The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Vol. 28, No. 3 ECONOMIC REVIEW 1992 Quarter 3 2 Comparing Central Banks’ Rulebooks by E.J. Stevens 16 Forbearance, Subordinated Debt, and the Cost of Capital for Insured Depository Institutions by William P. Osterberg and James B. Thomson An Introduction to the International Implications of U.S. Fiscal Policy 27 by Owen F. Humpage m m FEDERAL RESERVE BANK OF CLEVELAND DEE M I C REVI EW 1992 Quarter 3 Vol. 28, No. 3 Comparing Central Banks’ Rulebooks 2 by E.J. Stevens A central bank’s daylight overdraft and reserve requirement rules influence payments institutions and its own monetary policy operating practices. This article contrasts Federal Reserve rules with those of the Deutsche Bundesbank, the Bank of Japan, and the Bank of England. The fundamental lesson is that no unique set of regulations is necessary for the effective performance of a central bank’s monetary and payments system functions. How ever, adopting a different rulebook (by eliminating Federal Reserve daylight overdrafts or reserve requirements, for example) would entail some adaptation of payments institutions and monetary policy operating practices. Comparisons to the other central banks indicate what some of these adaptations might be. Forbearance, Subordinated Debt, and the Cost of Capital for Insured Depository Institutions 16 by William P. Osterberg and James B. Thomson Requiring banks to issue subordinated debt has been proposed as a way to reduce the deposit insurance subsidy and to increase market discipline. Using a modified cost of capital framework, this article develops an explicit pricing model for subordinated debt that con siders the possibility of Federal Deposit Insurance Corporation for bearances. The results reveal that forbearance alters the required rate of return on subordinated debt while increasing its value to debt holders. Moreover, the authors show that a policy of forbearance weakens the effectiveness of such debt in reducing deposit insurance premiums and as a source of market discipline. Economic Review is published quarterly by the Research Depart ment of the Federal Reserve Bank of Cleveland. Copies of the Review are available through our Public Affairs and Bank Relations Depart ment, 1-800-543-3489. Coordinating Economist: James B. Thomson Advisory Board: David Altig Erica L. Groshen William P. Osterberg Editors: Tess Ferg Robin Ratliff Design: Michael Galka Typography: Liz Hanna Opinions stated in Economic Review are those of the authors and not necessarily those of the Federal Reserve Bank of Cleveland or of the Board of Governors of the Federal Reserve System. Material may be reprinted provided that the source is credited. Please send copies of reprinted material to the editors. ISSN 0013-0281 An Introduction to the International Implications of U.S. Fiscal Policy 27 by Owen F. Humpage A commonly held belief is that aggregate U.S. fiscal policy meas ures— in particular, the federal budget deficit— are directly linked to U.S. interest rates, exchange rates, and the trade balance. Through the use of Engle—Granger cointegration tests and the development of simple two-period, two-country models, the author illustrates a complex relationship that depends on the distortionary nature of taxes and on relative differences between pub lic and private propensities to consume and to import. Fiscal policies can cause trade deficits, but this need not be the case. Comparing Central Banks’ Rulebooks by E.J. Stevens E.J. Stevens is an assistant vice president and economist at the Federal Reserve Bank of Cleveland. The author thanks Diana Dumitru for invaluable research assistance and Jeffrey C. Marquardt and David Van Hoose for useful comments. Introduction Banks’ account relationships with their Federal Reserve Banks are changing because account regulations are changing. The Board of Governors of the Federal Reserve System began a program in 1986 to limit banks’ use of daylight overdrafts, broadened the program in 1991, and beginning in 1994 will charge a fee for daylight overdrafts that exceed certain minimum amounts. The Board also reduced reserve requirements to zero on nontrans actions deposits in 1990 and cut the highest reserve requirement on transactions deposits from 12 percent to 10 percent in 1992. The purpose of this article is to examine how major changes in our central bank’s rulebook might affect Federal Reserve operations and U.S. monetary and payments institutions. To this end, I contrast Federal Reserve overdraft and reserve requirement regulations — and the insti tutional setting in which they are administered — with analogous rules and institutional settings at three of the world’s other leading central banks: the Deutsche Bundesbank, the Bank of Japan, and the Bank of England. Do the account regulations in a central bank’s mlebook matter? Central banks in industrialized countries all perfonn roughly the same functions, centered on controlling the issuance of base money and providing safe, final settlement of interbank payments. They do so, however, with apparently quite different regulations governing the accounts of their customer banks. Some cen tral banks allow daylight overdrafts while others do not; some have no reserve requirements; and some are more ready to lend than others. O f course, some central banks may perform better than others, with less inflation or more safety in their payments systems. Jiirg Niehans has observed that .. the effects and the effectiveness of central bank policy depend to a large extent on technical and institutional details that vary from one country to another and in the course of time.” (Niehans [1978], p. 263) Surely, however, major dif ferences in perfonnance have more to do with a central bank’s objectives, and with its institutional and political will to achieve them, than with its rulebook of account regulations. A central bank’s mlebook is important, none theless. In addition to any costs imposed on banks, account regulations influence the operating tech niques and involvement of the central bank in the money market. With unaltered objectives, sub stantial changes in the Federal Reserve’s mlebook w ould require associated modifications in both 3 its operating practices and the nation’s pay ments institutions. The curse of considering many questions labeled as “central banking” is the absence of an agreed-upon frame of reference within which to conduct the analysis. The grand perspective of monetary theory is too broad for this purpose; it says little about the mundane details of central bank operations. Likewise, marginal analysis of an individual bank’s decisions under a particu lar set of central bank rules is too narrow; it fails to capture systemic implications of the relevant market institutions. Comparison with other central banks is used here as a way to gain perspective on the Federal Reserve’s rulebook. With or without daylight overdrafts, and with high, low, or no reserve requirements, each central bank is able to per form similar day-to-day monetary and payments system functions. Differences in rules can be associated with differences both in market in stitutions and in the way a central bank interacts with financial markets and the banking system. The remainder of the article is divided into five sections. The first briefly reviews the unique monetary and payments system functions of any central bank. The next section compares the role of each of the four central banks considered here in financing customer banks’ clearing im balances during the course of a day. Two sections then contrast the four central banks’ techniques for maintaining policy-intended supplies of customer banks’ balances and the monetary base. These practices involve central bank operations that monetize and demonetize debt (covered in section III). Some central banks avoid lending directly to individual banks, tending instead to use open-market operations in securi ties to adjust the aggregate supply of base money and relying on markets to allocate funds among banks. Others are more willing to bypass credit markets by lending directly to individual banks when adjusting aggregate supply. The level and averaging features of reserve requirements (covered in section IV) influence the extent to which a central bank must respond to daily shocks to the aggregate supply of base money. Some rely more heavily than others on customer banks to absorb these shocks. The concluding section summarizes the international comparisons and extracts some apparent lessons about changing the Federal Reserve’s rulebook. I. Monetary and Payments Functions of a Central Bank As the monetary authority, a modern central bank controls the supply of “outside,” or base, money. This anchors the price level in the long run while allowing a central bank to respond to variations in the economy over the business cycle and to liquidity needs in the short run. As the banker for commercial and other banks operating in its country, a central bank is able to settle interbank payments because it is the unique common site of banks’ deposit accounts: A simple bookkeeping transfer from one account to another can settle payments involving any two banks. In the same way, a central bank is able to settle payments to and from its government or official foreign institutions that hold deposit accounts with it. In short, the central banks of most industrialized countries control the aggregate supply of base money while transferring owner ship of banks’ base-money balances to settle the daily clearing of payments. Monetary policy deals with the growth rate of the monetary base. Raising or lowering this rate has the immediate effect of, or is brought about by, changing the interest rate at which banks can acquire very short-run funding of their ac counts at the central bank. Ignoring completely any questions about monetary policy, the ques tion I address here is how a central bank recon ciles banks’ need for settlement with its own need to maintain a targeted level of base money. Rules about the account balances banks hold at the central bank are necessary if the central bank is to control the monetary base. Private banks have no earnings incentive to hold any substantial balance in their accounts with the central bank, because such balances typically earn no interest.1 If a bank foresees ending the day with a positive balance, it can lend that amount overnight in markets for funds with same-day payment. Moreover, banks are no dif ferent from their own customers: Absent penal ties, they have every incentive to use overdrafts as a dependable source of financing, not only during a day, but overnight. This means that in the absence of overdraft and reserve require ment rules, all banks would have an incentive to create balances at the central bank by over draft, but no incentive to hold all the balances being created. ■ 1 For a rare instance of interest-bearing reserve assets, see Dotsey (1991). Sources of Daylight Credit Enforcing a rule against overnight overdrafts allows a central bank to limit the supply of base money each day. Scarce base money is available only to those who pass a market test (by selling goods, services, or existing securities or by bor rowing). This does not, however, limit the intra day supply of base money created by temporary, “daylight” extensions of credit. Much of the daily activity of banks involves daylight credit, which must be repaid by day’s end to avoid overnight overdrafts. In modern industrialized economies, deposi tors draw checks and other payment orders on their bank accounts as the immediate quid pro quo for many market transactions, and banks use base money (or deposits at other banks) to settle interbank clearing imbalances. Even pay ments with same-day settlement can involve delays that make it possible for banks to “pay out” more money than they have on hand at the moment. They rely on daylight credit provided by those institutions that must wait for settle ment before being paid in safe base money. Clearinghouses are a common source of day light credit. Routine, standardized transactions within groups of banks, securities dealers, or members of exchanges can be covered by blan ket agreements about who can do how much business on credit from the other members of the group prior to settlement, and about how to apportion losses if one of its members is unable to settle. A central bank provides daylight credit if it makes final payments during the day for customer banks lacking sufficient balances to cover pay ments as they are made. The amount added to the supply of balances will be drained, all else equal, only when a borrowing bank repays the overdraft. Repaying Daylight Credit Repayment of daylight credit from either source should be routine even if banks hold zero bal ances at the central bank overnight. If all trans actions simply involve payments among banks, zero balances are sufficient: What some banks lose from adverse clearings during a day, other banks gain. The losers should be able to bor row or buy what they need from the gainers to cover their positions, as long as they have ac cess to markets with same-day payment. Difficulties may arise if there is an aggregate shortage of balances in the banking system as a whole. If banks normally hold no excess balances at the central bank, such a shortage will occur on any day during which banks’ balances at the central bank are drained into currency, govern ment or foreign accounts, or other miscellaneous accounts on the central bank’s balance sheet. With too few balances to go around, one or more banks will be unable to repay daylight credit ex tended by a clearinghouse or the central bank. Three mechanisms might allow banks to acquire the funds needed to repay daylight credit, despite uncontrolled factors draining bal ances. One is a central bank’s “defensive” mar ket operations. These are designed specifically to offset uncontrolled factors draining (or add ing) base money. Banks’ balances decline whenever a central bank reduces its assets or increases its other liabilities or capital, all else equal. Defensive operations are simply central bank actions taken to offset an undesired net change in all of the factors affecting the aggre gate amount of its constituent banks’ balances. The central bank supplies or drains balances in the aggregate, relying on the market to distrib ute balances to those banks that need them. Arbitrage associated with reserve require ments is a second mechanism that allows the banking system itself to absorb uncontrolled deviations of the aggregate supply of balances around a policy-intended level. A binding reserve requirement for an averaging period creates an aggregate average demand for cen tral bank balances on the part of banks. The average quantity demanded, however, can be deferred or brought forward on any day in re sponse to movements in interbank interest rates. This interday arbitrage both absorbs the “noise” from uncontrolled supply factors that are offset ting during a reserve averaging period and dampens associated variations in interest rates.2 A third mechanism is direct loans from the central bank, whereby it acts as a pure liquiditymotivated lender of last resort to the banking system. Direct lending is used primarily as an ■ 2 The power of a reserve requirement to produce noise-absorbing arbitrage has limits, however, at least in the short run. On the low side, problems can arise if payments are made by direct transfers of central bank balances, but the central bank limits the availability of daylight over drafts. Even though banks may be willing to postpone holding overnight balances, there may be too few balances to allow all banks to meet their payment needs during a day within the existing institutional environment. On the high side, some banks might inadvertently accumulate such large reserve positions early in an averaging period that they could avoid ex cess reserves for the whole period only by running overnight overdrafts, which are prohibited (Dumitru and Stevens [1991 ]). escape valve when defensive operations and reserve averaging fail to offset factors draining reserves from the nation’s banks, or when unex pected factors increase reserve demand. Banks normally are discouraged from relying on direct central bank loans, with lending rationed by administrative decision (Banks of England and Japan), by banks’ reluctance to borrow (Federal Reserve), or by a loan rate that generally ex ceeds market rates (Bundesbank). For these rea sons, individual banks typically do not plan to repay daylight credit by borrowing directly from the central bank. Instead, they try to acquire bal ances in the market — balances created deliber ately by the central bank and coaxed out of the holdings of other banks. Repayment problems might arise, however, even with no aggregate shortage of balances at the central bank. A bank may be unable to repay daylight credit because either its incipient failure or an operational problem (such as a computer breakdown) prevents borrowing from, or selling assets to, other banks with excess balances at the central bank. Central bank, commercial bank, and clearinghouse rules that prevent unlimited use of daylight credit protect against this problem. Conceptually, this source of repayment problems can be reserved for a fuller discussion of central banks’ lender-of-last-resort and bank supervision functions.3 As a practical matter, however, a central bank may have difficulty maintaining such a clean distinction between direct lending as a safety valve for aggregate shortages of reserve balances and the importun ing of either troubled banks or (in the United ■ 3 The daylight credit involved in making payments causes wellknown payments system risk problems for banks, their clearinghouses, and their central banks. In particular, a central bank needs to manage credit ris k . Daylight credit extensions on private net settlement systems can in volve systemic risk problems. If payment finality is guaranteed by a clearinghouse (a regulatory requirement advocated by the major central banks), failure of a bank to cover a negative position at settlement re quires other participants to make up the difference. Such guarantees are now explicit in the rules of the Clearing House Interbank Payments Sys tem (CHIPS) in the United States and the Gaiteme (Foreign Exchange) Yen System in Japan. An alternative structure makes payment finality contingent on suc cessful completion of the settlement process at the end of a day, as in the Towne Clearing of same-day paper check payments in London. In such instances, a bank's failure to cover a negative position precludes settle ment and implies disintegration of the day's payments, leaving their status subject to negotiation or litigation. Another possibility, explicit in the rules of some clearinghouses, is to “unwind" the settlement; that is, to exclude all payments to and from the offending bank and calculate a new settlement. However, at least in the case of large-value, same-day networks, the typical perception is that a central bank would prevent disin tegration, or an unwinding, by lending to ensure successful completion of the original settlement (Stevens [1989], Bank for International Settle ments [1990b]). http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis States) banks attempting to take advantage of a below-market discount rate. II. Daylight Credit Private clearinghouses operate in each of the four countries considered here, with same-day net set tlement on the books of the central bank. In addi tion, the Federal Reserve Banks, the Bundesbank, and the Bank of Japan operate their own on-line payment networks that enable banks to make im mediate payments throughout a day by transferring central bank balances directly to other banks.4 Central bank daylight overdrafts are pro vided only in the United States and Germany, from payments made on the Fedwire electronic network operated by the regional Federal Reserve Banks, and on the express electronic and paper transfer network operated by the Bundesbank. Both systems include thousands of participants and are dominated by largevalue payments. Flowever, the incidence of daylight overdrafts might be expected to be greater on Fedwire, where they average more than $100 billion daily, than on the Bundesbank system. This is because the value of payments relative to gross domestic product (GDP) made on Fedwire is more than five times greater than on the Bundesbank network, and the ratio of payments to balances is more than 30 times larger (Bank for International Settlements [1990a], Board of Governors [1989]). The Federal Reserve permits daylight over drafts for most banks within established limits, but will begin phasing in a fee of 25 basis points (an nual rate) in 1994? Compliance with limits is verified on an ex post basis, rather than by pre venting excess payments (as is now done, for instance, on the CHIPS large-value transfer sys tem). The Bundesbank, in contrast, apparently does not execute payments that would produce a daylight overdraft exceeding a bank’s preexisting collateral, and does not impose a fee. The Bank of Japan’s large-value same-day payments systems (Bank of Japan Cheque and Financial Network Systems) are comparable to those of the Federal Reserve Banks and the Bundesbank. However, the Bank of Japan will not execute payments that would result in a daylight overdraft. To acquire balances in time ■ 4 Descriptions of the four nations’ payment mechanisms can be found in Bank for International Settlements (1 9 8 5 ,1990a). ■ 5 A small number of banks with daylight overdrafts in excess of limits and arising from transfers of Treasury securities in the Federal Reserve book-entry system must post collateral. 6 T A B L E 1 Combined Balance Sheet of the Federal Reserve Banks3 Assets Gold and Special Drawing Rights Government securities: Outright Repurchase agreements Loans to banksb Denominated in foreign currencies All other assets Total Liabilities 11,068 241,431 18,354 190 32,633 23,901 327,577 Components of base money: Currency Banks’ balances Government balance Other deposits All other liabilities and capital accounts 267,657 38,658 8,960 6 ll 11,691 Total 327,577 a. As o f December 31, 1990. All figures are expressed in millions of dollars. b. 0.5 percent o f balances. SOURCE: Board of Governors o f the Federal Reserve System. to make payments, banks must manage their balances throughout a day, perhaps by borrow ing intraday or overnight, or by selling assets during a day for payment over the network. Whereas the Bank of Japan prohibits daylight overdrafts, the Bank of England does not provide them because it doesn’t operate any payments sys tems. Interbank payments take place entirely through private clearinghouse arrangements, not on the books of the central bank. Each day, only the net settlement position of a bank vis-à-vis all other banks in one or more clearinghouses is set tled using the bank’s account at the central bank. Even if a bank could settle its clearinghouse posi tion only with an overnight overdraft or loan from the Bank of England, the Bank has no formal responsibility to guarantee settlement. In the past, Federal Reserve provision of day light overdrafts clearly was more liberal than at the other three central banks. Until 1986, no limits were imposed and no collateral was re quired for healthy depository institutions. Pro vision began to move toward comparability in 1986, with the adoption of the potentially more restrictive current limits, based on a bank’s capi tal. With the imposition of a fee in 1994, Federal Reserve provision will become somewhat more like that of the other central banks. III. Defensive Operations The level of short-term interest rates is the effec tive policy instrument of each of the four central banks considered here. Defensive operations are deliberate actions taken to insulate the supply of base money, and thereby the level of directly affected short-term interest rates, from uncon trolled changes that are inconsistent with policy intentions. Most defensive operations take place within the daily market period in which shortestterm interest rates reflect the forces of demand and supply in the market for banks’ balances — what Niehans calls “the ultrashort-run liquidity of the banking system.” (Niehans [1978], chap ter 12) The length of the ultrashort run — from a few minutes to as much as a week — may be related to reserve requirement arrangements, which are discussed in section IV. All four central banks use one or both of two basic techniques in their defensive operations: 1) m anaging flows of banks’ balances to and from government and official foreign accounts at the central bank, and 2) using market transac tions and lending to offset unmanaged factors affecting the central bank balance sheet or inter est rates. In what follows, I discuss the use of these techniques by each of the four banks. Federal Reserve The Fed uses both techniques. Monetizing gov ernment securities through outright purchases in the secondary market or directly from foreign cus tomers is the dominant source of base money in the United States (see table 1). Fluctuations in the Treasury’s balance at the Reserve Banks, if not offset, change the supply of banks’ balances at the Fed. This is avoided, for the most part, by having the Treasury maintain two sets of deposit accounts: one with banks, to wrhich its receipts are paid, and another at the Federal Reserve D T A B L E 2 Balance Sheet of the Deutsche Bundesbank3 Assets Gold, Special Drawing Rights, and net claims on the European Monetary Cooperation Fund Securities: Outright Bills of exchangeb Other Repurchase agreements Lombard loans Liabilities Components of base money: Government balance 61,309 4,262 108,828 150,548 66,874 5,149 Other deposits 54,916 All other liabilities and capital accounts 31,083 5,187 Denominated in foreign currencies 58,308 All other assets 31,456 Total Currency Banks’ balances 39,219 308,570 Total 308,570 a. As o f December 31, 1989- All figures are expressed in millions o f marks. b. 92 percent o f balances. SOURCE: Deutsche Bundesbank. Banks, from which its payments are made. The Treasury can transfer funds from the receiving accounts to the paying accounts each morning to offset the day’s projected payments. This practice leaves a relatively constant projected target balance in the paying accounts, prevent ing Treasury operations from adding or draining banks’ balances.6 Defensive operations are used to offset shortrun variations in the public’s demand for curren cy and in banks’ demand for required balances, as well as in a host of miscellaneous items. The vehicle for temporary defensive operations is re purchase agreements (RPs) in the secondary mar ket for Treasury securities — that is, purchases (to add balances) with an agreement to resell, or sales (to drain balances) with an agreement to buy back, one or a few days later. Transactions are conducted by inviting bids from designated (pri mary) dealers and by accepting enough bids to fill the projected need on a best-bid basis. These fre quent, temporary adjustments can be used to finetune the supply of balances on a daily basis. When needed, transactions take place at about 11:30 a.m., based on projections of demand and of ■ 6 Banks’ balances at the Federal Reserve Banks could be completely insulated from the effects of Treasury operations (within projection er rors), were it not for occasional episodes when 1) paying accounts must move above the normal target because receipts exceed banks’ limited w ill ingness to hold Treasury deposits, or 2) receiving accounts are exhausted and the paying accounts must be drawn down below the normal target be http://fraser.stlouisfed.org/ cause the Treasury is constrained from issuing new debt. Federal Reserve Bank of St. Louis factors affecting supply. The banking system must accommodate any deviation of actual from projected balances for the day, although as noted above, a substantial shortfall could force banks to borrow at the discount window. Defensive operations are not always based on projected quantities. The Fed’s proximate mon etary policy target is perceived as a level of the overnight interbank (federal funds) rate. Devia tions of the funds rate from target can indicate projection errors or market expectations that are inconsistent with policy. Operations in the second ary market, therefore, may be intended to defend or to correct the market’s perception of the interest-rate policy target (Meulendyke [19891 ). Bundesbank In contrast to the Federal Reserve, the Bundes bank does not rely on outright purchases of government securities as its dominant means of supplying base money (see table 2). A large por tion of base money is supplied (within estab lished “refinancing” quotas) through purchases of domestic and foreign bills of exchange with maturities of several months. Banks sell these instruments to the central bank at the official discount rate, which is typically below market rates. An even larger source of base money orig inates from the continuous rollover of RPs of one- and two-month maturities. TABLE 3 Balance Sheet of the Bank of Japan3 Assets Gold Liabilities 140 Securities: Government bonds Bills and commercial paper Bills discounted 31,542 6,906 144 Loans5 6,160 Denominated in foreign currencies 2,996 All other assets 1,269 Total 49,157 Components of base money: Currency Banks’ balances Government balance Other deposits All other liabilities and capital accounts Total 39,798 4,881 521 424 3,533 49,157 a. As o f December 31, 1990. All figures are expressed in billions of yen. b. 126 percent of balances. SOURCE: Bank o f Japan. The Bundesbank adjusts the aggregate sup ply of banks’ balances weekly, typically by regu lating the volume of RPs accepted. Other means of adjustment include shifting federal govern ment deposits to banks, foreign exchange swaps or RPs, and sales of special short-maturity Treasury bills. But, for the most part, any re maining need for short-run adjustments must come at the initiative of the banks themselves, by varying their Lombard borrowing from the Bundesbank (collateralized by eligible securi ties) at the Bank’s Lombard rate. This rate always is higher than the discount rate and typically is higher than market rates (Deutsche Bundesbank [1985, 1990]). The Bundesbank also has an opportunity to indicate when the overnight interbank rate has been affected by either projection errors (under supply, for example, would be expected to drive the rate up toward the Lombard rate) or a market perception of rates inconsistent with actual policy intentions. Both the cutoff rate in accepting RPs and the volume accepted can provide short-run signals to the market. A more direct signal can be given by inviting tenders for RPs at a designated interest rate, rather than by simply accepting the best rates offered for a desired quantity. Bank of Japan Like the Federal Reserve, the Bank of Japan holds a large portfolio of government securities whose outright purchase is the dominant source http://fraser.stlouisfed.org/ of base money (see table 3). The Bank also can Federal Reserve Bank of St. Louis operate in a variety of other markets to adjust the monetary base and to influence conditions in specific markets. These actions include en gaging in Treasury bill and commercial paper RPs, purchases and sales of commercial bills (including sales of Bank of Japan bills), and sales of government bills with an RP. In addi tion, a pivotal group of large banks is continu ously indebted to the central bank, within established lines of credit, at the basic discount rate, which is typically below interbank lending rates (Tatewaki [1991]). The Bank of Japan has two daily oppor tunities to adjust the supply of balances. One is through operations in the market (typified by commercial paper RPs) aimed at the market rate on uncollateralized interbank call loans — the counterpart to the federal funds rate in the United States. The second is by a later daily decision about the quantity of loans the Bank will extend or collect. This lending decision is made shortly before 3:00 p.m., when same-day transactions in the call loan market must end (because the Bank’s same-day payments net work closes), but about an hour after same-day net positions on the Gaiteme foreign-exchange net settlement network have been calculated. Thus, Bank of Japan lending decisions can ac commodate a need for balances, or put upward or downward pressure on the call loan market, based on information accumulated during the day. The Bank assists the market in distinguish ing defensive operations from those with policy implications by releasing data, also at 3:00 p.m., showing demand and supply of funds and its 9 T A B L E 4 Balance Sheet of the Bank of England3 Assets Issue Department Securities: Government Other Total Liabilities Currency 10,021 5,009 Other 15,030 Total 15,021 9 15,030 Banking Department Securities: Government Bills discounted Loans 1 Banks’ operating balances 175 843 Government balance 454 540 Other deposits: Cash ratio Other 651 All other assets 1,302 All other liabilities and capital accounts Total 4,336 Total 1,491 1,288 928 4,336 a. As of February 28, 1990. All figures are expressed in millions of pounds. SOURCE: Bank of England. own market operations for that day, as well as an estimate for the next day (Nakao and Horii [199H, Bank for International Settlements [1986, 1990a], and Bank of Japan [1991]). Bank of England The Bank of England maintains an accounting distinction between two departments. The Issue Department supplies currency, which finances outright holdings of government and other securities. The Banking Department supplies the small amount of banks’ deposit balances (there are no reserve requirements), largely by dis counting (purchasing) eligible securities and through collateralized lending (see table 4). Weekly government bill tenders normally drain enough funds from the banking system to the government’s account to create a persistent shortage of balances, requiring daily defensive operations to add balances back into the system. Procedures for defensive operations are elaborate, because banks’ small cushion of desired “target” balances provides little room for error in draining or adding balances each day. Banks report their targets to the Bank of England, and at three times during the day, the Bank reports its estimate of the day’s shortage or surplus of balances relative to the aggregate of these targets. Open-market operations typi cally are carried out with the discount houses, which in turn provide banks with daily financ ing facilities. Operations might be conducted after publica tion of the first estimate of the day’s balance position, if the need is large. More often, the Bank operates after releasing its noon update of estimated need. A third round of operations may come after the Bank’s 2:00 p.m. update. A further opportunity to adjust comes through “late assistance” in the form of secured lending to discount houses and other money brokers, which may extend later into the afternoon. Operations at any of these times can do more than simply adjust the quantity of balances. The Bank has discretion over the type of operation (outright, RP, lending), whether it invites transac tions or responds to requests, and the terms on which it will engage in transactions (type of secur ity, maturity, and “stop rate”). Manipulation of these variables, in conjunction with the Bank’s published estimates of the day’s position, provides an oppor tunity for the Bank to clarify its policy intentions while engaging in defensive operations (Bank for International Settlements [1986,1990a], Bank of England [1988]). Summary All four central banks engage in defensive oper ations along the twin dimensions of quantity of balances and level of interest rates. Where con trol of the quantity of balances is not effective or, for some other reason, market expectations are not consistent with policy intentions, the central banks can manipulate the types and terms of their market operations to provide sig nals — interpreted on the basis of market tradi tions — about the level of interest rates thought to be consistent with policy intentions. No amount of such suasion can be effective, how ever, if not supported by control of the quantity of balances. Clear differences are visible in the degree to which any of the four central banks might be ex pected to seek precise control of the daily aggre gate supply of balances and relevant interest rates using defensive operations. The Bundes bank’s reliance on weekly RPs leaves the daily supply of balances subject to uncontrolled fac tors that might move interest rates within the ceiling provided by the Lombard rate. Federal Reserve reliance on morning open-market operations, guided only by projections, means that the actual daily supply of balances is sub ject to projection errors, although daily signals may be sufficient to maintain clarity about the level of interest rates consistent with policy in tentions. The Bank of Japan, by making decisions about lending and repayment late in the day — after one clearinghouse has closed and immediately before the close of another — is in a better position to avoid projection errors in its daily defensive operations. The Bank of England, relying on successive estimates, opera tions, and late assistance over the course of a single day, can minimize projection errors by using repeated updates of market information to estimate the need to adjust the aggregate supply of balances. IV. Reserve Requirements A banking system is in a better position to absorb day-to-day uncontrolled variations in the supply of balances when banks must meet reserve re quirements. The central bank must eliminate any net excess or deficiency of balances by the end of the reserve averaging period, but not every day. A bank calculates its required reserves by matching various reservable deposits with their respective reserve ratios. Specifications of both reservable deposits and reserve ratios differ in widely inventive ways among the four central banks. These computational features influence the net after-reserves marginal cost of bank lending financed by various types of deposits. They also might be germane to monetary policy operations. For example, predictability of de mand for reservable balances and the accuracy of projections underlying defensive operations are affected by shifts among deposit accounts having different reserve ratios.8 However, these features will not be considered here because they are not of foremost importance to the inter action of central banks’ reserve requirement rules with their monetary and payments system operations in the “ultrashort run.” Rather, of in terest here are 1) the average quantity of noninterest-bearing reserve balances that banks must hold and 2) the length of the averaging period over which banks can spread this artificial demand and over which the central bank can spread its supply. Three of the four central banks had reserve requirement regulations in 1990. The aggregate quantity of required balances in each country can be compared directly only by choosing ex change rates at which to convert to a common currency. Examining the ratio of required bal ances to a country’s GDP avoids this complica tion, while making a rough adjustment for differences in the scale of national economies.9 Both methods of comparison are shown in table 5, with required and excess balances converted to U.S. dollars, as well as scaled by each coun try’s nominal GDP. ■ ■ 7 "Large” projection errors occurred on 27 days in 1991, according to the Federal Reserve Bank of New York, but “large” is undefined. The New York Bank conducts a weekly Thursday press briefing that reviews in general terms the factors affecting banks’ reserve balances during the week ending the previous day. Among other items, the briefing indicates either 1) that there were no large net one-day deviations from projections, or 2) the days on which there were large deviations, giving their sign and source but not their dollar values. 8 A convenient comparison of the basis for computing required reserves in the four countries can be found in Kneeshaw and Van den Bergh (1989). The irrelevance of methods of computation for monetary policy implementation is discussed in Stevens (1991). ■ 9 An alternative scale adjustment is to take the ratio of required balances to total deposits (whether subject to requirements or not) of all institutions that are subject to requirements. The rank order is the same as for GDP. D T A B L E 5 Banks’ Deposit Balances at Central Banks3 Excess Required Millions of U .S . dollars Percentage of GDP Millions of U .S . dollars Percentage of GDP Days in Averaging Period Federal Reserve*7 33,843° 0.6lc 933 0.017 I4d Discount rate +2% Bundesbank! 29,782 2.52 189 0.016 30 Lombard rate +3% Bank of Japan 33,410 1.14 28 0.001 30 Discount rate +3-75% n.a. n.a. 232 0.024 1 Bank of England8 Penalty for Deficiency n.a. a. 1990 annual averages. Currency conversions are at the annual average exchange rate. b. Reserve requirements were cut substantially in December 1990 and April 1992. The average dollar am ount o f required plus clearing bal ances declined 25 percent between May 1990 and May 1992. c. Includes (after rounding) 0.59 percent of required balances and 0.03 percent o f clearing balances. d. Ninety-one days for small banks. e. 1989 values are used to avoid discontinuity caused by reunification. f. Holdings of vault cash cannot be deducted from required reserves in calculating required balances. g. Excludes “cash ratio” deposits. SOURCES: Bank of England, Board of Governors of the Federal Reserve System, Deutsche Bundesbank, Bank of Japan, and the International Monetary Fund. Federal Reserve In the United States, large banks must meet reserve requirements within a basic 14-day averaging period. Each bank can rely on daily market transactions to manage balances, aided by a provision for carryover of excesses or deficiencies into the next 14-day period that creates a limited 28-day averaging period.10 A bank’s holdings of vault cash, as well as its deposit balance at a regional Reserve Bank, are eligible to satisfy the legal reserve requirement. Even some of the largest institutions satisfy their entire requirement with vault cash. In addition to a legal reserve requirement, many banks contract to hold required clearing balances during a reserve maintenance period. These required balances are administered on the same basis as the legal requirement, but yield earnings credits at the level of the federal funds rate with which to pay for Reserve Banks’ ■ 10 The Federal Reserve appears to be unique in allowing this addi tional averaging between adjacent periods (see Kneeshaw and Van den Bergh [1989]). A deficiency or excess of up to 4 percent of required re serves (increased from 2 percent in 1992) can be carried into the next averaging period (but not beyond). Because many banks satisfy a large portion of their reserve requirement with vault cash, eligible carryover can be much larger than 4 percent of required balances. priced services. Failure to maintain at least the required amount of vault cash plus legal and clearing balances, after carryover, results in a fee levied on the deficiency at a rate of 2 per centage points above the discount rate. This charge, in addition to administrative oppro brium, makes deficiencies rare. Bundesbank Required balances in Germany are of an order of magnitude roughly comparable in dollar value to the aggregate quantity held by U.S. and Japanese banks, but are substantially higher rel ative to GDP. In addition, a long, 30-day averag ing period provides the German banking system with a substantial ability to absorb offsetting variations in the daily supply of balances. All in stitutions subject to reserve requirements must maintain a required deposit balance. Vault cash is eligible to meet requirements, but only up to 50 percent of the amount of a bank’s required reserve. Failure to satisfy the reserve require ment results in a penalty at a rate 3 percentage points above the Lombard rate (which itself is typically higher than market rates). Bank of Japan Required reserves in Japan, while of the same order of dollar magnitude as those in Germany and the United States, stand in an intermediate position when measured relative to GDP— al most twice the U.S. ratio, but only half the Ger man ratio. Like German banks, Japanese banks maintain required reserves within a 30-day averaging period, providing the banking system with a significant ability to absorb offsetting daily variations in the supply of balances. All re quired reserves must be held as balances with the Bank of Japan: Vault cash holdings do not satisfy reserve requirements. The penalty for a reserve deficiency is a rate 3.75 percentage points above the official discount rate. Bank of England The United Kingdom is unique among the four countries in having no reserve requirement regulation.11 Banks do target, and hold, selfdetermined levels of operating balances as a buffer against lower-than-expected clearing house net positions at the end of a day. In the aggregate, however, this practice has almost none of the shock-absorbing function asso ciated with a reserve requirement: It is impos sible for banks to accommodate daily variations in the aggregate supply of balances by postpon ing or accelerating the accumulation of bal ances. Extra balances today are worthless on future days, while an unexpected shortage today can be no greater than target balances. And target balances are quite small — about one one-hundredth of required balances in Ger many, and normally smaller than the size of daily defensive operations conducted by the Bank of England. National Differences in Required Balances There is no obvious rationale for the observed national differences in the level of balances a banking system is required to maintain on ■ 11 Institutions with more than minimum amounts of eligible liabilities must hold nonoperational, non-interest-bearing “cash ratio” deposits, fixed for six-month intervals at about one-half of 1 percent of both demand and term deposits (without averaging) to finance the Bank ing Department of the Bank of England. This arrangement is more nearly analogous to the Fed’s requirement that member banks hold stock in a Federal Reserve Bank than to a reserve requirement. deposit at its central bank. One striking associa tion can be detected: Less frequent defensive operations tend to be related to higher require ments that allow the banking system itself to ab sorb daily, offsetting variations in the supply of balances. The Bundesbank, with the highest level of required balances, tends to rely on weekly operations; the Federal Reserve, with an intermediate level of required balances, tends to rely on daily operations; the Bank of Eng land, with no required balances, may take action as frequently as four times a day. Association is not explanation, however. Are reserve requirements lower because a central bank is more assiduous in controlling the supply of balances, or does a central bank control the supply of balances more assiduously because reserve requirements are lower? Moreover, the association is not consistent across the four cen tral banks: The relatively high level of require ments in Japan would seem to allow the Bank of Japan to be less attentive than it is in conduct ing defensive operations. Other factors that might account for differences do not explain much, either. Longer averaging periods could be a substitute for higher require ments, but that is not the pattern actually observed (see table 5).12 Provision of daylight overdrafts could likewise substitute for higher balances, but no such pattern is evident. For example, while the Bank of Japan prohibits daylight overdrafts, the level of required balances relative to GDP is only half that of the Bundesbank, which does allow such transactions. Perhaps another factor is at work. A central bank might offset the “tax” of a relatively high reserve requirement with the “subsidy” of loans to banks at below-market rates. A perfect offset would leave the marginal cost of lending unaf fected by reserve requirements, but none of the four central banks operates this way. More likely is a partial offset to the total cost of operating within all the rules of the central bank. The na tional basis for this offset is indicated in the foot notes to tables 1-3, as measured by total central bank assets acquired from the banking system at subsidy rates, divided by banks’ required balances held with the central bank. The ratio of subsidized assets to required balances varies from about zero at the Federal Reserve Banks to a high of 126 percent at the Bank of Japan. These values provide some evidence that the cost of required balances may ■ 12 Extra days could replace extra balances in deferring and accel erating the accumulation of required balances while accommodating a given pattern of variations in supply; supply variations might be more likely to be offsetting over longer averaging periods. ES not be as unequal as their levels, but that the off set from the subsidy cannot equalize cost. For example, a simple calculation suggests that the Bundesbank would have to maintain a negative discount rate in refinancing bills of exchange if it were to offset the difference between the GDP-based measure of its required balances and that of the Federal Reserve.13 Moreover, even a plausible association between required balances and subsidized assets would not ex plain why nations might choose these different institutional patterns. Just as it is impossible to explain why reserve requirements differ, so, too, it is hard to explain variations in the average level of excess balances among the four banking systems (see table 5). One interpretation of excess balances might be that they measure the accuracy of a central bank’s defensive operations. If banks have no incentive to hold non-interest-bearing idle bal ances, defensive operations must aim at zero excess balances to prevent extreme volatility in interbank interest rates. Information about excess balances alone is not sufficient to justify this interpretation, how ever. Even if a central bank were able to achieve zero excess balances, normal practice would be to target a positive level, demanded by banks in the aggregate. Individual banks have an incen tive to target small excess balances on the last day of a reserve averaging period. This reflects the monetary and nonmonetary penalties for failing to meet requirements, coupled with each bank’s inevitable uncertainty about both the in cidence of unplanned, last-minute transactions and the accuracy of its record keeping. Observed excess balances may reflect actual demand, not inaccurate supply. A bank operating in the context of a positive requirement normally will have a cushion of balances so that it can operate closer to a zero target for excess balances than a bank with no balance requirement. With a requirement, the cushion is lacking only on the last day of a reserve maintenance period, when the bank can no longer postpone or accelerate the accu mulation of required balances; without a re quirement, the cushion is lacking every day. This may explain why excess balances are highest at the Bank of England — assuming, of 13 Let R equal the market rate forgone on reserve balances, and s equal the subsidy to that rate for central bank loans. In the case of the ■ Bundesbank, for example, from the values in tables 2 and 5, 0.92s =2.52/?-0.61/?. That is, s = 1.92/?. The level of the subsidy would be almost twice the level of the market rate, implying a negative loan rate at the central bank. http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis course, that actual balances are an indication of banks’ desired balances rather than of errors in defensive operations. Lower excess balances at the Bundesbank may reflect another difference: German banks may be willing to set targets closer to zero ex cess balances because they can rely on Lombard borrowing to round out their reserve position at the last moment on the last day of an averaging period, albeit at an above-market price (and not repeatedly). The Bank of England and the Fed eral Reserve Banks do not maintain lending facilities as hospitable to last-minute borrowing by individual banks. Bank of Japan lending might account for the minuscule level of excess balances in that country, either as a means of achieving precision in supplying balances or, given such precision, as a reflection of low de mand on the part of individual banks in antici pation of precise supply. V. Conclusion The central banks of the United States, Germany, Japan, and the United Kingdom perform the same basic functions. In the payments system, they pro vide safe, base money both as currency and as banks’ deposit balances, as well as a facility for set tling clearinghouse payments through bookkeep ing transfers of banks’ balances. In addition, the Federal Reserve, Bundesbank, and Bank of Japan all provide a system for making same-day pay ments by direct transfers of banks’ balances. Each attempts to provide the quantity of base money required to maintain short-term interest rates at policy-desired levels. To facilitate payments, some central banks (the Banks of Japan and England) rely entirely on clearinghouse organizations to supplement the supply of base money with daylight credit. Others (the Federal Reserve and Bundesbank) supple ment the supply of base money themselves during the day by providing daylight overdrafts. All four central banks engage in defensive operations designed to insulate the overnight supply of banks’ balances and the level of short term interest rates from the temporary effects of variations in currency holdings, government balances, and other uncontrolled factors. The four differ, however, in the extent to which reserve requirements enable the banking sys tem itself to accommodate day-to-day shocks to the supply of banks’ balances arising from these factors. The contrast is most apparent between the Bank of England, with no reserve require ments and multiple defensive operations each KB T A B L E 6 Summary Comparison of Central Bank Rules and Operations, 1990 Federal Reserve Bundesbank Bank of Japan Bank of England Yes Yes Yes Yes Yes, within line of credit, monitored ex post (fee begins in 1994) Yes, within limit of Lombard collateral; no fee No No Overnight overdraft Penalty Lombard loan Prevented Discretionary Central bank defensive operations Daily if needed; in morning, from projections of demand and supply Weekly or more frequently Twice daily if needed, before and after close of clearinghouse Four times daily if needed, before and after close of clearinghouse Sources of Daylight Credit Private clearinghouses Central bank Sources of Overnight Balances Reserve Requirement Level High, but falling Highest Higher None Averaging period 14-day, with limited 28-day 30-day 30-day None SOURCES: See references in text. day, and the Bundesbank, with high reserve re quirements and major reliance on weekly defen sive operations. The fundamental lesson of this study is that there is no unique set of rules a central bank must impose on the banking system (see table 6). Monetary and payments system functions can be carried out under a variety of rules and regulations whose relative costs would be enor mously difficult to establish. Applying this lesson to the Federal Reserve helps to clarify some recent issues. A common apprehension about limiting banks’ daylight overdrafts has been the possibility of payments system “gridlock,” which some fear would re quire banks either to hold costly excess balances at the Federal Reserve Banks or to develop a finely tuned system for trading and transferring balances on an hourly or partial-day basis. Ex perience in nations whose central banks do not provide daylight credit suggests another likely alternative: Banks will rely more extensively on private clearinghouses in making payments. Lowering, or even eliminating, reserve re quirements has considerable appeal in the United States, where their apparent burden on banks’ domestic and global competitiveness seems unrelated to their statutory monetary policy rationale. Deregulating the banking sys tem by removing reserve requirements, how ever, would have the seemingly paradoxical effect of increasing, rather than decreasing, the pivotal role of the central bank in the money market. As in the case of the Bank of England, assiduous defensive market intervention could be necessary each day simply to match the daily supply of banks’ balances with any residual precautionary demand. Alternatively, copying the Bundesbank’s Lombard facility, the Federal Reserve Banks’ discount w indow lending could play a larger defensive role if administrative and market discouragement of borrowing were abandoned in favor of a penalty discount rate. □ References Bank for International Settlements. Paym ent Sys tems in Eleven Developed Countries. Basle: BIS, 1985. ______ . Changes in Money-Market Instruments a n d Procedures: Objectives a n d Im plications. Basle: BIS, March 1986. ______ . Large-Value Funds Transfer Systems in the Group o f Ten Countries. Basle: BIS, 1990a. ______ . Report o f the Committee on Interbank Netting Schemes o f the Central Banks o f the Group o f Ten Countries. Basle: BIS, Novem ber 1990b. Bank of England. “Bank of England Operations in the Sterling Money Market,” Quarterly B ul letin, vol. 28, no. 3 (August 1988), pp. 391-409. Kneeshaw, J.T., and P. Van den Bergh. “Changes in Central Bank Money Market Operating Procedures in the 1980s,” Bank for Interna tional Settlements, Economic Paper No. 23, January 1989. Meulendyke, Ann-Marie. U.S. Monetary Policy a n d F in an cial Markets. Federal Reserve Bank of New York, 1989. Nakao, Masaaki, and Akinari Horii. “The Process of Decision-Making and Implementation of Monetary Policy in Japan,” Bank of Japan, Special Paper No. 198, March 1991. Niehans, Jiirg. The Theory o f Money. Baltimore: Johns Hopkins University Press, 1978. Stevens, E.J. “Removing the Hazard of Fedwire Daylight Overdrafts,” Federal Reserve Bank of Cleveland, Economic Review, vol. 25, no. 2 (1989 Quarter 2), pp. 2-10. Bank of Japan. A n n u a l Review. Tokyo: Bank of Japan, 1991. ______ . “Is There Any Rationale for Reserve Re quirements?” Federal Reserve Bank of Cleveland, Economic Review, vol. 27, no. 3 (1991 Quarter 3), pp. 2-17. Board of Governors of the Federal Reserve Sys tem. A n n u a l Report 1988. Washington, D.C.: Board of Governors, 1989. Tatewaki, Kazuo. B anking a n d Finance in Japan: A n Introduction to the Tokyo Market. New York: Routledge, 1991. Deutsche Bundesbank. “Recent Developments with Respect to the Bundesbank’s Securities Repurchase Agreements,” M onthly Report o f the Deutsche Bundesbank, vol. 37, no. 10 (October 1985), pp. 18-24. ______ . Report o f the Deutsche Bundesbank fo r the Year 1990. Frankfurt-am-Main: Deutsche Bundesbank, 1990, pp. 101-02. Dotsey, Michael. “Monetary Policy and Operat ing Procedures in New Zealand,” Federal Reserve Bank of Richmond, Economic Review, vol. 77, no. 5 (September/October 1991), pp. 13-19. Dumitru, Diana, and E.J. Stevens. “Federal Funds Rate Volatility,” Federal Reserve Bank of Cleveland, Economic Commentary, August 15, 1991. 16 Forbearance, Subordinated Debt, and the Cost of Capital for Insured Depository Institutions by W illiam P. Osterberg and James B. Thomson Introduction Among the proposals intended to prevent the commercial banking industry from suffering a fate similar to that of the nation’s savings and loans (S&Ls) is the requirement that banks issue subordinated debt. The claims of the holders of such debt are subordinate to the claims of the Federal Deposit Insurance Corporation (FDIC), which reduces the agency’s exposure to loss. Furthermore, the rates paid on subordinated debt theoretically reflect a bank's riskiness; thus, a subordinated debt requirement penalizes rela tively risky institutions by imposing market dis cipline. However, as is the case with competing regulatory proposals, the efficacy of a subordi nated debt requirement is directly affected by regulators’ adherence to stated guidelines. In this article, we emphasize that a subordi nated debt requirement interacts with other reg ulatory forces such as deposit insurance. The role of subordinated debt will also change when the risk-based capital system for U.S. banks be comes effective in December. Under the old sys tem of capital regulation, primary capital had to be at least 5.5 percent of on-balance-sheet assets and total capital had to be at least 6 percent of http://fraser.stlouisfed.org/ assets, with subordinated debt included in total Federal Reserve Bank of St. Louis W illiam P. Osterberg is an economist and James B. Thomson is an assistant vice president and economist at the Federal Reserve Bank of Cleveland. The authors are grateful to David Altig, Robert Avery, and Edward Kane for help ful comments and suggestions. capital but not in primary capital. Under the new system, subordinated debt is included in Tier 2 capital, and the total of Tier 1 and Tier 2 capital must be at least 8 percent of risk-weighted assets. Although the impact of subordinated debt will be affected by the process of risk-weighting, such debt is a relatively small component of total capital, amounting to only 10 percent of equity capital (the largest component of total capital) for FDIC-insured commercial banks in 1992:IQ (see FDIC [1992]). As background for understanding the issues surrounding a subordinated debt requirement, it is worth considering recent experience in the S&L in dustry. Several of the same factors that contributed to losses incurred in the bailout may also be behind the current deficit in the FDIC’s deposit in surance fund. These include fraud and misman agement, outdated regulations, and regulatory laxity. In addition, mispriced deposit insurance has provided incentives for S&L managers to maintain relatively risky portfolios. With fixed-rate deposit insurance, the riskiness of an institution’s portfolio does not impact the rate it must pay for deposits. Regulatory capital forbearance, which occurs when regulators supplement bank capital rather than adhering to stated guidelines, may have increased the incentives for insolvent S&Ls to D take on more portfolio risk in an attempt to regain solvency. In fact, these incentives can be come so perverse that speculative investments with little chance of paying off may be under written by insured institutions. The failure of deposit insurance premiums to correctly reflect risk and, to a lesser extent, regulatory forbear ance are unfortunately also present in the com mercial banking industry.1 Proposals to reform the current system of bank regulation can be described in terms of their reli ance on market mechanisms. At one extreme are calls to replace government deposit insurance with a private, market-based system. More widely discussed is the proposal to implement a system of risk-based government deposit insurance in which an individual bank’s premium would vary with the composition of its portfolio. The feasibil ity of this approach has been studied by Flannery (1991), Merton (1977,1978), Ronn and Verma (1986), and Pennacchi (1987b).2 An analogous system is the risk-based capital standard, which would reduce the subsidies to risk-taking embed ded in the current system. Some proposals are intended to lessen the exposure of the insurer. These include limiting coverage (by restricting coverage to one account per individual or by reducing the total dollar ■ 1 Many studies have analyzed the risk-taking incentives embedded in the current deposit insurance system (see Kane [1985,1989a, 1989b]). It deposit insurance were “fairly" priced, as discussed by Thomson (1987b), then the premium would set the value of the insurer’s claim to zero and would not distort the market incentives for risk-taking. It is not clear, on average, whether deposit insurance is fairly priced (see Pennacchi [1987b]). However, since all banks pay the same premium per dollar of deposits, rela tively risky banks are obviously being subsidized by relatively safe ones. Analyzing the impact of deposit insurance is also complicated by the presence of regulations. In fact, Buser, Chen, and Kane (1981) present a ra tionale for combining underpriced deposit insurance with capital regulation. ■ 2 The FDIC Improvement Act of 1991, which mandated that the agency do a sim ilar study, is to some degree the driving force behind its recent announcement of a risk-sensitive deposit insurance schedule. While this proposed premium schedule is a step in the right direction, it w ill only marginally alter the degree of mispricing and hence w ill have lit tle effect on adverse incentives. For a critical evaluation of the FDIC’s plan, see the statement of the Shadow Financial Regulatory Committee (1992). ■ 3 One alternative proposal is to institute depositor preference laws. Without such laws, uninsured deposits, insured deposits, nondeposit claims, and the claims of the insurer have equal priority in the event of bankruptcy. With such laws, depository claims, which are inherited by the insurer, have priority over nondeposit claims. Hirschhorn and Zervos (1990) analyze these laws empirically and note that their effectiveness can be seriously diluted if they lead to an increase in the amount of col lateralized claims. Another alternative is to require stockholders to post surety bonds, which would be used to offset creditors’ losses if a bank failed (see Kane [1987] and Osterberg and Thomson [1991]). This would effectively reestablish the double call provision that existed prior to the Banking Act of 1935. http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis amount insured) or changing banks’ capital structure through, among other techniques, a subordinated debt requirement.3 The maturity of subordinated debt generally exceeds that of uninsured deposits, so holders of such debt are less likely to “run.” Consequently, as we point out later in this paper, forbearance is more likely to be extended to uninsured depositors than to subordi nated debt holders, who receive principal and interest payments only after the claims of senior creditors are satisfied. Since subordinated debt claims are junior to those of the FDIC, the agency’s exposure would be reduced. In addition, by increasing the risk exposure of claimants subordinate to the FDIC, this pro posal would utilize market incentives; that is, rates on subordinated debt would presumably reflect a bank’s riskiness. Baer (1985), Benston et al. (1986, chapter 7), and Wall (1989) favor such an approach. Osterberg and Thomson (1991) analyze the theoretical impact of a subor dinated debt requirement on both the cost of capital and the value of deposit insurance. Un fortunately, the empirical evidence on using subordinated debt to enhance market discipline is mixed (see box 1). This article provides a theoretical analysis of the extent to which subordinated debt prices apply market discipline to banks. In theory, the required rate of return will vary positively with the bank’s riskiness, reducing the subsidy provided by deposit insurance and ensuring that the bank’s investment decisions will take risk into account. In addition, regulators could utilize the information contained in the secondary market prices of subor dinated debt. As is the case with other proposals that rely on market discipline, however, the effec tiveness of such an approach will depend on whether the government implicitly insures the claims of subordinated debt holders or other tech nically uninsured claims. Several studies (Allen and Saunders [1990], Osterberg and Thomson [1990], and Thomson [1987a, 1987b]) show how forbearance influences the values of deposit insur ance and insured institutions, as well as the rate of return on uninsured deposits. In this paper, we analyze the impact of forbear ance on the values of and required rates of return on subordinated debt, uninsured deposits, and deposit insurance. Our results are consistent with those of Gorton and Santomero (1990) in that we find ranges over which subordinated debt acts like either debt or equity. We also find a nonlinear relationship between asset risk and the rate of return required on subordinated debt. The manner in which we incorporate forbearance into our analysis is similar to techniques used by Allen 18 B O X I Empirical Evidence on Market Discipline In general, evidence regarding the extent to which mar ket prices reflect risk is mixed (see Gilbert [1990]). Except for Randall (1989), studies of bank equity prices show that they indeed reflect portfolio risk. Valid criticisms of Randall’s work can be found in Gilbert’s summary of this literature. Studies of rates paid on certificates of deposit and on subordinated debt are more ambiguous. The two most relevant studies for our purposes are those of Avery, Bel ton, and Goldberg (1988) and Gorton and Santomero (1990). Both papers examine the empirical relationship between risk premia on bank subordinated debt and balance-sheet measures of bank risk. Each finds weak evidence that market risk premia on subordinated debt are related to risk proxies constructed from accounting data in the current regulatory environment. These results contrast with those of earlier studies by Baer and Brewer (1986) and Hannan and Hanweck (1988), who find a sig nificant relationship between risk premia and risk prox ies in deposit markets. Gorton and Santomero develop an explicit pricing model for subordinated debt showing that sometimes it acts like equity and other times like debt. Specifically, when the bank’s asset value is expected to be above (below) the value of claims against it, subordinated debt acts like debt (equity). Also crucial in the analysis are assumptions about the overall regulatory environment. Many studies (see Mar cus [1984] and Pennacchi [1987a]) have emphasized the role that assumptions about closure policies play in analyz ing deposit insurance. Gilbert (1990) points out that the banks analyzed by Avery, Belton, and Goldberg were mainly large firms whose subordinated debt holders were likely to have been insured de facto. This again highlights the important role that de facto regulation plays in interpreting the informativeness of market prices and rates of return? a. The test for market discipline in Gorton and Santomero and in Avery, Belton, and Goldberg simultaneously examines the assumptions regard ing model specification, closure rules, and the accuracy o f accounting ratios as measures of risk. In addition, the results may be sensitive to the particular sample period used. Gorton and Santomero’s findings suggest that the weak relationship between the subordinated-debt risk premium and risk proxies constructed from accounting data in Avery, Belton, and Goldberg is not due to either m odel specification or closure rules. However, since the sample period encompasses the failure of Continen tal Illinois Bank, where the FDIC fully protected the subordinated debt holders of the parent holding company, it is not clear that these studies’ results generalize to other sample periods. and Saunders (1990) and others (see box 1). Our findings, which point out the need to specify carefully and correctly the regulatory en vironment in place when market performance is measured, are broadly consistent with those of Gilbert (1990). The model is presented in section I. Section II reports the results of an earlier, single-period analysis of a bank with uninsured deposits, in sured deposits, and subordinated debt (see Osterberg and Thomson [1991]). We show that subordinated debt affects the value of the in sured bank only through its impact on the size of the deposit insurance subsidy, and that the fair value of deposit insurance is a function of the subordinated debt requirement. In section III, we extend the analysis to include the possi bility of FDIC bailouts of uninsured liability holders. Section IV then investigates the effects of mispriced deposit insurance and FDIC for bearances on the values of subordinated debt capital and deposit insurance. We find that the usefulness of subordinated debt as an equity like buffer is reduced by FDIC forbearance pol icy and that investors’ required rate of return on subordinated debt is inversely related to forbear ance. Conclusions and policy implications are presented in section V. I. The Model To determine the effects of subordinated debt and surety bonds on the cost of banks’ debt and equity capital, we utilize the single-period capital asset pricing model (CAPM) as employed by Chen (1978) and Osterberg and Thomson (1991). The value of a bank equals the present value of its fu ture cash flows. Debt and equity values are in turn equal to the present value of these respective claims on the firm’s cash flows. Certain cash flows are discounted at the risk-free rate of return, while uncertain cash flows are converted to certaintyequivalent flows by deducting a risk premium from the expected cash flow. The CAPM implies that the risk premium is simply the market price of risk multiplied by nondiversifiable risk. Our primary assumptions are 1) the risk-free rate of return is constant, 2) capital markets are perfectly competitive, 3) expectations are homoge neous with respect to the probability distributions of risky asset yields, 4) investors are risk averse, seeking to maximize the utility of terminal wealth, and 5) there are no taxes or bankruptcy costs. In section II, we assume that at the end of the period, perfect “me-first” rules are enforced. That is, all claimants receive payment according to the 19 m Variable Definitions Bl = Total promised payments to insured depositors Bu = Total promised payments to uninsured depositors z — Total promised payments to the FDIC (= pB() d p = Deposit insurance premium per dollar of insured deposits S = Total promised payments to subordinated debt holders B = Total promised payments when subordinated debt (= Bi + Bu + z) is absent K = Total promised payments when subordinated debt is present (= Bt + Bu + z + 5) Yfn> Ybu, Ys, Ye, and YFDIC = End-of-period cash flows to insured depositors, uninsured depositors, subordi nated debt holders, stockholders, and the FDIC, respectively Vhi, Vbu, Vs, Ve, and VFDIC = Values of insured deposits, uninsured deposits, subordinated debt, bank equity, and the FDIC’s claim, respectively E(Rhi), E (Rhtl), E (RS), and E(Re) = Expected rates of return on insured and uninsured deposits, subordi nated debt, and equity, respectively Vj = Value of the bank r = Risk-free rate of return (/ ? = ! + r) X = End-of-period gross return on bank assets F (X ) = Cumulative probability distribution function for X CEQ(X) = Certainty equivalent of X (=£1X1 - X COV[X, R J ) X = Market risk premium Rm = Return on market portfolio XCOV iX, Rm) = Nondiversifiable risk priority of their claim. Realized cash flows are used to satisfy the claims of senior creditors (de positors and the FDIC) before junior creditors (subordinated debt holders) are paid. Equity holders receive any residual cash flow after all creditor claims are satisfied. In sections in and IV, forbearance by the FDIC occurs when the agency fails to enforce me-first rules and allows payments to other creditors (senior or junior) or equity holders at the expense of its own claim. Sections II through IV utilize the definitions in box 2. We assume that all debt instruments are dis count instruments, so that the end-of-period prom ised payments to depositors and subordinated debt holders include principal plus interest. We also assume that the deposit insurance premium is paid at the end of the period.4 II. No FDIC Bailouts In this section, we present results from Osterberg and Thomson (1991) for a bank with insured deposits, uninsured deposits, and subordinated debt. The FDIC charges a fixed premium of p on each dollar of insured deposits. Total liability claims against the bank, K, equal the sum of the end-of-period promised payments to uninsured depositors (Bu), to insured depositors CBf), to sub ordinated debt holders, S, and to the FDIC (z = pB;). We assume that on average the FDIC under prices its deposit guarantees and provides a sub sidy that reduces the cost of capital for banks as it increases their value.5 Given these assumptions, the end-of-period cash flow to insured depositors, Yhi, equals the promised payments, B(, in every state. Regard less of capital structure, the value and expected return of one dollar of insured deposits are Vhj = R~l Bi and E (R hi) = r, respectively. The cash flows to uninsured depositors depend on promised payments as well as on the total level of promised payments net of the subordinated debt, K - S: a. For simplicity, w e express the premium as a function of insured depos its. However, the results of interest here would not be materially affected by adopting the more realistic assumption that premiums are levied on the total of domestic insured and uninsured deposits. ■ 4 For simplicity, we view the premium as an end-of-period claim on the bank. This is equivalent to assuming that the premium is subor dinate to Bi and that, in effect, the bank receives coverage without neces sarily paying the full premium. Although this condition influences the size of the subsidy, it does not qualitatively affect the key results. ■ 5 Buser, Chen, and Kane (1981) introduce regulatory taxes into a sim ilar framework. 20 if X > K - S= B..+ B.. + z , Yb u = B . „ BUX / ( K - S ) if K - S > X > 0, 0 if 0 > X . Notice that although the total promised pay ments to debt holders and the FDIC equals K, the effective bankruptcy threshold equals K less the claims of subordinated debt holders. Assum ing that K - S is less than the previous threshold without subordinated debt, the value of unin sured deposits would rise with S. However, as we discuss below, whether or not this occurs depends on deposit insurance pricing, which in fluences z and thus K. The value of and the re quired rate of return on uninsured deposits are (1 ) Vbu= R~]K E (R ,J = 1 - F (K - S ) + [1 /(K - S ) ]E$~S(X) 1 - F (K - S ) + [\/(K- S) 1CEQfî- s (X ) - 1. Equation (2) shows that the cost of uninsured deposit capital is a function of the bank’s nondiversifiable risk, XCOV(X, Rm), total promised payments to depositors and the FDIC, K -S, the probability that losses will exceed the level of sub ordinated debt, F (K - 5), and the risk-free rate of return, r. As stated above, the cost of uninsured deposit capital, E(Rhli), is influenced by deposit insurance pricing. Specifically, Osterberg and Thomson (1990,1991) show that underpriced (overpriced) deposit guarantees lower (raise) both the effective bankruptcy threshold for senior claims, F (K - S), and the bankruptcy threshold, F(K ). Furthermore, underpricing (overpricing) in creases (reduces) uninsured depositors’ claims rel ative to both senior claims, Bu / (K - S), and total claims, B J K. The size of this effect depends on the FDIC’s pricing error per dollar of insured deposits and the deposit mix. The end-of-period expected cash flows accru ing to the subordinated debt holders are Ys= S (3) v;=i?-1{5[l-JF(/:-5)]-/:[F(^) - F (K - S ) } + C E Q *_S(X )} and (4) E(R s) = {(511 - F (K - S ) ] - K [F (K ) - F (K - S ) ] +E s K_s (X )} /{ S [ l - F (K - S )] - K [F(K ) - F (K - S ) ] + C E Q * _ S(X ) 1} - 1.0. ll- W - S ) ] + [Bn / ( K - S ) ] CEQ q ~ s (X ) } and (2) The value of the subordinated debt and the required rate of return on subordinated debt capital are if X > K , X+ S-K if K > X > K - S, 0 if K - S> X . Equations (3) and (4) show that the cost and value of subordinated debt capital depend on the probability of bankruptcy, F (K ), the face value of subordinated debt, S, total promised payments, K, and the probability that senior claimants will not be repaid in full, F (K - 5). Again, since K is influenced by insurance pric ing, so are Vs and E (R S). Note that the last two terms in equation (3) represent the claims of subordinated debt holders in states where they are the residual claimants. Our expression for E(RS) is consistent with Gorton and Santomero’s expression for the risk premium on subordinated debt. Here, senior claims, K - S , total claims, K and the variance of X (which influences F( ■) over the relevant ranges in equation [4]) have a nonlinear impact on the risk premium. The end-of-period cash flows accruing to stockholders are Ye —X — K if X > K , 0 if K > X . The value of equity and the expected return to stockholders are (5) (6) = R-1 { CEQk (X ) - K[ 1 - F {K ) ] } and Ek { X ) - K [ 1 - F (K ) R CEQk {X ) - K[l - F (K ) ] - 1 .0 . 21 The value of equity is unaffected by the sub ordinated debt requirement as long as total claims, K, remains unchanged. K, of course, is in fluenced by S and the pricing of the premium, 2". Equation (7) gives the total value of a bank with subordinated debt. (7) Vf = R ~ '{ c E Q 0 (X ) + B jF (K - S ) - z[ 1 - F (K - S ) ] - [(B i + z )/ (K - S )] C E Q « - s ( X ) }. Subordinated debt affects the bank’s value only through the last three terms on the right side of (7). As we show below, these terms equal the net value of deposit insurance to the bank. However, the definition of correct pricing of deposit insur ance implies that its net value is zero, and that a subordinated debt requirement has no impact on bank value. Note, however, that pricing deposit insurance correctly requires the premium to vary with the size of the subordinated debt require ment. In this case, the impact of such a require ment depends on insurance pricing. The net value of deposit insurance is simply the value of the FDIC’s claim on the bank. The end-of-period cash flows to the agency and the value of its position are (8) Yf d i c = z if X > K - 5, (B' + z ) X / ( K - S ) - B i if K - S > X > 0, —B i if 0 > X , and VFDIC = R ~ l i z [ 1 - F ( K ~ S ) ] + [ (B i + z ) / ( K - S ) ] C E Q *- S(X ) ~ Bj F (K - S ) } Bu if X > K - S = B i + B u + z BUX / ( K - S ) if K - S > X > Gh , Bu if Gb > X > G v 0 A Bu X / ( K - S ) if o Section II explained how subordinated debt affects a bank’s value through its influence on the deposit insurance subsidy. Here, we show how forbearance affects the value of an insured bank with subordinated debt in its capital structure. Pre vious empirical analyses of subordinated debt prices have failed to account for the possibility that the FDIC conditionally guarantees some uninsured liabilities, a practice defined here as forbearance. We consider two types of FDIC forbearances that differ in their assumed treatment of subordi nated debt holders versus uninsured depositors. In case A, the FDIC bails out all uninsured cred itors when earnings, X, fall between Gh and Gl and K - S > Gh. In other words, subordinated debt holders are paid in states where they would otherwise receive nothing. In the same states, uninsured depositors receive the balance of their promised claim from the FDIC. In case B, the FDIC extends forbearances to all uninsured creditors when earnings are less than Gh but greater than 6}, and K > Gh > K - S. Subordinated debt holders are paid off when they otherwise would have received partial payment, as well as when they would have received nothing without forbearance. We assume that the income range over which the FDIC forbears is known to market participants. For each case, we model only one set of bounds for FDIC bailouts of uninsured creditors. The analysis follows that in Osterberg and Thomson (1990) and also holds for multi ple and disjoint bailout states. Case A. For uninsured deposits, the intro duction of FDIC forbearances into the capital structure results in the following end-of-period cash flows: A Notice that the FDIC now receives the full premium z over a wider range, since K - S < K. Because the effective bankruptcy threshold has changed, equation (8) can be interpreted as show ing the impact of the equity-like buffer provided by subordinated debt. The subordinated debt re quirement affects the value of the FDIC’s position by changing the probability that the put options corresponding to the agency’s guarantee will be “in the money” at the end of the period. Equation (8) also makes clear that if deposit insurance is to be priced fairly ( VhDIC = 0), the premium must be influenced by the subordinated debt requirement. III. Banks’ Cost of Capital and the Value of the Insurance Fund: The Impact of Forbearance if 0 > X . Comparing equations (9) and (10), below, to (1) and (2) makes apparent the difference be tween the two scenarios: In some states where uninsured depositors had previously received BUX /(K - S), they now receive Bu. Thus, it is clear that Vhu will increase and E (Rhlt) will fall. 22 The value of and the required rate of return on un insured deposits are now functions of the size and probability of the FDIC bailout. The threshold K - S will be influenced by the impact of forbear ance on the insurer’s choice of premium, z. (9) (10) and (12) to (3) and (4). Failure to account for this effect could lead empirical investigators to conclude that risk premia for certain banks are too low to be consistent with market discipline. In Osterberg and Thomson (1990), we show that the impact of extending forbearance to uninsured creditors is entirely captured by those yhll = R~X \.BU\ \-F(K-S) + F(Gh) -F(Gt) ] creditors and that there is no effect on equity holders. However, forbearance influences the + [BU/ ( K - S ) ) [CEQ£~s (X ) - CEQ g*(X )] }. values of deposit insurance and the bank. Equations (13) and (14) indicate the value of the bank and of FDIC guarantees when the bailout occurs for X between Gh and Gt. E{Rhl) = /?{{1 -F(K-S) + F{Gh)~ F(Gj) + [1/{K-S) ][E*fs(X) -E%>(X) ]} (13) Vf = R~' { CEQq (X) - z [1 - F(K-S) ] + {l-F(K-S) + F(Gh)-F(G l ) - [(B. + z)/(K -S) ]ICEQq~s (X) + 11/(K-S)][CEQ^~s(X) - C E Q ^ (X )}- C E Q ^(X ) + B tF ^ - S ) -C EQ ^(X ) 11 } - 1.0. + (S+Bu)[F(Gh) -F (G ,)]}. z The end-of-period cash flows to the subordi nated debt holders are if X > K - S, ( B i + z )X /(K - - S )- B i if K - S> X > Gh, A if A S ^3 X - B u -Br X > K, (Bt+ z) X / ( K - S ) - B t if Gt> X > 0, X + S - K if K > X > K-S, ~Bi 0 if K - S > X > Gh 5 if Gh > X > Gh 0 if G ,> X . Ys = S if (14) if 0 > X, a n d VFD/C = R-1 {z[l - F(K - S )] + [(Bi + z ) / ( K - S )] The value of the subordinated debt and its required rate of return are [CEQ$-S(X )-C EQ g*(X)] + CEQGP(X) - B j F (K - S ) (11) Vs= R~l {5[1 -F(K-S) +F{Gh) -F(G ,)] - (S + B u )[F(G h ) - F ( G l )]}. -K[F(K) -F(K-S) ] + CEQK_s (X) } and (12) £(;?,) = tf{{S[l -F(K-S) +F(Gh)-F(G l )] -K[F(K) -F(K-S)] + E k_ s {X) \ -1511 -F(K-S) + F(Gh) -F (G l)} - K[F(K) -F(K-S)] + CE Q K _S{X) 1} - 1.0. In some states where X falls below K - S , S is now received instead of zero. Thus, Vs must rise and E(RS) must fall. We show this below through a formal comparison of equations (11) The crucial role of deposit insurance pricing in determining the impact of forbearance is most easily seen by noting that the bank’s value in equation (13) is simply the sum of the value of an all-equity firm and the net value of im plicit and explicit FDIC guarantees (from [14]): Vj- R~x CEQ q (X ) + Vpoic- O f course, if the FDIC prices its guarantees fairly, then VFDIC = 0 and Vf —R l CEQ0 ( X ) , the value of the all equity firm. The impacts of the subordinated debt requirement, forbearance, and capital struc ture are reflected in the value of the deposit insurance subsidy. In this case, the pricing of both the explicit and implicit guarantees will in fluence the impact of subordinated debt. 23 Case B. Introducing FDIC forbearances into the capital structure when X is less than Gh ( G) > K - S > Gh) results in the following endof-period cash flows to uninsured depositors: (18) £(/?s) = /?{l5[l -F(Gl )]-K [F(K )-F(G h)] + E *_S(X) 1/(511 -F(G,) ] —K[F(K) - F{G,)) if Ybu= Bu X > G, Bu X / ( K - S ) if G , > X > 0, 0 0>X. if Again, the value of and the required rate of return on uninsured deposits are functions of the size and probability of the FDIC bailout. However, unlike the previous case, when the uninsured de positors suffered some losses after the subordi nated debt was exhausted, this policy guarantees their claims for all values of X above Gl . Thus, Vhu will rise and E (R hlt) will fall. + CEQk _s {X)\}~ 1.0. Since Gt > K - S> Gh, a comparison with the no-bailout case shows that Vs rises and E (Rs) falls. Equations (19) and (20) indicate the value of the bank and of FDIC guarantees when the FDIC bailout occurs for X between Gh and Gl . (19) Vf = R ^ { c E Q 0{X )- z[l- F (K )] ~[{K-S)/{K)} C E Q ^(X )- C E Q ^(X ) + Bi F(K) + BU[F(K) -F(G,)] + S[F(Gb) (15) Vbu = /?-1 {bu [1 - F{G[) 1 -F{G ,)]-{ (B' + z )/(K - S ) ] CEQ$,(X)}. + 1BU/(K-S) 1[CEQfi(X) ]} and (16) if X > K, = z E (R bu) = R { il- F (G l ) K - S - B t -Bu if K > X > Gh X —Bu —Bj —S if Gh > X > G x (Bi + z) X / ( K - S ) -Bi if G1 > X > 0 , + 11/(K - S )]E fi(X )} -Bi if 0 > X , and +{l-F {G l ) + [l/(K -S)] CEQfr(X) I } -1.0. (20) VFDic= {z[l —F(K) ]- (Bu +Bi+ S) 1F(Gh)- F (G l)] - B i F(Gl) The end-of-period expected cash flows accruing to the subordinated debt holders are + [(Bt+z)/(K -S) ] CEQ$h(X) + [BU/(K -S)]C EQ % (X )}. Y = S X+S-K if X>K, if K > X > G. if Gh > X > G, if G ,> X > 0 . The value of subordinated debt and its re quired rate of return are (17) V ^^ jsil- F C G - )] - K[F(K) - F (G h) 1+ C EQ k g (X ) } and As in case A, the bank’s value depends on both the FDIC’s pricing of its explicit guarantees and the value of its implicit guarantees via forbearance. IV. The Effects of Mispriced Deposit Guarantees and Forbearance on the Value of Subordinated Debt Capital In this section, we use the results of sections II and III to analyze explicitly the impact of mis priced deposit insurance and FDIC forbearance policies on the value of, and hence the required return on, subordinated debt. Mispricing deposit insurance increases the value of subordinated debt. To see this, first 24 define D as total promised payments to liability holders and Ysd as the respective cash flows accruing to subordinated debt holders per dollar of promised payment when insurance is mis priced or fairly priced.6 In order to calculate the impact of mispricing on the value of subordinated debt, we construct a replicating portfolio for the one-dollar par-value subordinated debt claim when deposit guarantees are mispriced. This portfolio consists of one unit of a one-dollar par-value subordinated debt claim when deposit insurance is fairly priced, and a sec ond security AdYs (= ys- ysd) with the following cash flows: share of subordinated debt without FDIC for bearances and a security AaYs(AbYs) with the following cash flows: \ Ys = 0 if X > G h , 1 if G h > X > G l , 0 if G ,> X . = 0 (K - X ) / S if X > G h , if G b > X > K - S 1 if K - S > X > G , , 0 if G ,> X . In case A, subordinated debt holders receive payment from the FDIC equal to the par value of their claim for all values of X between Gh and Gl . (D - X ) / S if D > X > K , In case B, they receive a partial bailout when X is K > X > D S , if (D - K ) / S between Gh and K - S and a full bailout when X l + ( X - K ) / S if D - S > X > K - S is between K - S and Gt. The difference between if K - S > X . 0 the cash flows in the two cases reflects the differ The value of this security is ence in the assumed bailout policy. In case A, the FDIC extends forbearances only when losses ex ceed the value of the subordinated debt. In case (21) Ad Vs = { R S ) ~ l |dLF(Z)) - F ( D - S ) ]- C E Q ° ( X ) B, forbearances are extended before losses totally exhaust the subordinated debt. - K [ F ( K ) - F ( K - S ) ] + C E Q °Z S (X ) Equations (22) and (23) show that the value of the securities that replicates the value of for + 5[F(D-5)-F(/C-5)]}, bearance to subordinated debt holders is posi tive and that Ah Ys> Aa Ys.7 which is positive if 0 if X > D , C E Q ° Z S (X) > (K-S) [F ( D - S ) - F ( K - S ) ]. Equation (21) shows that mispricing deposit insurance affects the value of subordinated debt capital by altering the probability that subordi nated debt holders will be repaid in full. In effect, deposit insurance subsidies alter the ranges over which subordinated debt prices behave like equity and debt prices. Forbearance policies also affect the value of, and thus the rate of return on, sub ordinated debt. In either case, however, forbear ance both increases the value of subordinated debt and changes pricing. Following the procedure used above, we next construct a replicating portfolio for a onedollar par-value subordinated debt claim when the FDIC bails out liability holders. The replicat ing portfolio for case A (case B) consists of one ■ 6 When there are no FDIC forbearances and deposit insurance is fairly priced, the end-of-period expected cash flows accruing to the sub ordinated debt holders are Ys, d = S X+S-D 0 if X > D , if D > X > D - S , if D - S > X . (22) AaVs=R~1[F(Gh)- F (G ,)] > 0 . (23) A b Vs = C ^ ) -1 {a'[F(G^) -.F(/:-S)] - C E Q j£ h_ s ( X ) + S[F(K-S) - F(Gt) ] } > 0 . As noted by Gorton and Santomero, subordi nated debt is a hybrid instrument whose price and return behave like debt for high values of X, but like equity for low values of X . The pos sibility of FDIC bailouts when X is in the range for which subordinated debt would typically be have like equity complicates the pricing dynam ics. Specifically, without forbearance, there is a range of values for X such that subordinated debt prices switch from acting like debt to acting like equity as earnings increase. The introduc tion of FDIC forbearances may change the switch point or introduce multiple switch points. ■ 7 To see this, note that F { K - S) - F( Gi ) > F ( G h ) - F( Gi ) and ( K / S ) [F ( G b) - F ( K - S) ] > (1 / S ) CEQG Kb_s ( X ) . 25 Previous empirical studies of the relationship between subordinated debt prices and balance sheets by Gorton and Santomero and Avery, Belton, and Goldberg do not account for the possible impact of FDIC forbearance policy. The theory presented above provides one possible explanation of previous empirical findings that risk premia on subordinated debt are weakly related to risk proxies. V. Conclusion Using the cash-flow version of the CAPM devel oped by Chen (1978) and extended by Osterberg and Thomson (1990, 1991), we develop an explicit pricing model for subordinated debt that considers the possibility of implicit guarantees of nominally uninsured debt capital. Similar guarantees have been present during the sample periods of recent empirical studies of subordi nated debt prices. Our findings indicate that FDIC forbearance increases the value of subor dinated debt and thus alters investors’ required rates of return. Forbearance reduces the usefulness of subor dinated debt in two ways. First, the possibility of FDIC bailouts directly increases the deposit insurance subsidy. However, given the possi bility of such bailouts, the size of the subsidy is reduced by a subordinated debt requirement as long as there is some chance that subordinated creditors will realize losses. Second, forbearance reduces the rate of re turn required on subordinated debt of a given risk, a policy that may easily impede market dis cipline of bank risk-taking. This in turn reduces the amount of information in secondary market prices of subordinated debt. Forbearance thus introduces a potential source of specification error in empirical studies of the risk premium in subordinated debt markets. As we have emphasized previously (Osterberg and Thomson [1990,1991]), the impact of capital structure changes on insured banks de pends on deposit insurance pricing. If deposit insurance is fairly priced, neither subordinated debt requirements nor forbearance will impact overall bank value. However, in the more realis tic case of deposit insurance mispricing, the effects of expected capital structure changes are altered through their interaction with the overall regulatory environment. References Allen, Linda, and Anthony Saunders. "Forbear ance and Valuation of Deposit Insurance as a Callable Put,” Baruch College, Working Paper, December 1990. Avery, Robert B., Terrence Belton, and Michael Goldberg. “Market Discipline in Regulating Bank Risk: New Evidence from the Capital Markets,” Journal o f Money, Credit, a n d Bank ing, vol. 20 (November 1988), pp. 597-610. Baer, Herbert. “Private Prices, Public Insurance: The Pricing of Federal Deposit Insurance,” Federal Reserve Bank of Chicago, Economic Perspectives, vol. 9 (September/ October 1985), pp. 41-57. ______ , and Elijah Brewer. “Uninsured Depos its as a Source of Market Discipline: Some New Evidence,” Federal Reserve Bank of Chicago, Economic Perspectives, vol. 10, no. 5 (September/October 1986), pp. 23-31. Benston, George J., Robert A. Eisenbeis, Paul M. Horvitz, Edward J. Kane, and George G. Kauf man. Perspectives on Safe a n d Sound B ank ing: Past, Present, a n d Future. Cambridge, Mass.: MIT Press, 1986. Buser, Stephen A., Andrew H. Chen, and Edward J. Kane. “Federal Deposit Insurance, Regula tory Policy, and Optimal Bank Capital,” Jo u rn a l o f Finance, vol. 36, no. 1 (March 1981), pp. 51-60. Chen, Andrew H. “Recent Developments in the Cost of Debt Capital,”Jo u rn a l o f Finance, vol. 33, no. 3 (June 1978), pp. 863-77. Federal Deposit Insurance Corporation. Q uar terly B anking Profile, Quarter 1 1992. Flannery, M.J. “Pricing Deposit Insurance when the Insurer Measures Risk with Error,”Jo u r n al o f B anking a n d Finance, vol. 15, nos. 4/5 (September 1991), pp. 975-98. Gilbert, R. Alton. “Market Discipline of Bank Risk: Theory and Evidence,” Federal Reserve Bank of St. Louis, Review, vol. 72, no. 1 (January/February 1990), pp. 3-18. 26 Gorton, Gary, and Anthony M. Santomero. “Mar ket Discipline and Bank Subordinated Debt,” Jo u rn a l o f Money, Credit, a n d Banking, vol. 22, no. 1 (February 1990), pp. 119-28. Hannan, Timothy H., and Gerald A. Hanweck. “Bank Insolvency Risk and the Market for Large Certificates of Deposit,”Jo u rn a l o f Money, Credit, a n d Banking, vol. 20, no. 2 (May 1988), pp. 203-11. Hirschhom, Eric, and David Zervos. “Policies to Change the Priority of Claimants: The Case of Depositor Preference Laws "Jo u rn a l o f F i n an cial Services Research, vol. 4, no. 2 (July 1990), pp. 111-26. Kane, Edward J. The Gathering Crisis in Federal Deposit Insurance. Cambridge, Mass.: MIT Press, 1985. ______ . “No Room for Weak Links in the Chain of Deposit Insurance Reform, "Jo u rn a l o f F in an cial Services Research, vol. 1 (Sep tember 1987), pp. 77-111. ______ . “How Incentive-Incompatible DepositInsurance Funds Fail,” National Bureau of Eco nomic Research, Working Paper No. 2836, February 1989a. ______ . The S&L Insurance Mess: How D id It Happen? Washington, D.C.: The Urban In stitute, 1989b. Marcus, Alan J. “Deregulation and Bank Finan cial Policy "Jo u rn al o f Banking a n d Finance, vol. 8, no. 4 (December 1984), pp. 557-65. Merton, Robert C. “An Analytic Derivation of the Cost of Deposit Insurance and Loan Guarantees: An Application of Modern O p tion Pricing Theory,” Jo u rn a l o f Banking a n d Finance, vol. 1 (June 1977), pp. 3-11. ______ . “On the Cost of Deposit Insurance When There Are Surveillance Costs,”Journal o f Business, vol. 51 (July 1978), pp. 439-52. Osterberg, William P., and James B. Thomson. “Deposit Insurance and the Cost of Capital,” Re search in Finance, vol. 8 (1990), pp. 255-70. ______ , a n d ______ . “The Effect of Subordi nated Debt and Surety Bonds on the Cost of Capital for Banks and the Value of Federal Deposit Insurance,”Jo u rn a l o f B anking a n d Finance, vol. 15, nos. 4/5 (September 1991), pp. 939-53. Pennacchi, George G. “Alternative Forms of De posit Insurance: Pricing and Bank Incentive Issues "Jo u rn a l o f Banking a n d Finance, vol. 11, no. 2 (June 1987a), pp. 291-312. ______ . “A Reexamination of the Over- (or Under-) Pricing of Deposit Insurance," Jour n a l o f Money, Credit, a n d Banking, vol. 19 (August 1987b), pp. 340-60. Randall, Richard E. “Can the Market Evaluate Asset Quality Exposure in Banks?” Federal Reserve Bank of Boston, New E ngland Eco nom ic Review, July/August 1989, pp. 3-24. Ronn, Ehud I., and Avinash K. Verma. “Pricing Risk-Adjusted Deposit Insurance: An OptionsBased Model,”Journal o f Finance, vol. 41 (September 1986), pp. 871-95. Shadow Financial Regulatory Committee. “The FDIC’s Proposed Schedule of Risk-Sensitive Premiums,” Statement No. 83, June 1, 1992. Thomson, James B. “FSLIC Forbearances to Stockholders and the Value of Savings and Loan Shares,” Federal Reserve Bank of Cleve land, Economic Review, Quarter 3 1987a, pp. 26-35. ______ . “The Use of Market Information in Pricing Deposit Insurance,”Jo u rn a l o f Money, Credit, a n d Banking, vol. 19, no. 4 (November 1987b), pp. 528-37. Wall, Larry D. “A Plan for Reducing Future Depos it Insurance Losses: Puttable Subordinated Debt,” Federal Reserve Bank of Atlanta, Eco nomic Review, vol. 74 (July/August 1989), pp. 2-17. 27 An Introduction to the International Implications of U.S. Fiscal Policy by Owen F. Humpage Introduction A predominant characteristic of U.S. macroeco nomic developments in the 1980s was the simulta neous emergence of large federal budget deficits and unprecedented international trade deficits. Many economists, relying on open-economy vari ants of the standard income-expenditure model, have linked these deficits in a causal chain that also ties them to high U.S. interest rates and to the dollar’s appreciation earlier in the decade (see Hutchison and Pigott [1984]). The description has now become part of popular economic lore, but as is often the case with legend or myth, many of the intricacies of and important quali fications to a fundamentally plausible story have been lost in its common transmittal. Moreover, a paucity of hard empirical support for the simple and direct relationship offered by this popular view has done little to curtail its telling.1 This paper acknowledges that fiscal policies can create trade deficits, but argues that this need not be the case and typically has not been the case ■ Owen F. Humpage is an eco nomic advisor at the Federal Reserve Bank of Cleveland. The author thanks David Altig, Janice Boucher Breuer, and Joseph Haubrich for their helpful com ments and Diana Dumitru for her research assistance. in the United States. Section I offers a simplified version of the two-period, representative-agent model found in Frenkel and Razin (1987).2 Un like the standard income-expenditure approach, this model does not assign a predominantly causal role to government budget deficits, but it does allow that, under certain circumstances, fis cal policies can influence the trade balance, real interest rates, and real exchange rates. The out come depends on how the government’s propen sities to import and to consume out of current income compare with those of the private sector, and on the distortionary effects of taxes. Section II offers an empirical investigation of U.S. fiscal policy during the floating-exchangerate period, using Engle-Granger (1987) co integration techniques. The empirical tests search for common long-mn trends between economic variables suggested by the theoretical analysis and aggregate measures of U.S. federal fiscal policy. The results do not support the common contention of simple, direct relation ships among these measures and U.S. trade balances, interest rates, or exchange rates. As 1 The popular accounts derive from the open-economy version of the income-expenditure (or Keynesian) model. Frenkel and Razin (1987, http://fraser.stlouisfed.org/ part II) offer an unabridged account of this model. Federal Reserve Bank of St. Louis ■ 2 See also Aschauer (198B), Hill (1990), and Koenig (1989). 28 noted in the concluding section, however, such tests are subject to important qualifications and do not preclude the possibility of short-term relationships. constraint that the present value of private inter temporal consumption equals the present value of his two-period after-tax endowments. The consumer maximizes 1 CD [ / = £ p 'u,(c,) t= 0 I. A Simple Model A nation running a current account deficit absorbs more real economic resources than it produces. Its citizens accommodate differences between their desired consumption and production by purchas ing additional goods from abroad, and they fi nance their activity by borrowing in world money markets. Because government spending and tax policies affect consumption and production deci sions, a nation’s fiscal policies can strongly influ ence its international trade patterns. Frenkel and Razin (1987) show that the rela tionship is often similar to that described in in ternational economics as the transfer problem. Because fiscal policies typically involve a trans fer of funds from the private sector to the gov ernment sector, their international implications depend on a comparison of both the govern ment’s and the private sector’s propensities to save and to import. Moreover, when govern ment activities are deficit financed, the outcomes depend more on the existence of tax distortions than on public borrowing per se. Following Frenkel and Razin, this section develops a sim ple model to illustrate these points. To appre ciate the argument, however, one must first understand the motives for international trade and the intertemporal nature of trade deficits. Two-Period Trade and the Nature of a Deficit Consider a hypothetical economy consisting of two countries (home and rest-of-w^orld), each possessing and consuming quantities of two goods over two time periods. Each country con sists of a single representative consumer and a government, which taxes and spends. Assume that no production takes place, but that both countries start each time period with a specific endowment of the two goods. Let a single consumer with homothetic prefer ences represent each country.3 Each consumer maximizes utility over two periods, subject to the ■ 3 Homothetic preferences are such that, for constant relative prices, any given percentage change in income results in the same percentage change in the consumption of all goods. Homothetic preferences cause the http://fraser.stlouisfed.org/ Income expansion curves in figures 1 through 5 to be straight lines. Federal Reserve Bank of St. Louis subject to (2) C\(1 + C 0(l + t 0)+ T0 + -— + Tx + (l + rv) ) Yx F°+ (l + rv)' Here, Ct refers to private after-tax consumption in time t(= 0,1), such that (3) Ct = c x t+ p c mt, where cx t and cm t represent consumption of goods X and M in specific time periods. The terms of trade, p, expresses units of M in terms of units of X, (3'is a subjective discount factor applied to fu ture utility, and rx Is the real interest rate. I express each in terms of good X, but the following arbitrage condition makes measurement arbitrary: (4) ( l + rv) = P,/Po( l + r j . With two goods and two time periods, how ever, unanticipated changes in the terms of trade within any period can affect intertemporal decisions.4 The Tt terms represent lump-sum taxes, whereas the tt terms are tax rates applied to private consumption. At the beginning of each period, consumers receive an endowment, Yt, of the two goods, such that (5) Yt= qxt+pq mt, where q it( i =x, m ) refers to quantities of the two goods, X and M. I assume that consumers seek to smooth consumption over the two periods by borrowing or lending through inter national credit markets. The government uses tax revenue to finance expenditures, Gt , subject to the constraint that the present discounted value of government ex penditures over the two periods equals the present discounted value of tax revenue: 1 4 For a discussion, see Frenkel and Razin (1987), pp. 168-71. 29 F I G U R E 1 Optimization over Time and the Trade Deficit c; Y, -c: over the two periods must equal the present value of the endowments. The trade account must balance, and the countries must extinguish all international debts. Equation (1) assumes that utility is intertemporally separable. Each consumer desires an optimal expenditure over the two periods. Within each period, the consumer chooses an optimal con sumption bundle of the two goods, one that maxi mizes Ut . Although this choice is constrained by the overall level of expenditure within a period and by relative prices, the choice of a consump tion bundle in any period is otherwise independ ent of the choice in any other period. Intertemporal Consumption Assuming no government sector for the moment, the representative individual allocates his con sumption over the two time periods until the following condition holds: SOURCE: Author. (6) Go + (\+ r ) ~ To +to Co T, txCx (1 + rx) (1 + rx) ' Solvency requires that the government retire any budget deficit incurred in the first period during the second period. For each nation as a whole, the first-period budget constraint is (7) C0 + G0~ Yq + ^ 0 • Any nation can absorb, through private con sumption and government spending, more or less than its current endowment, as equation (5) shows, but if it absorbs more than its endow ment, the nation must borrow (B0 > 0 ) , and if it consumes less, it will lend the excess (B0 < 0 ). The second-period budget constraint is given by (8) q +G ^ - d + r , ) ^ . Since this model contains only two periods, each country must retire any first-period debts in the second period. Therefore, solvency re quires that over the two periods, (9) c\ C0 + G0 + ~ ~ 0 0 (1 + r ) F° + (1 + r ) Accordingly, the present value of private after http://fraser.stlouisfed.org/ tax consumption plus government spending Federal Reserve Bank of St. Louis m + where Ut is the marginal utility derived from consumption in period t. The first term in equa tion (10), the consumer’s marginal rate of substi tution between present and future consumption, measures his willingness to trade current for fu ture consumption. The higher his subjective dis count factor, the more the consumer prefers present to future consumption. The second term, one plus the real interest rate, is the inter temporal terms of trade— the market terms at which a consumer can trade current for future consumption. As equation (10) indicates, the utility-maximizing consumer will allocate his consumption over the two periods until his will ingness to substitute between them equals the terms offered for this exchange in the market. If at any time this condition is not met, an exchange of resources can enhance the consumer’s utility. In figure 1, this maximization process is illus trated with an Edgeworth-box diagram, which shows the home country’s origin in the lower left comer and the foreign country’s origin in the upper right comer. (An asterisk designates foreign variables.) The utility curves I and II show, for a given level of utility, the willingness of the home country and the rest of the world to trade current for future consumption.5 The ray extending from each origin, the income expansion path, shows ■ 5 See also Hill (1990) and Koenig (1989). 30 FI GURE 2 Intratemporal Consumption Optimization across Goods and Trade at Time t * qm t * ^m l ' / After allocating consumption across time, each representative consumer will choose quantities of the two goods that maximize utility at each point in time. Consumers will choose among the two goods X and M until ( 11) < lx t ............... cx t \ ^ \ \ \ \ r-- ! SOURCE: Author. the respective country’s optimal level of con sumption for changing levels of income and a fixed real interest rate. The slopes of these two rays indicate that the home country prefers cur rent consumption relatively more than does the foreign country. Point A, at the center of the diagram, marks initial endowments and shows that each coun try receives equal consumption bundles in each period, Yt = Y* (t= 0,1). At point A, however, the countries’ subjective temporal preferences for consumption differ. The home country val ues present consumption more highly than does the foreign country. Consequently, both can in crease their utility by agreeing to trade at some rate of intertemporal exchange passing within the ellipse formed by their utility curves. The line passing through points A and E, whose slope is - (1 + rx) , is one such rate of exchange. Given the real interest rate rx, the nations will trade to point E, at which the conditions for op timal consumption, given by equation (10), hold.6 The home country now consumes more than its initial endowment in the first period, running a trade deficit, B 0, but it will run a surplus, (1 + rx) B(), in the second period. At point E, each country is on a higher utility curve than at point A In fact, point E is a Pareto optimum; no country can be made better off without making the other worse off. ■ 6 The home and foreign countries will negotiate the optimal interest rate. uX, t p- The term on the left side of equation (11) gives the marginal rate of substitution, the rate at which each consumer is willing to substitute be tween goods X and M. The term on the right side is the market-based relative price of the two goods, or the temporal terms of trade. If during any time period the condition depicted in equation (11) is not fulfilled, an opportunity exists for welfare-enhancing trade. I again illustrate the maximization process by reproducing the Edgeworth box in figure 2 with appropriate changes in the axis and in the termsof-trade line. I depict the home country as favor ing consumption of good M , the importable good. At the initial endowment point, A, the home coun try values consumption of this good more than does the foreign country, and both countries can gain from exchange along the terms-of-trade line (with slope -p) to point E, where the condition given in equation (11) holds. At point E, the home country consumes the importable good in excess of its initial endowment, but it consumes less than its initial endowment of the exportable good. Nature of Trade and Trade Deficits Despite the simplicity of the model, figures 1 and 2 offer important insights into the nature of inter national trade and the causes of trade imbalances. Trade takes place in this model because of 1) dif ferences in nations’ time preference for consump tion at the initial endowment point, or 2) differ ences in the relative preferences for the two goods in any time period given endowments.7 A trade imbalance results when a country desires a consumption profile that differs from its endowment profile. A country that consumes more (less) than its current endowment will run ■ 7 I do not include comparative advantage as a motive for trade, despite its predominance in the literature, because the model does not Inelude production. 31 a trade deficit (surplus) 8 Changes in the real in terest rate act to clear the intertemporal imbal ance between endowments and consumption. This suggests that factors that influence decisions about intertemporal consumption— including government policies— also affect the trade balance. Hill (1989), for example, argues that a country’s demographic profile influences its trade balance because younger households tend to save less than older households. Moreover, because this model specifies the interest rate in terms of good X, and as a result of the arbitrage condition (4), factors that cause an unexpected change in the terms of trade can also influence the interest rate, intertemporal decisions, and the trade balance. The relation ship between changes in the terms of trade and the trade balance depends on whether these changes are permanent or temporary, on the ini tial position of the trade balance, and on the parameters of the model (see Frenkel and Razin [19871, pp. 176-82). The analysis in figure 1 also helps to dispel the notion that a trade deficit represents a state of economic disequilibrium or a deterioration in the economic well-being of the deficit country. Instead, the model illustrates that both the sur plus and the deficit countries improve their economic welfare by running trade imbalances. A developing country, for example, might run a trade deficit in order to acquire capital goods, with the intention of eventually financing the acquisition by running a trade surplus. Such strategies are typically considered welfare en hancing. Nevertheless, the figure does illustrate that the deficit country must eventually finance its debts though a reduction in future consump tion. In the comparative static model presented here, the reduction is absolute. In a dynamic model, with growing economies, any change in future consumption is measured relative to where it would have been in the absence of trade. In such a model, it is not necessarily the case that a trade deficit must lower future standards of living.9 ■ 8 In the National Income and Product Accounts, gross national product (GNP) equals consumption (C) plus Investment (I) plus govern ment purchases (G) plus exports (X) minus imports (M): GNP = C + 1+ G + X - M. Rearranging this expression, one obtains GNP - C - 1- G = X - M, which shows the relationship between national savings on the left side and the trade balance on the right side. ■ 9 See Anderson and Bryan (1989). Government Fiscal Policy and the Trade Deficit Much of the recent concern about U.S. fiscal policy centers on the impact of federal budget deficits on real interest rates, exchange rates, and the trade balance. The theoretical analysis of fiscal policy, therefore, begins by considering the effects of deficit-financed tax reductions, in cluding 1) a lump-sum tax cut, and 2) a reduc tion in the tax rate on consumption. Because many politicians and economists favor a balanced-budget amendment, I next consider the effects of balanced-budget fiscal policies in the form of 1) temporary and per manent balanced-budget spending, and 2) balanced-budget spending on the exportable commodity. As we shall see, different types of policies can have different combinations of ef fects on real interest rates, the terms of trade, and the trade balance. Deficit-Financed Cut in Lump-Sum Taxes With the introduction of taxes into the model, equation (12) gives the condition for optimal intertemporal consumption: ( 12) u' (i + o p i/; (i + o (1 + r ). In maximizing welfare, the representative con sumer now chooses an intertemporal consump tion pattern that equates his marginal rate of substitution between current and future con sumption to intertemporal terms of trade that include taxes on current and future consump tion as well as on real interest rates. As is well known, lump-sum taxes in the consumer’s bud get constraint (equation [2]) do not affect the choice of the optimal consumption pattern, and therefore will have no effect on real interest rates or on the trade balance. According to the principle of Ricardian equivalence, the intertemporal path of private consumption is invariant with respect to whether the government finances a given level of expendi ture via lump-sum taxes or via borrowing. If consumers understand that the issuance of gov ernment debt implies a future tax liability to retire that debt, and if they also desire a smooth inter temporal consumption path, then a deficitfinanced cut in taxes will not cause them to increase their present consumption. Instead, they 32 FI GURE 3 Deficit-Financed Reduction in Consumption Taxes c^ E o SOURCE: Author. will save the additional purchasing power result ing from the tax cut in order to meet the future tax liabilities associated with retiring the govern ment debt. The method of financing will, there fore, leave the interest rate unaffected. The simple two-period model outlined above incorporates Ricardian equivalence in that the single representative agent must retire any government debt in the second period. The real-world application of Ricardian equivalence, however, seems more problematic given that taxes are distortionary, that the present genera tion might push the burden of retiring the debt onto future generations, or that the tax cut redis tributes income to segments of the population with high marginal propensities to consume, while leaving the burden of servicing the debt spread across all citizens.10 Deficit-Financed Reduction in Consumption Taxes When I allow a deficit-financed reduction in consumption taxes, equation (12) indicates that it will distort that optimal intertemporal con sumption choice. This can be seen in figure 3, which illustrates the effects of a deficit-inducing reduction in taxes on cunent consumption. ■ 10 For an empirical application to the twin deficit issue, see Enders and Lee (1990). Point A represents an initial equilibrium, at which present and future taxes on consumption are equal at home and abroad. Now consider a temporary tax reduction on current domestic consumption in time period 0. The line for taxadjusted intertemporal terms of trade for the home country shifts from that designated as a in figure 3 to that designated as (3. (The foreign country continues to face intertemporal terms of trade given by line a . ) As the figure shows, the deficit-inducing tax cut encourages current domestic consumption and results in an excess demand for current out put given by (C Q- C0) . The real interest rate will subsequently rise, causing the world termsof-trade line a to become steeper, until the markets for current and future consumption clear at a point such as E. Because at point E the home country is consuming more than its initial endowment, it runs a trade deficit amounting to (C q - C0). At point E, the home country con sumes less than its endowment of the future goods, thereby running a trade surplus in period 1, given by (C x - C f ). Point E is also on a lower indifference curve because the higher interest reduces the present value of future in come. Although not shown, the foreign country might share part of this effect. At the new market-clearing point E, the tax creates a distortion between the market intertem poral terms of trade, given by line a ', which the foreigner faces, and the tax-adjusted intertemporal terms of trade, given by line (3', which the home country faces. The resulting lens between the two utility curves, which pass through point E, repre sents the welfare costs of the tax distortion.11 Figure 3 shows that a deficit-inducing tax reduction that encourages current consumption over future consumption will raise the real interest rate and create a trade deficit in the home country. Although the model does not include production, extrapolating from its underlying logic, one would expect that a deficit-financed tax reduction (for ex ample, a payroll tax cut or a lower capital gains tax that stimulated current production relative to cur rent consumption) could lower real interest rates and generate a trade surplus. As the model suggests, no simple relationship exists among government budget deficits, real in terest rates, and the trade deficit. In comparing the results of this section with those of the previous one, I find that it is the distortionary nature of the ■ 11 Although not drawn as such, the slope of line ( 3 'will be higher than that of line (3 because of the rise in the world Interest rate. 33 F I G U R E 4 Balanced-Budget Spending on Current Output SOURCE: Author. tax that is crucial and not the deficit per se (see Frenkel and Razin [1987], p. 223).12 Balanced-Budget Spending The preceding suggests that the relationships among fiscal policy, real interest rates, and the trade deficit depend on the distortionary nature of taxes rather than on the use of deficit financ ing. This section extends the investigation by considering balanced-budget spending meas ures. If the observed correlations between defi cits and the trade balance in the early 1980s stemmed from specific tax and spending poli cies, then a balanced-budget amendment would be of little avail in lowering real interest rates or eliminating the trade deficit. Assume that the economy is initially in equi librium with a balanced trade account. Point A in figure 4, which is similar to figure 1 in its ini tial construction, depicts such a situation, with the home country consuming C0 in the current period and Cx in the future period. In equi librium at point A, the intertemporal terms of trade are given by line / with slope - (1 + rx) . Now allow a temporary rise in home-country government spending, financed entirely with a lump-sum tax on the home-country con- ■ 12 I do not consider taxes on specific commodities (such as tariffs); they are a standard topic of trade theory. sumers.13 The model depicts this as an increase in government spending on the current good only. The government’s fiscal action reduces the amount of current output available for both domestic and foreign private consumption, which figure 4 shows as a shortening by G0 in the horizontal dimensions of the Edgeworth box. Two other adjustments follow: First, for the foreign country only, point A shifts to point A*, where both current and future consumption are unaffected by the home government’s fiscal policy. Second, because of the tax, T0, homecountry consumption shifts from point A to point B. (Notice that the horizontal distance measured by T0 equals the horizontal distance G0.) As its after-tax income falls, the homecountiy private sector reduces its consumption of both C0 and Cx, but because individuals will attempt to smooth consumption over both periods, the reduction in current consumption will not match the increase in the government’s current consumption. Taking account of all of these initial effects in figure 4, we see that balanced-budget government spending initially creates an excess demand for current output, designated by (C 0 - C 0 ), and an excess supply of fuaire output, designated by (C j - C j ). These imbalances will cause the real interest rate to rise, increasing the attractiveness of future private consumption relative to current private consumption. Graphically, the rise in the real interest rate will pivot the intertemporal termsof-trade line to a position such as that shown by /' until a new equilibrium, as defined by equation (10), obtains. Figure 4 shows such an equilibrium at point E. Here, the home country records a current-period trade deficit equal to (C% - C Q ). The model indicates that a temporary increase in home-govemment, balanced-budget spending reduces both domestic and foreign private con sumption and causes a home-country trade defi cit. Intertemporal aspects of these resource trans fers are accommodated through a rise in real interest rates. Extending the analysis to consider the effects of a permanent increase in balanced-budget spending helps to illustrate more clearly the na ture of the relationship between government spending and the trade deficit. One can show the effects of a permanent increase in govern ment spending in an Edgeworth-box diagram by altering both its horizontal and vertical dimensions. W hen both dimensions change, ■ 13 Assume that the propensity of the government to spend on goods Xand M exactly matches that of the private sector, so that the terms of trade do not change. This assumption is discussed below. 34 F I G U R E 5 Balanced-Budget Spending on Exportables SOURCE: Author. however, many different configurations of results are possible, depending on the propen sities of the government to spend on current and future consumption (see Frenkel and Razin [1987], pp. 195-98). If, for example, the govern ment’s propensities to consume current and future output exactly match those of the private sector, as indicated by the slope of the diagonal running from 0 to 0* in figure 4, then no trade imbalance or change in real interest rates would result from government spending. The equilib rium point would simply slide down the diago nal from A toward 0. Frenkel and Razin argue that international re percussions of government spending are similar to those typically discussed in the literature as the transferproblem. Balanced-budget spending transfers resources from the home-country private sector to the government sector. If the homecountry government’s intertemporal preference for consumption differs from that of the private sector, the transfer will alter the overall world equi librium for intertemporal consumption. If the over all propensity to spend on current output rises, as depicted in figure 4, real interest rates will increase and a home-country trade deficit may ensue. Conversely, if the overall world propensity to con sume current output falls, real interest rates will decline and the home country may experience a trade surplus. According to the model, one must know more to predict the effects than simply that government spending increased. Government Spending on Export Goods The effects of government spending on a par ticular commodity within a specific time period are analytically similar. Assume that the private sector has obtained the optimal pattern of con sumption over both time periods and across both goods. Figure 5 depicts the optimal domes tic and foreign private consumption of the ex portable and importable goods for a given time period at point A. I assume that the government has the same rate of time preference as does the private sector. The initial effects of government balancedbudget spending on the export good are depicted as shifting the initial foreign position to point A' and as shifting the initial domestic private-sector position to point B for reasons paralleling those offered in the explanation of the similar shift in figure 4. The tax and spend ing patterns then create an excess demand for the export good given by (X Q- X Q ) and create an excess supply for the import good equal to (M 0 - M 0). The terms of trade will improve (the relative price of the exportable good will rise) until an equilibrium such as point E obtains. The example outlined above is not a general case. I have assumed that domestic and foreign propensities to spend on the importable good are exactly the same and less than one, but I have set the government’s propensity to spend on this good at zero. Allowing the government to spend on both the exportable and the import able good, additional outcomes are possible and reasonable. Frenkel and Razin (1987, pp. 202-03) explain this, again following the argu ments that underlie the transfer problems. In general, the terms of trade will deteriorate (im prove) if the government’s propensity to import exceeds (is less than) the home country’s pro pensity to import. The terms of trade will be un changed when the propensities are exactly alike. As noted earlier, with the interest rate defined in terms of the exportable good, unanticipated changes in the terms of trade can affect intertem poral decisions and, hence, the trade deficit. This results because of the arbitrage condition depicted in equation (4). II. Empirical Evidence The simple theoretical model shows that fiscal policy can be related to trade deficits, real interest 35 rates, and real exchange rates, but that the con nection need not necessarily hold. Whether, as is often asserted, a simple, direct relationship be tween U.S. fiscal policies and the U.S. trade balance exists seems largely a matter for empiri cal analysis. Using Engl e-Granger cointegration techniques, this section tests for a long-term relationship among various measures of U.S. fis cal policy, the trade balance, exchange rates, and interest rates.14 Because cointegration looks for long-term relationships, one might view this exercise as testing the effects of the permanent component of fiscal policies. Cointegration Many macroeconomic time series are not sta tionary; that is, their mean, variance, and co variance can change over time. Intuitively, this suggests that, given a random shock, these series will move off to new time paths instead of returning to their original ones. The presence of nonstationarity can invalidate many standard statistical techniques for hypothesis testing, mak ing it difficult to determine if two nonstationary series, such as government spending and inter est rates, are related. Economists often model time series as ARIMA (p, d, q) processes, where d is the number of times the series must be differ enced to achieve stationarity.15 For most economic time series, d-\. Economists refer to such series as containing a unit root or as being integrated of order 1, and designate such series 1(1). Engle and Granger (1987) propose a method by which one can determine whether two 1(1) times series tend to move in tandem or drift; apart over time. In the former case, even though the in dividual series are nonstationary, their joint rela tionship is stationary. Engle and Granger refer to such series as being cointegrated. The Engle-Granger cointegration test is simi lar to the Dickey-Fuller (1979) test for unit roots. One must perform the latter tests as a first step in the cointegration test to see if the relevant series are each 1(1), because time series that are integrated of different orders generally are not cointegrated. The Dickey-Fuller (DF) test in volves regressing a time-series variable Y on its ■ 14 Boucher (1991) uses similar cointegration tests to study the relationship between the nominal current account balance and a set of variables either related by virtue of the savings—investment identity or commonly held to “cause” the current account. Included among Boucher’s causal variables is the nominal federal budget deficit. ■ 15 ARIMA (p , d q ) refers to Autoregressive Integrated Moving http://fraser.stlouisfed.org/ Average (see Box and Jenkins [1970]). Federal Reserve Bank of St. Louis past value to see if the resulting coefficient is equal to 1. As is common, I specify the DF test with a constant and a time trend (13) r ,= p0 + p , / + p 2 + where ut is the error term. Failure to reject the null hypothesis that (3, = 1 indicates that Y is I (1). One calculates the DF test statistic exactly like a standard t statistic, but the DF statistic does not have a t distribution. TSP version 4.20 provides critical values based on the appropriate distribution. Fuller (1976, table 8.5.2) also provides critical values. The presence of serial correlation in the error terms greatly weakens the power of the DF test, but one can correct for serial correlation by aug menting the above specification with lagged first differences of the dependent variable.16 The augmented Dickey-Fuller (ADF) test is (14) J',= P0 + P1/ + P2 J',.1 p ^ + Yt- i-\+ i- 0 where et is the error term. The null hypothesis remains the same: (3, = 1. According to Engle and Granger, two 1(1) time series, Y and X, are cointegrated if a linear combination of these two variables is stationary. Such a combination can be obtained from an or dinary least squares regression of Y on X , called the cointegrating regression. In what fol lows, I consistently specify the cointegrating regression to include a constant term ((30): (15) r ,= p0 + p2x , + e,. The error term, e ,, from the cointegrating re gression is then a linear combination of X and Y, and one can use the DF procedures to test for a unit root in the error term. Following con vention, I specify the test as p (16) E,= P , E , _ , + X P / + 2A e ,-i-l. 1= 0 including lagged first differences of the error term when necessary to adjust for possible serial correlation. The null hypothesis is (3, = 1. Failure to reject the null hypothesis indicates that the error term is not stationary and that it tends to drift away from its expected value, zero, over the sample period. ■ 16 DF tests are robust to heteroscedasticity. 36 ï Data Description Description (Code) Source Trade-weighted dollar (TWD) Board of Governors of the Federal Reserve System 10-year Treasury bill (LTR) DRI/McGraw-Hill, Inc. Trade balance: Net exports of goods and services (NEX) National Income and Product Accounts Government deficit: Change in publicly held federal debt (DEF) Flow of Funds Government spending: Federal expenditures (FEXP) Federal purchases (FPUR) National Income and Product Accounts National Income and Product Accounts NOTE: All series are inflation adjusted. I deflated LTR, DEF\ and FEXP using the Consumer Price Index. Others are published in an inflation-adjusted format. W m em m m t a b l e i Unit Root Tests Dickey-Fuller Statistic Augmented DickeyFuller Statistic TWD LTR -1.11 -3.06 -2.17 -2.10 NEX DEF FEXP FPUR -1.31 -6.14 -2.41 -2.74 -2.75 -3.14 -1.66 Variables Data Most popular discussions of the international ram ifications of U.S. fiscal policy focus on the federal budget deficit and federal spending, so my meas ures of fiscal policy exclude the state and local sec tors. I test for cointegration between either the federal budget deficit (DEF), federal government spending (FEXP), or federal government pur chases of goods and services (FPUR), and long term interest rates (LTR), the trade-weighted dollar (TWD), and net exports of goods and services (NEX). Box 1 describes the data sources. Consistent with the theoretical analysis, all vari ables are in real, or inflation-adjusted, form. If an individual series was unavailable in this form, I deflated the nominal series with the Consumer Price Index. I initially ran all tests from 1973:IVQ through 1991 :IIIQ to include 74 observations, but because augmented versions include four lagged variables, the tests run from 1974:IVQ to 1991:IIIQ and include 70 observations. -2.05 Critical values for: a = .01, DF = -4.09 a = .05, DF = -3.47 a = .10, DF = -3.16 NOTE: All variables are inflation adjusted. All series start in 1973:IVQ and end in 1991 :IIIQ. Dickey-Fuller tests include a constant and a time trend. Aug mented Dickey-Fuller tests include four lagged first-differences o f the depend ent variables, w hich shorten the estimation period by four quarters. SOURCE: Author’s calculations on TSP version 4.20. This, in turn, implies that the two time series Y and X do not share a common underlying trend; they tend to drift apart over the sample period. One can extend the approach to consider cointegration among three or more time-series http://fraser.stlouisfed.org/ variables, each of which is 1(1). In such a case, Federal Reserve Bank of St. Louis one adds the additional variables to the right side of the cointegrating regression (equation [151) and proceeds with the DF tests described above. The test statistic, however, is sensitive to the number of right-side variables (exclusive of the constant) in the cointegrating equation. TSP version 4.20 provides appropriate critical values, based on work by MacKinnon (1990). Causality is not an issue in cointegration tests. Consequently, the designation of depend ent and independent variables for both bivariate and multivariate tests is arbitrary. Often, how ever, the results are sensitive to the ordering of the variables in the cointegrating regression. One should test all possibilities. Results Because cointegration presumes that the series under consideration are 1(1), table 1 shows the results of applying DF and ADF tests to the indi vidual time series. All of the series except FEXP and FPUR were serially correlated, so ADF tests were appropriate in most cases. None of the re sults, after any necessary adjustments for serial correlation, reject the null hypothesis of a unit root. Cointegration is an appropriate way to pro ceed with these data. Table 2 presents the results of bivariate Engle-Granger cointegration tests. The first col umn lists the two relevant variables. The second column shows the ADF test statistics. The first sta- B T A B L E sent only the results for ADF tests. The tests find no evidence of cointegration. 2 Bivariate Engle-Granger Cointegration Tests Variables Augmented Dickey-Fuller Statistic (1974:IVQ-1991:IIIQ) DEF, LTR DEF, TWD DEF, NEX -3.55 -2.46 -3.31 -2.44 -3.19 -2.50 FEXP, LTR FEXP, TWD FEXP, NEX -0.84 -2.11 -0.84 -2.27 -1.35 -2.76 FPUR, LTR FPUR, TWD FPUR, NEX -0.83 -2.36 -0.37 -2.24 -1.14 -2.64 Critical values for: a = .01, DF = -4.56 a = .05, DF = -3.92 a = .10, DF = -3.60 NOTE: All variables are inflation adjusted. The first statistic in each pair is for the regression o f the first variable o n the second. The second statistic in each pair is for the regression of the second variable o n the first. Because serial cor relation was present in nearly all cases, I conducted ADF tests with four lagged first-differences o f the dependent variables. In the few cases where serial cor relation was not present, using ADF tests did not change any conclusions reached with a simple DF test. SOURCE: Author’s calculations on TSP version 4.20. tistic in each set is for the cointegrating regres sion (equation [1]) of the first variable from column 1 on the second variable, and the second statistic is for the cointegrating regres sion of the second variable on the first variable. Because serial correlation was a problem in nearly every case, table 2 presents only the results of the ADF test. In the few cases where serial correlation was not present, using the ADF tests did not alter any conclusions reached with the DF test. The bivariate results indicate that neither the federal deficit (DEF) nor federal expenditures (FEXP) nor federal purchases (FPUR) is cointe grated with real long-term interest rates (LTR’), with the real effective dollar exchange rate (TWD), or with real net exports (NEX). Moreover, the re sults are robust to the designation of the depend ent variable in the cointegrating regression. Table 3 presents the results of multivariate cointegration tests. In these cases, I regressed the first variable listed in the table (to the left of the semicolon) on a constant and on the remain ing three variables. Because serial correlation was again a problem in nearly all cases, I pre Interpretation of Empirical Results The empirical test found no evidence that the U.S. trade balance, long-term U.S. interest rates, and the real trade-weighted dollar have shared a common trend with the U.S. federal budget deficit or with alternative measures of federal spending during the floating-exchange-rate regime. Such results, of course, do not preclude the existence of a relationship between fiscal policies and these economic variables. Cointegration tests search for a stationary linear combination of hypothetically related vari ables. The inclusion of other variables could reveal a linear combination that is stationary. I did not, for example, include foreign variables, such as interest rates. Moreover, I did not scale the deficit relative to GNP, as many researchers do, nor have I attempted to take direct account of the level of public debt. Deficit-financed fis cal policies, when the level of public debt is very high, could have substantially different ef fects on real interest rates, exchange rates, and the trade balance than would similar policies at a low level of public borrowing. Similarly, the relationship between fiscal policy measures and the trade deficit might not be linear, and a linear approximation of that relationship might fail to show any connection at all. For these reasons, cointegration tests of times series may be sensi tive to the time period investigated. Although cointegration tests reveal long-term relationships among the hypothetically related variables, they may not find a shorter-term re lationship. I have interpreted the cointegration tests as measuring the effects of the permanent components of U.S. fiscal policies. The tempo rary aspects, as the theoretical model shows, can have different and profound effects on im portant economic variables. Boucher (1991), for example, concludes that nominal U.S. current accounts and nominal U.S. government budget deficits are not cointegrated, but using Granger causality tests, she finds evidence that U.S. government budget deficits do help to predict current account deficits. Similarly, Abell (1990) considers the twin deficit relationship in a VAR model estimated strictly over the period of the dollar’s rapid appreciation: February 1979 to February 1985. Although he does not find that budget deficits Granger-cause trade deficits over this period, he does conclude that deficits m TABLE 3 III. Conclusion Mulfivariate Engle-Granger Cointegration Tests Augmented Dickey-Fuller Statistic (1974:IVQ-1991:IIIQ) Variables DEF; LTR, TWD, NEX LTR; TWD, NEX, DEF TWD; NEX, DEF, LTR NEX; DEF, LTR, TWD -3.77 -2.94 -2.17 -2.53 FEXP; LTR, TWD, NEX LTR; TWD, NEX, FEXP TWD; NEX, FEXP, LTR NEX; FEXP, LTR, TWD -1.22 -3.27 -2.50 -2.16 FPUR; LTR, TWD, NEX LTR; TWD, NEX, FPUR TWD; NEX, FPUR, LTR NEX; FPUR, LTR, TWD -1.53 -3.77 -2.75 -2.53 Critical values for: a = .01, DF = -5.29 a = .05, DF = -4.63 a = .10, DF = -4.30 NOTE: All variables are inflation adjusted. Because serial correlation was pres ent in nearly all cases, I conducted ADF tests with four lagged first-differences o f the dependent variables. In the few cases where serial correlation was not present, using ADF tests did not change any conclusions reached with a simple DF test. SOURCE: Author’s calculations on TSP version 4.20. affect interest rates, which then influence ex change rates, which then alter the trade bal ances. 17 Hence, one should interpret the results here as a general conclusion about die relation ship between federal fiscal policies and the trade deficit during the period of floating exchange rates, rather than as a comment on fiscal policy over a subperiod, such as the early 1980s, or as a prediction about possible future effects of U.S. fis cal policies. ■ 17 Because of the enormous volume of empirical studies on the relationships among measures of fiscal policy and interest rates, ex change rates, and the trade deficit, I do not survey the literature. The over whelming conclusion from even a cursory review is that the results are mixed, with no clear pattern as to the source of the differences among the studies. In addition to articles cited in the text, other avenues for pursuing the empirical literature are the following: For results from large structural models, see Hooper and Mann (1987) and Throop (1989a, 1989b). For articles using VAR techniques, see Darrat (1988) and Rosenswelg and Tallman (1991). For some cross-country results, see Bernhelm (1988) and Laney (1984). For a look at deficits and interest rates, see Evans (1985) and Hoelscher (1986). On deficits and exchange rates, see Evans (1986) and Hutchison and Throop (1985). This paper challenges the commonly held belief that aggregate U.S. fiscal policy measures, notably the federal budget deficit, bear a simple and direct causal relationship with U.S. trade deficits in par ticular, and with U.S. interest rates and exchange rates. The simple two-period, two-country models developed here from earlier work by Frenkel and Razin (1987) illustrate a complex relationship that is dependent, in terms of both degree and direction, on the distortionary nature of taxes and on relative differences between public and private propensities to consume and to import. Although fiscal policies and the trade balance can be related, they need not be. The Engle-Granger cointegration tests, which this paper employs, find no evidence of a long term relationship between common aggregate measures of U.S. fiscal policy and real long-term interest rates, real dollar exchange rates, and real net exports. This does not mean that the large U.S. federal budget deficits of the 1980s did not contribute to the sharp deterioration of U.S. trade in the early 1980s; nor does it imply that a rising federal deficit in the 1990s will not prevent further improvements in the U.S. trade balance. The findings, however, do serve to strengthen my main proposition, that the com mon story about the simple and direct relation ship between federal fiscal policies and the trade balance is largely economic folklore. References Abell, John D. “The Role of the Budget Deficit during the Rise in the Dollar Exchange Rate from 1979-1985,” Southern Economic Jo u r n al, vol. 57, no. 1 (July 1990), pp. 66-74. Anderson, Gerald H., and Michael F. Bryan. “Foreign Capital Inflows: Another Trojan Horse?” Federal Reserve Bank of Cleveland, Economic Commentary, November 1, 1989. Aschauer, David Alan. “Fiscal Policy and the Trade Deficit,” Federal Reserve Bank of Chi cago, Economic Perspectives, vol. 10, no. 3 (May/June 1986), pp. 15-22. Bemheim, B. Douglas. “Budget Deficits and the Balance of Trade,” in Lawrence H. Summers, ed., Tax Policy a n d the Economy, vol. 2. Cam bridge, Mass.: MIT Press and National Bureau of Economic Research, 1988, pp. 1-31. 39 Boucher, Janice L. “The U.S. Current Account: A Long and Short Run Empirical Perspective,” Southern Economic Journal, vol. 58, no. 1 (July 1991), pp. 93-111. Hoelscher, Gregory. “New Evidence on Deficits and Interest Rates "Jo u rn a l o f Money, Credit, a n d Banking, vol. 18, no. 1 (February 1986), pp. 1-17. Box, George E.P., and Gwilym M. Jenkins. Time Series Analysis: Forecasting a n d Control. San Francisco: Holden-Day, 1970. Hooper, Peter, and Catherine L. Mann. “The U.S. External Deficit: Its Causes and Persist ence,” Board of Governors of the Federal Reserve System, International Finance Dis cussion Paper No. 316, November 1987. Darrat, Ali F. “Have Large Budget Deficits Caused Rising Trade Deficits?” Southern EconomicJour nal, vol. 54, no. 4 (April 1988), pp. 879-87. Dickey, David A., and Wayne A. Fuller. “Distri bution of the Estimators for Autoregressive Time Series with a Unit Root "Jo u rn a l o f the Am erican Statistical Association, vol. 74, no. 366 (June 1979), pp. 427-31. Enders, Walter, and Bong-Soo Lee. “Current Ac count and Budget Deficits: Twins or Distant Cousins?” Review o f Economics a n d Statis tics, vol. 72, no. 3 (August 1990), pp. 373-81. Engle, Robert F., and C.W.J. Granger. “Co integration and Error Correction: Representa tion, Estimation, and Testing,” Econometrica, vol. 55, no. 2 (March 1987), pp. 251-76. Evans, Paul. “Do Large Deficits Produce High In terest Rates?” Am erican Economic Review, vol. 75, no. 1 (March 1985), pp. 68-87. ______ . “Is the Dollar High Because of Large Budget Deficits?”Jo u rn a l o f Monetary Economics, vol. 18, no. 3 (November 1986), pp. 227-49. Hutchison, Michael M., and Charles Pigott. “Budget Deficits, Exchange Rates, and the Current Account: Theory and U.S. Evidence,” Federal Reserve Bank of San Francisco, Economic Review, Fall 1984, pp. 5-25. Hutchison, Michael M., and Adrian W. Throop. “U.S. Budget Deficits and the Real Value of the Dollar,” Federal Reserve Bank of San Francisco, Economic Review, Fall 1985, pp. 26-43. Koenig, Evan F. “Recent Trade and Exchange Rate Movements: Possible Explanations,” Federal Reserve Bank of Dallas, Economic Review, September 1989, pp. 13-27. Laney, Leroy O. “The Strong Dollar, the Current Account, and Federal Deficits: Cause and Ef fect,” Federal Reserve Bank of Dallas, Eco nom ic Review, January 1984, pp. 1-14. MacKinnon, James G. “Critical Values for Co integration Tests,” UCSD Economics Discus sion Paper 90-4, 1990, pp. 1-14. Rosensweig, Jeffrey A., and Ellis W. Tallman. Frenkel, Jacob A., and Assaf Razin. Fiscal Policies a n d the World Economy: An Intertemporal Approach. Cambridge, Mass.: MIT Press, 1987. Fuller, Wayne A. Introduction to Statistical Time Series. New York: John Wiley and Sons, 1976. Hill, John K. “Demographics and the Trade Bal ance,” Federal Reserve Bank of Dallas, Eco nom ic Review, September 1989, pp. 1-11. ______ . “The Trade Balance and the Real Ex change Rate,” Federal Reserve Bank of Dal las, Economic Review, November 1990, pp. 1-15. “Fiscal Policy and Trade Adjustment: Are the Deficits Really Twins?” Federal Reserve Bank of Atlanta, Working Paper 91-2, March 1991. Throop, Adrian W. “Fiscal Policy, the Dollar, and International Trade: A Synthesis of Two Views,” Federal Reserve Bank of San Francisco, Economic Review, Summer 1989a, pp. 27-44. ______ . “Reagan Fiscal Policy and the Dollar,” Federal Reserve Bank of San Francisco, Eco nom ic Review, Summer 1989b, pp. 18-26. m Third Quarter Working Papers Current Working Papers of the Cleveland Federal Reserve Bank are listed in each quarterly issue of the Economic Review. Copies of specific papers may be re quested by completing and mail ing the attached form below. Single copies of individual papers will be sent free of charge to those who request them. A mailing list service for personal subscribers, however, is not available. Institutional subscribers, such as libraries and other organiza tions, will be placed on a mail ing list upon request and will automatically receive Working Papers as they are published. ■ 9208 Generational Accounting: The Case of Italy ■ 9209 Capital Forbearance and Thrifts: An Ex Post Examination of Regulatory Gambling ■ 9210 Debt, Collateral, and U.S. Manufacturing Investment: 1954-1980 by Daniele Franco, Jagadeesh Gokhale, Luigi Guiso, Laurence J. Kotlikoff, and Nicola Sartor by William P. Osterberg by Ramon P. DeGennaro and James B. Thomson Please complete and detach the form below and mail to: Research Department Federal Reserve Bank of Cleveland P.O. Box 6387 Cleveland, Ohio 44101 Check item(s) requested Please send the following Working Paper(s): □ 9208 □ 9209 □ 9210 Send to: Please print Name Address City State Zip