain vanilla swap, however, and depends on the assumed arbitrage-free interest rate model. In pricing our IAR swap, we find the fixed rate consistent with the pre determined amortization schedule, the assumed distribu tion of one-year interest rates, and our binomial represen tation of the model. The cash flows are functions of the fixed rate F, the current rate, and the path of previous rates. Because we have only four possible cash flow paths, we can solve explicitly for the fixed rate, F, that makes the average present value over these possible cash flow paths equal to zero. In this way, we obtain a fixed rate of 10.26 percent.5 In this example, with its virtually flat 10 percent term structure, the fixed rate on a plain vanilla swap is approxi mately 10 percent. The 26 basis point premium in the IAR swap fixed rate is the value of the embedded options that the fixed rate payer implicitly purchases. Table 3 shows the fixed rate payer’s cash flows over the four paths and the three time steps, given the 10.26 percent fixed rate. Notice that when the interest rate is 10 percent at time 2 (paths 2 and 3), the cash flows depend on the inter est rate at time 1. This difference illustrates the pathdependent nature of the IAR swap. 5 Let R t be the one-year interest rates and let CFpt be the cash flows for the four possible paths, p, and the three time periods, t. We solve for the fixed rate that sets the present value of the cash flows, or p= 1 t=0 Hedging Fixed rate payers (usually swap dealers) may wish to hedge their highly variable payments. For example, if rates rise in the first period, dealers receive $.663, but if rates fall, the dealer pays $.631. In the second period, dealers face a similarly variable outcome that depends on the path of interest rates. We show that if fixed rate payers hedge the uncertain cash flow s every period, they w ill earn exactly the additional 26 basis points that they pay as option premium. Although there are many ways to implement hedges, all methods involve calculating changes in the swap’s value given sm all changes in the underlying interest rates. Because our interest rate model involves only one factor, we need only one instrument to hedge the swap. For sim plicity of exposition, we choose to replicate the IAR swap’s payoffs using forward contracts instead of the more typi cally used futures contracts. In our example, the forward rate implied by the initial term structure is 9.991 percent on one-year contracts maturing at time 1. We choose the first hedge at time 0 to offset the two pos sible time 1 swap values. The time 1 swap values are com posed of two elem ents: the actual cash flow s paid or received on the swap and the expected value of the time 2 payments or receipts. The actual cash flows from the swap are the value of the time 0 payment (-$.263) at time 1 plus the time 1 amount (+$.663 in the up-state, or -$.631 in the Table 2 Fixed Rate Payer’s Cash Flows from the IAR Swap Cash Flows Path Time 0 Time 1 Time 2 1 2 3 4 $100'(10%-F) 100*(10%-F) 100*(10%-F) 100*{10%-F) $90*(11%-F) 90*(11%-F) 50*(9%-F) 50*(9%-F) $90’ (12%-F) 72*(10%-F) 40*(10%-F) 0*(8%-F) C \+ R p ,q ) <7=0 equal to zero. Table 3 Fixed Rate Payer’s Cash Flows from the IAR Swap with a Fixed Rate of 10.26 Percent Table 1 Amortization Schedule Interest Rate (Percent) Notional Amortization (Percent) 12 11 10 9 8 0 10 20 50 100 Digitized 72 for FRASER FRBNY Quarterly Review/Winter 1993-94 Cash Flows Path Time 0 Time 1 Time 2 1 2 3 4 $-0,263 -0.263 -0.263 -0.263 $0,663 0.663 -0.631 -0.631 $1,563 -0.189 -0.105 0.0 allows for nonparallel shifts in the yield curve, it implicitly assumes multiple sources of risk; it thus requires multiple hedging instruments. Bucket hedging is useful if interest rate dynamics are more complicated than the single factor model assumes. down-state).6 The expected remaining value of the swap is $.612 in the up-state, and -$.048 in the down-state. If the dealer combines the $100 swap with -$97.5 of the forward contract, the portfolio’s value will be equal to zero at time 1 whether rates rise to 11 percent or fall to 9 per cent.7 At time 1, the dealer follows the same type of calcula tion, keeping track of the time 2 values of the swap and the previous hedge. The new hedge amounts are -$87.0 if we are in the up-state or +$5.2 if we are in the down-state. The process of readjusting hedges through time is known as “dynamic hedging.” If we adopt these hedge amounts, the outcome from hedging the swap along each path offsets the payoffs from the swap along that path. Table 4 illustrates the calcula tions of the hedged swap’s value along the first path. The hedged swap’s value along the other three paths will also equal zero at time 2. Another hedging method computes the change in the sw ap ’s value fo r changes in each forw ard rate. This “bucket” hedge method involves (1) the initial purchase of a series of forw ard contracts in amounts that offset the recomputed swap’s value and (2) the dynamic adjustment of the hedge through purchases or sales of additional for ward contracts in the future.8 Because bucket hedging Issues In this example, the hedges perfectly offset the swap if any of the four modeled interest rate paths is realized. Although it is simplistic to assume that interest rates will follow one of these four paths, the example illustrates potential issues that can arise when valuing and hedging interest-ratedependent derivatives. In particular, the pricing and hedg ing of any interest rate derivative security depend on deci sions at several levels concerning: • the interest rate model: How many factors are rele vant? W hat type of process do they follow — for example, normal, lognormal? • the parameters of the model: What are the volatili ties? If the model includes more than one factor, what are the correlations? • the implementation of the model: How small are the time steps? Is it a binomial or trinomial tree? How many simulations are used? 6 The total cash flow from the swap in the up-state is therefore $.372, which equals -$.263*( 1.11) + $.663. If assumptions about the model and the parameters of the model are incorrect, the hedging cannot offset realized gains and losses. In our example, hedging depends on the forward rates implied by our interest rate tree. If these rates are not realized, the cash flows from the hedges cannot perfectly offset the cash flows from the swap. These rates can be wrong because the short rate process is in fact not well represented by a single factor normal distribution with constant volatility. Valuing the swap using other interest rate models— for example, a two factor lognormal interest 7 The hedge amounts are essentially (the negative of) the derivative of the swap's value with respect to interest rates. In our example, the first hedge amount, $97.5, equals $.983 (the swap's value in the up-state) less $-.966 (the value in the down-state), divided by .02 (the difference in the interest rates). 8 In our example, we would initially sell $80.1 of the forward contract maturing at time 1 and $19.1 of the contract maturing at time 2. If rates rise to 11 percent, we would sell $68.1 of the contracts maturing at time 2 at the new forward rate; if rates fall to 9 percent, we would buy $24.5 of the time 2 forward contracts. Table 4 Payment Stream for the First Path gfiSS Swap at time 0 Cash Flows Time 1 Time 2 -.263 -.263(1.11) -.263(1.11)(1.12) = -.327 .663(1.12) = -.743 1.563 = 1.563 -97.5(.11-.0999)(1.12) = -1.102 Swap at time 1 .663 Swap at time 2 Hedge entered at time 0 Hedge entered after up-jump at time 1 Value of the hedged swap at the end of time 2 Future Value Time 0 -97.5(.11-.0999) MB jjjj -87 .0(.12-.1099) -.878 0.0 FRBNY Quarterly Review/ Winter 1993-94 73 rate model—can give a different fixed rate and different hedges. Different assumptions about the parameter values also affect the fixed rate. In our example, if the volatility is 1.5 percentage points instead of 1.0, the fixed rate will increase from 10.26 percent to 10.60 percent. The differences across models and parameter values can be considerable, and careful judgment should be used when testing the sen sitivity of the results to different assumptions. The fixed rate and the subsequent hedging also depend on how the model (with its assumptions) is implemented. The goal in implementing the model is to approximate numerically a stochastic process. If we shorten the time steps, we will find a different fixed rate than we find with annual time steps. The appropriate time step for valuation Digitized 74 for FRASER FRBNY Quarterly Review/Winter 1993-94 is the one in which the fixed rates have converged on a value. In our example, the hedge ratios at time 1 are sig nificantly different when the rates rise to 11 percent than they are when the rates fall to 9 percent. If we shorten the time steps and update the hedge ratios more often, the hedging will change more gradually than is illustrated by our example. Actual models are more complex than our example at all levels: volatilities are not necessarily constant, the initial term structure is not conveniently flat, and models are implemented with higher frequencies. Adjustments need to be incorporated for nonparallel shifts in the yield curve because nonparallel shifts will affect the swap’s value. Making errors at any of these levels will potentially result in a misvalued instrument. Recent Ttends in Commercial Bank Loan Sales by Rebecca Demsetz Loan sales represent an important departure from the tra ditional bank activity of originating credit to be held until maturity. The dollar volume of commercial and industrial (C&l) loan sales rose rapidly in the mid-1980s but has declined equally rapidly over the past few years. Previous studies have discussed these aggregate trends;1 however, aggregate data mask some interesting differences between the loan sales activities of the largest sellers and those of all other banks. This article seeks to provide insight into recent loan sales declines by examining the sales activities of two distinct groups of institutions. The first group includes the top few sellers only and is referred to as the market’s “first tier.” All other institutions are labeled “second tier.” The article finds that recent declines in loan sales appear to reflect a drop-off in the origination of loans likely to be traded in the secondary market, rather than a disruption of the secondary market process. Diminished origination of salable loans reduces the volume of “inputs” available for secondary market transactions. This pattern seems to have characterized the sales activities of both first-tier and sec ond-tier banks. Second-tier sales first fell during the 199091 recession and have continued to decline with the persis tent weakness of C&l lending since the recession. First-tier trends also reflect recession-related origination declines and the ongoing weakness in C&l lending, but appear to be 1 For example, see Joseph Haubrich and James Thomson, “The Evolving Loan Sales Market," Federal Reserve Bank of Cleveland Economic Review, July 1993; Richard Cantor and Rebecca Demsetz, “Securitization, Loan Sales, and the Credit Slowdown,” Federal Reserve Bank of New York Quarterly Review, Summer 1993; and Allen Berger and Gregory Udell, “Securitization, Risk, and the Liquidity Problem in Banking,” in Michael Klausner and Lawrence White, eds., Structural Change in Banking (Homewood, III.: Irwin Publishing, 1992), pp.227-91. driven mainly by a decrease in large credits related to cor porate acquisitions, leveraged buyouts (LBOs), and recapi talization. The following three sections examine loan sales trends; the role of corporate acquisitions, LBOs, and recap italization; and the importance of economic conditions. Aggregate, first-tier, and second-tier sales trends Chart 1 tracks the aggregate loan sales activity of all insured domestic commercial banks from the first quarter of 1986 through the first quarter of 1993. The data in Chart 1, drawn from banks’ Reports of Income and Condition (“Call Reports”), measure the dollar volume of C&l loans origi nated and sold without recourse during each calendar quarter.2 Quarterly loan sales flows attributable to insured domestic commercial banks peaked at $285 billion in 1989. By 1993, these flows had dropped to $89 billion, a decline of almost 70 percent. The aggregate trends revealed in Chart 1 mask important differences in loan sales trends associated with first-tier and second-tier sellers. Chart 2 adds two additional series describing the loan sales activities of these subsets of the insured domestic commercial bank population. The first-tier subset includes the top five sellers in each quarter exam ined. All other banks belong to the second-tier subset. The size of the first-tier subset may seem arbitrarily small; how ever, loan sales attributable to the second five sellers are much smaller than those attributable to the top five sellers and follow trends similar to those experienced by the 2 The term “loans sold” refers to the sale of entire loans or portions of loans; “loans originated" refers to loans made directly by the reporting bank and does not include loans purchased from other institutions. When a loan is sold “without recourse,” the risk of the loan is transferred to the buyer. FRBNY Quarterly Review/Winter 1993-94 75 remainder of the loan sales market.3 Two first-tier sellers are especially important, accounting for an average of 42 percent of aggregate sales between first-quarter 1986 and first-quarter 1992.4 Chart 2 shows that first-tier banks, all very large institu tions, account for a substantial fraction of both the level of aggregate sales and trends in aggregate sales. Sales by first-tier banks increased rapidly through 1988, fell sharply from the third quarter of 1989 through the first quarter of 1990, and then fell more gradually over subsequent years. These first-tier sales account for 81 percent of the aggre gate loan sales increase between first-quarter 1986 and third-quarter 1989 and 80 percent of the subsequent loan sales decline. Second-tier sales trends differ from first-tier trends but also show a rise and subsequent decline over the 1986-93 period. Second-tier sales rose gradually in the mid-1980s, peaked in the first quarter of 1990 (after the peak in first-tier sales), and then fell at a rate similar to the rate of decline in loan sales by first-tier banks. 3 In addition, the composition of the first-tier subset is quite stable over time. A total of eight institutions appear in the seven first-tier subsets corresponding to the first quarters of 1986-92. The role of corporate acquisitions, LBOs, and recapitalization For firs t-tie r banks, loan syndication a ctivity provides insight into trends in the origination of salable loans. Syndi cations are large credits shared by a group of banks upon origination. Since syndications are commonly parceled into smaller credits that are sold in the secondary market, syndi cated loan volume gives some indication of the strength of secondary market loan supply by large sellers. The table reports annual syndicated loan volume by purpose from 1987 through 1992. These data reflect lines of credit as well as actual loan originations, so they overestimate the vol ume of syndicated loans available for secondary market sale. Nevertheless, they do clarify the trends in the origina tion of salable loans by large banks. The table shows that total syndicated loan volum e increased between 1987 and 1989 and then fell abruptly. Furthermore, the sharp drop in total syndicated loan vol ume was driven by loans in the “leverage” category. Syndi cated loans extended for leverage purposes finance corpo rate acquisitions, LBOs, and recapitalization. They are unlikely to represent lines of credit, because investmentgrade borrowers generally use credit lines to support com- 4 Berger and Udell first noted the importance of these two institutions, Security Pacific and Bankers Trust. See “Securitization, Risk, and the Liquidity Problem in Banking.” Chart 2 Chart 1 Quarterly Sales of Commercial and Industrial Loans by Insured Domestic Commercial Banks: Aggregate Trends Quarterly Sales of Commercial and Industrial Loans by Insured Domestic Commercial Banks: Subsample Trends Billions of dollars Billions of dollars 3 0 0 ------;------------------------------------------------------ -------------------- / 250 / / 1//r / / y I/ 200 150 100 / ♦'V* ' J / \ \ ' V \ \ \ First-tier sales Source: Federal Financial Institutions Examination Council, Reports of Condition and Income. 76 FRBNY Quarterly Review/Winter 1993-94 Aggregate sales \ v. V " Second-tier sales 50 0I I1986' ' I ' 87M I M88 I I M89 ' I M90 ' ' M91 ' I M92 I I93I ^ 1 1 i i. i . l 11 i 1 i i i 1 i ; i i i i i 1986 87 88 89 90 — \ V 111 1. 1I 91 I l l 92 93 Source: Federal Financial Institutions Examination Council, Reports of Condition and Income. Note: "First-tier" banks are defined as the top five sellers in each quarter. All other banks are defined as "second-tier.” mercial paper issuance. Data from the Bank Loan Report, a publication of Investment Dealers Digest, confirm a dra matic drop in syndications related to acquisitions, LBOs, and recapitalization between 1989 and 1990. Chart 3 compares trends in loan sales by first-tier sellers with trends in the volume of syndicated loans in the lever age category. In the chart, loan sales are reported at quar terly intervals and syndicated loan volume is reported at annual intervals, so comparisons should be made with cau tion. It is clear, however, that trends displayed by the two series are similar. Several authors have noted a positive correlation between aggregate sales and corporate merger Syndicated Loan Volume by Purpose activity.5 Chart 3 demonstrates a strong correlation between first-tier sales and lending related to acquisitions, LBOs, and recapitalization.6 Conversations with market partici pants confirm the importance of these activities in explain ing aggregate trends in the secondary market volume of top loan sellers. In the future, the effect of such activities on the loan sales market will depend on the extent to which they involve bank financing. The importance of economic conditions Trends in lending related to corporate acquisitions, LBOs, and recapitalization appear to be less important in explain ing loan sales by second-tier banks, which continued to rise through the first quarter of 1990. The timing of the secondtier sales decline suggests that the recession-related slow down in C&l loan origination was a key underlying factor. As Chart 4 shows, a composite index of four coincident Billions of Dollars 1987 Purpose Leverage (acquisition, LBO, recapitalization) Debt repayment Specialty finance General purpose Total 1989 1988 1990 1991 1992 66.1 11.5 17.0 42.5 162.7 42.3 8.6 70.7 186.5 44.4 7.1 95.3 57.9 42.6 17.4 123.4 20.9 46.5 16.6 150.4 39.9 58.5 23.0 215.1 137.1 284.4 333.2 241.3 234.4 336.5 5 See, for example, Berger and Udell, “Securitization, Risk, and the Liquidity Problem in Banking,” and Haubrich and Thomson, “The Evolving Loan Sales Market.” 6 Other authors have attempted to explore the relationship between bank loan sales and merger-related lending using Call Report data on “highly leveraged transactions,” or “HLTs.” (See Joseph Haubrich and James Thomson, “Loan Sales, Implicit Contracts, and Bank Structure,” in Proceedings from a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, 1993.) The main drawback of these data is that they were introduced in the Call Report only after the dramatic declines in corporate merger activity and aggregate loan sales. Other important limitations are that the HLT data measure the existing stock of highly leveraged transactions rather than the flow of new HLT originations and that all credits extended to an HLT borrower are considered HLT transactions, regardless of their particular purpose. Source: Loan Pricing Corporation. Chart 3 Chart 4 First-Tier Sales and Volume of Syndicated Loans in the Leverage Category Second-Tier Sales and Index of Coincident Indicators Billions of dollars 250 --------------------------------------------------------- Billions of dollars Index Index of ^ dent in d ic a to rs / ' Scale — ► / Q uarterly first-tier sales ____ A —A __________ 200 110 \ Annual syndicated X_____ loan volume, _j V leverage category 150 100 V 100 \ ' y t / • t Qua rterly ' second- ier sales \ M___ Scale » t * \ / \ 1986 87 88 89 90 91 92 Sources: Federal Financial Institutions Examination Council, Reports of Condition and Income; Loan Pricing Corporation. 93 1 1 1 1 1..I I ...!..I l l 1986 87 88 1.. M l ...I I I ..1 89 90 i i i 1 m 91 M 90 92 93 Sources; Federal Financial Institutions Examination Council, Reports of Condition and Income; Survey of Current Business. FRBNY Quarterly Review/Winter 1993-94 77 indicators of business conditions was highly correlated with second-tier sales from 1986 through the recent recession, peaking just one quarter after the peak in second-tier sales.7 The correlation between second-tier sales and the index of coincident indicators weakened after the reces sion. Economic conditions improved, but loan sales by both second-tier and first-tier banks continued their decline. This divergence of trends may be attributable to the persistent weakness in borrowing by large corporations after the recent recession. An econometric analysis confirms the importance of eco nomic conditions and lending opportunities in explaining cross-sectional variation in loan sales activity.8 This analy sis divides the country into fourteen geographical regions and investigates the effects of regional economic condi tions and a variety of bank characteristics on the loan sales activities of individual banks. Five of the geographical regions are identical to Census regions; the remaining four Census regions are divided into smaller geographical areas, with states that experienced similar economic cir cumstances over the relevant period grouped together. Variables used to measure economic conditions in each of the fourteen regions include the unemployment rate (a measure of current conditions) and consumer confidence (a measure of expected future conditions). Results of this empirical analysis suggest that regional economic condi tions have been relatively important determinants of loan 7 The four coincident indicators included in the composite index measure employment on nonagricultural payrolls, personal income less transfer payments, the index of industrial production, and manufacturing and trade sales. The composite index is available on a monthly basis from the Survey of Current Business. Values reported are from March, June, September, and December. 8 See Rebecca Demsetz, “Economic Conditions, Lending Opportunities, and Loan Sales,” Federal Reserve Bank of New York, working paper, 1994. 78 FRBNY Quarterly Review/Winter 1993-94 sales activity in recent years. In addition, the positive effect of economic conditions can be attributed, at least in part, to the relationship between economic conditions and loan origination opportunities. Declines in the dollar volume of loan sales are consistent with either a drop in secondary market supply or a drop in secondary market demand. Price data can help determine whether the recession-related reduction in sales was sup ply- or demand-driven. Data from the Asset Sales Report of the American Banker magazine reveal that secondary mar ket yields on C&l loans to investment grade borrowers fell relative to commercial paper yields during 1990 and 1991.9 In conjunction with the decrease in loan sales volume over the same time period, this yield decline (price increase) suggests a drop in secondary market supply. Data from the Board of Governors’ Senior Loan Officers Opinion Survey are consistent with this interpretation of secondary market dynamics in the early 1990s. The majority of banks included in the August 1992 and August 1993 surveys reported either increased demand or little change in demand by typical loan purchasers over the previous year. In summary, while the initial decline in first-tier sales is associated with a sharp drop in lending to finance corporate acquisitions, LBOs, and recapitalization, the turning point for second-tier sales can be linked to recent cyclical slow downs in C&l lending. The continued declines in both firsttier and second-tier sales since the recent recession have coincided with persistent weakness in C&l originations and the resulting reduction in secondary market supply. Together, these relationships suggest that recent drops in loan sales volume may be best explained by declines in the origination of salable loans. 9 Reported yields are on loans and commercial paper to borrowers rated A1 (Standard and Poor’s)/P1 (Moody’s) and borrowers rated A2/P2. Treasury and Federal Reserve Foreign Exchange Operations August-October 1993 During the August-October period the dollar appreciated 3.7 percent against the Japanese yen, depreciated 3.2 percent against the German mark, and was little changed on a trade-weighted average basis,1declining 0.4 percent. On August 19, the U.S. monetary authorities purchased $165 million against yen in the period’s only intervention operation. The yen appreciates, then reverses against the dollar During early August, the yen strengthened against the cur rencies of all major industrialized countries, reaching record highs against the dollar, mark, Swiss franc, the pound sterling, and the Canadian and Australian dollars. On August 11, the release of data indicating a wider than expected 28 percent year-on-year expansion of Japan’s merchandise trade surplus to $11.84 billion triggered sharp yen appreciation, and the yen traded to a new high against the dollar of ¥103.50. Continuing weakness in domestic economic indicators was perceived as evidence that reduc tion of Japan’s current account surplus was unlikely in the near term, and the yen moved to several new daily highs against the dollar, peaking at a postwar high against the dollar of ¥100.40 on August 17. From August 16 to 18, conditions in the Japanese money This report, presented by Peter R. Fisher, Senior Vice President, Federal Reserve Bank of New York, and Manager for Foreign Operations, System Open Market Account, describes the foreign exchange operations of the U.S. Department of Treasury and the Federal Reserve System for the period from August 1993 through October 1993. Frank Keane was primarily responsible for preparation of the report. 1 The dollar’s movements on a trade-weighted basis are measured using an index developed by staff at the Board of Governors of the Federal Reserve System. markets were eased. On August 19, the Japanese cabinet met and agreed to try to devise additional measures to stimulate domestic demand. The dollar was trading at ¥102.50 in early New York dealing on August 19, but then declined quickly to ¥101.35 following the release of the worse than expected $12.1 billion U.S. merchandise trade deficit for June; at the same time, the dollar abruptly declined one pfennig against the mark. The U.S. monetary authorities intervened shortly after the release of the trade data. During the day they purchased a total of $165 million against the yen, shared equally between the Federal Reserve and the Treasury’s Exchange Stabilization Fund. This operation was coordinated with another monetary authority. Initially, the operations surprised market participants, and the dollar promptly rose. During the morning, Treasury Under Secretary Summers released a statement welcom ing the decline in Japanese money market rates and expressing concern that further yen appreciation could retard growth in the Japanese and world economies. Oper ations continued after Under Secretary Summers’ state ment but ceased before noon. Market participants subse quently continued to cover short positions throughout the afternoon, and the dollar reached a high of ¥106.75 before closing the day at ¥105.95. In the month following the operation, the dollar-yen exchange rate largely traded between ¥103.00 and ¥106.00 as market participants increasingly focused on the appar ent weakness of the Japanese economy. A series of Japan ese data releases showed continued weak business senti ment, deteriorating corporate profits, and a 0.4 percent decline in second-quarter GDP. Consequently, when the Bank of Japan lowered the official discount rate (ODR) FRBNY Quarterly Review/Winter 1993-94 79 on September 21 by a greater than expected 75 basis points to 1.75 percent, the action was perceived as an appropriate supplement to the government’s efforts to stim ulate the economy, not as a device to avoid further yen appreciation. Favorable reactions by senior U.S. officials to the Bank of Japan’s action led to a perception that tensions between the U.S. and Japan on trade issues had given way to greater cooperation, and the yen declined about 1.5 per cent, closing on September 21 at ¥106.18. The dollar firmed gradually over the latter half of the three-month period while expectations of near-term volatil ity in the dollar-yen exchange rate dwindled substantially. The implied one-month option volatility fell from about 14 percent in mid-September to around 10 percent in late October. The period closed with the dollar-yen exchange rate trading steadily above ¥108.00 in late October. Mark appreciates against dollar in wake of ERM crisis The European Community finance ministers and central bank governors agreed, effective Monday, August 2, to per Chart 1 The Dollar against the Japanese Yen Japanese yen per U.S. dollar Note: Inset panel shows the six-month exchange rate movement. 80 FRBNY Quarterly Review/Winter 1993-94 mit currencies participating in the Exchange Rate Mecha nism (ERM) to fluctuate within 15 percent of their central parities. However, a uth orities from Germ any and the Netherlands agreed to maintain their bilateral exchange rate within 2.25 percent of their central parity. During the uncertainty created by the currency turmoil in Europe, mar ket participants had aggressively accumulated dollar posi tions in late July. When widely anticipated European inter est rate reductions failed to m aterialize in the first few weeks of August, the mark began to appreciate against the dollar. The negative sentiment toward the dollar during this period was reinforced by market reports of dollar sales by European central banks to adjust reserve positions after July’s currency turmoil, and by a widening of interest rate differentials in the mark’s favor implied by Eurocurrency futures contracts. The Bundesbank C ouncil’s decision on August 26 to leave official rates unchanged disappointed market expec tations of an interest rate cut, and banks were caught short of funds at the end of a reserve period. When the Council did lower the discount and Lombard rates by 50 basis points to 6.25 percent and 7.25 percent, respectively, on September 9, the concurrent sm aller than expected 10 basis point reduction in the Bundesbank’s money market repurchase rate, to 6.70 percent, led to continued tightness in short-term German money markets. These develop ments resulted in continued mark strength against the dol lar. Although the mid-September political unrest in Russia caused the dollar to appreciate briefly against the mark, the dollar again drifted lower against the mark when the crisis was resolved, closing at DM 1.6013 on October 13. On O ctober 21, the B undesbank Council surprised exchange markets by again reducing its discount and Lom bard rates by 50 basis points to 5.75 percent and 6.75 per cent, respectively. The Council also announced that it would conduct the following week’s fourteen-day repurchase agree ment at a fixed rate of 6.40 percent, a 27 basis point reduction from the prior day’s variable rate repurchase agreement. The dollar, which had begun rising gradually against the mark before the announcement, rose steadily over the remainder of the period, closing at DM 1.6857 on October 29. Other operations The Federal Reserve and the Treasury’s Exchange Stabi lization Fund (ESF) each realized profits of $22.1 million from the sales of Japanese yen in the market. Cumulative valuation gains on outstanding foreign currency balances as of the end of October were $3,368.5 million for the Fed eral Reserve and $2,839.0 million for the ESF. The Federal Reserve and the ESF regularly invest their foreign currency balances in a variety of instruments that yield market related rates of return and have a high degree of liquidity and credit quality. A portion of the balances is invested in securities issued by foreign governments. As of the end of October, the Federal Reserve and the ESF held either directly or under repurchase agreements $10,004.3 million and $10,276.6 million, respectively, in foreign gov ernment securities valued at end-of-period exchange rates. FRBNY Quarterly Review/Winter 1993-94 81 Chart 3 Doliar-Yen Interest Rate Differential Implied by the Three-Month Eurodeposit Futures (December Contract) Interest rate Interest rate differential 1.50 1.25 1.00 0.75 0.50 I . ............HIM iiiiiiiiiiimmni iiiiniumi.... October August 1993 September 1993 Chart 4 German Mark-Dollar Interest Rate Differential Implied by the Three-Month Eurodeposit Futures (December Contract) Interest rate differential 3.00 Interest rate 2.75 2.50 2.25 2 . 0 0 '................ ............................. , August 1993 Digitized for 82 FRASER FRBNY Quarterly Review/Winter 1993-94 i i i i i i i i i m i i i i i i i " I M I I I I ...................... September 1993 October .. Chart 5 Table 1 Federal Reserve Reciprocal Currency Arrangements Short-Term Interest Rates for Selected Countries Percent Millions of Dollars Amount of Facility Institution October 31, 1993 Austrian National Bank National Bank of Belgium Bank of Canada National Bank of Denmark Bank of England Bank of France Deutsche Bundesbank Bank of Italy Bank of Japan Bank of Mexico Netherlands Bank Bank of Norway Bank of Sweden Swiss National Bank Bank for International Settlements: Dollars against Swiss francs Dollars against other authorized European currencies Total 250 1,000 2,000 250 3,000 2,000 6,000 3,000 5,000 700 500 250 300 4,000 600 1,250 30,100 1993 Table 2 Net Profit (+) or Losses (-) on United States Treasury and Federal Reserve Foreign Exchange O perations Millions of Dollars Valuation profits and losses on outstanding assets and liabilities as of July 31, 1993 Realized profits and losses August 1-October31, 1993 Valuation profits and losses on outstanding assets and liabilities as of October 31, 1993 Federal Reserve U.S. Treasury Exchange Stabilization Fund +3,226.6 +3,005.5 +22.1 +22.1 +3,368.5 +2,839.0 Note: Data are on a value-date basis. FRBNY Quarterly Review/Winter 1993-94 83 Recent FRBNY Unpublished Research Papers* Single copies of these papers are available upon request. Write Research Papers, Room 901, Research Function, Federal Reserve Bank of New York, 33 Liberty Street, New York, N.Y. 10045. 9328. Abate, Joseph, and Michael Boldin. “The Money-Output Link: Are F-Tests Reliable?” September 1993. 9329. McCarty, Jonathan. “Does the Household Balance Sheet Affect Durable Goods Expenditures?” November 1993. 9330. Brauer, David. “Why Do Services Prices Rise More Rapidly Than Goods Prices?” December 1993. 9331. Malz, Allan M. “New Varieties of Foreign Currency Operations.” December 1993. 9332. Boldin, Michael D. “Econometric Analysis of the Recent Downturn in Housing Construction: Was It a Credit Crunch?” December 1993. 9401. Packer, Frank. “Venture Capital, Bank Shareholding, and the Certification of Initial Public Offerings: Evidence from the OTC Market in Japan.” January 1994. 9402. Osier, C. “Policy Stablization and Exchange Rate Stability." January 1994. 9403. Demsetz, Rebecca S. “Economic Conditions, Lending Opportunities, and Loan Sales.” February 1994. + Single copies of these papers are available upon request. Write Research Papers, Room 901, Research Function, Federal Reserve Bank of New York, 33 Liberty Street, New York, N.Y., 10045. FRBNY Quarterly Review/Winter 1993-94 . FEDERAL RESERVE BANK OF NEW YORK Studies on Causes and Consequences of the 1989-92 Credit Slowdown The persistent weakness of the U.S. economy between early 1989 and late 1992 was accompanied by an unprecedentedly sharp slowdown in credit growth. Economists have debated the linkages between these events, particularly the possibility that credit supply constraints may have played a significant role in weakening economic activity it period. Studies on Causes and Consequences o f the 1989-92 Credit Slowdown, a collec tion of papers by staff members of the Federal Reserve Bank of New York, explores this and other issues relating to the credit slowdown. The volume considers the relative importance of credit supply and credit demand factors, the role of bank and nonbank credit sources, the impact of credit supply shifts on the economy, and the implications of those shifts for monetary policy. Postpaid $5.00 U.S., $10.00 foreign. Orders should be sent to the Public Information Department, Federal Reserve Bank of New York, 33 Liberty Street, New York, N.Y. 10045. Checks should be made payable to the Federal Reserve Bank of New York. FRBNY Quarterly Review/Winter 1993-94 85 Single-copy subscriptions to the Quarterly Review (ISSN 0147-6580) are free. Multiple copies are available for an annual cost of $12 for each additional subscription. Checks should be made payable in U.S. dollars to the Federal Reserve Bank of New York and sent to the Public Information Department, 33 Liberty Street, New York, N.Y., 10045-0001 (212720-6134). Single and multiple copies for U.S. subscribers are sent via third- and fourthclass mail. Subscriptions to foreign countries, with the exception of Canada, are mailed through the U.S. Postal Service’s International Surface Airlift program (ISAL) from John F. Kennedy International Airport, Jamaica, New York. Copies to Canadian subscribers are handled through the Canadian Post. Quarterly Review subscribers also receive the Bank’s Annual Report. Quarterly Review articles may be reproduced for educational or training purposes, provid ed that they are reprinted in full and include credit to the author, the publication, and the Bank. Library of Congress Card Number: 77-646559 FRBNY Quarterly Review/Winter 1993-94