The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Finance and Economics Discussion Series Division of Research and Statistics Division of Monetary Affairs Federal Reserve Board, Washington, D.C. Working Studies 1: Part 1 OPERATING PROCEDURES AND THE CONDUCT OF MONETARY POLICY: CONFERENCE PROCEEDINGS Special issue Editors: Marvin Goodfriend and David H. Small March 1993 NOTE: Working Studies are collections of staff studies organized around specific themes and issued as occasional supplements to the Finance and Economics Discussion Series. Working Studies are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staffs, by the Board of Governors, or by the Federal Reserve Banks. Upon request, single copies of the paper will be provided. References in publications to the Finance and Economics Discussion Series (other than acknowledgement by a writer that he has access to such unpublished material) should be cleared with the author to protect the tentative character of these papers. Operating Procedures and the Conduct of Monetary Policy: Conference Proceedings ABSTRACT The Federal Reserve System, through its Committee on Financial Analysis, sponsored a conference on monetary policy operating procedures and strategies which was hosted by the Federal Reserve Bank of St. Louis on June 18-19, 1992. An update on the academic and System staff views of such issues seemed desirable because it had been a decade since the last broad examination of operating procedures, which appeared in "New Monetary Control Procedures". Since that time, the Federal Reserve has shifted from reserve-based operating procedures tied closely to movements in transactions money to discretionary changes in reserve conditions keyed to a variety of indicators of developments in broad money and credit, financial markets more generally, and the economy. Meanwhile, the operating procedures of foreign central banks have also been evolving, perhaps with implications and lessons for the Federal Reserve. The conference was intended to review these developments and relate them to achievement of longer-term objectives for the United States economy, and to stimulate further thinking and research on topics related to the design and execution of monetary policy. To these ends, System economists prepared and presented the papers in this volume, which were reviewed by discussants. Professor Bennett McCallum of Carnegie-Mellon University and Professor John Taylor of Stanford University were invited to discuss specific papers and provide overviews of the entire conference proceedings. Their overviews appear at the end of this volume. The conference was organized by Al Broaddus (then Director of Research and currently President) and Marvin Goodfriend of the Federal Reserve Bank of Richmond and by David Lindsey (Deputy Director) and David Small of the Division of Monetary Affairs of the Board of Governors of the Federal Reserve System. pr7'*»*jmf<*. /7**L Donald L. Kohn Director Division of Monetary Affairs Board of Governors of the Federal Reserve System April 2, 1993 1. Board of Governors, 1981. A more narrowly focused study was conducted by the Federal Reserve Bank of New York in "Intermediate Targets and Indicators for Monetary Policy: A Critical Survey", 1990, TABLE OF CONTENTS VOLUME 1 Session 1. Historical Overview Ann-Marie Meulendyke "Federal Reserve Tools in the Monetary Policy Process in Recent Decades" Comments: Robert L. Hetzel Marvin Goodfriend "Interest Rate Policy and the Inflation Scare Problem: 1979-1992" Comments: R. Alton Gilbert Session 2. International Comparisons John Morton and Paul Wood "Interest Rate Operating Procedures of Foreign Central Banks" Bruce Kasman "A Comparison of Monetary Policy Operating Procedures in Six Industrial Countries" Comments (on both John Morton and Paul Wood and on Bruce Kasman) Stephen A. Meyer Robert B. Kahn and Linda S. Kole "Monetary Transmission Channels in Major Foreign Industrial Countries" Comments: Craig S. Hakkio Session 3: Time Series Econometric Issues William Roberds, David Runkle, and Charles H. Whiteman "Another Hole in the Ozone Layer: Changes in FOMC Operating Procedure and the Term Structure Comments: Glenn D. Rudebusch Charles Evans, Steven Strongin, and Francesca Eugeni "A Policymaker's Guide to Indicators of Economic Activity" Comments: Richard W. Kopcke -2- Session 4: Operating Issues Related to Banking Wilbur John Coleman II, Christian Gilles, and Pamela Labadie "Discount Window Borrowing and Liquidity" Comments: Michael Dotsey Kenneth N. Kuttner "Credit Conditions and External Finance: Interpreting the Behavior of Financial Flows and Interest Rate Spreads" Comments: David Wilcox VOLUME 2 Session 5: Reserve Targeting, Interest Rate Targeting, and Term Structure Volatility Joseph E. Gagnon and Ralph W. Tryon "Price and Output Stability Under Alternative Monetary Policy Rules" Comments: Satyajit Chatterjee Steven Russell "Monetary Policy Experiments in a Stochastic Overlapping Generations Model of the Term Structure" Comments: Eric M. Leeper Jeffrey Fuhrer and George Moore "Inflation Persistence" Comments: John B. Taylor Session 6: Adapting to Regulatory Change Allan D. Brunner and Cara S. Lown "Implementing Short-Run Monetary Policy with Lower Reserve Requirements" Comments: Edward J. Stevens John Wenninger and William Lee "Federal Reserve Operating Procedures and Institutional Change" Comments: Daniel L. Thornton -3- Session 7: Feedback Rules for Monetary Policy John P. Judd and Brian Motley "Controlling Inflation with an Interest Rate Instrument" Comments: Evan F. Koenig Gregory D. Hess, David H. Small, and Flint Brayton "Nominal Income Targeting with the Monetary Base as Instrument: An Evaluation of McCallum's Rule" (with an appendix by Richard D. Porter) Comments: Bennett T. McCallum Summary and Overview John B. Taylor "New Directions in Monetary Policy Research: Comments on the Federal Reserve System's Special Meeting on Operating Procedures" Bennett T. McCallum "Concluding Observations" LIST OF CONTRIBUTORS CONTRIBUTORS AFFILIATION Brayton, Flint Board of Governors Broaddus, Al Federal Reserve Bank of Richmond Brunner, Allan D. Board of Governors Chatterjee, Satyajit Federal Reserve Bank of Philadelphia Coleman, Wilbur John, II Board of Governors Dotsey, Michael Federal Reserve Bank of Richmond Eugeni, Francesca Federal Reserve Bank of Chicago Evans, Charles Federal Reserve Bank of Chicago Fuhrer, Jeffrey Federal Reserve Bank of Boston Gagnon, Joseph E. Board of Governors Gilbert, R. Alton Federal Reserve Bank on St. Louis Gilles, Christian Board of Governors Goodfriend, Marvin Federal Reserve Bank of Richmond Hakkio, Craig S. Federal Reserve Bank of Kansas City Hess, Gregory D. Board of Governors Hetzel, Robert L. Federal Reserve Bank of Richmond Judd, John P. Federal Reserve Bank of San Francisco Kahn, Robert B. Board of Governors Kasman, Bruce Federal Reserve Bank of New York Koenig, Evan F. Federal Reserve Bank of Dallas Kohn, Donald L. Board of Governors Kole, Linda S. Board of Governors Kopcke, Richard W. Federal Reserve Bank of Boston Kuttner, Kenneth N. Federal Reserve Bank of Chicago Labadie, Pamela Board of Governors Lee, William Federal Reserve Bank of New York Leeper, Eric M. Federal Reserve Bank of Atlanta Lindsey, David E. Board of Governors Lown, Cara S. Federal Reserve Bank of New York McCallum, Bennett T. Carnegie-Mellon University Meulendyke, Ann-Marie Federal Reserve Bank of New York Meyer, Stephen A. Federal Reserve Bank of Philadelphia Moore, George Board of Governors Morton, John Board of Governors Motley, Brian Federal Reserve Bank of San Francisco Porter, Richard D. Board of Governors Roberds, William Federal Reserve Bank of Atlanta Rudebusch, Glenn D. Board of Governors Runkle, David Federal Reserve Bank of Minneapolis Russell, Steven Federal Reserve Bank of St. Louis Small, David H. Board of Governors Stevens, Edward J. Federal Reserve Bank of Cleveland Strongin, Steven Federal Reserve Bank of Chicago Taylor, John B. Stanford University Thornton, Daniel L. Federal Reserve Bank of St. Louis Tryon, Ralph W. Board of Governors Wenninger, John Federal Reserve Bank of New York Whiteman, Charles H. University of Iowa Wilcox, David Board of Governors Wood, Paul Board of Governors FEDERAL RESERVE TOOLS IN THE MONETARY POLICY PROCESS IN RECENT DECADES Ann-Marie Meulendyke1 Most students of money and banking in the United States would identify open market operations, reserve requirements, and the discount rate as the basic tools of monetary policy. They would add that open market operations are the primary, most actively employed tool because of their flexibility and ease of use. The historical roles of open market operations in the conduct of monetary policy under the guidelines established by the Federal Open Market Committee (FOMC) were examined in some detail in an earlier article by the author.2 This article provides parallel treatment for reserve requirements and the discount window. Both articles focus on the years since the 1951 Treasury- Federal Reserve Accord, an agreement that freed the Federal Reserve from the obligation to peg interest rates on U.S. Treasury debt and enabled it to resume an independent monetary policy. Before beginning the review of reserve requirements and the discount window, it may be helpful to summarize the main findings on open market operations. Since the Accord, the FOMC has used various money and credit measures, as well as assessments of the underlying economic and price picture, as intermediate objectives to guide the settings of its operating instruments. Reserve measures and interest rates have alternated as the FOMC's primary guide for day-to-day operations. In the first two decades after the Accord, the Trading Desk at the New York Federal Reserve Bank carried out the FOMC's instructions 1. Manager and Senior Economist, Open Market Department, Federal Reserve Bank of New York. Ted Tulpan provided excellent research assistance. The author wishes to thank Peter Sternlight, Betsy White, John Wenninger, Bruce Kasman, and Spence Hilton of the New York Federal Reserve and Robert Hetzel of the Richmond Federal Reserve for helpful comments on an earlier draft. 2. Ann-Marie Meulendyke, "A Review of Federal Reserve Policy Targets and Operating Guides in Recent Decades," Intermediate Targets and Indicators for Monetary Policy: A Critical Survey, Federal Reserve Bank of New York, July 1990. Reprinted from Federal Reserve Bank of New York Quarterly Review, vol. 13, no. 3 (Autumn 1988), pp. 6-17. Meulendyke for achieving the desired average behavior of various measures of bank credit. Operating decisions were keyed to free reserves--reserves in excess of those needed to meet reserve requirements less reserves borrowed at the discount window--and to the tone and feel of the money markets. By the 1970s, the monetary aggregates had replaced credit measures as intermediate targets and the day-to-day emphasis shifted toward controlling the overnight interbank rate, called the federal funds rate. During the 1970s, adjustments to the federal funds rate were generally small, and at times there was a reluctance to make necessary increases in the rate. Partly as a result, money growth persistently exceeded its targets, and inflationary pressures reached clearly unacceptable levels by the latter part of the decade. changed its approach to policy. In 1979, the FOMC Under the new procedures, it targeted levels of nonborrowed reserves, a measure that was closely linked through reserve requirement ratios to desired growth rates of a narrowly defined measure of money, Ml. In addition, it allowed the federal funds rate to move over a much wider range than before to increase the likelihood that money growth would be brought under control. Although these procedures contributed to increased fluctuations in both money and interest rates, they did help to bring down average money growth and inflation. At the same time, however, the creation of money substitutes and the deregulation of interest rates were making Ml a less reliable guide to future behavior of economic activity and prices. Consequently, the FOMC changed procedures once again late in 1982, adopting a borrowed reserve procedure resembling the free reserve technique of the 1960s. The degree of reserve pressure--defined as the volume of reserves that banks as a group were forced to borrow at the discount window--was adjusted judgmentally when developments in the economy, money, or prices suggested that a change was appropriate. Over time, the borrowing relationship that underpinned this approach has become less dependable. Consequently, the Desk has once again come to rely more closely on the behavior of the federal funds rate, although the rate has not become a formal target. -2- Meulendyke RESERVE REQUIREMENTS This section reviews the various roles of reserve requirements in the monetary policy process. It describes how the monetary authorities, charged with determining appropriate reserve requirements, have responded to the distinct and sometimes conflicting interests of the Federal Reserve, the banks, and the Treasury. Particular attention is given to the different parties' views of the optimal level of reserve requirements. sought to minimize reserve requirements. Historically, banks have Because the reserves that banks must hold against their deposits do not pay interest, the requirements act as an implicit tax on deposit creation. By contrast, the Treasury has sometimes resisted efforts to lower requirements because reserves provide it with an indirect source of revenue. The effective tax is sensitive to both the level of required reserves and of interest rates and has consequently been subject to considerable variation over time. The Federal Reserve, approaching the issue from a somewhat different perspective than either the Treasury or the banks, has viewed requirements as a mechanism that can help to stabilize the demand for reserves. It has sought to make them high enough to promote that stability but low enough to minimize the distortions in resource allocation that inevitably accompany any tax. The Board's most recent cuts in requirements were intended to reduce the implicit tax on banking. The lowered requirements reduced the effective tax to less than $1 billion; it helped depositories improve earnings and deal more effectively with both strains on their capital and dramatically increased insurance premia. Along with their desirable effects, however, the recent reductions brought required reserves to levels that no longer met many banks' reserve needs for clearing purposes. Consequently, the total demand for reserves became more difficult to predict, and the use of open market operations became more complicated. The history of reserve requirements since the 1951 Accord encompasses numerous regulatory changes and legislative initiatives that attempted to address these conflicting interests. Effective required reserve ratios have been cut substantially on balance over the years, -3- Meulendyke both to reduce the distorting impact of the implicit tax on the behavior of banks and their customers and to change reserve pressures. Reserve requirements since the Accord are shown in Chart 1. Required reserve balances at the Federal Reserve are currently very similar in level to those of the early 1950s despite the massive growth in deposits over the intervening decades. The Roles of Reserve Requirements Over the years, analysts have attributed several different roles to reserve requirements in the policy process. The literature since World War II has most commonly cited two--money control and revenues for the Treasury.3 Reserve requirements could affect the process of monetary control by their existence and through changes in the mandated ratios of reserves to deposits. The existence of requirements provides the linkage that allows changes in reserve levels, accomplished through open market operations, to encourage a change in monetary deposits. In theory, in a system where required reserves are a specified fraction of deposits, an increase in the amount of reserves provided to the banking system should be associated with an increase in reservable deposits in an amount that is a multiple of the reserve increase. The size of the multiple would be the inverse of the required reserve ratio, as in the classic textbook reserve multiplier process. In practice, the relationships linking reserves and deposits are far from precise, partly because not all deposits are subject to the same reserve requirement ratios and partly because excess and borrowed reserve levels can vary. The primary direction of causality linking deposits and reserves will depend upon the Federal Reserve's guidelines for reserve provision. Regardless of its operating procedures, the Fed has found the existence of reserve requirements to be a valuable tool of monetary policy because 3. See Marvin Goodfriend and Monica Hargraves, "A Historical Assessment of the Rationales and Functions of Reserve Requirements," Federal Reserve Bank of Richmond Review, March-April 1983, for an excellent review of the rationales for reserve requirements. -4- Meulendyke requirements contribute to a stable demand for reserves.4 A number of observers have argued that reserve requirements are not essential because banks would demand reserves in any case to settle transactions with other banks and to avoid overdrafts.5 Many Federal Reserve commentators have rejected this claim, contending that the voluntary demand for reserves would probably not be stable in the absence of requirements because the banks would always be trying to minimize excess reserves but would have varying degrees of success depending on each period's reserve flows.6 The Board of Governors of the Federal Reserve System may also change reserve requirement ratios to influence monetary policy. To force a contraction in deposits, the Board can raise requirements; to encourage more expansion, it can lower requirements. Although such measures may accomplish desired adjustments in reserve availability, they tend to be a blunt instrument, not well suited to fine tuning. The Federal Reserve discovered that problem in the 1930s, when legislation first gave it the power to change reserve requirements. In recent decades, it has generally used open market operations to cushion the immediate impact of a reserve requirement change. As noted earlier, reserve requirements have also been seen as a source of revenue for the Treasury since they represent an implicit tax 4. Gordon H. Sellon, Jr., "The Instruments of Monetary Policy," Federal Reserve Bank of Kansas City Economic Review, May 1984, pp. 3-20, discusses this issue. 5. For examples, see Deane Carson, "Is the Federal Reserve System Really Necessary?" Journal of Finance, vol. 19, no. 4 (December 1964), pp. 652-61; and Robert E. Hall, "A Free Market Policy to Stabilize the Purchasing Power of the Dollar," in Barry Seigel, ed., Money in Crisis: The Federal Reserve, The Economy, and Monetary Reform, Pacific Studies in Public Policy (Cambridge, Massachusetts: Ballinger, 1984), pp. 30321. Thomas Mayer, Monetary Policy in the United States (New York: Random House, 1968), pp. 39-43, discusses the theoretical arguments against requirements but concludes that they are useful, giving reasons similar to those cited in the text. 6. Richard D. Porter and Kenneth J. Kopecky, "The Role of Reserve Requirements as a Public Policy Tool," Conference on Reserve Requirements and the Role of the Federal Reserve System, Washington, D.C., January 18-19, 1979. -5- Meulendyke on deposit creation. Required reserves on which no interest is paid reduce bank earnings--at least to the extent that the level of reserves exceeds what banks would hold voluntarily. They also enhance the revenues of the Federal Reserve because the Fed buys interest bearing Treasury debt when it supplies the reserves. The Treasury benefits indirectly because the Federal Reserve turns its profits over to the Treasury. How burdensome a given level of requirements will be for banks depends on several factors, but especially on the level of nominal interest rates: the higher the rates, the greater the earnings forgone. Mindful of the "tax" effects of increasing reserve requirement ratios, the Federal Reserve has often turned to other tools when it wanted to tighten policy. Policy Responses to Conflicts between Treasury Revenues and Money Control. Federal Reserve and government policies toward reserve requirements from the end of World War II through 1980 were significantly influenced by ongoing strains arising from the different reserve objectives of the government, the Federal Reserve, and the banks. Membership in the Federal Reserve was voluntary for state- chartered banks, so they could escape the tax by dropping their membership. (State requirements were lower and generally could be met by maintaining balances at other banks, for which services were provided, and sometimes by holding Treasury bills, which paid interest.) The Federal Reserve wanted reserve requirements to be broad based enough to facilitate money control.7 The Fed believed that reserve requirements could be set in a way that would strengthen the linkages between reserves and money and between reserves and short-term interest rates. The existing structure encouraged departures from Federal Reserve membership that weakened those linkages. The Federal Reserve proposed two solutions to this conflict during the 1970s. First, it called for universal membership so that all banks would be subject to the Fed's reserve requirements. Second, it proposed paying interest on required reserves to offset the banks' 7. G. William Miller, "Proposals on Financial Institution Reserve Requirements and Related Issues," testimony before the House Committee on Banking, Finance and Urban Affairs, July 27, 1978. -6- Meulendyke revenue loss and to make membership in the Federal Reserve System attractive.8 The generally high nominal interest rates prevailing during the 1970s made requirements particularly onerous and increased the incentive to surrender membership. Negotiations to address these issues culminated in the Monetary Control Act of 1980 (MCA). The act extended reserve requirements to all depository institutions while allowing membership to remain voluntary. It also lowered required reserve ratios to reduce the implicit tax on member banks. Although the lower requirements helped to ease the implicit tax on banks, the exceptionally high interest rates of the early years of the 1980s lifted the implicit tax so that the potential earnings of many depositories were significantly diminished and their ability to pay competitive rates thereby constrained. Wide spreads between market rates and deposit rates encouraged depositors to move funds into instruments exempt from reserve requirements. The Federal Reserve continued to ask for the right to pay interest on required reserve balances (in conjunction with allowing interest on demand deposits) but its appeals were not successful.9 The eight-year phase-in period for the new reserve requirement structure mandated by the MCA discouraged the Fed from making changes in requirements for monetary policy purposes. The role of requirements in money control continued to be discussed; it was especially important between 1979 and 1982 when the Fed was seeking to control Ml by adjusting nonborrowed reserves.10 Thereafter, as the Fed moved away 8. Both the Federal Reserve's proposals for legislation and some alternative proposals appear in Miller, "Proposals on Financial Institution Reserve Requirements." 9. See statement by J. Charles Partee before the Subcommittee on Financial Institutions Supervision, Regulation and Insurance of the House Committee on Banking, Finance and Urban Affairs, October 27, 1983, reported in the Federal Reserve Bulletin, November 1983, pp. 850-51. 10. To improve the linkage between reserves and deposits, the Federal Reserve did switch from lagged reserve accounting to almostcontemporaneous reserve accounting, a change that was announced in 1982 but not put into effect until 1984. -7- Meulendyke from Ml control, the reserve-Ml linkage received less attention. Nevertheless, even now the linkage is used to forecast required reserves and banks' demand for reserves. Required Reserves and Their Role in Bank Liquidity. In the nineteenth and early twentieth centuries, most analysts believed that an important function of required reserves was providing liquidity to the banks. Most postwar commentary on reserve requirements has, however, downplayed the idea. Many writers have pointed out that if banks have to hold reserves to meet requirements, they cannot simultaneously use those reserves to make loans or handle unexpected withdrawals.11 That conclusion is almost certainly appropriate when the object is to provide liquidity over time. Nonetheless, reserve balances do provide a very important form of liquidity for periods shorter than the time interval over which requirements must be met on average (one or two weeks in recent decades). These balances constitute a clearing mechanism for interbank check and wire transfers. Far from being sterile balances sitting idly at the Federal Reserve, as they are described in many textbooks, reserves actually flow from one depository institution's account to another's many times a day. The short-run liquidity role of reserve requirements garnered some attention within the Federal Reserve during the 1980s. At that time, the Fed was seeking an explanation for observed increases in excess reserves.12 Understanding the importance of the Fed's findings 11. Before the founding of the Federal Reserve, there was no regular mechanism to produce extra reserves to meet seasonal credit needs. Small banks kept part of their reserves in the form of deposits at large banks and used those reserves to meet their seasonal needs. The withdrawal of interbank deposits from the large cities actually extinguished reserves, forcing interest rates to climb sharply higher at those times. These liquidity problems have been widely discussed. See, for instance, Thomas Mayer, James S. Duesenberry, and Robert Z. Aliber, Money, Banking, and the Economy, 3d ed. (New York: W.W. Norton and Company, 1987), pp. 28-29. 12. The large volumes of daylight overdrafts also alerted the Federal Reserve to some banks' heavy dependence on reserve balances for clearing activities. -8- Meulendyke requires a brief review of the composition and uses of required reserves.13 Since 1959, banks have been able to satisfy reserve requirements by holding vault cash and reserve balances at the Federal Reserve. Beginning in 1968, the vault cash applied to meeting reserve requirements in the current period was the vault cash banks had held in an earlier period. Consequently, vault cash could not play a role in meeting the banking system's marginal reserve requirements once a reserve maintenance period began. Since the reserve requirement restructuring of the 1980s, many depository institutions, including small commercial banks, thrifts, and credit unions, were able to meet their reserve requirement with vault cash alone. It does not appear, however, that the requirements determine the institutions' holdings of vault cash; instead these institutions base their holdings on anticipated customer demands for currency and a strong preference not to be embarrassed by shortages of cash. For institutions that consistently meet or more than meet their reserve requirements with vault cash ("nonbound" institutions), reductions in the level of the requirements are of no consequence.14 Those medium and large depository institutions that do not cover their whole requirement with vault cash ("bound" institutions) have to hold on average during each reserve maintenance period sufficient reserve balances at the Federal Reserve to meet the remainder of their requirement (called required reserve balances). But those reserve balances also serve as the means of payment for the clearing and settlement process. Any depository that does even a portion of its own clearing of checks or funds wires has to maintain a reserve balance to facilitate that clearing. The volume of transactions executed each day using reserve accounts as a means of payment has long been high relative to the 13. The following discussion draws heavily from Ann-Marie Meulendyke, "Monetary Policy Implementation and Reserve Requirements," internal working paper, September 1992, pp. 3-5. 14. The Federal Reserve excludes surplus vault cash from its measures of total and nonborrowed reserves. -9- Meulendyke balances held in the accounts. turn over many times a day. For many depositories, reserve balances That turnover rate has had an upward trend. The trend reflects cuts in reserve requirements that occurred between 1980 and 1984, and again in 1990 and 1992, and increases in the volume of transactions being processed by the Federal Reserve.15 3 show recent patterns in these measures.16 Charts 2 and The daily flows have a large predictable component, but considerable potential for surprise remains. The Federal Reserve generally processes instructions to pay out reserve balances even if the action puts the sending bank into overdraft. The Fed imposes a penalty charge on any institution that ends the day overdrawn. Consequently, depository institutions have to aim for a significant positive end-of-day balance to minimize the risk of an inadvertent overdraft, regardless of their reserve requirements. When reserve requirements were reduced, it became more common for precautionary needs to exceed required reserve balances. Depository institutions can deal with these additional precautionary reserve needs by holding excess reserves, but this strategy is costly since no interest is paid on reserves. When required reserve balances declined in the early 1980s and again at the end of 1990, depositories continued to try to minimize excess reserve holdings, but they were restricted in their ability to do so as the difficulties in avoiding overnight overdrafts became more severe. If they ended up with excess reserves, they might not be able to work them off later in the same maintenance period without risking being overdrawn. In trying to cope with the narrowing of ranges of reserve balances that were 15. Since 1980, depositories have been able to establish required clearing balances to provide some reserve management flexibility. These are additional reserve balances that depositories agree in advance to hold. In return, they receive credits to pay for priced Federal Reserve services. The level of priced services used by a depository provides an effective maximum demand for required clearing balances. Required clearing balances were fairly small until after the 1990 cut in reserve requirements, when many large banks started to hold them. 16. Fedwire transactions have the largest impact on reserve balances, but other wire transfer operations and check processing transactions also lead to reserve transfers. These other transactions raise the turnover rate for reserve balances even further. -10- Meulendyke acceptable in the management of reserves, depositories devoted considerable resources to monitoring internal reserve flows. In the process, they became less tolerant of excess reserves early in maintenance periods because of their diminished ability to work them off in subsequent days. These developments restricted the depositories' day-to-day flexibility in managing reserves, caused more frequent bulges in excess reserves, and added to end-of-day volatility in the federal funds rate. Reserve Requirements in the 1950s and Early 1960s At the time of the Treasury-Federal Reserve Accord of 1951, reserve requirement ratios on demand deposits of Federal Reserve member banks were 24 percent for banks located in "central reserve cities" (New York and Chicago), 20 percent for member banks in "reserve cities" (other cities with Federal Reserve Banks or branches), and 14 percent for "country banks" (the term for all other member banks). The reserve ratio for time and savings deposits was 6 percent for member banks in all locations. During the fifteen years between 1951 and 1966, requirements were raised on five occasions and were lowered ten times.17 The changes in reserve requirements were sometimes made in conjunction with complementary changes in the discount rate, while at other times the moves were made independently. Open market operations were used to cushion the changes in reserve requirements, so that hardly any of the immediate impact of the reserves released or absorbed was felt as a change in excess or borrowed reserves. In those years, the Federal Reserve formally described reserve requirements as a policy tool used to make reserves more or less plentiful so as to alter credit availability and money market interest 17. Reserve requirement ratios were changed for several reasons over these years. Although many of the changes were undertaken to make reserves more or less costly as part of the monetary policy process, changes were also made to meet seasonal reserve demands and to implement the 1959 legislation aimed at equalizing reserve ratios at central reserve and reserve city banks. In addition, ratios were slightly modified in 1966 when tranches were introduced for both demand and time deposits. At the same time, savings accounts were separated from time deposits for required reserve calculations. -11- Meulendyke rates--the near-term policy goals of the time.18 Its decisions about reserve requirements were, in practice, constrained by the exodus of small banks from the Federal Reserve System in the 1950s. Legislation passed in 1959 addressed an apparent inequity between large and small banks in an attempt to make membership more attractive for the small banks. Country banks had lower nominal reserve requirements, but they often had to tie up relatively large sums in non-interest-earning balances that did not serve any other purpose. generally handled payment clearing for them.) (A reserve city bank Because of their customer bases, most country banks had to hold relatively high amounts of vault cash, but they could not use these holdings to satisfy requirements. The 1959 act permitted the Fed to count vault cash toward meeting reserve requirements. That change--implemented in three steps during 1959 and 1960--reduced effective requirements, especially for country banks. It was hoped that the lower requirements would encourage those banks to remain members of the Federal Reserve. Contemporary Views of Reserve Requirements. A commonly held view about reserve requirements was expressed by a presidential commission appointed in 1963 to study financial institutions. The commission concluded that "there is, within broad limits, little basis for judging that in the long run one level [of reserve requirement ratios] is preferable to another in terms of facilitating monetary policy."19 The commission felt that the effects of requirements on bank earnings and Treasury revenues should be the primary factor considered in choosing reserve ratios. While it saw the advantages to bank profitability of a significant cut, it believed that the cost to the Treasury would be too great. Some academic literature of the time offered other views on reserve requirements and monetary control. Several articles and books dealt with the concept of fractional reserve requirement ratios and 18. Board of Governors of the Federal Reserve System, Report, various years. Annual 19. Report of the Committee on Financial Institutions to the President of the United States. Walter W. Heller, Chairman. Washington, D.C.: Government Printing Office, 1963, p. 12. -12- Meulendyke described the strengths and weaknesses of that structure. analyzed the tax implicit in reserve requirements. 20 Tolley He suggested that the level of reserve requirement ratios and hence of the amount of the tax had come about by accident. for such a tax. He then tried to establish a rationale He believed that under a gold standard, a system in which real resources had to be devoted to producing money, a fee was appropriate to encourage people to economize on the use of money. But when the cost of producing money is trivial, as it is with fiat money, the only justification for a charge is that the government could benefit from the revenues arising from the Federal Reserve's provision of reserves. Tolley went on to observe, however, that the government's gains would cause misallocation of resources as banks took actions to reduce the effect of the tax. low reserve requirements. Such a distortion would argue for very But Tolley thought very low requirements might make monetary control difficult because shifts between currency (which is effectively subject to a 100 percent reserve requirement) and deposits would have a large impact on the amount of money created, as would mistakes in estimating reserve provision. Hence, he recommended that interest be paid on required reserves so that requirements would not need to be reduced. Friedman also discussed how shifts in preferences between currency and deposit holdings could ease or tighten reserve conditions.21 He reiterated the arguments from the 1930s for 100 percent reserve requirements. Such requirements had been proposed as a solution to the unpredictable multiplier effects of fractional reserve accounting arising from the differential treatment of deposits and currency. Friedman also recognized the undesirable tax effect of 100 percent requirements and described the inevitable incentive for money and credit provision to move outside the regulated area of banking. combat that problem, he recommended paying interest on reserves. To Later, 20. George S. Tolley, "Providing for Growth of the Money Supply," Journal of Political Economy, December 1957, pp. 477-85. 21. Milton Friedman, A Program for Monetary Fordham University Press, 1959), pp. 65-76. -13- Stability (New York: Meulendyke the Federal Reserve seriously considered the proposal to pay interest on reserves; it has periodically requested authority to do so from the Congress. Reserve Requirements in the Latter Part of the 1960s and 1970s Reserve requirements continued to be raised and lowered to reinforce tightening or easing moves implemented with other tools during the rest of the 1960s and 1970s. Requirements were increased four times and decreased seven times during these years.22 Sensitivity to the membership problem sometimes made the Federal Reserve Board hesitant to raise requirements. On occasion, the Board raised them just on large time deposits--deposits mostly issued by the large banks, which were the least able to give up the services provided by Fed membership. The combination of higher inflation and higher interest rates that emerged during these years drew increasing attention to the tax burden of reserve requirements and the related question of differential treatment of member and nonmember banks. The Federal Reserve appointed a study group headed by Robert Black to review reserve requirement ratios. The group reported its recommendations in 1966.23 The primary result of that study was the decision to move from near-contemporaneous reserve requirements with one-week reserve maintenance periods for reserve city banks and two-week periods for country banks to weekly reserve periods for all member banks with a two-week lag between the computation and maintenance periods. This change was believed to make calculating requirements easier for the banks and the New York Fed's Trading Desk.24 22. The count does not include the 1972 restructuring that raised requirements for some banks and lowered them for others, as described later in the text. 23. Proposals, Robert P. Black, Report of the Ad Hoc Subcommittee on Reserve May 13, 1966. 24. The other change was to permit banks to carry forward reserve excesses up to 2 percent of required reserves for one reserve period. (Banks already had the authority to carry forward 2 percent of reserve deficiencies.) -14- Meulendyke Lagged reserve requirements weakened the direct linkage between reserves and money, making it harder, in theory, to manipulate reserves as a means of controlling money. For the most part, the Federal Reserve did not see any reason to be concerned because it was not attempting to control money in this way. Instead, the Fed was attempting to affect money growth indirectly by influencing the demand for money. It altered the cost of obtaining reserves and hence the cost at which credit was provided.25 In 1972, another Federal Reserve reform addressed the problem of retaining member banks. For both reserve city and country banks, reserve requirement ratios were to be graduated on the same schedule by volume of deposits. The change represented a significant cut in reserve requirements for small banks in Federal Reserve cities and caused some large banks outside of Federal Reserve cities to face higher requirements. The series of graduated steps in the required reserve schedule further weakened the relationship between required reserves and monetary deposits, an outcome that distressed those economists who wanted to see the Federal Reserve control reserves in order to control money growth. At the time, the Federal Reserve was targeting the federal funds rate and reserve requirements were lagged, so the concerns were not immediately relevant to operations.26 Nonetheless, Federal Reserve membership continued to decline. The Federal Reserve proposed paying interest on reserves on a couple of 25. Lyle E. Gramley and Samuel B. Chase, Jr., "Time Deposits in Monetary Analysis," Federal Reserve Bulletin, October 1965, pp. 13801404. 26. Nonetheless, shortly afterwards the Federal Reserve did take limited steps to use reserve targeting when it experimented with reserves on private deposits. See Ann-Marie Meulendyke, "A Review of Federal Reserve Policy Tatgets and Operating Guides in Recent Decades," Intermediate Survey, Targets and Indicators for Monetary Policy: A Critical Federal Reserve Bank of New York, July 1990, pp. 463-64. -15- Meulendyke occasions in the 1970s to halt the decline, but the revenue loss to the Treasury engendered strong congressional opposition.27 The Monetary Control Act and Reserve Requirements in the 1980s At the end of the 1970s, the Federal Reserve once again tried to achieve universal membership. Although it did not literally accomplish that, it did achieve, through the 1980 MCA, the most important goal associated with expanded membership: the extension of reserve requirements to all depository institutions. Furthermore, the Fed was permitted to collect deposit data on an ongoing basis from all but the smallest depositories, enabling it to improve both estimates of actual money and forecasts of future money. Reserve requirement ratios for member banks were cut over a four-year period from a top rate of 16 1/4 percent to a top rate of 12 percent on transactions deposits. A low reserve tranche was also established of 3 percent on the first $25 million of deposits, with the amount allowed to rise over time.28 Nonmember banks and thrifts that faced the increases in requirements were given an eight-year phase-in period to reach the final levels of requirements specified in the act. The Federal Reserve Board retained the option to adjust reserve ratios within specified bands. The MCA was directed toward improving the Fed's ability to control money. It focused on deposits in Ml, the primary intermediate policy variable at the time. It did not, however, provide any scope for using reserves to control M2, a secondary target at the time the act was passed but the primary monetary target later in the decade. Money market mutual fund balances remained exempt, and MCA actually took away from the Federal Reserve the power to impose reserve requirements on personal time and savings deposits. Aside from the changes to reserve requirements mandated by the legislation, only minor modifications were made to reserve requirements 27. Specific proposals to pay interest on reserves were introduced in the Congress in 1977 and 1978. See Stuart E. Weiner, "Payment of Interest on Reserves," Federal Reserve Bank of Kansas City Economic Review, January 1985, pp. 20-21. 28. In 1982, the Gam-St Germain Act modified the reserve requirement structure further to introduce a zero requirement tranche. -16- Meulendyke during the 1980s.29 Because the structure of requirements had been set within specified limits by the MCA, it was generally felt that there was little point to considering policy-related changes in the ratios. Such changes would have been difficult to implement during the eight-year phase-in period. Since the legislation had not given the Federal Reserve the option to pay interest on reserve balances, the Board might have hesitated to raise requirements because of the implied increase in the tax burden.30 Furthermore, the Federal Reserve believed it could achieve its objectives just as well through open market operations and discount window policy. Excess Reserve Behavior and Potential Problems with Reserve Requirements. The Federal Reserve saw increasing evidence during the 1980s that depository institutions were having difficulty managing reserves. These observations suggested that reserve requirements might be inadequate for smooth monetary operations. Normal levels of excess reserves rose fairly steadily in the years following passage of the MCA. Some of the increase was the inevitable result of extending reserve requirements to nonmember depository institutions.31 But member bank excess reserves were also rising, in a pattern that contrasted with their behavior during much of the 1970s, when they generally hovered in a range near $200 million. discoveries. The search for explanations led to several It was observed that excess reserves tended to move inversely to required reserves not met by vault cash, both period to 29. In March 1983, the Board eliminated reserve requirements on time deposits with an initial maturity of two and one-half years or more. In September 1983, it reduced the minimum maturity for exemption from requirements to eighteen months. 30. The MCA did provide for payment of interest on supplemental reserve requirements if such requirements were needed for monetary control. The provision has not been used. 31. At some point during the phase-in period, vault cash no longer met all of the larger nonmember institutions' requirements, and they opened reserve accounts at the Federal Reserve. Only then could these institutions have excess reserves. (Previously, they may have had excess reserves from their own perspective in the form of surplus vault cash and deposits at correspondents, but the Federal Reserve does not count these in its reserve measures.) -17- Meulendyke period and over time, as balances held at Federal Reserve Banks trended lower.32 The sharp drop in required reserve balances between 1980 and 1984 occurred as lower reserve requirements were being phased in for member banks under MCA and the spread of automatic teller machines was encouraging rapid expansion of vault cash holdings (Chart 1 ) . Average required reserve balances rose again in the next few years, but excess reserves continued to expand at member banks as well as at nonmember banks. Conversations with officials at a number of banks underscored the growing role of large payments flowing through their reserve accounts. The volume of wire transfers over Fedwire--the Federal Reserve's wire transfer system--grew rapidly (Chart 3), making it increasingly difficult for banks to predict reserve balances. Since the Federal Reserve penalized end-of-day overdrafts, banks had to be careful not to aim for too low a reserve balance lest an unexpected late day outflow (or an expected receipt that did not arrive) should leave them overdrawn. These discoveries suggested that for a number of banks, reserve balances needed to meet requirements were not very different in size from those needed to manage clearing and settlement and to avoid overdrafts. These factors were taken into account by the Federal Reserve in estimating the aggregate demand for excess reserves.33 But they did not lead to serious discussions of the structure of reserve requirements during the 1980s. Cuts in Reserve Requirements in the 1990s The Federal Reserve Board eliminated reserve requirements on nontransaction deposits at the end of 1990. In explaining its action, the Board indicated that the existing structure had been designed 32. David Jones, "Excess Reserves under MCA," November 10, 1983, and David Small and Brian Madigan, "An Analysis of Excess Reserves," July 1, 1986, internal memoranda, Board of Governors of the Federal Reserve System. 33. Markets, Ann-Marie Meulendyke, U.S. Monetary Policy and Financial Federal Reserve Bank of New York, 1990, chap. 6. -18- Meulendyke "primarily to permit greater precision of monetary control when policy focused on reserve aggregate targeting." It went on to describe the changing conditions that had prompted its move: In subsequent years, as the Federal Reserve, moved away from the procedures in effect in the early 1980s, which required a broad reserve base, reserve requirements on nonpersonal time accounts have become somewhat of an anachronism. Moreover, the current 3 percent requirement has placed depository institutions at a disadvantage relative to other providers of credit, spawning efforts to circumvent the requirement. The Board took action at this time also in response to mounting evidence that commercial banks have been tightening their standards of creditworthiness [a development that] has in recent months begun to exert a contractionary influence on the economy. . . . Lower reserve requirements at any given level of money market interest rates will reduce costs to depository institutions, providing added incentive to lend to creditworthy borrowers.3A The reduction in reserve requirements boosted earnings for some depository institutions but, as indicated earlier, it had the undesirable side effect of complicating reserve management for many institutions. With lower routine levels of required reserve balances, their ability to accept reserve variability from day to day within a two-week reserve maintenance period without either incurring an expensive overdraft or being stuck with unusable excess reserves was reduced. Depositories found they had to use considerable resources to hold down excess reserves. The action also complicated operations of the Open Market Trading Desk at the New York Federal Reserve Bank. Relatively modest reserve excesses often inspired sharp declines in the federal funds rate, even on days that were not the ends of maintenance periods. Depositories had less ability to absorb and make use of the excess reserves because they could not run large deficiencies in subsequent days without ending overdrawn. When a number of depositories discovered toward the end of a day that they had excess reserve positions and tried to sell the funds into the interbank federal 34. Federal Reserve Bulletin, February 1991, p. 95. -19- Meulendyke funds market, their efforts often pushed the funds rate down sharply, sometimes almost to zero. At that time of day, it is too late for open market operations to be undertaken to affect that day's reserves. Hence, depositories as a group could not eliminate the excesses except by repaying discount window loans. In 1991, routine borrowing from this source was already at very low levels, so little could be repaid. Low reserve balances also increased the likelihood of an incipient overdraft. Depositories that discovered they were overdrawn late in the day generally tried to cover the overdrafts by borrowing in the federal funds market. If funds were scarce systemwide, sufficient reserves might not be available. Depositories could obtain reserves from the discount window, but in the early months of 1991, many banks were unusually reluctant to borrow for fear that such a step could be read as a sign that they were in trouble. That reluctance to borrow often caused federal funds to be bid to very high levels before some banks finally turned to the window to cover the shortages. The Desk's Approach to Managing Reserves in this Environment. At the time of the 1990 reserve requirement cut, the Desk was formally targeting borrowed reserves.35 Because the relationship between borrowing and the funds rate remained unreliable, however, the Desk was also taking considerable guidance from the federal funds rate. The Desk still attempted to achieve the levels of nonborrowed reserves believed consistent with demands and the desired degree of reserve pressure, but demands became harder to gauge after the cut in requirements.36 In choosing its reserve management strategy, the Desk had traditionally focused on two-week average reserve levels that banks had to hold over 35. Ann-Marie Meulendyke, "A Review of Federal Reserve Policy Targets and Operating Guides in Recent Decades," Intermediate Targets and Indicators for Monetary Policy: A Critical Survey, Federal Reserve Bank of New York, July 1990, describes the formal procedures. Recent modifications are discussed in "Monetary Policy and Open Market Operations during 1990," Federal Reserve Bank of New York Quarterly Review, Spring 1991, pp. 66-74. 36. See "Monetary Policy and Open Market Operations during 1991," Federal Reserve Bank of New York Quarterly Review, Spring 1992, pp. 80-88. -20- Meulendyke the maintenance period as a whole, although it had also made efforts to avoid extreme movements in daily reserve levels. Once reserve requirements were reduced, the Desk had to pay increased attention to daily levels because of the depositories' diminished tolerance for being short or long relative to their requirements. The Desk often found that the funds rate in the morning was not a good guide to reserve availability; the rate sometimes plunged or rose sharply late in the day when the depositories finally discovered that reserves were plentiful or scarce. Because market participants judged the Fed's policy stance by the behavior of the federal funds rate, the signaling of policy intentions sometimes conflicted with the desired reserve management strategy. If, for instance, an estimated reserve shortage coincided with a funds rate level below that perceived to be the target, the Desk had to decide whether to meet the estimated reserve need. If it met the need, it would risk giving a misleading indication that the stance of policy had been eased. But not meeting the need would increase the chances of a sharp rise in the funds rate late in the day, possibly accompanied by heavy discount window borrowing. Such greater than desired reserve pressure imposed an unintended cost on the banks and involved a risk that observers could be misled about policy. Although these conflicts had been a periodic feature of reserve management for years, they increased in frequency once levels of required reserve balances fell. Reserve balances rose during 1991, helping to ease somewhat the difficulties of reserve management. However, another cut in reserve requirement ratios in April 1992 once more lowered the range of flexibility in day-to-day management of reserves, although typical reserve balance levels remained above those of the early part of 1992.37 37. A series of papers prepared by the staff of the Federal Reserve Bank of New York after the 1990 cut in required reserve ratios considers the operational difficulties of low required reserve ratios and evaluates possible solutions. Overall, the papers suggest that the best solution to the reserve management problems encountered with low -21- Meulendyke THE ROLE OF THE DISCOUNT WINDOW IN POLICY IMPLEMENTATION Like reserve requirements, the discount window has played a supporting role to open market operations in the monetary policy process. This section describes the guiding principles for discount window borrowing. It reviews the two main features of that borrowing, the rules that govern the use of the facility and the rate or rates that are charged. It then provides a chronological review of developments in the behavior of borrowing from the 1950s to the present. The Philosophy behind the Discount Window Mechanism Federal Reserve views of the discount window's roles changed considerably between the founding of the Federal Reserve in 1914 and the 1930s as open market operations gradually replaced discount window borrowing as the primary source of Federal Reserve credit. Then, between 1934 and 1950, the discount window fell into disuse, and there was little consideration of the roles of the window as a policy tool. The Federal Reserve's concept of the policy role of the discount window was reexamined after the 1951 Accord and again in the latter half reserve balances would be to pay interest on reserves so that requirements could be increased without raising the costs to depository institutions. The collection of papers also evaluates other alternatives. A return to more routine use of the discount window would provide the banking system with valuable flexibility, but overcoming the current strong reluctance to borrow appears to be a difficult challenge. In the absence of such changes, only one of the other alternatives could provide more than modest help to the reserve management process: permitting banks to end the day overdrawn. Nonetheless, permitting overdrafts would have significant drawbacks. If this approach were to be seriously considered, permitted overdrafts would have to be collateralized and made subject to a modest charge. Even so, it seems to go against the thrust of efforts to reduce daylight overdrafts and could be seen as weakening the essential discipline of a reserve requirement structure. Other approaches deserving consideration include expanding reserve carryovers and shortening the vault cash lag, variants of which have recently been introduced by the Board of Governors. These approaches, however, would raise reserve management flexibility only slightly. -22- Meulendyke of the 1960s. Both studies led to some modifications in the rules for borrowing but did not change the underlying philosophy. Most of the rule changes since the early 1970s have been small and have addressed specific concerns. Since the Accord, the Federal Reserve's discount window policy has discouraged persistent reliance on borrowing. That stance has ensured that borrowed reserves generally represent only a modest share of total reserves. The Fed believes that the discount window should serve as a safety valve, a temporary source of reserves when they are not readily available from other sources.38 The window in recent decades has been available to healthy banks for occasional, but not continuous, use.39 Borrowing has been rationed through a variety of means that have encouraged a "reluctance to borrow." The degree of reluctance shown by the banks has varied considerably over the years, even in the absence of changes in the guidelines for borrowing. At the same time, the Fed has counted on there being some amount of borrowing because borrowing is an element in the reserve adjustment process. In this context, the window has played a vital role in meeting unexpected reserve needs. Various open market operating procedures depend on some degree of stability in the banks' demand for borrowed reserves, but the administrative guidelines and changing bank attitudes have made this stability difficult to achieve. For much of the time since the mid-1960s, the discount rate has been below competing market rates, in particular the overnight federal funds rate. Consequently, administrative restrictions rather than the rate have had the biggest role in limiting the amount of borrowing. Banks have responded to the profit incentive to borrow, but in doing so they have had to factor in 38. All borrowing from the Federal Reserve must be fully collateralized. 39. At times, the Fed also provides extended credit at market-based rates to banks whose financial difficulties have cut them off from regular sources of financing. Banks using the facility must work with their regulators toward a solution. That type of borrowing is not a monetary policy tool, and thus is not a focus of this piece. -23- Meulendyke some nonprice costs--such as potential loss of future access to the window--that are difficult to estimate. During the 1980s, increasing financial difficulties and bank failures led banks to become more reluctant to borrow, even under conditions that would formerly have led them to borrow. The rise in banking crises made many banks fearful that if they borrowed, rumors would start that they were in financial trouble. Thus, the demand for borrowing became even less predictable, reducing the value of the relationship between borrowing and the spread between the Federal funds rate and the discount rate. The direct cost represented by the rate charged for discount window borrowing has also played some role in the policy process. Changes in the rate have normally attracted general attention to the state of monetary policy, giving rate changes the potential for an announcement effect. The extent of the announcement effect has varied over time, depending on the verbal message given with the rate change and the way borrowing was being used in the policy process. Sometimes the Fed has sought to signal policy changes when it changed the rate. At other times it deliberately downplayed the significance of the move. Changes in the discount rate are voted by the Boards of Directors of the twelve Federal Reserve Banks and approved by the Board of Governors. The governors generally approve changes in the rate when they want to signal a change in the stance of policy or when market rates have moved significantly away from the discount rate, so that the discount rate is "catching up" with the changes. Rate changes have normally complemented the guidelines established by the FOMC for the conduct of open market operations. The discount rate per se has not, in the post-Accord period, been regarded as a primary means of influencing the amount of discount window borrowing. Indeed, because short-term interest rates have frequently exceeded the discount rate since the mid-1960s, rationing of the use of the window has had to be accomplished through means other than the rate. There have been numerous recommendations over the years that the rate be given the primary role in rationing credit, either because the approach was more straightforward and less arbitrary than -24- Meulendyke rationing administratively or because the use of a below-market rate implied a subsidy. The specifics of the relationship between the discount rate and open market policy changed modestly when the techniques of policy implementation were changed but have throughout relied on administered disincentives to borrow. The Discount Window in the 1950s Through the mid-1960s Borrowing jumped dramatically in the early 1950s. It rose from an average of $130 million in 1950 to an average of $800 million in 1952. By December 1952, it had reached $1.6 billion. Interest rates rose after the Accord, and the discount rate lagged behind. (Chart 4 shows borrowed reserves and their share of total, reserves between 1950 and 1965, along with the discount rate and short-term interest rates.) The cost structure made borrowing attractive for the first time since the early 1930s. An excess profits tax instituted in 1951 increased the incentive to use the discount window because borrowings served as an offset in computing the tax. A Federal Reserve System committee was established in 1953 to examine the history of the rationales for borrowing. The committee concluded that the established "tradition against borrowing" should be encouraged because it contributed to the soundness of individual banks and the banking system.40 The committee report served as the basis of the 1955 revisions to Regulation A, the regulation governing use of the window.A1 The report observed that the founders of the Federal Reserve had expected the discount window to be the primary source of Federal Reserve credit. In the early years of the Federal Reserve, many member banks borrowed a substantial portion of the reserves they needed from the window; indeed, it was not unusual for a bank to borrow continuously. By contrast, in the years before the founding of the Federal Reserve, a bank that was heavily dependent on borrowed funds, rather than on its own capital and deposits, was believed to be more vulnerable to failure. 40. System Committee on the Discount and Discount Rate Mechanism, "Report on the Discount Mechanism," March 12, 1954. 41. Federal Reserve Bulletin, January 1955, pp. 8-14. -25- Meulendyke The committee noted that the development of open market operations during the 1920s as an alternative source of Federal Reserve credit made possible a gradual move to discourage heavy borrowing. Once again, banks that borrowed persistently came to be seen as more likely to fail, and this view was reinforced during the early 1930s when the number of bank failures soared. Mindful of this negative image, the banks themselves became reluctant to borrow and instead built up holdings of excess reserves during the latter part of the 1930s. This course of action was simplified by the monetization of the vast gold inflows inspired by the revaluation of gold in 1934 and by the approach of war in Europe in the latter years of the decade.42 By the early 1950s, however, a decade and a half with low numbers of bank failures had apparently reduced the banks' own reluctance to borrow to such an extent that many banks were inclined to return to the window when doing so became profitable. felt this behavior should be discouraged. The committee It reiterated the belief that a bank that used its own resources to meet increased demands for credit was healthier than one that was dependent on borrowed funds. In its 1954 report, the committee recommended that routine reserve provision be accomplished almost entirely through open market operations. The report also recommended limiting the term of borrowing to fifteen days under normal circumstances. It noted that most banks had emerged from the war with substantial portfolios of government securities that could be sold to raise additional funds for seasonal or other purposes. The regulations that were subsequently adopted guided discount officers in distinguishing between appropriate and inappropriate borrowing. Borrowing was considered inappropriate when the funds were used for normal business activities. In particular, the committee disapproved of borrowing to profit from interest rate differentials. The role of the discount window during the rest of the 1950s and early 1960s generally followed the pattern set out by the committee's 42. Markets, Ann-Marie Meulendyke, U.S. Monetary 1990, Chap. 2. -26- Policy and Financial Meulendyke guidelines. There was some debate about whether the reluctance to borrow was motivated by the banks' own caution or by Federal Reserve restrictions. Some banks almost never borrowed, suggesting an internally generated reluctance. Many banks, however, apparently took account of the full cost of borrowing, including potential loss of future access, and borrowed when it was profitable. borrowing was rarely a large bargain. In that context, In fact, the discount rate was often slightly above short-term Treasury bill rates, although both borrowing and the incentive to borrow varied cyclically. Normally, borrowing was only a modest share of total Federal Reserve credit. The Board of Governors approved periodic adjustments to the discount rate and issued a statement of purpose with each adjustment. Often the changes lagged market rates, and the Board explained its action as an effort to catch up with market rates. When the discount rate was low relative to other short-term rates, borrowing often rose. (The primary alternative rate was the Treasury bill rate in the 1950s; the federal funds market grew in importance during the 1960s.) Some academic economists criticized the discount mechanism. They did not like the fact that banks were given mixed signals about borrowing, with the relatively low discount rate often encouraging use of the window while the administrative guidelines were discouraging it. They felt that the rules made it difficult to judge whether policy was tight or easy.43 The authors preferred a rate that was set above market rates--a penalty rate--but urged that no administrative restrictions be placed on borrowing. Discount Window Policy in the Late 1960s and 1970s Higher interest rate levels in the latter half of the 1960s, especially the "tight money" episode of 1966, encouraged more borrowing (Chart 5). The decline in membership was also garnering attention, and there was concern the discount window was not sufficiently available to small 43. See Milton Friedman, A Program York: Fordham University Press, 1959), Reserves and the Money Supply for Monetary Stability (New pp. 38-41; A James Meigs, Free (Chicago: University of Chicago Press, 1962); and Warren Smith, "The Discount Rate as a Credit-Control Weapon," Journal of Political Economy, April 1958, pp. 171-77. -27- Meulendyke member banks. A series of studies were undertaken during the late 1960s under the guidance of a steering committee of Federal Reserve Governors and Presidents.44 The studies reviewed the history of the discount mechanism, compared the discount window with the tools and techniques of foreign central banks, evaluated some of its problems, and presented several possible reforms. The steering committee endorsed the practice of permitting banks to borrow only intermittently. It wanted to continue the administrative disincentives to frequent borrowing, but it was troubled that some banks seemed to get little or no benefit from the window. The summary report recommended some changes to make borrowing more convenient, especially for small unit banks with large seasonal swings in loan demand and limited access to the national credit markets. The report's recommendation of a special seasonal borrowing privilege for small member banks was adopted in 1973 and remains in effect, although it has been modified somewhat in recent years.45 The report also proposed that one form of adjustment credit should consist of a basic borrowing privilege that would give all (member) banks some access at reasonable cost to Federal Reserve credit based on published guidelines for amount and frequency of borrowing. Even the proposed basic borrowing privilege did not envision continuous borrowing: if a bank needed additional credit, its borrowing would be subject to scrutiny. The approach was not adopted, although the proposed frequency schedule did influence the informal guidelines used by the discount officers in subsequent years. Finally, the study brought to light considerable inconsistencies in the administration of the window by the different Federal Reserve Banks. Efforts were made to improve coordination in order to minimize those differences. During the 1970s, Federal Reserve monetary policy focused on adjusting the federal funds rate to respond to deviations in money 44. Board of Governors of the Federal Reserve System, of the Federal Reserve Discount Mechanism, Reappraisal 1971. 45. The seasonal borrowing privilege was extended to nonmember banks under the MCA. In 1992, the Board began charging a market rate on seasonal borrowing tied to the federal funds rate and certificate of deposit rates. -28- Meulendyke growth from desired ranges. The discount window generally played a subsidiary role in the process.46 Changes in the discount rate were often motivated by changes in market rates, as they had been in earlier decades, although occasionally changes were intended to create an announcement effect.47 The amount of borrowing generally increased as the federal funds rate rose relative to the discount rate, a relationship that suggested that banks were seeking to maximize profits through their borrowing decisions. The Open Market Trading Desk took that relationship into account when choosing how many nonborrowed reserves to provide, since the amount of desired borrowing affected the reserve levels consistent with the desired funds rate. Relation between Discount Policy and Reserve Targeting from 1979 to 1982. Borrowing took on increased importance after the October 1979 changes to reserve operating procedures. Under the new procedures, the Trading Desk provided only the level of nonborrowed reserves estimated to be consistent with targeted Ml. If depositories needed additional reserves to meet their requirements because Ml was above target, they would have to borrow them at the discount window. In practice, the system was structured so that there was some borrowing even when Ml was on target. Only when Ml was far below target for a while in 1980 was 46. Economists have debated the importance of the discount rate as a mechanism for changing policy. Sometimes Federal Reserve announcements indicated that the rate was changed to catch up with market rates. Other times they cited monetary policy concerns. At issue is whether these announcements had an impact beyond that of open market operations. See Cook and Hahn, "The Information Content of Discount Rate Announcements and Their Effect on Market Interest Rates," Journal of Money, Credit, and Banking, vol. 20, no. 2 (May 1988), pp. 168-80; Lombra and Torto, "Discount Rate Changes and Announcement Effects," Quarterly Journal of Economics, February 1977, pp. 171-76; and Daniel L. Thornton, "The Market's Reaction To Discount Changes: What's Behind The Announcement Effect? Federal Reserve Bank of St. Louis, Working Paper Series, November 1991, pp. 2-23. 47. In November 1978, reserve requirements, the discount rate, and the funds rate target were all raised simultaneously as a dramatic gesture to attack the rising rate of inflation and weakening exchange value of the dollar. -29- Meulendyke borrowing allowed to drop to frictional levels, leading the federal funds rate to fall below the discount rate. The adjustment mechanism depended heavily on the enforced reluctance to borrow. When banks borrowed to satisfy their reserve requirement, they reduced their future access to the discount window. Consequently, when the banking system as a whole had to borrow a higher volume of reserves to meet requirements, individual banks would bid up the federal funds rate as they tried to avoid being one of the banks that turned to the window. The process gave banks the message to cut back on deposit-expanding activities. Chart 6 gives key borrowing and interest rate relationships during these years. The move to the new procedures inspired discussion of the appropriate guidelines for setting and changing the discount rate. Some Board members initially had expected that the discount rate would be changed more frequently than before to keep it more closely aligned with market rates. In practice, the basic discount rate was changed fairly frequently--sixteen times between October 1979 and October 1982--but it still moved much less than the funds rate. At times, unprecedented weekly average spreads developed between the funds rate and the discount rate. During two periods of exceptionally restrictive provision of nonborrowed reserves, in 1980 and again in 1981, the volume of borrowing ran very high. The Board introduced a surcharge on frequent borrowing by large banks as part of the Administration's credit restraint program in March 1980.A8 The frequency limits for access at the basic rate were similar to those that had been proposed a decade earlier for the basic borrowing privilege. In addition, banks did not have unlimited access to the discount window even when they paid the surcharge. The 48. A more detailed discussion of the rationale underlying the program of credit restraint is given in a statement by Frederick H. Schultz, Vice Chairman, Board of Governors of the Federal Reserve System, before the Subcommittee on Access to Equity Capital and Business Opportunities of the House Committee on Small Business, April 2, 1980. It is reprinted in the Federal Reserve Bulletin, April 1980. -30- Meulendyke funds rate often exceeded even the combined basic rate and surcharge-which reached a high of 18 percent in 1981.A9 Borrowed Reserve Targeting in the 1980s and Early 1990s Borrowed reserve targeting replaced nonborrowed reserve targeting in 1983 as the primary guide for choosing desired reserve levels. The shift in emphasis removed the automatic linkage between reserves and money targets. Borrowed reserve targeting made more formal use of the relationship between the amount of borrowing and the spread between the federal funds rate and the discount rate that arises from the restrictions on heavy use of the discount window. As was the case under the previous procedures, forcing increased borrowing tended to lead the banks to bid up the federal funds rate relative to the discount rate as they sought to avoid having to borrow. Reduced borrowing encouraged less aggressive bidding for Federal funds and the rate would fall. The FOMC raised borrowed reserve objectives when it wanted to tighten policy and lowered them when it wanted to ease policy.50 Chart 7 shows key borrowing and rate relationships during these years. A change in the discount rate was viewed as a substitute for a change in the borrowing assumption. Whenever the discount rate was raised or lowered, the FOMC made an explicit decision whether that action by itself accomplished the desired policy adjustment. On some occasions, the amount of assumed borrowing was left unchanged so that the average federal funds rate would be expected to rise or fall by the same amount as the discount rate move. At other times, the borrowing allowance was changed in a direction that lessened the impact of the discount rate change. For example, the FOMC would raise the borrowing 49. The surcharge was initially imposed in March 1980. It was then removed in May of that year, only to be reimposed in September. In 1981, the surcharge underwent further changes. It was increased in May, reduced in September, reduced again in October, and finally eliminated in November. 50. Marvin Goodfriend, "Discount Window Borrowing, Monetary Policy, and the Post-October 6, 1979 Federal Reserve Operating Procedure," Journal of Monetary Economics, September 1983, pp. 343-56, offers a critique of that relationship and suggests that it will inevitably be unreliable. -31- Meulendyke assumption when the discount rate was lowered so that the average funds rate would not fall by as much as the discount rate. Increased Reluctance to Borrow in the 1980s and Early 1990s. A series of banking crises and failures beginning in 1982 reintroduced a source of reluctance to borrow that had largely disappeared after the 1930s. Once again, banks became concerned that borrowing at the discount window might be interpreted as a sign that they were so weakened financially that they could not borrow funds from normal sources. The concern was especially high in 1984, when Continental Illinois National Bank suffered a crisis of confidence, experienced runs by its large depositors, and was forced to borrow massive amounts from the Federal Reserve to keep operating. Continental's experience made many other banks more hesitant to borrow, and wider spreads of the funds rate over the discount rate emerged for a given amount of borrowing fostered by the Federal Reserve. As more banking crises developed and then were resolved, the reluctance to borrow became alternately more and less severe, but it never returned to its pre-1984 pattern. By the fall of 1987, the borrowing relationship became sufficiently uncertain that the Federal Reserve felt compelled to reduce its reliance on it as a guide to policy. Since that time, the Fed has given greater weight to indicators of money market conditions such as the federal funds rate. Nonetheless, the extreme reluctance to borrow and the resulting uncertainty about how banks will respond to changing levels of reserve availability have also introduced some volatility of the funds rate. When banks have not wanted to borrow, they have reacted to a reserve shortage by bidding up the funds rate to very high levels before they finally turn to the discount window. Indeed, on one occasion in 1990, the funds rate reached 100 percent, a level not seen even when interest rates and borrowing levels were routinely much higher a decade earlier. While efforts have been made to explain to the banks and the public that occasional borrowing is an appropriate action to relieve temporary shortages of reserves, the message has so far had limited impact. The reluctance to borrow has compounded the reserve management difficulties associated with low reserve requirements, described in the -32- Meulendyke previous section. The low requirements reduced depositories' ability to handle normal day-to-day variation in reserve flows because the range of reserve levels that fell between excess reserves and overdrafts narrowed. The extreme reluctance to borrow weakened one means for banks to recover from an unexpected reserve shortage. The problems that arise when borrowing and required reserves do not behave as desired underscore the importance of these tools. The policy process benefits when both reserve requirements and the discount window can play their assigned supporting roles in the monetary policy process. Open market operations can be hard pressed to achieve policy goals without their help. 1. Required Reserves and Applied Vault Cash 1951 -1992* 70,000 60,000 [- 50,000 (0 c o I 40,000 m 30,000 20,000 10.000 0 * All figures are quarterly averages. Note: Before December 1959, the Federal Reserve did not allow vault cash to count towards the fulfillment of reserve requirements. 2. Required Balances and Excess Reserves 45.000 40,000 35,000 30,000 * 25,000 «/» 20.000 15,000 Required Reserve Balances Required Reserve Balances plus Required Clearing Balances Required Reserve Balances plus Required Clearing Balances plus Excess Reserves 10,000 5,000 0 J i i i i 1980 i i i 1981 j _ 1982 1983 L- I 1984 1985 1986 1987 1988 1989 1990 1984 1985 1986 1987 1988 1989 1990 _J I I I I I I i 1 I 1 J I—J I 1 ,.J. I I 1 1991 1992 2.000 Excess Reserves 1,500 CO C o 1,000 «/> 500 0 1980 1981 1982 1983 Note: All figures are quarterly averages. 1991 1992 * o < '% CD T3 0) U_ E CD 0) 03 CO O O CO o o o o CO o o to suojino $ o o o o CO 4. Borrowed Reserves and Selected Interest Rates 1950-1965 Borrowed Reserves as a Percentage of Total Reserves 1,600 1,400 1,200 Borrowed Reserves 1,000 800 600 400 200 0 N nrflFlh HI^I^IJWI^ 4 A 3 2 1 0 \ Discount Rate NN NNN y ^- ^\ " ^ X * LUI Effective Federal Funds Rate (starting in 1960) 3-month new bill rate 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 Note: Quarterly averages except for discount rate. Discount rate is the rate in effect on the last day of the quarter. 5. Borrowed Reserves and Selected Interest Rates 1966-1979 6 Borrowed Reserves as a Percentage of Total Reserves '5 c a> 4 a. 3 2 1 0 2,500 Borrowed Reserves 2,000 CO I 1,500 N KJ N N N N SrTlnNf^lrpnrTirnKTiF] ^ 12 c a> <D Q. 10 Effective Federal Funds Rate 8 6 4 1966 1967 1968 1969 1970 1971 1972 Note: Quarterly averages except for the discount rate. Discount rate is the rate in effect on the last day of the quarter. 1973 1974 1975 1976 1977 1978 1979 6. Borrowed Reserves and Selected Interest Rates 1979-1982 Borrowed Reserves as a Percentage of Total Reserves 1 o Borrowed Reserves 2,000 51 1,500 1,000 51 500 0 I KWM ISSSH R KWN K NVsN K Effective Federal Funds Rate Discount Rate 8 79Q4 80Q1 81Q1 Note: Quarterly averages except for the discount rate. Discount rate is the rate in effect on the last day of the quarter. 82Q1 82Q4 7. Borrowed Reserves and Selected Interest Rates 1983-1992 c <D O k- <D Q. (0 C o I 4fy 12 10 c Effective Federal Funds Rate 8 2 <D Q. 6 4 1983 1984 1985 1986 1987 Note: Quarterly averages except for the discount rate. Discount rate is the rate in effect on the last day of the quarter. 1988 1989 1990 1991 1992 Meulendyke REFERENCES Black, Robert P. Report of the Ad Hoc Subcommittee on Reserve Proposals, May 13, 1966. Board of Governors of the Federal Reserve System. years. Annual Report, Board of Governors of the Federal Reserve System. various years. Annual Statistical Board of Governors of the Federal Reserve System. various months. Federal Board of Governors of the Federal Reserve System. Reappraisal Reserve Discount Mechanism, various Reserve of the Digest, Bulletin, Federal vols. 1-3, August 1971. Carson, Deane. "Is the Federal Reserve System Really Necessary?" Journal Finance, vol. 19, no. 4 (December 1964), pp, 652-61. of Cook, Timothy, and Thomas Hahn. "The Information Content of Discount Rate Announcements and Their Effect on Market Interest Rates," Journal of Money, Credit, and Banking, vol. 20, no. 2 (May 1988), pp. 167-80. Feinman, Joshua. "Estimating the Open Market Desk's Daily Reaction Function." Federal Reserve Board, Division of Monetary Affairs, Washington, D.C., August 1991. Friedman, Milton. A Program for Monetary University Press, 1959. Stability. New York: Fordham Froyen, Richard T. "The Determinants of Federal Reserve Discount Rate Policy," Southern Economic Journal, vol. 42, no. 2 (October 1975), pp. 193-200. Goodfriend, Marvin. "Discount Window Borrowing, Monetary Policy, and the Post-October 6, 1979 Federal Reserve Operating Procedure," Journal of Monetary Economics, September 1983, pp. 343-56. Goodfriend, Marvin, and Monica Hargraves. "A Historical Assessment of the Rationales and Functions of Reserve Requirements," Federal Reserve Bank of Richmond, Economic Review, vol. 69, no. 2 (March/April 1983), pp.3-21. Gramley, Lyle E., and Samuel B. Chase Jr. "Time Deposits in Monetary Analysis," Federal Reserve Bulletin, vol. 51, no. 10 (October 1965), pp. 1380-1404. Hall, Robert E. "A Free Market Policy to Stabilize the Purchasing Power of the Dollar," in Barry Seigel, ed., Money in Crisis: The Federal Reserve, the Economy, and Monetary Reform, Pacific Studies in Public Policy. Cambridge Massachusetts: Ballinger, 1984, pp. 303-21. Meulendyke Heller, Walter W. Report of the Committee on Financial President of the United States, April 1963. Institutions Jones, David. "Excess Reserves under MCA," internal memorandum. Governors of the Federal Reserve System, November 10, 1983. to the Board of Lombra, Raymond E. , and Raymond G. Torto. "Discount Rate Changes and Announcement Effects," Quarterly Journal of Economics, vol. 91, no. 1 (February 1977), pp. 171-76. Mayer, Thomas. 1968. Monetary Policy in the United States. New York: Random House, Mayer, Thomas, James S. Duesenberry, and Robert Z. Aliber. Money, Banking, and The Economy, 3d ed. New York: W.W. Norton and Company, 1987. Meigs, A. James. Free Reserves Chicago Press, 1962. and the Money Supply. Chicago: University of Meulendyke, Ann-Marie. "A Review of Federal Reserve Policy Targets and Operating Guides in Recent Decades," in Intermediate Targets and Indicators for Monetary Policy: A Critical Survey. Federal Reserve Bank of New York, July 1990. Reprinted from Federal Reserve Bank of New York, Quarterly Review, vol. 13, no. 3 (Autumn 1988), pp. 6-17. Meulendyke, Ann-Marie. "Monetary Policy Implementation and Reserve Requirements," internal working paper. Federal Reserve Bank of New York, July 1992. Meulendyke, Ann-Marie. U.S. Monetary Policy Reserve Bank of New York, 1990. and Financial Markets. Federal Miller, William G. "Proposals on Financial Institution Reserve Requirements and Related Issues," testimony before the House Committee on Banking, Finance and Urban Affairs, July 27, 1978. "Monetary Policy and Open Market Operations during 1990," Federal Reserve Bank of New York, Quarterly Review, vol. 16, no. 1 (Spring 1991), pp. 66-74. "Monetary Policy and Open Market Operations during 1991," Federal Reserve Bank of New York, Quarterly Review, vol. 17, no. 1 (Spring 1992), pp. 72-95. Partee, J. Charles. "Statement before the Subcommittee on Financial Institutions Supervision, Regulation and Insurance of the House Committee on Banking, Finance and Urban Affairs," Federal Reserve Bulletin, November 1983, pp. 850-51. -2- Meulendyke Porter, Richard D., and Kenneth J. Kopecky. "The Role of Reserve Requirements as a Public Policy Tool," Conference on Reserve Requirements and the Role of the Federal Reserve System, Washington, D.C., January 18-19, 1979, pp. 863-80. Schultz, Frederick H. "Statement before the Subcommittee on Access to Equity Capital and Business Opportunities of the House Committee on Small Business," Federal Reserve Bulletin, vol. 66, no. 4 (April 1980), pp. 309-11. Sellon, Gorden H. Jr. "The Instruments of Monetary Policy," Federal Reserve Bank of Kansas City, Economic Review, vol. 69, no. 5 (May 1984), pp. 3-20. Small, David, and Brian Madigan. "An Analysis of Excess Reserves," internal memorandum. Board of Governors of the Federal Reserve System, July 1, 1986. Smith, Warren L. "The Discount Rate as a Credit-Control Weapon," Journal Political Economy, vol. 66, no. 2 (April 1958), pp. 171-77. System Committee on the Discount and Discount Rate Mechanism. Discount Mechanism," March 12, 1954. of "Report on the Thornton, Daniel L. "The Market's Reaction To Discount Changes: What's Behind the Announcement Effect?" Federal Reserve Bank of St. Louis, Working Paper Series, November 1991. Tolley, George S. "Providing for Growth of the Money Supply," Journal Political Economy, vol. 65, no. 6 (December 1957), pp. 465-85. of Waud, Roger N. "Public Interpretation of Federal Reserve Discount Rate Changes: Evidence on the 'Announcement Effect,'" Econometricaf vol. 38, no. 2 (March 1970), pp. 231-50. Weiner, Stuart E. "Payment of Interest on Reserves," Federal Reserve Bank of Kansas City, Economic Review, vol. 70, no. 1 (January 1985), pp. 20-21. -3- A Note on Theories of Money Stock Determination Robert L. Hetzel1 The purpose of this brief review of theories of money stock determination is to encourage economists to once again work on such models. The current lack of interest in them probably explains the irrelevance of textbook discussions to the actual monetary arrangements that determine the money stock. A discussion of money stock determination needs to make explicit whether the central bank is using an interest rate or the quantity of reserves as its policy instrument. This choice possesses different implications for the way in which the central bank gives the money stock and the price level welldefined equilibrium values. It also yields different implications for which of the behavioral relationships of the public are key for determining the money stock. For example, textbook discussions, which do not clearly identify the policy instrument, confuse the roles of the "three tools" of monetary policy: open market operations, reserve requirements, and the discount window. For example, textbooks do not mention that in the 1970s when the Fed targeted the funds rate directly, changes in required reserves ratios and in the discount rate had no first-order effects on the money stock. EARLY BANK-RATE THEORIES Henry Thornton in his book Paper Credit (1802) and in speeches before Parliament (1811) formulated the first theory of money stock determination with rate targeting by the central bank. (See Hetzel 1987.) Thornton's model is in the quantity theory tradition, which explains the determination of the The author is Vice President and Economist at the Federal Reserve Bank of Richmond. Hetzel price level through the interaction of a money supply and money demand function. The key element of his theory is that the public cares only about real variables, that is, real quantities and relative prices. In particular, the real rate of interest is only transitorily affected by changes in fiat money creation. hypothesis.) (This idea is now referred to as the natural rate Money creation allows the central bank to set the market rate at a different value than the real rate adjusted for expected inflation, but only transitorily. The flip side of the natural rate hypothesis is that only the central bank can give nominal variables like the money stock and the price level well-defined values. Thornton argued that the Bank of England, despite its real-bills rhetoric, gave nominal variables determinate values by targeting the exchange rate. Thomas Joplin (1823) gave Thornton's idea of a natural rate of interest its modern meaning as the real rate of interest that equates saving and investment. With the Resumption Act of 1819 that returned Britain to the gold standard, the idea of a central bank that creates fiat money virtually disappeared. Joplin thought that the banking system, through variation in its reserves-deposits ratio, created changes in money that caused transitory divergences in the market and the natural rate. This idea reappeared later in the work of Knut Wicksell (1898) and Irving Fisher (1918). With the supremacy of the gold standard in the nineteenth century, the idea of central bank money creation practically disappeared and, along with it, the idea of a natural rate of interest. David Hume's price-specie flow mechanism became the basic model of money stock determination in the nineteenth century. Wicksell was unique in reinventing the idea of a natural 2 Hetzel rate of interest, but he had no central bank in his model. He also made no distinction between real and nominal variables. Abandonment of the gold standard in World War I, as in the Napoleonic Wars, led to the reemergence of theories of fiat money creation. Gustav Cassel (1928) developed the market rate-natural rate theory of Thornton and Wicksell. He pointed out that achieving price level determinacy required more than equality of the market rate and the natural rate. An infinite number of price levels are consistent with equality of these two rates. Cassel argued that the central bank should vary its discount rate in order to keep the price level at a targeted value. Interest in theories of money stock determination in which the central bank uses an interest rate as its policy variable disappeared in the Depression with the prevalence of elasticity pessimism. THE RESERVES-MONEY MULTIPLIER THEORY The reserves-money multiplier theory emerged at the end of World War I. It had been advanced occasionally in the nineteen hundreds, but had never caught on (Humphrey 1987). United Kingdom. Pigou (1917) and Keynes (1923) advanced it in the In the United States, Phillips (1921) built up the reserves- money multiplier formula from a deposit expansion process whereby a reserve injection creates deposits as it passes from bank to bank. This deposit creation process continues until the newly-created reserves are absorbed into required reserves. The Phillips' description of the deposits creation process was flawed from the beginning. Even if changes in aggregate reserves are exogenous, bank deposit creation is constrained by the interest rate on reserves in the 3 Hetzel interbank market for reserves, not the quantity of reserves a bank holds. By the 1920s, there was a Fed funds and a call money market that allowed banks to buy and sell reserves. Quantity theorists, however, liked the idea of the reserves-deposits expansion process because it provided an easy refutation to real bills proponents who argued that banks cannot create deposits. Quantity theorists could use the Phillips7 story to argue that real bills proponents failed to understand that what is true for the individual bank is not necessarily true for the banking system. The revival of models of money stock determination in the 1950s centered on reserves-money multiplier models. Why did quantity theorists turn to these models given that the Fed was targeting free reserves, which is an indirect procedure for targeting the interest rate? In the 1950s, economists had not yet rediscovered the natural rate hypothesis and the idea of a natural rate of interest. Without these concepts, quantity theorists could not model money stock determination with central bank rate targeting in a way that made the money supply function differ from the money demand function by depending crucially on central bank behavior. The reserves-money multiplier theory offered an easy explanation of how central banks control the price level by controlling the supply of money. In the 1970s, Fed economists working on a monthly model for use at FOMC meetings rejected the reserves-money multiplier framework. Pierce and Parry 1975 and Davis 1974.) (See Thomson, Given the Fed's target for the funds rate, they viewed the money stock as being demand determined by the public's demand for money function. The problem with this view was that the price level was taken as determined outside the model. 4 The model then did not Hetzel distinguish between the determination of nominal and real money. With the price level exogenously determined, the Fed can vary the real quantity of reserves to control the (market and real) rate of interest and the nominal and real quantity of money demanded by the public. If the price level is endogenously determined, however, these models say nothing about the nominal quantity of money. While there is a determinate relationship between the interest rate and the real quantity of money, there is no determinate relationship between the interest rate and the nominal quantity of money. RATIONAL EXPECTATIONS MODELS WITH RATE TARGETING The key conceptual issue in models of money stock determination with rate targeting by the central bank is how nominal variables are rendered determinate, that is, given a well-defined equilibrium value? Patinkin (1965) pointed out that the central bank must "concern" itself with some nominal variable. In his model, that nominal variable is bank reserves. The price level is then rendered determinate through a real balance effect. Because an arbitrary rise in the price level reduces the real value of a variable the public cares about, bank reserves and money, it causes the public to reduce its real expenditure, and the price level is returned to its equilibrium value. When the central bank targets an interest rate, however, all nominal variables including bank reserves are determined endogenously. There is no real balance effect. In the early 1980s, a number of economists figured out how to explain nominal determinacy with interest rate targeting by the central bank. The key papers were by Dotsey and King (1983); Canzoneri, Henderson and Rogoff (1983); 5 Hetzel and McCallum (1981, 1986). Their models embodied the natural rate hypothesis so that only the central bank, not the public, could give nominal variables a well-defined equilibrium value. The initial models achieved nominal determinacy by causing the central bank to behave in such a way that the public's expectation of the future price level remained fixed. Where Patinkin fixed reserves, these models fixed the expected future value of the price level. They did so by not allowing base drift in money. Goodfriend (1986) brought these models closer to actual central bank behavior by allowing the public's expectation of the future price level to vary in response to macroeconomic shocks. He did so by allowing base drift in the money stock where the amount of such drift depends upon the extent to which the central bank desires to smooth interest rates. Goodfriend gave the central bank a cost function that made it averse to large "jumps" in the price level relative to expectations. By making the central bank care both about the difference between the contemporaneous price level and the prior period's expectation of the contemporaneous price level and about the difference between the contemporaneous price level and the expected future price level, he imposed a level and a change constraint that made the public's expectation of the future price level well defined, while still allowing that expectation to change in response macroeconomic shocks. Instead of a real balance effect, these models make use of a relative price effect. Given the public's expectation of the future price level, an arbitrary change in the contemporaneous price level changes the interest rate by changing expected inflation. Changes in the interest rate then affect the demand for money and the reserves-supplying behavior of the central bank in a 6 Hetzel way that returns the price level to an equilibrium value. The simplifying assumption that makes these models tractable analytically is rational expectations. This assumption allowed expectations to be determined in a simple enough way that the models could highlight how the central bank makes nominal values well defined by the control it exercises over the way the public forms its expectation of the future values of nominal variables. AN EXAMPLE AND SOME BASIC PRINCIPLES This section draws on the models briefly described above to highlight some of the basic concepts in a theory of money stock determination. Consider an example where the central bank implements monetary policy by setting an interest rate. The natural rate hypothesis implies that the central bank cannot set its interest rate target arbitrarily. It must have procedures that allow it to set its rate target in a way that tracks on average the economy's equilibrium interest rate. The central bank, however, is assumed to smooth changes in the economy's equilibrium interest rate by supplying reserves when market rates rise and withdrawing reserves when market rates fall. Changes in reserves and the money stock then emerge in response to the macroeconomic shocks that impinge upon the economy. It is also necessary to make some assumption about the central bank's subsequent behavior toward the random changes in money introduced by rate smoothing. The central bank can either offset these changes subsequently, in part or in full, or incorporate them permanently into the level of the money stock. In practice, central banks follow the latter "let bygones-be-bygones" policy of base drift in money and prices. 7 Hetzel Assume that a persistent real shock raises the equilibrium real rate of interest. For example, a new technology leads to increased investment. When the market rate rises initially, the central bank buys government securities. The monetary base and the money stock increase. Because the real rate of interest is ultimately determined solely by real factors like investment opportunities and the public's thrift, the real rate of interest must eventually rise to a higher equilibrium value that is independent of the increase in money. Similarly, the real quantity of money desired by the public will ultimately be unaffected by the actions of the central bank. The rise in market rates will make currency and bank reserves more costly to hold, but the return on bank deposits that pay interest will rise. The equilibrium real quantity of money may then either decrease or increase. In either event, the supply of money changes in a way that is largely unrelated to any change in the public's demand for real money. For this reason, it will be convenient to assume that the rise in market rates leaves the real quantity of money demanded by the public unchanged. Ultimately, the change in the nominal quantity of money will not affect any of the new equilibrium values of the real variables, the real rate of interest and the real quantity of money. At the original price level, however, the public is now holding a larger quantity of real money balances. The price level must rise to return real money balances to their lower, equilibrium value. This example can be used to elucidate a number of key concepts in a theory of money stock determination. 2 2 For a somewhat different treatment, see any of the expositions by Milton Friedman that feature a helicopter drop of money, for example, Friedman (1992, Chapter 2 ) . 8 Hetzel First, the example illustrates the distinction between nominal and real variables. In the short run, macroeconomic shocks originating in the private sector produce changes in the nominal money stock. In the long run, however, the central bank exercises complete control over the nominal money stock. The central bank determines the amount of base money to create in response to such shocks. It also determines the extent to which the money created in response to shocks will affect permanently the level of the money stock. In contrast, the public largely determines the real quantity of money. The qualification "largely determines" is added in recognition that the central bank can indirectly influence the real quantity of money in that a higher rate of inflation increases the cost of holding real money. of the example, however, is unaffected by this qualification. The result The increase in the stock of money does not ultimately increase the real quantity of money. Second, it is important to keep separate the different popular meanings of the word "money." A theory of money stock determination concerns the quantity of money, defined as some monetary aggregate like the monetary base, Ml or M2. Money is often also used to mean credit. In this example, the increase in investment demand and higher real rate of interest will increase real saving and credit. Money is also often used to mean income. In the example, real income is probably largely unchanged, although the composition of output generating income will change to include more investment and less consumption. In the example, the real quantity of money, real credit, and real income can all behave differently. Third, the example illustrates the quantity theory approach to analyzing the determination of the price level through the interaction of a money demand 9 Hetzel and a money supply function. In the example, the money supply function shifts as the central bank buys government securities in response to a rise in interest rates. Because no corresponding shift in the money demand function occurs, the price level rises. The nominal money demand function is the product of the price level and the real quantity of money desired by the public. The nominal money supply function depends upon the reserve-supplying behavior of the central bank. In the short run, shifts in this function depend upon the extent to which the central bank smooths the interest rate, that is, the extent to which it supplies reserves when the interest rate changes. In the long run, shifts in the money supply function depend upon the extent of base drift, that is, the extent to which, if at all, the central bank subsequently offsets the changes in money produced by changes in the interest rate. Finally, shifts in this function depend upon the trend rate of growth of money and inflation the central bank accepts and the public expects. A consequence of the natural rate hypothesis is that only the central bank can determine the trend rate of growth of money and prices. Milton Friedman gave the quantity theory a particular empirical expression by arguing that shifts in the money supply function have historically been large relative to shifts in the money demand function. For this reason, over long periods of time, the price level and the nominal quantity of money move together. Friedman summarizes this view by saying that inflation is everywhere and always a monetary phenomenon. Note, however, that the analytical usefulness of the quantity theory only requires that unpredictable changes in money demand are small relative to shifts in the 10 Hetzel money supply function and to predictable shifts in money demand. Equivalently, unpredictable changes in money demand should be small relative to changes in nominal expenditure or output. Note that with rate targeting the key behavioral relationships of the money supply function are not the reserves-currency and reserves-deposits ratios discussed in textbooks. Fluctuations in these ratios are automatically offset at the prevailing funds rate target. For example, if currency flows out of banks or if banks increase the desired level of excess reserves, the funds rate rises. In order to maintain its funds rate target, the central bank supplies reserves, thereby accommodating changes in these ratios and avoiding a change in bank deposits. Fourth, what appears true for the individual is not necessarily true for individuals collectively. In the example, after the increase in the money stock, individuals believe they can reduce their money holdings to a desired lower level. The public, however, cannot reduce its nominal money holdings. The individuals who sold government securities to the central bank did so because they were offered a good price, not because they wanted to reduce their holdings of assets. After selling securities to the central bank, individuals allocate their increased cash among different assets to replace the securities they sold. Temporarily, the increased demand for assets keeps the interest rate below its new, higher equilibrium rate. As a consequence, real expenditure rises until the price level increases sufficiently to return real money balances to their original level. The interest rate then rises to its new, higher equilibrium value. More generally, economic fallacies often arise out of inappropriate 11 Hetzel generalization of individual experience. To the individual, it appears that the cause of inflation is the rise in prices of individual commodities. The cause of inflation then is sought for in the determinants of the relative prices of individual commodities, rather than in the behavior of money. Fifth, the central bank must ensure that the price level and money stock possess equilibrium values. The central bank, however, must do more than simply set an interest rate target that is consistent with the economy's natural rate of interest (augmented by expected inflation). At the central bank's prevailing rate target, an arbitrary perturbation in the price level will produce a corresponding change in the demand for bank credit and in bank deposits and money. together. All nominal magnitudes can then wander off aimlessly The central bank must keep some nominal value steady. How do central banks impart this nominal steadiness to equilibrium values? Central banks dislike "large, unusual jumps" in nominal prices. 3 This concern imparts an "inertia" to the public's expectation of the future price level. Arbitrary changes in the contemporaneous price level, therefore, produce changes in the contemporaneous price level relative to the expected future price level. These changes create a relative price effect analogous to the real balance effect as the mechanism for eliminating arbitrary changes in the price level. Specifically, given an arbitrary change in the price level, some real or relative variable must change to produce an inverse change in the public's real expenditure. That change is the change in the price level relative to the expected future price level, which produces an inverse 3 This statement is given content by defining jumps relative to expected values. See the discussion of Goodfriend (1987) above. 12 Hetzel movement in the interest rate. The movement in the interest rate affects the reserve supplying behavior of the Fed and the public's demand for money in a way that returns the price level to its equilibrium value. This relative price effect can be explained by analogy. Consider how nominal determinacy is achieved when the central bank of a country targets its exchange rate with another country. For the sake of argument, assume that the Fed targets the Deutsche Mark price of a dollar. As shown in equation (1), the DM/$ exchange rate equals the product of the ratio of the German price level (DM/German good) to the U.S. price level (S/US good) and the real terms of trade (German good/US good). German price level. The nominal benchmark for the dollar is the If the U.S. price level rises arbitrarily, the foreign exchange value of the dollar falls, and the Fed buys dollars with Deutsche marks. The monetary base and the money stock fall and the price level returns to its equilibrium level. DM ... DM (l) — German good = $ German good • $ US good US good In the case of an interest rate target, the Fed targets the price of today's dollars ($t) in terms of tomorrow's dollars ($ t + 1 ), or one plus the interest rate. As shown in equation (2), this price equals the product of the ratio of the expected future price level to the contemporaneous price level and the real terms of trade with the future. With a rate target, the nominal benchmark is the expected future price level. Now, an arbitrary rise in the contemporaneous price level produces a fall in the ratio of the expected future price level to the contemporaneous price level, the first factor on the 13 Hetzel right side of (2). The fall in this ratio produces a decline in the market rate of interest by reducing the inflation premium. A decline in the market rate of interest produces an increase in the demand for money. prompts the central bank to sell securities. It also The demand for money increases, while the monetary base and the money stock fall, and the price level returns to its equilibrium level. S^ (2) $f tK = good)f" (-L.) 9 (good),., (good), good ' Sixth, the public is forward looking. It must form an expectation of the future price level in order to determine a market rate of interest. In making saving and investment decisions, individuals care about the price of today's goods in terms of tomorrow's goods. They contract, however, in terms of the price of today's dollars in terms of tomorrow's dollars. Individuals must, therefore, form an expectation of the future purchasing power of the dollar. The central bank determines how the public forms that expectation. Consider again the example of the real shock with rate smoothing by the central bank that causes money and prices to increase. If the public expects that the central bank will reverse the increase in money and prices in the future, then the public will expect a subsequent fall in prices, or at least a temporary reduction in inflation relative to trend. A temporary reduction in the inflation premium will for a while moderate the rise in the market rate produced by the rise in the equilibrium real rate. (This situation probably obtained in World War II.) Alternatively, if the public expects the central bank to incorporate 14 Hetzel each period's random change in money and prices permanently into their future levels, the interest rate will rise immediately by the amount of increase in the equilibrium real rate (apart from a temporary liquidity effect). The public might even expect that the central bank will allow the rate of inflation to rise permanently, in which case the market rate will rise by more than the increase in the real rate. In this case, the price level will also rise by more than the increase in money. As this discussion illustrates, the response of the public to today's action of the central bank depends upon what the public expects the central bank to do tomorrow. (The public may not actually watch the actions of the central bank, but it will respond to contemporaneous changes in the price level in a way that is consistent with the behavior of the price level that the central bank has produced over time.) For this reason, at least since the 1970s, economists have generally formulated their recommendations for the central bank as strategies to be maintained over time, rather than as particular policy actions. The idea is that the policymaker can predict the consequences of a policy action taken as part of a known strategy because he has some basis for predicting what the public anticipates in the way of subsequent policy actions (Lucas 1975). Seventh, the example involves both monetary policy and fiscal policy. The monetary policy action undertaken by the central bank is the increase in the monetary base. Monetary policy, that is, the systematic behavior of the central bank, is described by the extent to which the central bank changes the monetary base when interest rates change and the extent to which such changes become permanently incorporated into the level of the monetary base. 15 The Hetzel fiscal policy side of the central bank's action is the reduction of the government debt held by the public due to the purchase of the government security. The revenue the government must collect in the future to pay off its debt falls. Taxpayers can increase their consumption. Who pays for this windfall to taxpayers? When the price level rises, whoever holds money must add to his money holdings in order to maintain their real purchasing power. The money holder must refrain from consumption while restoring the real value of his cash balances to their former level. (The additional dollars held are like receipts showing payment of the tax.) There is a wealth transfer from holders of cash balances to taxpayers in general. Inflation is a tax levied on whoever holds money. Pressure for inflation comes from confusion between money and wealth creation. This confusion turns on ignorance of all the principles listed above: the difference between money and credit or income; the difference between nominal and real money; the fallacy of generalizing on the basis of particular examples; failure to understand the central responsibility of the central bank for the behavior of the price level; and failure to realize that the public is forward looking in the way it forms its expectations of central bank behavior. Pressure for inflation, however, also comes from the fact that while money creation does not augment wealth, it can redistribute it. A central fact of the political economy of money creation, and of its eternal appeal, is that the tax money creation imposes does not have to be explicitly legislated. 16 Hetzel REFERENCES Canzoneri, Matthew B., Dale W. Henderson and Kenneth S. Rogoff, "The Information Content of the Interest Rate and Optimal Monetary Policy," The Quarterly Journal of Economics, November 1983, 545-65. Cassel, Gustav, "The Rate of Interest, The Bank Rate, and the Stabilization of Prices," Jhe Quarterly Journal of Economics, August 1928, 511-29. Davis, Richard G., "Implementing Open Market Policy with Monetary Aggregate Objectives," in Federal Reserve Bank of New York Monetary Aggregates and Monetary Policy, October 1974. Dotsey, Michael and Robert G. King, "Monetary Instruments and Policy Rules in a Rational Expectations Environment," Journal of Monetary Economics, September 1983, 357-82. Fisher, Irving, Elementary Principles of Economics, New York: The Macmillan Co., 1918. Friedman, Milton, Money Mischief, New York: Harcourt Brace Jovanovich, 1992. Goodfriend, Marvin, "Interest Rate Smoothing and Price Level TrendStationarity," Journal of Monetary Economics, May 1987, 335-48. Hetzel, Robert L., "Henry Thornton: Seminal Monetary Theorist and Father of the Modern Central Bank," Federal Reserve Bank of Richmond Economic Review, July/August 1987, 3-16. Humphrey, Thomas M. "The Theory of Multiple Expansion of Deposits: What It Is and Whence It Came," Federal Reserve Bank of Richmond Economic Review, March/April 1987, 3-10. Joplin, Thomas. Outlines of a System of Political Economy (1823), New York: Augustus M. Kelley, 1970. Keynes, John Maynard, A Tract on Monetary Reform (1923), in lire Collected Writings of John Maynard Keynes, vol. 4, London: The Macmillan Press, 1971. Lucas, Robert E., "Econometric Policy Evaluation: A Critique," in The Phillips Curve and Labor Markets, edited by Karl Brunner and Allan Meltzer, pp. 19-46. Carnegie-Rochester Conference Series no. 1. New York: North Holland, 1975. McCallum, Bennett T., "Price Level Determinacy with an Interest Rate Policy Rule and Rational Expectations," Journal of Monetary Economics, November 1981, 319-29. , "Some Issues Concerning Interest Rate Pegging, Price Level Determinacy, 17 Hetzel and the Real Bills Doctrine," Journal of Monetary Economics, January 1986, 135-60. Pigou. A. C. "The Value of Money," Quarterly Journal of Economics, November 1917, 38-65. Phillips, Chester A. Bank Credit, New York: The Macmillan Co., 1921. Thomson, Thomas D., James L. Pierce, and Robert T. Parry, "A Monthly Money Market Model," Journal of Money Credit and Banking, November 1975, 41131. Thornton, Henry, An Enquiry into the Nature and Effects of the Paper Credit of Great Britain (1802) and Iwo Speeches (1811), ed., F. A. v. Hayek, New York: Rinehart and Co., 1939. Wicksel1, Knut, Interest and Prices (1898), New York: Augustus M. Kelley, 1965. 18 INTEREST RATE POLICY AND THE INFLATION SCARE PROBLEM: 1979-1992 Marvin Goodfriend1 U.S. monetary policy since the late 1970s is unique in the post-war era in that rising inflation has been reversed and stabilized at a lower rate for almost a decade. The inflation rate of 3 to 4% per year, representing a reduction of 6% or so from its 1981 peak, is the result of a disinflationary effort that has been long and difficult. This paper analyzes the disinflation by reviewing the interaction between Federal Reserve policy actions and economic variables such as the long-term bond rate, real GDP growth, and inflation. The period breaks naturally into a number of phases, with the broad contour of events as follows. A period of rising inflation was followed by disinflation which, strictly speaking, was largely completed in 1983 when inflation stabilized at around 4% per year. But there were two more "inflation scares" later in the decade when rising long-term rates reflected expectations that the Fed might once more allow inflation to rise. Confidence in the Fed was still relatively low in 1983, but the central bank has acquired more credibility since then by successfully resisting the inflation scares. 1. The author is Vice President and Associate Director of Research at the Federal Reserve Bank of Richmond. The paper has benefitted greatly from discussions with Timothy Cook and Robert King, and from presentations at the 1992 NBER Summer Institute and the Board of Governors of the Federal Reserve System. Comments by John Boschen and George Moore were also very helpful. Goodfriend I analyze the conduct of monetary policy using a narrative approach that pays close attention to monthly movements of long and short-term interest rates. My approach is intended to complement existing studies such as the VAR-based analyses by Bernanke and Blinder (1992) and Sims (1991), and the more conventional studies of the period by Friedman (1988) and Poole (1988). The goal is to distill observations to guide future empirical and theoretical analysis of monetary policy with the ultimate objective of improving macroeconomic performance. Based on a familiarity with the Fed over this period and the work of Fed economists, I interpret policy actions in terms of the Federal funds rate rather than a measure of money. I view the paper as a case study of the Federal Reserve's interest rate policy. The Fed's primary policy problem during the period under study was the acquisition and maintenance of credibility for its commitment to low inflation.2 I measure credibility by movements of inflation expectations reflected in the long term interest rate. For much of the period the Fed's policy actions were directed at resisting inflation scares signalled by large sustained increases in the long rate. A scare could take well over a year of high real short-term interest rates to contain. Moreover, just the threat of a scare appears to have made the Fed tighten aggressively in one instance and probably made it more cautious when pushing the funds rate down to encourage real growth on a number of occasions. 2. See Rogoff (1987) for a theoretical survey of credibility, reputation, and monetary policy. - 2 - Goodfriend Inflation scares are costly because resisting them requires the Fed to raise real short rates with potentially depressing effects on business conditions. Hesitating to react is also costly, however, because by revealing its indifference to higher inflation the Fed actually encourages workers and firms to ask for wage and price increases to protect themselves from higher expected costs. The Fed is then inclined to accommodate the higher inflation with faster money growth. Inflation scares present the Fed with a fundamental dilemma whose resolution has decided the course of monetary policy in the post-war Prior to the 1980s, the Fed generated an upward trend in the period. inflation rate by reacting to inflation scares with a delay. The more prompt and even preemptive reactions since the late 1970s have been a hallmark of the recent disinflationary era. The plan of the paper is as follows. First, I introduce and discuss the premises that underlie my interpretation of monetary policy in the body of the paper. presented next. The chronological analysis of policy is Finally, I summarize the main empirical findings in a series of observations chosen to sharpen further our interpretation and evaluation of the conduct of monetary policy. A brief conclusion follows. PREMISES UNDERLYING THE INTERPRETATION OF POLICY The first step in any study of monetary history is to choose an indicator of the stance of policy. For example, in their study of U.S. - 3 - Goodfriend monetary history Friedman and Schwartz (1963) focus on the monetary base because it summarizes monetary conditions whether or not a country is on the gold standard and whether or not it has a central bank. Focusing on the base allowed them to tie together a long period marked by many institutional changes, making possible their famous empirical findings about money, prices, and business conditions* For my purposes, however, the base is not a good choice of indicator. Although the Fed could have used the base as its instrument by controlling it closely in the short-run, it has not chosen to do so. Instead, the Fed has chosen to use the Federal funds rate as its policy instrument. Hence this study, which seeks to investigate the short-run interactions between Fed policy and other economic variables, interprets policy actions as changes in the Federal funds rate. The remainder of this section discusses the premises underlying my interpretation of policy. Interest Rate Targeting Throughout its history the Fed's policy instrument has been the Federal funds rate or its equivalent. At times, notably from the mid to late 1970s, it has targeted the funds rate in a narrow band commonly 25 basis points wide (Cook and Hahn 1989). More often, it has targeted the funds rate indirectly, using the discount rate and borrowed reserve targets. Although the funds rate appears noisier under borrowed reserye targeting than under direct funds rate targeting, it is nevertheless tied relatively closely to a chosen Federal funds rate target (Goodfriend - 4 - Goodfriend 1983). Since a borrowing target tends to be associated with a particular spread between the funds rate and the discount rate, targeting borrowed reserves lets a discount rate adjustment feed through one-for-one to the funds rate. Forcing banks to borrow more reserves at a given discount rate also raises the funds rate (Goodfriend and Whelpley 1986). The Fed has used the borrowed reserve procedure to help manage the funds rate since it ended its experiment with nonborrowed reserve targeting in October 1982 (Wallich 1984, Thornton 1988). Significant Federal funds rate movements since then should be viewed as deliberate target changes. It is less obvious that Federal funds rate changes in the period of the New Operating Procedures from October 1979 to October 1982 should be interpreted as deliberate. Under those procedures, the Fed was to fix the path of nonborrowed reserves available to depository institutions so that increases in the money stock would force banks to borrow more reserves at the discount window and thereby automatically drive up the funds rate and other short term interest rates. Despite the widespread emphasis on automatic adjustment in the description of the post-October 1979 procedures, however, it was wellrecognized at the time that movements in the funds rate would also result from purely judgmental actions of the Federal Reserve (Levin and Meek 1981, Annual Reports of Open Market Operations 1981-83). These actions included: (1) judgmental adjustments to the nonborrowed reserve path taken at FOMC meetings that changed the initially expected reserves banks would be forced to borrow at the discount window (in effect, a - 5 - Goodfriend funds rate target change by the FOMC), (2) judgmental adjustments to the nonborrowed reserve path between FOMC meetings, (3) changes in the discount rate, and (4) changes in the surcharge that at times during the period was added to the basic discount rate charged to large banks. Cook (1989) presents a detailed breakdown of policy actions affecting the funds rate during this period showing that two-thirds of the funds rate movement was due to judgmental actions of the Fed and only one-third resulted from automatic adjustment. Moreover, as we shall see below, the large Federal funds rate movements in the nonborrowed reserve targeting period are overwhelmingly attributable to deliberate discretionary actions taken by the Fed to manage short-term interest rates. In light of this, it is more accurate to refer to the period from October 1979 to October 1982 as one of aggressive Federal funds rate targeting than one of nonborrowed reserve targeting. The Role of Money The Federal Reserve was established with a mandate to cushion short-term interest rates from liquidity disturbances. Between the Civil War and the creation of the Fed, such disturbances caused short rates to rise suddenly and sharply from time to time. While generally trading in a range between 4 and 7 percent, the monthly average call loan rate reported by Macaulay (1938) rose roughly 5 percentage points in one month on 26 occasions between 1865 and 1914. Moreover, as a result of banking crises, sudden changes of over 10 percentage points occurred 8 times during the same period. These episodes were distinctly temporary, - 6 - Goodfriend ranging from one to four months, with many lasting for no more than one month. Such extreme temporary spikes are absent from interest rates since the founding of the Fed (Miron 1986, Mankiw, Miron, and Weil 1987). In line with its original mandate, the Fed has routinely accommodated liquidity disturbances at a given targeted level of shortterm interest rates. Furthermore, by giving banks access to the discount window the Fed has been careful not to exert excessively disruptive liquidity disturbances when changing its interest rate target.3 It follows that easing or tightening has mainly been accomplished by changing the level of short rates to set in motion forces slowing the growth of money demand in order to allow a future reduction in money growth and inflation. To view the Federal Reserve's policy instrument as the Federal funds rate is thus to set money to the side, since at any point in time money demand is accommodated at the going interest rate. This does not say, however, that money can be left out of account altogether. The Fed, the markets, and economists alike recognize that trend inflation is closely connected to trend money growth, and that achieving and 3. Total reserve demand is not very interest elastic in the short run. So whenever the Fed cuts nonborrowed reserves to support a higher Federal funds rate target, it allows banks to satisfy a roughly unchanged reserve demand by borrowing the difference at the discount window. The negative relation between nonborrowed reserves and the funds rate in part reflects the administration of the discount window, which creates a positive relation between bank borrowing and the spread between the funds rate and the discount rate. Christiano and Eichenbaum (1991) emphasize the importance of this mechanism in understanding the liquidity effect. - 7 - Goodfriend maintaining price stability requires controlling money. During the period under study, money growth was often viewed as an important indicator of future inflation or disinflation by both the Fed and the markets. Furthermore, we know from the work of McCallum (1981) and others that an interest rate policy just describes how changes in interest rates correspond to changes in the money stock. At a deeper level, then, there is an equivalence between talking in terms of interest rates or money. The important difference is that simple interest rate rules descriptive of policy have implications for how money and prices actually evolve over time (Goodfriend 1987, Barro 1989). We should keep this in mind when reviewing the current period for clues about how policy influences the inflation rate. Ultimately we seek to understand what it is about interest rate policy that turns one-time macroeconomic shocks into highly persistent changes in the growth of money and prices. Interpreting Comovements Between Short and Long Rates The Fed targets the funds rate in order to stabilize inflation and real economic growth as best it can. Output and prices, however, do not respond directly to weekly Federal funds rate movements but only to longer-term rates of perhaps six months or more. Hence, the Fed targets the funds rate with the aim of managing longer-term money market rates. It exercises its leverage as follows. The market determines longer-term rates (abstracting from a time varying term premium and default risk) as the average expected level of the funds rate over the relevant horizon. - 8 - Goodfriend To see why, consider the pricing of a three-month bank loan. A bank could fund the loan with a three-month CD, or it could plan to borrow Federal funds overnight for the next three months.. Cost minimization and competition among banks keep the CD rate in line with the average expected future funds rate; competition in the loan market links loan rates to the CD rate and expected future funds rates. Finally, arbitrage among holders of money market securities links Treasury bill and commercial paper rates to CD rates of similar maturity. Since simplicity is crucial in communicating policy intentions, the Fed manages its funds rate target to maintain an expected constancy over the near-term future. Target changes are highly persistent and seldom quickly reversed, so that a target change carries the expected level of the funds rate with it and thus longer-term money market rates too.4 In this way, interest rate policy as practiced by the Fed anchors the short end of the term structure of interest rates to the current Federal funds rate. By the above argument, the interest rate on long bonds too must be determined as an average of expected future short rates. At best, the Fed affects short-term real interest rates temporarily, so average future short rates over the horizon of a 30-year bond should sum to a real interest rate that varies in a range perhaps 1 or 2 percentage 4. Goodfriend (1991) contains a discussion of evidence consistent with this view reported in Fama (1984), Fama and Bliss (1987), Mankiw, Miron, and Weil (1987), Hardouvelis (1988), and Cook and Hahn (1989). - 9 - Goodfriend points around 3% per year--pius the expected trend rate of inflation.5 From this perspective, we can view fluctuations in the long-term rate as driven by: (1) a component connected with the current Federal funds rate target that anchors short maturity rates, and (2) a component driven by expectations of inflation. Because the present discounted value of coupon payments far out in the future is smaller at higher interest rates, we should expect a given funds rate target change to exert a greater effect on the long bond at higher rates of interest.6 It is 5. Consider a bond paying nominal interest (i) taxable at rate (r), when the expected inflation rate is (*e). The real after-tax ex ante return on such a bond is then r » (l-r)i-ire, so the expected inflation rate over the life of the bond may be expressed as ?re * [i - r/(l-r)](l-r). Woodward (1990) reports market expectations of the after-tax real rate of interest on long-term bonds using quarterly data on British index-linked gilt-edged securities from 1982:2 to 1989:1. The ex ante post-tax real rate ranged from 1.5% to 3.2% per annum with a mean of 2.6%. Assuming investors keep after-tax ex ante rates on long-term government bonds in the U.S. and U.K roughly equal, we can set r =.026 in the above expression to infer long-term expected inflation in the U.S. A tax rate in the U.S. of .20, for example, yields 7re = [i-3.2](.8). If we take i as the yield to maturity on a 30-year U.S. government bond, then 7re is the average per annum inflation rate expected over the 30-year horizon. The tax rate in the above expression is the marginal rate that applies to the relevant marginal investor, e.g., individual, corporation, or foreigner. The rate is difficult to determine. Its exact value, however, is not important for the analysis in the text. The analysis relies on the view that significant changes in the long-term nominal rate primarily reflect proportional movements in inflation expectations, a view supported by the narrow range of ex ante post-tax real rates reported by Woodward. 6. A given Federal funds rate target change will exert a greater effect on the long-term bond rate the shorter the average life of the security as measured by its duration. The duration of a coupon bond may be thought of as the term to maturity of an equivalent zero coupon bond that makes the same total payments and has the same yield. The duration of a 30 year coupon bond selling at par is approximately 1/r, where r is the yield to maturity. See Moore (1989). Thus, the duration of the 30 year government (coupon) bond discussed in the text is only about 12.5 years at an interest rate of 8% and 7.1 years at a 14% interest rate. - 10 - Goodfriend useful to distinguish three sources of interaction between the Federal funds rate and the long-term rate: Pure Cyclical Funds Rate Policy Actions. The Fed routinely lowers the funds rate in response to cyclical downturns and raises it in cyclical expansions. I call such policy actions purely cyclical if they maintain the going trend rate of inflation. Even purely cyclical policy actions exert a pull on longer rates, however, so they are a source of positive comovement between the funds rate and the long rate. But because cyclical actions strongly influence only the first few years of expected future short-term interest rates, only a relatively small fraction of purely cyclical funds rate changes are transmitted to the long rate. Long-Run Inflation. Changes in the trend rate of inflation are a second source of positive comovement between the funds rate and the long rate. While the long rate moves automatically with inflation expectations, the funds rate does not unless the Fed makes it do so. Nevertheless, the Fed can hold short-term real rates relatively steady in the presence of rising or falling inflation by moving the funds rate up or down to allow for a rising or falling inflation premium. In so doing, it causes short and long rates to move relatively closely together. Aggressive Funds Rate Policy Actions. The Fed occasionally takes particularly aggressive funds rate policy actions to encourage real growth or to stop and reverse a rising rate of inflation. Aggressive - 11 - Goodfriend actions combine a purely cyclical effect with a potential change in the long-run rate of inflation. The cyclical effect moves the long rate in the same direction as the funds rate, while the inflation effect moves the long rate in the opposite direction. Thus the net effect of aggressive actions on the long rate is somewhat complex. Consider an aggressive reduction in the funds rate to encourage real growth. Initially, funds rate actions taken to fight recession pull the long rate down too. However, excessive easing that raises inflation can cause the long rate to reverse direction and begin to rise, even as the Fed continues to push short rates down. Thus we might expect to see the long rate move in the opposite direction from the funds rate near cyclical troughs. A sharp funds rate increase during the ensuing recovery exerts two conflicting forces. It tends to raise the long rate by reversing the cyclical funds rate decline, but it also reverses somewhat the expected rise in inflation, tending to lower the long rate. For a relatively brief recession with little excessive easing, the cyclical funds rate effect would dominate the inflation effect, so the long rate would tend to rise with the funds rate during the recovery. Thus, the long rate would move opposite from the funds rate for only a few months near a recession trough. Now consider an aggressive increase in the funds rate intended to bring down the trend rate of inflation. Such a tightening potentially shifts both components of the long rate since short rates rise and expected long-run inflation may fall. One expects the first effect to dominate initially, however, because a large aggressive increase in - 12 - Goodfriend short rates exerts an immediate significant upward pull on the long rate, while the public may not yet have confidence in the disinflation. If the Fed persists with sufficiently high short-term real rates, however, inflation and real growth eventually slow and the Fed can tentatively bring rates down somewhat. A declining long rate, at this point, would suggest that the Fed's disinflation has acquired some credibility. Inflation Scares I call a significant long rate rise in the absence of an aggressive funds rate tightening an "inflation scare", since it reflects rising expected long-run inflation.7 Inflation scares are costly because the higher inflation that they signal reduces the efficiency of the payments system, with negative consequences for employment, productivity, and economic growth. Moreover, scares are costly because they present the Fed with a difficult dilemma. Resisting them requires the Fed to raise real short rates with potentially depressing effects on business conditions. But failing to respond promptly creates a crisis of confidence that encourages the higher inflation to materialize: workers and firms ask for wage and price increases to protect themselves from higher expected costs. In short, by hesitating, the Fed sets in motion higher inflation that it is then inclined to accommodate with faster 7. Since short maturity rates are anchored to the Federal funds rate target, they cannot convey as clear a signal of inflation expectations as the long rate. See Dotsey and King (1986) for more on the informational implications of interest rate rules. - 13 - Goodfriend money growth. The record of rising inflation and disinflation reviewed below contains examples of scares that resulted in higher money growth and inflation, as well as those that were successfully resisted by the Fed.8 A REVIEW OF INTEREST RATE POLICY This study focuses on the period of disinflation beginning in October 1979. Nevertheless, I begin my review by briefly describing conditions in the immediately preceding years. For the most part, the data discussed throughout come from charts and tables included at the back of the paper. Rising Inflation: the Late 1970s Inflation was rising gradually in the late 1970s, with rates of 6.9%, 7.9%, and 8.6% in 1977, 1978, and 1979 as measured by fourth quarter over fourth quarter changes in the GDP deflator. The corresponding real GDP growth rates were 4.5%, 4.8%, and 2.5%. The persistent inflation scare throughout the late 1970s carried the 30-year government bond rate from 7.8% in early 1977 to 9.2% by September 1979. Over the same period, the Fed steadily increased the Federal funds rate from around 4.7% to 11.2%, raising short-term real rates from a range between 0 to -2% to between 0 and +2%. The negative short-term real rates at the 8. An inflation scare may be consistent with either a positive or a negative association between money or prices, on one hand, and unemployment or real growth on the other, depending on the nature of the underlying macroshock that sets it off. . H . Goodfriend beginning of the period suggest that initially the Fed was actively encouraging inflation in order to stimulate real growth, though the steady increase in real short rates indicates a modest effort to resist inflation. Aborted Inflation Fighting: October 1979 to July 1980 By the time Paul Volcker became Fed Chairman in August 1979, oil price increases following the Iranian revolution in November 1978 greatly worsened the inflation outlook. Oil prices were to double by early 1980 and triple by early 1981 from November 1978 levels, and by the fall of 1979 the Fed felt that more drastic action was needed to fight inflation. The announcement on October 6, 1979 of the switch to nonborrowed reserve targeting officially opened the period of disinflation policy. The first aggressive policy actions in this period took the monthly average funds rate from 11.4% in August 1979 to 17.6% in April 1980. Cook (1989) reports that only 1 percentage point of this 6 point rise can be attributed to automatic adjustment. Virtually all of it represented deliberate policy actions taken by the Fed to increase short-term interest rates. It was the most aggressive series of actions the Fed had taken in the post-war period over so short a time, although the 5 percentage point increase from January to September of 1973 was almost as large. For its part, the 30-year rate rose sharply from 9.2% in August to a temporary peak of 12.3% in March after which it fell back to 11.4% - 15 - Goodfriend in April. rise. A closer look reveals the sources of this sharp long rate The sharp 2 percentage point monthly average funds rate rise from September to October pulled the long rate up about 0.6 percentage points. The monthly average funds rate then held in a range between 13.2% and 14.1% through February. January 1980 later turned out to be an NBER business cycle peak, and evidence of a weakening economy caused the Fed to pause in its aggressive tightening. But with the funds rate relatively steady, the long rate jumped sharply by around 2 percentage points between December and February, indicating a very serious inflation scare. The scare was probably caused in part by the ongoing oil price rises, but the Fed's hesitation to proceed with its tightening may have contributed to the collapse of confidence. In any case, the Fed reacted with an enormous 3 percentage point increase of the monthly average funds rate in March, 1 percentage point of which was due to the automatic adjustment. The long rate hardly moved in response, suggesting that the positive effect of the aggressive rise was offset by a decline in expected inflation. Moreover, the long rate actually came down by 0.9 percentage points in April even as the Fed pushed the funds rate up another 0.4 percentage points, suggesting that the Fed had already begun to win credibility for its disinflation policy. When one considers that business peaked in January, there is reason to believe that inflation would have come down as the recession ran its course in 1980 if the Fed had sustained its high interest rate policy. The imposition of credit controls in March, however, forced the - 16 - Goodfriend Fed to abort that policy. Schreft (1990) argues persuasively that by encouraging a decline in consumer spending the credit control program was largely responsible for the extremely sharp -9.9% annualized decline in real GDP in the second quarter of 1980. Supporting her view is the fact that personal consumption expenditures accounted for about 80% of the decline in real output, more than twice its average 35% contribution in post-war U.S. recessions. Accompanying the downturn in economic activity was a sharp fall in the demand for money and bank reserves that, according to Cook (1989), caused a 4.2 percentage point automatic decline of the funds rate from April to July. The Fed enhanced the automatic easing with judgmental actions, e.g., reducing the discount surcharge, that reduced the funds rate by an additional 4.3 percentage points over this period. The sharp interest rate decline coupled with the lifting of credit controls in July led to strong 8.4% annualized real GDP growth in the fourth quarter of 1980. In spite of the credit controls, or more accurately, because the credit controls caused the Fed to interrupt its inflation-fighting effort, inflation rose through the year from an annual rate of 9.8% in the first quarter to 10.9% in the fourth quarter as measured by the GDP deflator. Aggressive Disinflation Policy: August 1980 to October 1982 It was clear in late summer and early fall of 1980 that inflationary pressures were as strong as ever. After being pulled down roughly 2 percentage points by the aggressive funds rate easing from April to - 17 - Goodfriend June, the 30-year rate rose by about 50 basis points between June and July as the Fed continued to push the funds rate down another 50 basis points. The reversal signalled an inflation scare induced by the excessively aggressive easing, and the Fed began an unprecedented aggressive tightening. Of the roughly 10 percentage point rise in the monthly average funds rate from July to December 1980, Cook (1989) attributes only about 3 percentage points to the automatic adjustment. Thus, the runup of the funds rate to its 19% peak in January 1981 marked a deliberate return to the high interest rate policy. As measured by the GDP deflator, which was rising at nearly a 12% annual rate in the first quarter of 1981, real short-term rates were a high 7% at that point. As soon as the funds rate peak had been established, however, very slow growth in Ml and bank reserves automatically put downward pressure on the funds rate. According to Cook (1989), about 3.4 percentage points of the 4 percentage point drop in the funds rate between January and March was attributable to the automatic adjustment. Since the automatic adjustment had correctly signalled weakness in the economy in the second quarter of 1980, the Fed was initially inclined to let rates fall in early 1981. However, real GDP actually grew at a 5.6% annual rate in the first quarter, and when the strength of the economy became clear, the Fed took deliberate actions to override what it took to be a false signal that disinflation had taken hold. Reversing field, it ran the funds rate back up to 19% by June, using a series of deliberate tightening actions to supplement what Cook (1989) reports - 18 - Goodfriend would only have been a 0.8 percentage point automatic funds rate rise. It was not long before the aggressive disinflationary policy began to take hold. Annualized real GDP growth was -1.7% in the second quarter of 1981. The third quarter posted 2.1% real growth, but an NBER business peak was reached in July and real growth fell to -6.2% in the fourth quarter of 1981 and -4.9% in the first quarter of 1982. Meanwhile, the quarterly inflation rate as measured by the GDP deflator fell from 11.8% in the first quarter of 1981 to the 4.5% range by early 1982. The Fed brought the funds rate down from 19% at the business peak in July to 13.3% in November and held the funds rate in the 13 to 15 percent range until summer 1982 when it brought short rates down another 4 percentage points to around 10%. The funds rate reduction through November 1981 was large in nominal terms, but when one considers that inflation had declined to the 4.5% range by early 1982, the funds rate decline actually represented a 1 or 2 percentage point rise in shortterm real rates. Thus, one should still view policy as aggressively disinflationary in early 1982. As calculated by Cook (1989), automatic adjustments accounted for only 1 percentage point of the final 9 percentage point funds rate decline in the nonborrowed reserve targeting period, which ended formally in October of 1982. This last great decline should be seen as a deliberate funds rate easing calculated to achieve a sustained reduction in inflation without excessive harm to real growth. The long rate provides a picture of the Fed's progress over the - 19 - Goodfriend nonborrowed reserve inflation. targeting period in reducing the trend rate of The 30-year rate rose about 5 percentage points from a trough in June of 1980 to its 14.7% peak in October 1981. About 2 percentage points of that rise appears to be connected with the rundown and runup of the funds rate in 1980, the remaining 3 point gain through October 1981 reflected a continuing serious inflation scare. In fact, the sharp rise in the long rate after the funds rate had reached its peak in early 1981 may have contributed to the Fed's inclination to persist with its 19% funds rate until August 1981. Moreover, the discernable declining trend in the long rate from October 1981 to August 1982 indicates that the policy was still exerting disinflationary pressure. When the Fed finally decided to relax its disinflation policy by dropping the funds rate by over 4 percentage points in the summer of 1982, the long rate fell by around 3.5 percentage points along with it. We can decompose this last decline in the long rate into a purely cyclical component and an inflation expectations component using evidence from earlier in the aggressive funds rate targeting period. The sharp 2 percentage point funds rate rise from September to October 1979 pulled the long rate up 0.6 percentage points; and the sharp 8.5 percentage point funds rate reduction between April and July 1980 pulled the long rate down 2 percentage points. Taking 25% as the fraction of cyclical funds rate policy actions transmitted to the long rate, about 2.5 percentage points of the 3.5 percentage point fall in the long rate in the summer of 1982 reflected a reduction of inflation expectations. - 20 - Goodfriend Establishing Credibility: November 1982 to Spring 1986 Real GDP growth was still poor in the second half of 1982, running -1.8% and 0.6% in the third and fourth quarters, respectively. Consequently, the Fed continued to ease after relaxing its disinflationary policy, pushing the monthly average funds rate down to 8.5% by February 1983. November 1982 turned out to be an NBER business cycle trough, and real GDP growth was 2.6% in the first quarter of 1983. But the Fed kept the funds rate around 8.5% through May while the long rate remained steady at around 10.5%. It gradually became clear, however, that a strong recovery had begun. Real GDP grew at a spectacular 11.3% annual rate in the second quarter of 1983 and at rates of 6.1%, 7.0%, 7.9%, and 5.4% in the following four quarters. The Fed reacted to the recovery by raising the funds rate from 8.6% in May to 9.6% in August 1983. But the long rate rose simultaneously from 10.5% to 11.8%, initiating a serious inflation scare only a year after the Fed had relaxed its disinflation policy. Annualized quarterly inflation as measured by the GDP deflator was 4.8% or below throughout 1983 and 1984 with the exception of the first quarter of 1984, when it was 6%. Nevertheless, the long rate embarked on a spectacular rise to a 13.4% peak in June 1984. Amazingly, this was only about a percentage point short of its October 1981 peak, even though by 1984 inflation was 4 or 5 percentage points lower than in 1981. The Fed tightened in an effort to resist the inflation scare, raising the funds rate to an 11.6% peak in August of 1984. The long - 21 - Goodfriend rate began to decline in June 1984, indicating that the scare had been contained: The 7% real short rates needed to contain the scare ultimately brought quarterly real GDP growth down to the more normal 2 to 3 percent range in the second half of 1984. The Fed then lowered the funds rate rapidly by 3.2 percentage points from August to December and held it around 8% through 1985. Meanwhile, the long rate fell about 6 percentage points from its June 1984 peak to the mid-7% range by the spring of 1986. By then, the long rate was 3 percentage points below where it had been at the start of the 1983 scare. The Fed's containment of the scare apparently made the public confident of another 3 percentage point reduction in the trend rate of inflation. Maintaining Credibility: Spring 1986 to Summer 1990 Real GDP growth weakened considerably in the second quarter of 1986 to -0.3% from the strong 5.4% rate in the first quarter. With inflation appearing to have settled down in the 4% range, the Fed moved to encourage real growth by dropping the funds rate to the mid-6% range. Strong real growth in 1987 was accompanied by still another inflation scare in which the long rate rose about 2 percentage points from around 7.5% in March to 9.6% in October. Although real GDP growth was very strong throughout the year, this time the Fed responded to the scare with only a relatively modest increase in the funds rate. As it happened, the October stock market crash contained the scare somewhat, but the long rate remained above 8%. - 22 - Goodfriend With real growth still reasonably strong in 1988, the Fed proceeded to raise the funds rate sharply from the 6 to 7% range in early 1988 to a peak of 9.8% in March 1989. Though there was some evidence of a modest rise in inflation in 1988, the sustained funds rate tightening during the year is unique in that it was undertaken without a rise in the long rate. A preemptive tightening may have been needed to rewerse the perception that policy had eased permanently following the stock market crash. At any rate, the result was an increase in credibility reflected in a further decline in the long rate in 1989. Though that fall was partially reversed in early 1990, a gently declining trend in the long rate was discernable by then, indicating growing confidence on the part of the public in the Fed's commitment to low inflation. The 1990-91 Recession The period of weak real growth in 1989 ending in an NBER business cycle peak in July 1990 may have been partly due to the high real short rates. Temporary oil price increases following the invasion of Kuwait, however, also helped account for the near zero real growth in the third quarter of 1990, -3.9% real growth in the fourth quarter, and -2.5% in the first quarter of 1991. The Fed responded to the recession by bringing the funds rate down from slightly above 8% in the fall of 1990 to around 3% today. is remarkable that this sustained easing has not yet caused the long rate to rise, even though real short rates are now around zero. Real - 23 - It Goodfriend short rates were also about zero when excessive easing sparked the inflation scare in July 1980, but they were around 4% when excessive easing triggered the June 1983 scare, and around 3% at the time of the scare in April 1987.9 The real short rate floor at which easy policy becomes excessive no doubt varies to an extent with underlying real economic conditions such as government tax and spending policy, productivity shocks, or shifts in investment and consumer demand.10 But long rates may also be more tolerant of aggressive funds rate easing when the public is more confident of the Fed's commitment to maintain a low trend rate of inflation. OBSERVATIONS The record of interest rate policy reviewed above contains a number of empirical findings that are important for interpreting and evaluating monetary policy. This section summarizes the main findings in a series of observations. 1) Inflation scares appear to be central to understanding the Fed's management of short-term interest rates. The gradual funds rate rise from 1977 to October 1979 was undertaken in an environment of slowly rising long rates. The sharp long rate rise in early 1980, during a 4 month pause in the funds rate rise, was probably an important 9. The effect of the credit control program on consumer spending may account for the real rate getting as low as it did in 1980 before triggering a scare. 10. See, for example, the discussions in Campbell and Clarida (1987) and Poole (1988). - 24 - Goodfriend factor inducing the Fed to undertake its enormous 3 percentage point tightening in March. Sharply rising long rates in the first nine months of 1981 indicated that the Fed had yet to win credibility for its disinflationary policy, and probably contributed to the Fed's maintaining very high real short rates for as long as it did. On the other hand, the declining long rate from October 1981 to October 1982 encouraged the Fed to ease policy by indicating the public's growing confidence in the disinflation. The serious inflation scare set off in the summer of 1983 largely accounts for the runup of the funds rate to August 1984. The credibility acquired by the Fed in containing that scare yielded a 3 percentage point reduction in the long rate that allowed the funds rate to come down further too. There was no inflation scare per se when the Fed raised the funds rate in 1988. Nevertheless, that series of actions may be understood as preemptive, taken to reverse a public perception that policy had permanently eased following the stock market crash. The current funds rate easing has yet to trigger a sustained rise in the long rate, but the possibility of an inflation scare has probably limited the funds rate decline somewhat. 2) One might reasonably have expected the aggressive disinflation policy beginning in late 1979 to reduce long-term interest rate volatility by quickly stabilizing long-term inflation expectations at a low rate. Yet the reverse was true initially. Long rates turned out to be surprisingly volatile due to a combination of particularly aggressive funds rate movements and inflation scares. Amazingly, it took until - 25 - Goodfriend 1988 for the unusual long rate volatility to disappear. 3) One might also have expected the aggressive funds rate actions beginning in 1979 to be accompanied by opposite movements in the long rate. Again, the result was just the reverse. The aggressive actions moved the long rate in the same direction, apparently influencing the long rate primarily through their effect on short maturity rates. Only at funds rate peaks and troughs did the long rate move in the opposite direction. The long rate appeared to be influenced by a change in expected inflation only after sustained aggressive funds rate actions. 4) The long rate reached its peak in October 1981, indicating that it took two years for policy to reverse the rise in the trend rate of inflation. It would be a mistake, however, to conclude that acquiring credibility necessarily takes so long. On the contrary, a close look reveals that the long rate had already turned down in April 1980 while the funds rate was still rising, indicating that some credibility had been won by then. Credibility might even have been achieved sooner if the Fed had not hesitated temporarily between December 1979 and February 1980 to continue the aggressive funds rate tightening begun in October. In any case, the credit control program interrupted the disinflation policy in May 1980 and high interest rates were restored fully only in early 1981. The automatic adjustment feature of the nonborrowed reserve operating procedure then caused a sharp decline in the funds rate between January and March of 1981 that was only fully reversed by June. Thus, three interruptions account for - 26 - Goodfriend the delay in the Fed's acquisition of credibility for its disinflation policy. 5) Interestingly enough, the long rate was roughly in the same 8% range in the early 1990s as it was in the late 1970s, in spite of the 4 or 5 percentage point reduction in the inflation rate. Apparently, investors then perceived the 7 to 9% inflation rate as temporarily high, while, if anything, they perceive the current 3 to 4% rate as a bit below trend. The slowly declining long rate in the current period is indicative of the steady acquisition of credibility, but the high long rate indicates a lingering lack of confidence in the Fed. 6) The Fed appears to have remarkable latitude to push the Federal funds rate down in the recent recession without triggering a rise in the long rate. On three occasions when trying to encourage real growth in the 1980s (July 1980, June 1983, and April 1987) it could not push the funds rate more than 1 or 2 percentage points below the long rate before triggering an inflation scare; yet it pushed the funds rate 4 percentage points below the long rate in 1992. The greater flexibility to reduce short rates evident in the current recession is reminiscent of that in early post-war recessions when the Fed presumably had more credibility. Chart 2 shows that the funds rate was pushed almost 3 percentage points below the long rate during the August 1957 - April 1958 recession before the long rate began to rise. The funds rate came down over 2 percentage points below the long rate in the April 1960 - February 1961 recession without much of a - 27 - Goodfriend rise in the long rate.11 7) The preceding observation suggests a powerful argument in favor of a Congressional mandate for price stability. By reducing the risk of inflation scares, such a mandate would free the funds rate to react more aggressively to unemployment in the short run. Thus, a mandate for price stability would not only help eliminate inefficiencies associated with long-run inflation, but the added flexibility conferred on the funds rate might improve countercyclical stabilization policy as well. 12 CONCLUSION The paper used institutional knowledge of Fed policy procedures, simple economic theory, and the inflation scare concept to analyze and interpret interest rate policy as practiced by the Fed since 1979. It focused on the primary policy problem during the period: the acquisition and maintenance of credibility for the commitment to low inflation. We saw that the Fed might have acquired credibility for its disinflation relatively quickly in early 1980 had it been able to sustain a high interest rate policy then. After all, long term rates were roughly equal to the inflation rate in 1979, indicating that the public believed 11. Kessel (1965) contains a good description and analysis of the cyclical relation between long and short rates. 12. See Black (1990) for a discussion of the benefits of price stability. Hetzel (1990 and 1992) discusses a proposal that the U.S. Treasury issue indexed bonds to provide a better indicator of long-run inflation expectations. - 28 - Goodfriend inflation was only temporarily high at the time. Unfortunately, a series of interruptions delayed the actual disinflation for two years, probably raising the cost in terms of lost output of acquiring credibility. Soon after relaxing its disinflation policy in 1982, the Fed's credibility was again challenged with a serious inflation scare that carried the long rate up from 10.5% to 13.4%. It took 11 months and 7% real short rates to contain the scare, indicating how fragile the Fed's credibility was in 1983 and 1984. The long rate decline to the 7.5% range by the spring of 1986 reflected a big gain in credibility. Yet the Fed was tested by another scare in 1987 that ended with the stock market crash. The crash itself, however, then set in motion expectations of excessive easing that the Fed resisted with a 3 percentage point funds rate rise in 1988 and 1989, a tightening that probably weakened real growth somewhat in 1989 and 1990. Reviewing the policy record makes one understand how fragile the Fed's credibility is and how potentially costly it is to maintain. Even after inflation had stabilized at around 4% in 1983, inflation scares and the Fed's reaction to them were associated with significant fluctuations in real growth. With that in mind, one cannot help but appreciate the potential value of a Congressional mandate for price stability that would help the Fed establish a credible commitment to low inflation. In fact, there is evidence that an interest rate policy assisted by such a mandate would work well. Both the Bundesbank and the Bank of Japan follow interest rate policies resembling the Fed's and - 29 - Goodfriend yet, for the most part, they have achieved better macroeconomic performance. Perhaps it is because they each enjoy a stronger mandate for price stability than does the Fed. - 30 - Goodfriend REFERENCES Annual Reports of Open Market Operations, Federal Reserve Bank of . New York Economic Review. 1981-83. Barro, Robert J., "Interest Rate Targeting." Journal of Monetary Economics. January 1989, 23, 3-30. Bernanke, Ben and Alan Blinder, "The Federal Funds Rate and the Channels of Monetary Transmission," American Economic Review. September 1992, 82, 901-21. Black, Robert P., "In Support of Price Stability," Federal Reserve Bank of Richmond Economic Review. January/February 1990, 3-6. Campbell, John Y. and Richard H. Clarida, "The Dollar and Real Interest Rates," Carnegie-Rochester Conference Series on Public Policy. 1987, 27, 103-40. Christiano, Lawrence and Martin Eichenbaum, "Liquidity Effects, Monetary Policy, and the Business Cycle," Federal Reserve Bank of Minneapolis, June 1991. Cook, Timothy, "Determinants of the Federal Funds Rate: 1979-1982," Federal Reserve Bank of Richmond Economic Review. January/February 1989, 3-19. Cook, Timothy and Thomas Hahn, "The Effect of Changes in the Federal Funds Rate Target on Market Interest Rates in the 1970s," Journal of Monetary Economics. November 1989, 24, 331-51. Dotsey, Michael and Robert G. King, "Informational Implications of Interest Rate Rules," American Economic Review. March 1986, 76, 33-42. Fama, Eugene F., "The Information in the Term Structure," Journal of Financial Economics. 1984, 13, 509-28. Fama, Eugene and R. Bliss, "The Information in Long Maturity Forward Rates," American Economic Review. 1987, 77, 680-92. Friedman, Benjamin M., "Lessons on Monetary Policy from the 1980s," The Journal of Economic Perspectives. Summer 1988, 2, 5172. Friedman, Milton and Anna J. Schwartz, A Monetary History of the United States Princeton: Princeton University Press, 1963. - 31 - Goodfriend Goodfriend, Marvin, "Discount Window Borrowing, Monetary Policy, and the Post-October 6, 1979 Federal Reserve Operating Procedure," Journal of Monetary Economics. September 1983, 12, 343-56. , "Interest Rate Smoothing and Price Level Trend-Stationarity," Journal of Monetary Economics. May 1987, 19, 335-48. , "Interest Rates and the Conduct of Monetary Policy," Carnegie-Rochester Conference Series on Public Policy. Spring 1991, 34, 7-30. Goodfriend, Marvin and William Whelpley, "Federal Funds: Instrument of Federal Reserve Policy," Federal Reserve Bank of Richmond Economic Review. September/October 1986, 3-11. Hardouvelis, G., "The Predictive Power of the Term Structure During Recent Monetary Regimes," Journal of Finance. 1988, 43, 339-56. Hetzel, Robert L., "Indexed Bonds as an Aid to Monetary Policy," Federal Reserve Bank of Richmond Economic Review. January/February 1992, 13-23. , "Maintaining Price Stability: A Proposal," Federal Reserve Bank of Richmond Economic Review. March/April 1990, 53-55. Kessel, Reuben A. The Cyclical Behavior of the Term Structure of Interest Rates. New York: National Bureau of Economic Research, 1965. Levin, Fred J. and Paul Meek, "Implementing the New Operating Procedures: The View from the Trading Desk," in New Monetary Control Procedures, edited by Stephen H. Axil rod, Washington: Board of Governors of the Federal Reserve System, 1981. Macau!ay, Frederick R., Bond Yields. Interest Rates, and Stock Prices. Cambridge, MA: National Bureau of Economic Research, 1938. Mankiw, N. Gregory, Jeffrey Miron, and David N. Weil, "The Adjustment of Expectations to a Change in Regime: A Study of the Founding of the Federal Reserve," American Economic Review. June 1987, 77, 358-74. McCallum, Bennett T., "Price Level Determinacy with an Interest Rate Policy Rule and Rational Expectations," Journal of Monetary Economics. November 1981, 8, 319-29. - 32 - Goodfriend Miron, Jeffrey, "Financial Panics, the Seasonality of the Nominal Interest Rate, and the Founding of the Fed," American Economic Review, March 1986, 76, 125-40. Moore, George, "Forward-Looking Bond Yield Approximations," Board of Governors of the Federal Reserve System, September, 1989. Poole, William, "Monetary Policy Lessons of Recent Inflation and Disinflation," The Journal of Economic Perspectives, Summer 1988, 2, 73-100. Rogoff, Kenneth, "Reputation, Coordination, and Monetary Policy," Carnegie- Rochester Conference Series on Public Policy, Spring 1987; reprinted in Robert J. Barro, editor, Modern Business Cycle Theory, Cambridge,MA: Harvard University Press, 1989, 236-64. Schreft, Stacey L., "Credit Controls: 1980," Federal Reserve Bank of Richmond Economic Review, November/December 1990, 25-55. Sims, Christopher, "Interpreting the Macroeconomic Time Series Facts: The Effects of Monetary Policy," Yale University, August 1991. Thornton, Daniel, "The Borrowed-Reserves Operating Procedure: Theory and Evidence," Federal Reserve Bank of St. Louis Economic Review, January/February 1988, 30-54. Wallich, Henry C , "Recent Techniques of Monetary Policy," Federal Reserve Bank of Kansas City Economic Review, May 1984, 21-30. Woodward, G. Thomas, "The Real Thing: A Dynamic Profile of the Term Structure of Real Interest Rates and Inflation Expectations in the United Kingdom, 1982-89," Journal of Business, July 1990, 63, 373-98. - 33 - Chart 1 FEDERAL FUNDS RATE AND 30-YEAR BOND RATE January 1977 - April 1992 20 30-Year Bond Rate 3 '|Mlllllttll|IIIUttltMIIIIIIIIIIII|HMttlMII|IIHMIMM|llMltlllll|IIMIIIIIIt|IIIHIIIIII|MIIIIIIIM|lltlHttMIIIIIIMIIIM|IIIIMIMIt|IMn 77 P«rc«nt Chang*. 40 to 40: Rttl OOP 4.4 I l l i c i t Prlct Doflitor 7.3 78 79 80 6.0 0.0 -0.2 6.4 0.7 10.1 81 82 83 84 85 86 87 88 89 90 91 -0.1 -i.l 6.7 4.9 3.3 2.2 4.5 3.3 1.7 -0.1 0.3 0.4 4.4 4.0 4.3 3.6 2.6 3.3 4.2 4.2 4.2 3.0 92 Chart 2 FEDERAL FUNDS RATE AND 20-YEAR BOND RATE August 1954 - December 1964 4.5 4.0 + 20-Year Bond Rate 3.5 + E | 3.0 c < c a 2.5 -M C 0) u t- * ^ cu 2 . 0 CL 1.5 + 1.0 + 5 ' I » III |I »I U M M M \ t1 M 111 M 1111 M 111 1 III11 I I 111 111 M 11 M t M I i M M |»It < M M N I ( M M M I M M | M I M II M H 11 M M I H 1 11II 55 56 57 58 59 60 61 62 63 i 0 C O o i N i n i n s N < ) n c D o o i n N N O N N N < f < r H O ) a ) a ) M n o o o ) a ) OCOOKIO)(OOOCO(DOOOOCOCOC08) 000)0)0)(BC0(D(DC0C0NNOC0C0 O O H O O N O H H ) < W 0 2 O ' , ) h Z < I 1 1 < e i 0 Z O ' , ) & . r < r n ' l < W 0 Z Q 1 b . I < (DcoeotooacoNNiotfxo H < n e o e o c o m N O ) o o o i n « ( M N N N H N H H N H c o n o ) N H a ) ^ 0 ) o o i o ^ N c o « o o o ) h o>a>CBO>CDa>cncoa>coeoeococoi 1 l i 4 l < S ^ r\rsNNrNr»cooo(orsr<>(OiOiO(Oiotntn(Oioiou)(OiOiO(OiO(ONr«(OtoiOiO(l)(ONNrsoooo(D(Ooo N N N i f l N o m m o o i r i N O H n n N m r*. i-t r»* M n m o u > « H O t f > ^ o ( 0 ( o a > N r ^ r ^ ( n c M o c ^ ^ a > < « > t n i n G o r s . ^ f ^ . a > t H t n c 4 c n r ) c n sr «e co <» . . . . o « < n i n i o i n o i n « a a ) n t n i n N n ( O N t n n n « n i n N N i n ( D a > i n a ) 0 ) H c o < r < o ( 0 0 ( 0 0 ( 0 0 ) 0 ) 9 >(00(0<DO)0)OIIO)0(00)0) OOOOOOCBCDflDi CO OO moeoNNnfflONO) mN<(oeooiiflNiOKO>ifl*HC)0(OMn(n m m o N O) t n eo o> CD o> O N H ( o * o i ( O O i i n H ( o e o o o ) * H H r ) O N 5 sss: § 51 ,3t 9 I a?. •-)U43:<2:,-5,n<woa5Q'-ju«s<z,^,n<wozQ'->ui3:<2:«^n<wozQ»^u.3:<i:»->t-5<wozQ n o n n o N n ( o m c o m i n ( O i n n « H « 0 ) ( O i o N (0^ma><a-cotDmr«.aorva>ro(04'^c4iocMcointn N N N n n N n 4 « « n n « « n n n n n N N H o o o o o o o o H H H H H H H H c s N i n n n N N H H H ) rs» m o ra o co n H to 0 ) O O O N N H O 0 I O H H H N N Z < I ^ n < M O 2 Q n h Z < Z ^ n < « O Z O i N N ( 0 ( o < f m i n o ) N H H o i n t o H N o n c o N i o i n ( D 4 N i o o ) H O ) N i o n ^ o o ) n en oj rv.(© Q i ^ H i n H i n N N a u j i n N s i o c n n i n ^ ^ n ^ i n i n o t N n o N t t n a ^ 0 ) i n « n o a o ) N < n n n N n ^ 4 < 7 ^ < f N o o o ) o > o « o c o e o e o ( o c o o ) a ) a a > o ) 0 ) 0 ) C D O ) o o H H H H O ) 0 ) 1 h I < l n i < I Q O 2 D n t u I < I ' m < t 0 O S Q ^ U COCOCOaOa0CDCDC0eOC00DCOCOO)CDO>O)COCOCO •-•<oo>fn«no>c>JO^,r^»Hto (OiO(DN(on40)H^inini ^^^^inininintf>(DU>(C^tO(0(0^^rs.(DeococoooooooooOf^<r>r)cnr>^-r^r^ocococoor4w^a> •^b«z<z,^,-)<woaBQ,->ui2:<zn«-)<coozQr)u*2:<s:^)r)<wo*Q^u*2:<s:^,^<wozQ Table 2 FEDERAL FUNDS HATE AHD 20-YEAR GOVERIMEBT BOND SATE August 1954 - December 1964 Federal 20-Year Funds Govt. Band Rate Rate (Percent per Annum) 1954: 1955: 1956: 1957: 1958: 1959: A S 0 N D J F M A M J J A S 0 N D J F M A M J J A S 0 N D J F M A M J J A S 0 N D J F M A M J J A S 0 N D J F M A M J J A S 0 N D 1.21 1.07 0.90 0.91 1.26 1.37 1.29 1.35 1.43 1.43 1.62 1.66 1.90 2.18 2.24 2.35 2.48 2.44 2.50 2.50 2.62 2.75 2.71 2.74 2.74 2.95 2.96 2.88 2.94 2.93 3.00 2.96 3.00 3.00 3.00 2.99 3.24 3.50 3.50 3.22 2.98 2.72 1.67 1.20 1.26 0.63 0.93 0.68 1.53 1.76 1.80 2.27 2.42 2.48 2.40 2.80 2.96 2.90 3.39 3.44 3.50 3.76 3.98 4.00 3.99 2.58 2.60 2.61 2.65 2.67 2.75 2.83 2.84 2.85 2.87 2.86 2.94 3.01 3.00 2.93 2.93 2.98 2.94 2.91 2.99 3.14 3.06 3.00 3.08 3.22 3.28 3.26 3.37 3.45 3.41 3.30 3.32 3.40 3.49 3.65 3.72 3.75 3.73 3.76 3.61 3.38 3.27 3.31 3.29 3.17 3.17 3.23 3.39 3.65 3.80 3.81 3.76 3.86 3.95 3.96 3.99 4.06 4.13 4.14 4.16 4.15 4.29 4.19 4.20 4.33 Federal 20-Year Funds Govt. Bond Rate Rate (Percent per Annum) 1960: 1961: 1962: 1963: 1964: J F M A M J J A S 0 N D J F M A M J J A S 0 N D J F M A M J J A S 0 N D J F M A M J J A S 0 N D J F M A M J J A S 0 N D 3.99 3.97 3.84 3.92 3.85 3.32 3.23 2.98 2.60 2.47 2.44 1.98 1.45 2.54 2.02 1.50 1.98 1.73 1.16 2.00 1.88 2.26 2.62 2.33 2.14 2.37 2.70 2.69 2.29 2.68 2.71 2.93 2.90 2.90 2.94 2.93 2.91 3.00 2.98 2.90 3.00 2.99 3.02 3.49 3.48 3.50 3.48 3.38 3.48 3.48 3.43 3.47 3.50 3.50 3.42 3.50 3.45 3.36 3.52 3.85 4.42 4.28 4.14 4.23 4.20 4.04 3.91 3.84 3.86 3.92 3.96 3.91 3.90 3.84 3.81 3.81 3.74 3.89 3.93 4.04 4.04 4.01 4.00 4.07 4.10 4.12 4.04 3.93 3.92 3.96 4.05 4.01 4.00 3.94 3.93 3.92 3.94 3.97 3.98 4.03 4.02 4.02 4.06 4.03 4.09 4.12 4.16 4.19 4.19 4.17 4.22 4.24 4.20 4.17 4.16 4.18 4.20 4.20 4.17 4.18 Table 3 QUARTERLY CHANGES IH REAL GDP AHD GDP IMPLICIT PRICE DEFLATOR (Seasonally Adjusted Compound Annual Rates) 1Q 1977 - 1Q 1992 Real GDP (Pareant) 1977: 1978: 1979: 1980: 1981: 1982: 1983: 1984: 1985: 1986: 1987: 1988: 1989: 1990: 1991: 1992: 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 6.0 6.9 5.7 -0.8 2.8 13.5 3.1 4.8 0.1 0.4 2.5 0.7 1.7 -9.9 0.1 8.3 5.6 -1.7 2.1 -6.2 -4.9 1.6 -1.8 0.6 2.6 11.3 6.1 7.0 7.9 5.4 2.2 2.7 2.7 3.2 5.2 2.3 5.4 -0.3 2.3 1.3 3.0 5.1 4.0 5.9 2.6 4.3 2.5 3.9 2.5 1.9 1.1 1.2 1.7 1.6 0.2 -3.9 -2.5 1.4 1.8 0.4 2.4 IaaiUcit Price Deflator (Percent) 6.1 8.4 7.4 7.3 5.7 10.7 8.3 8.8 8.6 8.4 9.6 8.1 9.8 9.6 10.0 10.9 11.8 7.5 9.6 8.8 4.5 5.5 4.4 3.4 4.8 2.8 4.2 4.2 6.0 4.1 4.5 2.6 4.9 3.0 2.6 3.9 2.1 2.1 2.9 3.3 3.3 2.9 3.3 3.6 3.6 4.4 5.1 3.9 5.4 4.2 3.4 3.7 4.4 4.4 4.7 3.2 5.0 3.1 2.1 1.7 3.1 COMMENTS ON "INTEREST RATE POLICY AND THE INFLATION SCARE PROBLEM: 1979-1992" R. Alton Gilbert Marvin Goodfriend states his goal in writing this paper as follows: M The goal is to distill observations to guide future empirical and theoretical analysis of monetary policy with the ultimate objective of improving macroeconomic performance." expect this goal to be realized. I I expect references to this paper as justification for using the federal funds rate as the measure of monetary policy actions and references by people who do theoretical and empirical work on credibility of central bank commitments to the control of inflation. One reason the paper will be attractive to others is that it provides simple indicators of monetary policy actions and the credibility of monetary policy. The federal funds rate is the indicator of monetary policy actions; a rise (decline) in the federal funds rate indicates a tightening (easing) of policy. The long-term government bond rate is a measure of the credibility of Federal Reserve policy to control inflation. Changes in the long-term rate are interpreted as changes in long-term inflation expectations. A rise in the long-term government bond rate indicates that the commitment of the Federal Reserve to contain inflation in less credible to investors, and a fall in the rate indicates greater credibility. The paper uses these indicators of policy actions and credibility of policy to examine the conduct of monetary policy from October 1979 to 1992. The purpose of my comments is to illustrate some problems in applying these simple indicators of monetary policy in interpreting specific events. Problems with using the federal funds rate as a measure of monetary policy actions are well known. Monetary policy actions may be inflationary even if the federal funds rate is rising, and policy actions may be deflationary even if the federal funds rate is falling. 1 In many situations changes Gilbert in interest rates must be supplemented with data on monetary aggregates to avoid errors in interpreting monetary policy actions. INDICATORS OF MONETARY POLICY IN 1983-84 The problem of judging whether a rise in the federal funds rate indicates a tightening of monetary policy can be illustrated by referring to events in 1983-84. Goodfriend refers to the rise in the federal funds rate from 8.6 percent in May 1983 to 9.6 percent in August 1983 as a tightening of monetary policy. This rise in the federal funds rate was accompanied by a rise in the long-term interest rate. Thus, while the rise in the federal funds rate from May to August of 1983 is characterized as a tightening of monetary policy, it was not effective in ending the inflation scare. It took an additional rise of 2 percentage points in the federal funds rate in the following year to begin reversing the rise in the long-term rate. Table 1 supplements the data on interest rates from Goodfriend's paper with growth rates of Ml. Growth rates of Ml reflect the policy actions of the Federal Reserve, through reserve requirements and the effects of policy actions on reserves. During the spring and summer of 1983, Ml was rising rapidly. was not a period of restrictive monetary policy. This The rise in the federal funds rate from May to August of 1983 reflects the effects of an economic expansion on interest rates, rather than the" effects of restrictive policy actions of the Federal Reserve. The federal funds rate peaked in the summer of 1984, when the Federal Reserve brought money growth to a halt. Reserve begin tightening monetary policy? So when did the Federal Adding information on Ml growth indicates that June 1983 is too early. 2 Gilbert INDICATORS OF MONETARY POLICY IN 1979-81 Table 2 presents the same data for the years 1979-81. I am going backwards in time in examining 1979-81 because interpretation of movements in interest rates in this period is more complex than in the 1983-84 period. Goodfriend's Analysis First I present my understanding of Goodfriend's analysis. The Federal Reserve began tightening monetary policy in August 1979, but the large increase in the federal funds rate in March 1980 was necessary to gain credibility for the anti-inflation policy of the Federal Reserve. Declines in long-term rates after March 1980 reflect greater credibility of anti-inflation policy. Declines in the federal funds rate in May, June and July of 1980, however, undermined the credibility of the Federal Reserve's anti-inflation policy, causing the long-term rate to begin rising again in July 1980. The Federal Reserve began tightening policy again in August 1980, but long-term rates did not peak until October 1981, two years after the Federal Reserve began its anti-inflation policy. Because the Federal Reserve temporarily abandoned its tightening of policy in the spring and summer of 1980, it took longer than it might have to reverse the inflation scare in 1981. An Alternative View My alternative explanation for movements of interest rates in 1980 that focuses on the effects of the credit control policy, which was imposed in March 1980 and removed in July 1980. This alternative explanation does not require assumptions about the Federal Reserve gaining and losing credibility for its 1. For information on the credit control program and its implications for the conduct of monetary policy in 1980, see Gilbert and Trebing (1981). 3 Gilbert anti-inflation policy over periods of a few months. This alternative explanation does not require labeling monetary policy as easing when the money stock was falling sharply and tightening and the money stock was rising rapidly. Imposition of credit controls caused a sharp drop in the demand for credit, which caused the declines in short-term and long-term interest rates. This decline in the demand for credit was accompanied by a sharp decline in the money stock, especially in April 1980, because the operating procedure used at the time 2 tended to be procyclical. These declines in interest rates were reversed in the summer of 1980, when the credit controls were removed. Again, with a procyclical operating procedure, the money stock rose rapidly after credit controls were removed. Using Ml as the indicator of monetary policy actions, there is a much shorter lag between the tightening of monetary policy and the peak of long-term interest rates in 1981. During much of the period from August 1980 through October 1981, Ml growth was rapid. The Federal Reserve did not consistently slow money growth until May 1981, and long-term interest rates peaked in October 1981. It is difficult to determine the degree to which the decline in long-term interest rates that began in the fall of 1981 reflected lower expectations of long-run inflation. The economy was in a severe recession by the fall of 1981, and the decline in demand for credit must have depressed long-term rates to some extent. CONCLUSIONS I conclude my comments by considering their implication for interest rates as indicators of monetary policy. I find the the 2. See Gilbert (1985) for a general description of the nonborrowed reserves operating procedure used from the fall of 1979 to the fall of 1982. 4 Gilbert federal funds rate to be an unreliable indicator of monetary policy actions. In some cases the federal funds rate rose (fell) while the money stock rose rapidly (declined sharply). When money growth is included as an indicator of monetary policy, there is a shorter lag between the tightening of monetary policy and the following peaks of long-term interest rates. Finally, changes in long-term interest reflect forces in addition to changes in the credibility of Federal Reserve anti-inflationary monetary policy, including credit controls and recessions. 5 Gilbert REFERENCES Gilbert, R. Alton. "Operating Procedures for Conducting Monetary Policy," Federal Reserve Bank of St. Louis Review (February 1985), pp. 13-21. Gilbert, R. Alton and Michael E. Trebing. "The FOMC in 1980: A Year of Reserve Targeting," Federal Reserve Bank of St. Louis Review (August/September 1981), pp. 2-22. 6 Gilbert Indicators of Monetary Policy, 1983-84 Month 1983 January February March April May June July August September October November December 1984 January February March April May June July August September October November December Federal funds rate 8.68% 8.51 8.77 8.80 8.63 8.98 9.37 9.56 9.45 9.48 9.34 9.47 9.56 9.59 9.91 10.29 10.32 11.06 11.23 11.64 11.30 9.99 9.43 8.38 30-year government bond rate Annual growth rate of Ml 10.63% 10.88 10.63 10.48 10.53 10.93 11.40 11.82 11.63 11.58 11.75 11.88 8.94% 14.15 16.53 11.34 14.21 9.78 12.59 6.31 6.28 12.59 3.29 3.28 11.75 11.95 12.38 12.65 13.43 13.44 13.21 12.54 12.29 11.98 11.56 11.52 9.35 3.72 8.76 8.45 5.29 9.08 1.57 0.44 8.04 -1.10 8.00 10.80 M ** tno I ro M oo <r» A oo oo 0 > M O O t O P » 0 0 0 0 * > 0 > C h t f l V J O N J S I O O O O M C O U I H P C O O vo CJ O O l ^ M v l C O ^ C O v J C D H f O C h O C D t n K J t o e o O N O U J t J O Ul H M H | k U U 1 U O K ) O O O J ^ U H O t M K) M P» I fO CO *> CO ** *> -J K J I O H M M O ^ D O M I O I O O • ^ U } O > ^ H U 1 V O O \ t O O \ 0 0 M U i U i O O « O < * 4 v 0 O i O O t 0 O 4 * N J H O ) N 1 K ) ^ O K ) K ) O U Q ) U U ^ ^ ^ U ( O U U K ) K } K ) O t A > t 4 U X > X * t < - 4 o o ) c c c p » ^ d p » a ) p ) c t » d » Q H - 0 K M H C r 3 O f t C < » P- O h C I T ( D CO H ? C H ID p rt pi h CD >< O O U I K J O ^ V O V O O S I N J ^ U (D CD && (D O o < <D CD O 5S 10ODOOOOO>O**IOC»MP»OO O U l H v J H l J v J 0 0 P \ O U I O (D 0> c p» »d P* CD HO h C h- 3* C Pi pi h > c c x >x O W O CD f t H3 c c O f t C *< t r i D CD (0 3 H & «D U U O C O O O O H U l s J v J ^ O O Z O < (D g & CD h 00 O K ) t J U 1 ( f l v J v O U ) 0 0 U 1 ^ U l i O D CD O CD g & ID h 00 J C A I C J M rt C K ID « 3 ct & CD *1 O O CD O O O O O O p- O h C p ? d » pi h H *< *< n ( D C c c p i » d p i c D p i O tr ID * v O L J t / i o o r o i o o o o o t t J O o t O H U l N j ^ ^ U O K J O O ^ i O dp H O I O S J N J K I O ^ O O V O ^ M M U J O O M t O t O v O M O O O t O t O O U 1 < J O O C j r o \ 0 O O U l O 4 * dP OOlOV0OOOOOOVOv0\0lOOO dp « O M - J * > \ 0 ^ K > r O O O O O C D 55 CD O O < CD w CD w a 3 & & CD CD •1 h M O ft 3 CD C pi O M 0* io O O l*i a <: o a (D i h »< n 3 • Pi 3 P» f t CD * 1 CD 3 ft hi pi PI H» ft CD 3 CD a a n CD -J vO I 00 SO o rPo o 3 CD ft P» n •< X M 3 a Po pi ft o h a ro CD n ft tr o p- INTEREST RATE OPERATING PROCEDURES OF FOREIGN CENTRAL BANKS John Morton and Paul Wood The interest rate operating procedures employed by central banks in the major industrial countries in implementing monetary policy over the past decade have varied considerably. This variation has not only been among countries at any moment in time, but also often within countries over time. The main aim of this paper is to investigate the experience of these countries with various interest rate operating procedures, both descriptively and in terms of statistical measures, in light of their stated intentions and goals in terms of monetary policy implementation. The first section of the paper lays out some possible criteria for selection of an interest rate operating procedure, and some of the implications of choosing different monetary policy targets. The second section describes in more detail the recent experiences of six industrial countries (Japan, Germany, France, the United Kingdom, Switzerland, and Canada) with respect to interest rate operating procedures. The third section presents and discusses some statistical measures of this experience. INTEREST RATE OPERATING PROCEDURES Questions regarding the desirable properties, characteristics, and criteria of selection of an interest rate operating procedure can be usefully divided into two broad categories. The first category involves the choice of target variable, or variables, at which interest rate policy is aimed. These targets could be either ultimate macroeconomic targets, such as output, unemployment, or inflation, or intermediate targets, such as growth of one or more monetary aggregates or exchange rates. The second general category of factors involves the choice of a 1. Division of International Finance, Board of Governors of the Federal Reserve System. We would like to thank Karen Johnson and Ralph Smith for useful comments, and James Heil for research assistance. Morton and Wood particular interest rate operating procedure, given the selection of a particular target. In practice, of course, both categories are interrelated. The choice of a policy target, which then becomes the focus of the interest rate operating procedure, can depend on a variety of factors. On a theoretical level, these would include prominently the likely source of economic disturbances, focussing on whether disturbances originate mainly from real or monetary factors, or from domestic or foreign sources. The possible choice of a monetary target would additionally depend importantly on the stability of monetary relations, in particular the demand for money. Political or institutional constraints can also be important in deciding on an exchange rate target, the most prominent recent example being possible membership in the European Monetary System (EMS). The individual country experiences described in the next section, while exhibiting substantial variations, do appear to show some general trends with regard to choice of policy targets. Over the past decade there appears to have been a general, though not universal, movement away from monetary targets and, for some countries, a movement towards exchange rate targets. This trend is most evident, of course, among EMS members, other than Germany. There also appears to have been a general tendency to adopt a more flexible, ad hoc approach to targets, with a variety of targets having shifting relative weights under different circumstances. The likely relationship between choice of a particular policy target and the variability of interest rates used to achieve that target is unclear. The outcome would depend on such factors as the main source of economic disturbances and the strictness with which policy targets are adhered to. In general, an intermediate target, such as monetary aggregate growth, might be more strictly followed, in the sense of a change in the target variable triggering a prompt and at times large change in interest rates. In this case, adoption of such a procedure might be expected to result in a more variable interest rate path. However, such a procedure might, over time, bring greater stability, - 2 - Morton and Wood eliminating possible discretionary swings in policy, and reducing interest rate variability. Given the choice of a particular target, the question arises as to the interest rate procedures used to achieve that target. Assuming that at any moment a particular interest rate level could best achieve some desired level of the target, an interest rate implementation procedure would seem to be desirable if it could achieve that interest rate level, and undesirable if it could not. Thus, it would appear desirable to have an interest rate procedure that was "flexible," in the sense of allowing prompt and, if needed, large changes in interest rates. Conversely, an "inflexible" system, which somehow hindered interest rate changes, would appear undesirable. As demonstrated in more detail in the next section, several countries have, over the past decade, moved to more flexible--in the sense defined above--interest rate operating procedures. These changes have sometimes been confined to the interest rate operating procedures themselves (Germany and the United Kingdom) or have taken place as part of a wider change to a more market-oriented monetary policy framework (Japan and France) . Movement to a more flexible interest rate operating procedure may involve moving to less of a reliance on the discount rate, since discount rate changes may be hindered by concerns over announcement effects. Despite the seeming general desirability of a "flexible" interest rate operating procedure, monetary authorities may at times be reluctant to adopt procedures which are seen to lead to "unstable," or overly volatile interest rates, showing a preference for a more stable interest rate path. INDIVIDUAL COUNTRY EXPERIENCES Japan The Japanese financial system has traditionally been characterized by a high degree of government control and restrictions. In terms of monetary policy, authorities have relied heavily on discount rate changes and quantitative controls. Over the past decade, this system has - 3 - Morton and Wood undergone substantial liberalization, featuring new financial instruments, more international openness, interest rate decontrol, and less reliance on the discount rate and more reliance on open market operations. A variety of pressures have encouraged the move to financial liberalization. An increase in government deficits starting in the mid- 1970s eventually led to the breakup of the system whereby banks were forced to accept government debt at below-market rates. The first real open market to emerge in the 1970s was the gensaki market, a repurchase market for government bonds. Market pressures to break down restrictions in domestic financial markets also came from abroad, particularly the United States. Various foreign exchange restrictions were reduced starting in 1980, and the Euroyen market grew rapidly in subsequent years. The process of financial liberalization, which started later in Japan than in any other of the major industrial countries, with the possible exception of France, gained real momentum after the mid-1980s. Banks had been permitted to issue negotiable certificates of deposit in 1979. In 1985, money market certificates, yen-denominated bankers acceptances, and large denomination time deposits were introduced. Treasury bills first appeared in 1986, and commercial paper in 1987. At the beginning of the 1980s, the Bank of Japan relied heavily on the discount mechanism to regulate credit conditions. Most interest rates were tied, formally or informally, to the discount rate. The Bank supplied credit to the market almost exclusively through changes in discount window lending. The discount rate was kept well below market interest rates, meaning that there was always an excess demand for discount borrowing. The Bank of Japan decided each day how much discount lending to make, both in total and to individual banks, effectively rationing credit. lending. The Bank also imposed ceilings on the growth in bank However, as more and more financial transactions took place at market-determined interest rates, and non-bank sources of credit grew in - 4- Morton and Wood importance, this system became increasingly less efficient. In response,, starting about 1988, Japanese authorities adopted a policy of relying increasingly on open market operations and less on discount window lending as a way of supplying reserves to the banking system. They also looked to day-to-day control of the overnight rate as the main interest rate control mechanism, giving less importance to the discount rate. These shifts in operating procedure are still incomplete and 3 ongoing. The Bank of Japan appears to have maintained an interest rate target rather than a monetary aggregate target over the past decade, i.e., monetary authorities appear to have varied interest rates in response to changes in macroeconomic targets, such as output and inflation, rather than money supply growth. The Bank of Japan announces H forecasts'* for M2+CD growth, but does not appear to treat these as targets. For one thing, the forecast is for four-quarter growth rates but is only announced at the beginning of the end-point quarter, meaning that much of the forecast is already history. Also, in recent years, as money growth has fluctuated sharply, M2+CD forecasts have been varied in line with actual data, rather than changing more slowly, as would more likely be the case were they treated as targets. Germany Entering the 1980s, the Bundesbank relied primarily on the Lombard window to extend credit to banks. Borrowing at the Lombard window is done at the initiative of banks. The rate on those loans, the Lombard rate, is a rate set by the Bundesbank and adjusted infrequently, often in conjunction with an equal movement in the discount rate. During the first half of the decade, the call money rate tended to be near the Lombard rate. When bank reliance on Lombard borrowing became too heavy, the Bundesbank would attempt to gain control over Lombard borrowing by 2. By the end of 1991, over 60 percent of city bank deposits carried market rates of interest. 3. The Bank of Japan still relies mainly on varying the amount of discount window lending, rather than open market operations, for daily adjustments of credit availability. - 5 - Morton and Wood putting quantitative limits on what banks could borrow at the Lombard window or by suspending the Lombard rate and substituting a special Lombard rate that could be changed daily. The Bundesbank began to use repurchase agreements (RPs) in the early 1980s and increasingly met bank liquidity needs with RPs beginning in 1983. The rates on RPs generally exceeded the Lombard rate during this time. In 1985, in response to excessive use of the Lombard window, the Bundesbank raised the Lombard rate above the RP rate to make the Lombard window the borrowing source of last resort for banks. Since 1985, the call money rate has tended to track the RP rate, somewhere between the discount rate and the Lombard rate. The RP rate has been moved around more than the Lombard rate, which is only adjusted a few times a year. The Bundesbank effectively targets the call money rate. Funds are injected into and withdrawn from the market through weekly auctions at which RPs are tendered with a maturity of approximately one month, with two-month RPs also offered on occasion. Sometimes RP funds are offered at a fixed rate announced in advance. More commonly, the funds are auctioned at a rate sufficient to clear the market. The Bundesbank sets the quantity tendered after it observes the bids, so it has some control over the repurchase rate. The Bundesbank can also inject funds through emergency short-term RP tenders and by moving Treasury funds held at the Bundesbank into commercial banks. Germany has had a monetary target continuously since 1975. For most of this period, the target was stated in terms of central bank money (CBM), a weighted sum of the components of M3. itself became the targeted aggregate. Starting in 1988, M3 In terms both of success in achieving targeted aggregate growth and the apparent importance attached to this goal, the Bundesbank seems to have been relatively committed to monetary aggregate growth as a policy goal. A complication to monetary targeting was introduced by the monetary union of eastern and western Germany in 1990. Since then, interpretation of monetary aggregate changes has become considerably more difficult and ambiguous. - 6- Morton and Wood France More than any other major industrial country, with the possible exception of Japan, France had tightly regulated financial markets moving into the 1980s. There were extensive foreign exchange controls, a number of financial instruments were either officially or effectively prohibited, many interest rates (including bank deposit rates) were regulated, and monetary control was exerted largely through ceilings on bank credit growth. Starting in the early 1980s, this highly regulated system began.to be liberalized and deregulated. The liberalization process was prompted partly by market pressures, arising from financial innovations and foreign competition, and partly from a deliberate government policy aimed at increasing market efficiency. Major events in this liberalization process included the 1982 introduction of short-term bond mutual funds (SICAVs), which provided strong competition to regulated-rate bank deposits, the introduction of negotiable certificates of deposit and commercial paper in 1985, and opening of the short-term treasury bill market to non-financial corporations and individuals in 1986. Although the process of financial liberalization has continued at various rates throughout the past decade, a key change in the procedures for implementing monetary policy took place in January 1987. The previous system of quantitative controls on bank asset growth was abolished, to be replaced by reserve requirements on liabilities, and the daily setting of the call-money market rate was also ended. The interest rate operating procedure established in 1987, which still in substance prevails, involves two key official interest rates. These are the intervention rate and the 5- to 10-day repurchase rate. Under normal circumstances, the interbank rate is between these two 4. The "encardrement du credit" credit ceiling system is described in Marc Quintyn, "From Direct to Indirect Monetary Policy Instruments: The French Experience Reconsidered," IMF Working Paper. March 1991, pp. 5-7. 5. Technically, credit growth ceilings were ended in 1985. However, they were replaced by a marginal reserve requirement system that in large part served as a functional equivalent. This system was abolished in January 1987. - 7 - Morton and Wood rates, with the intervention rate acting as a lower bound and the repurchase rate an upper bound. The main instrument used to influence the interbank interest rate is the intervention rate (the rate at which repurchase funds are offered approximately once a week by the Bank of France). The Bank of France also controls the quantity of reserves allocated at the 5-10 day repurchase rate, though borrowing at that facility is done at the initiative of individual private banks. French monetary authorities have maintained monetary targets continuously since 1977. The strength of the French commitment to these targets appears to have varied over time, but in general has not been as strong as in some other countries, such as the United Kingdom. More important, especially in recent years, has been the French commitment to an exchange rate target, formally an EMS parity, in practice the mark. France joined the EMS in 1979, but devalued the franc within the EMS three times in the 1981-1983 period. A key event in France's commitment to an exchange rate target was the 1983 decision of the Mitterrand government, after much internal debate, to remain within the EMS. 1987. The last realignment of the franc within the EMS was in January Since then, maintaining a stable franc-mark exchange rate has clearly been the paramount goal of French monetary policy. United Kingdom Over the past decade, monetary policy operating procedures have varied in the United Kingdom. At times, authorities appear to have operated mainly with an interest rate target, varying interest rates in response to macroeconomic goals, such as output or inflation. At other times, monetary aggregate targeting has been given priority, and, more recently, an exchange rate target has been most important. The Bank of England's interest rate operating procedures are conducted mainly through its money market dealing rates. These are the rates at which the Bank supplies liquidity daily to the market, primarily through open market transactions in commercial bills with the discount 6. The targeted aggregate has alternated between M2 to M3, with the aggregate undergoing a major redefinition in 1987. - 8- Morton and Wood houses. The Bank operates in four maturity bonds, ranging from 1-14 days to 64-91 days. In practice, the money market dealing rates operate much like discount rates in that they are set by the Bank and changed only infrequently. It should be noted that in 1981, when the Bank instituted the system described above, "it was hoped that this mechanism would provide scope for a reasonable degree of flexibility in short-term interest rates." Under the previous system, a Minimum Lending Rate, at which the Bank provided funds to the market, was officially set and remained unchanged for long periods of time. It was hoped that the new money market dealing rates would be partially set by market forces, and vary from day to day. Thus, the Bank's interest rate intentions would be revealed more indirectly, and the "announcement effect" of official lending rate changes would be lessened. As actual events unfolded, this o goal could not be achieved. The Bank remained the dominant force in the bill market, supplying significant funds each day, with its lending rates achieving a de facto discount rate status. British monetary authorities first adopted a monetary target in 1976. The importance attached to this target increased sharply with the start of the Thatcher government in 1979. However, over time, various problems with monetary targeting--in particular financial innovations which seemed to distort monetary aggregate growth--led to a de-emphasis of monetary targets, although they still officially remain in place. Important stages in the process of moving away from monetary targets included the use of more than one monetary aggregate as a target (starting in February 1982), use of a MO as a target, MO consisting almost entirely of currency, and thus not under the active control of monetary authorities (starting in February 1984), and significant and persistent overshooting of the original M3 target (starting in March 1985). 7. A. L. Coleby, "Change in Money-Market Instruments and Procedures in the United Kingdom," in Changes in Monev-Market Instruments and Procedures: Objectives and Implications. Bank for International Settlements, March 1986, p. 200. 8. For a description of this episode, see Coleby, pp. 201-204. - 9 - Morton and Wood U.K. monetary authorities have also at various .times used interest rate changes in order to achieve an exchange rate objective. In the late 1970s and early 1980s, the exchange rate of greatest interest to authorities was that of the pound against the U.S. dollar. However, increasingly, the exchange rate of the pound in terms of EMS currencies, and particularly the German mark, became the main objective. For a period in 1987-1988, British authorities maintained an unofficial but strong effort to keep the pound-mark exchange rate in a narrow range. In retrospect, this experiment was judged to have had an unfavorable outcome. Upward exchange market pressure on the pound against the mark led to an easing of monetary policy and reduction in interest rates that provided an undesirable inflationary stimulus to the domestic economy. Since October 1990, the pound has been an official member of the EMS's exchange rate mechanism. Experience in this later period has generally been more successful. Switzerland The Swiss National Bank's policy has been geared to two main objectives over the past decade, monetary targets and the exchange rate of the franc (mainly against the dollar early in the period, and against the mark in recent years). The relative importance of these two objectives appears to have varied over time. The Swiss have maintained monetary targets since 1975. The target has been stated in terms of adjusted central bank money (essentially the monetary base) since 1980. Between 1982 and 1987, Swiss monetary authorities came quite close to achieving their targeted monetary growth rate. However, both before and after this interval, substantial deviations from targets occurred. In the 1978-1979 period, there developed a clear conflict between exchange rate and monetary target objectives. There was strong upward pressure of the franc (mainly against the dollar), which would have required a significant easing of monetary policy and lowering of interest rates to counter. However, such an easing was likely to lead to a significant overshooting of the monetary target. Swiss authorities chose to put greater weight on the exchange rate objective in this instance. - 10 - Morton and Wood This resulted in a surge of money growth, a temporary abandonment of the monetary target in 1979, and a subsequent increase in inflation to what was, by Swiss standards, alarming levels. complicated by several factors. Analysis of this episode is In particular, the oil price shock of 1979, which contributed to inflationary pressures, and an upward shift in money demand, much of it coming from abroad. Nonetheless, Swiss authorities appear to have concluded that it was a mistake to allow an overshooting of the monetary target to the extent that took place. The more recent period of substantial deviation from monetary targets was triggered by two changes introduced in 1988 which substantially shifted the demand for base money. First, a new electronic interbank payments system was established which sharply lowered commercial banks' need for clearing balances at the Swiss National Bank. Secondly, the authorities modified cash liquidity requirements, moving from an end-of-month to monthly average measure and lowering overall requirements. The net result of these changes was a substantial undershooting of the central bank money target over the 1988-1990 period, difficulty in interpreting the actual degree of ease or tightness of monetary policy, and an increase in inflationary pressures in 1991-1992. Despite this recent difficulty with monetary targeting, there remains a reluctance by some in Switzerland to move fully to an exchange rate 9 target. This decision is increasingly taking the form of possibly joining the EMS, effectively pegging the Swiss franc to the mark. This possibility appears strengthened by the recent referendum vote in favor of IMF membership, and the subsequent announcement by the Swiss government that it would apply for EC membership. The Swiss National Bank has two means of affecting market liquidity. The first, and most important, involves foreign exchange swaps, usually of 1- to 3-month maturity. The second involves moving government balances into and out of the commercial banking system. The Swiss National Bank maintains both a discount rate, set at the Bank's 9. See, for example, G. Rich. "The Orientation of Monetary Policy and the Monetary Policy Decision-Making Process in Switzerland," Swiss National Bank, 1991. - 11 - Morton and Wood discretion and changed only infrequently, and a Lombard rate which, since May 1989, has been computed daily by a formula which sets the rate 200 basis points (rounded to the nearest 1/8 percent) above a reference call money rate. This latter change appears to have been motivated by a desire to make the Lombard rate a penalty rate, and Lombard lending truly an exceptional source of bank liquidity, as well as avoiding announcement effects from Lombard rate changes. Canada Canadian monetary policy has been dominated over the past decade mainly by the important influence of, and need to respond to, conditions in the United States. Given the close integration of U.S. and Canadian financial markets, this has usually meant that Canadian short-term interest rate changes mirror those in the United States fairly closely. Unlike the situation of EMS members, however, there has never been a formal or official commitment to keep the Canadian dollar-U.S. dollar exchange within some specified narrow range. It has been a general policy of the Bank of Canada to pursue a "leaning against the wind" intervention policy, moderating but not totally resisting exchange rate movements. In addition, monetary authorities have explicitly recognized the trade-off between exchange rate and interest rate changes, with, for example, a potentially inflationary depreciation of the Canadian dollar's foreign exchange value leading to some compensating tightening of monetary policy through higher Canadian interest rates. Canada first adopted an official monetary target in November 1975. However, various problems--including financial innovations which distorted monetary aggregates growth, and conflicts with other important targets, particularly the exchange rate--led to the abandonment of monetary targeting in November 1982. More recently, since 1990 targeting of another type has been introduced. Canadian financial authorities have announced a multi-year series of declining inflation targets. Although money aggregate growth (of M2) is to be used as one indicator guiding policy, it has been made clear that actual inflation is the main target. - 12 - Morton and Wood In March 1980, the Bank of Canada adopted a formula for determining its discount rate. More specifically, the discount rate is determined by a pre-announced rule based on the outcome of the weekly 3month Treasury bill auction. Since this procedure eliminates the "announcement effect" of discount rate changes, it provides the maximum degree of flexibility in implementing interest rate policy. STATISTICAL MEASURES OF VOLATILITY The preceding two sections suggest that, both in theory and in the views of central bank officials, a desirable property of an interest rate operating procedure is to be "flexible.H Although it is difficult to arrive at a simple, unambiguous definition of this term, its core meaning appears to involve the ability to change interest rates promptly and fully when needed. This could involve both day-to-day changes, and cumulative adjustments over time. On the other hand, central bankers also may wish to have a relatively smooth path of key interest rates, either for political reasons or in order for policy to be more easily predicted by market participants. Therefore, it is ambiguous whether interest rate volatility is good or bad. The discussion in the previous sections suggests several hypotheses as to the relative variability of interest rates under differing operating procedures and different policy regimes, although in some cases expected results are ambiguous. The switch to an interest rate operating procedure that is more market-oriented and less tied to official rates, such as that undertaken by Germany in 1985, might be expected to lead to a somewhat more variable interest rate path. Similarly, a general liberalization of financial market structure and monetary policy operating procedures, such as that which took place* in Japan and France around the mid-1980s, might also be expected to increase interest rate variability. Adoption by Canada of a formula discount rate, where official interest rate changes are unhampered by announcement effects, might be expected to result, other things being equal, in relatively greater interest rate variability than in other countries. - 13 - Morton and Wood The implications for interest rate variability of differing monetary policy target variables is a priori more uncertain. Thus, the expected relationship between interest rate variability in countries with a relatively strong commitment to a monetary aggregate target, such as Germany and Switzerland, and countries without monetary aggregate targets, such as Japan and Canada, is unclear, although there might be a weak presumption of somewhat greater interest rate variability in countries following a monetary aggregate target. There is a similar uncertainty about the role of exchange rate targets. Here, the main division would be between EMS members (France and, since 1990, the United Kingdom) and countries with no formal exchange rate commitment (Japan, Canada, and Switzerland). The situation here is further complicated by the fact that, even without a formal exchange rate arrangement, some countries still tie their monetary policies strongly at times to exchange rate targets (Canada to the U.S. dollar and Switzerland to the mark). In tables 1, 2, and 3, we show measures of daily interest rate volatility for the six countries in our study. For five of the countries, we divided the sample where there was a significant change in the operation of monetary policy. for Canada). (We did not find any break in policy In Japan, the break we chose is at the beginning of 1985, which marked the approximate beginning of rapid financial liberalization. For Germany, the break is in February 1985, when the Bundesbank shifted to relying on RP agreements rather than the Lombard window as the primary source of liquidity for the banking system. In France, the break is at the beginning of 1987, when French financial markets were liberalized and the Bank of France switched from direct credit allocation to open market operations. In the United Kingdom, the break is in October 1990, when sterling entered the exchange rate mechanism of the EMS. For Switzerland, the break is in 1988, when Swiss cash liquidity requirements were changed to a monthly-average basis from a month-end basis. The measure of daily volatility shown in Table 1 is the standard deviation of interest rates on a daily basis. For each of these countries, the interest rate path over time has either a substantial trend or prominent cycles, so that standard deviation measures over long - 14 - Morton and Wood periods are dominated by the large differences from mean, rather than day-to-day changes, and are thus relatively invariant for different frequencies. Because of this, we computed the standard deviation of daily data around the monthly mean for each month, then averaged the monthly standard deviations for each sub-period, and that is shown in Table 1. In Table 2, we show the standard deviation of daily changes in interest rates for each sub-period. The measure shown in Table 3 is the average absolute daily change in interest rates. The first comparison we can make is between volatilities of overnight and three-month interest rates. For each of our measures and for every country and every time period, overnight interest rates are more volatile than three-month interest rates. That would be expected if overnight rates reflect temporary liquidity pressures in addition to changes in monetary policy. We also find that countries with the least volatility in overnight rates also tend to have the least volatility in three-month interest rates. Comparing across countries, we find that Japanese interest rates have been less volatile at both the overnight and three-month maturities. German interest rates have generally been the next least volatile, followed closely by those of France, which has set its monetary policy to stabilize the franc-mark exchange rate especially in the more recent period. Interest rate volatility in the United Kingdom tends to be somewhere in the middle of this group of countries, while volatility has been the highest in Switzerland and Canada. We can thus find no clear division in interest rate volatility along the lines of countries that have monetary aggregate targets versus those that do not, because Japan and Canada (two countries without monetary targets) are on opposite ends of the volatility spectrum. Likewise, there is no clear division between EMS and non-EMS countries. Comparing across time periods, we see that standard deviations around monthly means have declined for all the countries and all maturities except for the Japanese three-month interest rate. However, the three-month rate used here (the CD rate) is only available starting June 1984, so the first sub-period has only seven months of data for that - 15 - Morton and Wood rate. The standard deviation for the German overnight -rate was almost unchanged. The most striking decline in interest rate volatility is also the most predictable. Swiss overnight interest rates became much less volatile after the switch to monthly-average liquidity requirements which reduced the sharp increases in overnight rates that tended to occur at the end of each month under the previous regime of month-end reserve requirements. The measure of average absolute change shown in Table 3 shows declines in volatility in the latter sub-periods except for Germany and the United Kingdom. It is somewhat surprising that interest rate volatility has decreased in the more recent sufc-periods for Japan and France, which moved to more flexible interest rate operating procedures. However, it is possible that the deepening of financial markets in the latter period of financial liberalization has contributed to lower interest rate volatility. In Germany, the move to a more flexible operating procedure in 1985 has been accompanied by slightly more volatility by the measure in Table 3. - 16 - Morton and Wood REFERENCES Bank for International Settlements. Changes in Money-Market Instruments and Procedures: Objectives and Implications. BIS, March 1986. Batten, Dallas S. and Michael Blackwell et. al. Policy in the Major Industrial Countries," 10, July 1990. "The Conduct of Monetary IMF Occasional papers. No. Bernanke, Ben and Frederic Miskin. "Central Bank Behavior and the Strategy of Monetary Policy: Observations from Six Industrialized Countries," Mimeo, 1992. Crow, John W. "The Bank of Canada and the Money Market," April 1989. Bank of Canada, Deutsche Bundesbank. "The Deutsche Bundesbank: It's Monetary Policy Instruments and Functions, 3rd Edition," Deutsche Bundesbank Special Series. No. 7, July 1983. Frowen, Stephen F. and Dietman Kuth ed. Monetary Policy and Financial Innovations in Five Industrial Countries: The U.K.. the USA. West Germany. France and Japan. St. Martin's Press, 1992. Kasman, Bruce and Anthony Rodrigues. "Financial Reform and Monetary Control in Japan." FRBNY Quarterly Review. Autumn 1991. Kneeshaw, J. T. and P. Van den Bergh. "Changes in Central Bank Money Market Operating Procedures in the 1980s," BIS Economic Papers. No. 23, January 1989. Osugi, K. "Japan's Experience of Financial Deregulation Since 1984 in an International Perspective," BIS Economic Papers. No. 26, January 1990. Quintyn, Marc. "From Direct to Indirect Monetary Policy Instruments: French Experience Reconsidered," IMF Working Papers. March 1991. The Rich, G. "The Orientation of Monetary Policy and the Monetary Policy Decision-Making Process in Switzerland," Swiss National Bank, 1991. Suzaki, Yoshio, Akio Kuroda and Hiromichi Shirakawa. "Monetary Control Mechanism in Japan," Bank of Japan Monetary and Economic Studies. Vol. 6 No. 2, November 1988. Temperton, Paul. U.K. Monetary Policy: Martin's Press, 1991. - 17 - The Challenge for the 1990s. St. - 18 - Data Series Overnight Germany: Japan: Frankfurt Interbank Call Money Rate Tokyo Unconditional Lender Rate France: Paris Day to Day Money Rate United Kingdom: Canada: U.K. Call Money Rate Canadian Day to Day Money Rate Switzerland: Zurich Call Money Rate Three-Month Germany: Japan: France: Frankfurt Interbank Loan Rate Rate on Certificates of Deposit (Secondary Market) Paris Interbank Rate United Kingdom: Canada: Canadian Finance Company Paper Switzerland: Interbank Sterling Interest Rate Swiss Interbank Rate - 18 - Morton and Wood Table 1 Average of Standard Deviations of Daily Data Around Monthly Means Overnight rate Three-month rate 1980-85 0.25 0.16 1985-92 0.24 0.08 1980-84 0.18 0.01 1985-92 0.13 0.07 1980-86 0.30 0.22 1987-92 0.23 0.14 1980-90 0.43 0.25 1990-92 0.36 0.18 1980-92 0.58 0.23 1980-87 3.56 0.24 1988-92 0.20 0.18 Germany Japan France United Kingdom Canada Switzerland 19 - Morton and Wood Table_2 Standard Deviations of Daily Changes in Interest Rates Overnight rate Three-month rate 1980-85 0.38 0.13 1985-91 0.32 0.06 1980-84 0.13 0.00 1985-91 0.10 0.04 1980-86 0.23 0.14 1987-91 0.20 0.09 United Kingdom 1980-90 0.40 0.18 1990-91 0.35 0.11 1980-91 0.58 0.16 1980-87 6.98 0.20 1988-91 0.18 0.11 Germany Japan France Canada Switzerland - 20 Morton and Wood Table 3 Average Absolute Change of Interest Rates (Daily Data) Overnight rate Three-month rate 1980-85 0.110 0.063 1985-92 0.118 0.035 1980-84 0.073 0.047 1985-92 0,063 0.018 1980-87 0.121 0.063 1987-92 0.114 0.051 1980-90 0.226 0.093 1990-92 0.233 0.063 1980-92 0.536 0.087 1980-87 0.884 0.105 1988-92 0.375 0.059 Germany Japan France United Kingdom Canada Switzerland - 21 - Overnight and Official Discount Rates for Japan (Monthly Average) Percent - = OVERNIGHT - = OFFICIAL DISCOUNT 8 H6 to -i 4 H 3 1 1982 1 I 1983 1984 1 1985 1 1 1986 1987 1 1988 1 I 1989 1990 1991 1992 Overnight, Discount, Lombard, and RP Rates for Germany (Monthly Average) Percent = OVERNIGHT = DISCOUNT = LOMBARD = RP 12 11 10 9 8 rt o o rt 3 7 6 5 3 v 1 1982 1 1983 1 X 1984 1985 J 1 1986 i 1987 1 1988 1 1989 1 1990 1 1991 1992 2 *: o o Overnight and Money Market Intervention Rates for France (Monthly Average) Percent rr O O ST 0 rt 3 ho o o ,1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 Overnight and Money Market Dealing Rates for the United Kingdom (Monthly Average) Percent o 03 0 o o 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 Overnight and Lombard Rates for Switzerland (Monthly Average, Month End, Respectively) Percent -OVERNIGHT = LOMBARD 15 14 13 12 t 11 M H 10 & »i rt O O ST 0 0) rt 3 a Ln ON O O a 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 Overnight and Discount Rates for Canada (Monthly Average) Percent »i rt O O rt 3 <^ 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 s: o o a- A COMPARISON OF MONETARY POLICY OPERATING PROCEDURES IN SIX INDUSTRIAL COUNTRIES Bruce Kasman1 The institutional environments in which the central banks of the industrial world operate have changed substantially since the mid1970s. Financial market liberalization, along with regulatory and technological change, has altered the relationships between central bank policy tools and objectives. Authorities have responded to these changes by revising the techniques and procedures they use to implement monetary policy. In Japan and France, where far-reaching reforms of the financial system have taken place, central bank operating procedures have been substantially transformed. In countries where well-developed capital markets existed earlier, the revisions in monetary policy operating procedures have been considerably less dramatic. As financial liberalization and innovation proceed, the institutional settings of the central banks have become more uniform. Although arrangements still vary across countries, this convergence suggests that a comparison of central bank operating procedures is now likely to be of greater relevance to policy makers than at any time in the past. An assessment of foreign practices may provide a particularly useful perspective on the changing conditions affecting the operations of the Federal Reserve's Open Market Desk. A noticeable increase in banks' reluctance to borrow at the Federal Reserve's discount window in recent years has at times contributed to large daily fluctuations in the Federal funds rate. Moreover, reductions in reserve requirements in 1990 and April of this year have led to occasional conflicts between the Desk's reserve management strategy and more volatile day-to-day conditions in the funds market. With other central banks offering a wide variety of alternative techniques for implementing policy 1. Federal Reserve Bank of New York. The author is grateful to Andre Bartholomae, Kevin Clinton, Spencer Dale, David Longworth, Ann-Marie Meulendyke, Michel Peytrignet and George Rich for providing useful comments and information. Valuable assistance in preparing this paper was provided by Matthew Maring. Kasman and a number currently operating in an environment of low, nonbinding reserve requirements, an examination of operating procedures followed by foreign central banks seems timely.2 This article describes monetary policy operating procedures in six industrial countries — the United States, Germany, Japan, the United Kingdom, Canada, and Switzerland. The object is to shed light on central bank strategies elsewhere in the industrial world and to compare them with the practices of the Federal Reserve. As part of this review, particular attention is given to the institutional environments in which central banks operate. The intermediate and ultimate objectives of a central bank, while important in an overall survey of monetary policy transmission, are not discussed in any detail. Our review suggests that basic central bank intervention strategies are currently quite similar across the industrial world. Nearly all the central banks analyzed use interest rate operating objectives to guide their daily activities. In addition, although the central banks employ different instruments, they all implement policy principally through daily operations supplying or absorbing reserves at market-determined prices. The Federal Reserve and several foreign central banks are also alike in having chosen to lower their reserve requirements in recent years. In most cases, the foreign monetary authorities have adjusted their operating procedures to accommodate this change. Specifically, they have provided a more elastic intraday supply of central bank reserves, largely through their credit facilities. In this way, they limit any tendency for reduced reserve margins to lead to higher day-to-day interest rate volatility. Our analysis suggests that some of the practices observed abroad might be helpful in limiting the short-run volatility of the federal funds rate in the United States. However, our analysis also indicates that the volatility of the federal funds rate, although higher since the 1990 cut in reserve requirements, 2. A good discussion of Federal Reserve operating procedures following the reduction in reserve requirements can be found in "Monetary Policy and Open Market Operations during 1991." This Quarterly Review, Spring 1992, pp. 72-95. -2- Kasman remains low relative to that of comparable rates in most other countries. Moreover, we find no evidence that federal funds rate variability, within its current range, is transmitted to other money markets. Thus, the rise in interest variability that has accompanied the reduction in reserve requirements in the United States has probably not materially affected the monetary policy transmission mechanism. COMPARING OPERATING PROCEDURES IN SIX INDUSTRIAL COUNTRIES Key Features of Central Bank Operating Procedures A central bank must choose implementation procedures that enable it to achieve its macroeconomic goals. Although the six central banks considered in this article have different objectives and operate under varied institutional environments, the key features of their implementation strategies are currently quite similar. All six central banks implement policy by controlling the aggregate level of reserves available to the banking system. Although they are not in a position to control movements in all components of their balance sheets, particularly those related to their function as banker to the government and their holdings of foreign currency reserves, these banks currently have sufficient information and operational leeway to neutralize the effects of other activities and regulate the aggregate supply of reserves with a high degree of control. In managing the reserve position of the banking system, central banks generally pursue short-run operating objectives. Operating objectives link reserve management activities to the intermediate and ultimate goals of policy and, in most countries, are also used to signal central bank policy intentions to market participants. Ideally, the authorities exert close control over operating objectives. Bank reserves have served as an operating objective but the relationship between reserves and economic activity generally has been viewed as too volatile for reserves to function as an effective short-run guide to policy. Most of these central banks -3- Kasman have instead geared their reserve management activities toward short-term interest rate objectives.3 A wide variety of money market interest rates are employed as operating objectives. Nonetheless, influence over overnight interest rates is a goal common to the daily activities of all six of these central banks. Each of these countries has a well-functioning interbank money market where individual banks trade reserves on deposit at the central bank.4 If the aggregate supply of banking system reserves does not correspond to demand, the cost of overnight funds in this market is immediately affected. Although central banks' reserve management activities give them considerable control over short-term interbank rates, their influence on interest rates must extend to maturities well beyond overnight rates to affect economic activity. Central bank influence over longer term rates is indirect and principally determined by market forces. Through arbitrage, longer term rates reflect market expectations of future short-term rates. A central bank's leverage over longer term rates is obtained largely through its influence on these expectations. By taking steps to communicate credible intentions about the range in which overnight and other short-term interest rates should trade in the future, central banks can transmit their interest rate policies throughout the money market term structure and beyond. To this end, most of these central banks limit themselves to infrequent adjustments in their operating objectives. Targeted interest rates are generally changed in small steps and only after 3. The notable exception is the Swiss National Bank, which has maintained bank reserve operating targets for most of this period. In addition, the Federal Reserve experimented briefly with nonborrowed reserve objectives from 1979 to 1982. The choice of monetary policy operating targets has been the subject of considerable debate. William Poole provides the seminal discussion of these issues ("Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Macro Model," Quarterly Journal of Economics, vol. 84 (1970), pp. 197-216. For a recent discussion of interest rate operating objectives in the United States, see Marvin Goodfriend, "Interest Rates and the Conduct of Monetary Policy" and the accompanying comments by William Poole in Carneaie-Rochester Conference Series on Public Policy, no. 34 (1991), pp. 7-39. 4. In Japan and the United Kingdom, nonbank financial intermediaries participate in the interbank market. In Canada, an important overnight market in call loans, used by both banks and investment dealers, exists alongside the interbank market. -4- Kasman a sufficient amount of new information has accumulated to warrant a change in policy. By encouraging expectations of interest rate stability over a medium-term horizon, policy makers' gain influence over rates throughout the term structure. Although interest rate operating objectives have been prevalent among these central banks over the past two decades, the type of implementation strategy employed has, in many countries, evolved considerably.5 Puring the 1970s, the central bank of Japan and several European central banks relied heavily on a system of administered interest rates to implement policy. Banks' marginal reserve demand in these countries was largely met through central bank credit facilities, often at below-market rates.6 "Official" or tightly controlled money market rates served as anchors for regulated deposit and lending rates. Together with other controls over financial activity, official rate changes were transmitted largely through their direct effect on bank credit availability. This approach came under pressure in the late 1970s. The delays by some central banks in adjusting interest rates to counter a buildup of inflation in the late 1970s raised concerns about the inflexibility of interest rate determination. Many observers believed that the use of highly visible official rates constrained banks from adjusting policy in a timely fashion. More important, however, rising inflation helped spur the liberalization of financial markets, which in turn substantially increased the importance of competitive forces in determining interest rates. Domestic financial markets also became more closely integrated with foreign markets. As a consequence, market-determined interest rates and exchange rates played an increasingly central role in private agents' expenditure decisions.7 5. An excellent discussion of how monetary policy procedures have evolved can be found in J.T. Kneeshaw and P. Van den Bergh, "Changes in Central Bank Money Market Operating Procedures in the 1980s", BIS Economic Papers, no. 23, January 1989. 6. Reliance on subsidized central bank credit sources for bank reserve needs characterized German, Japanese, and Swiss monetary policy. 7. A detailed analysis of financial innovation and its effect on the monetary policy transmission mechanism can be found in Financial Innovation and Monetary Policy, Bank for International Settlements, (Basle, Switzerland 1984). -5- Kasman Although procedural changes have been greatest in those countries where financial change has been most significant, the central banks under review have in general moved towards marketoriented methods for implementing monetary policy. As noted earlier, authorities increasingly rely on market-determined interest rates both as operating objectives and as key elements in the transmission mechanism. At the same time, market operations, in which central banks intervene in financial markets at freely determined prices, have gradually replaced lending and regulatory controls as the principal instrument for altering reserve supplies in most countries. The shift toward market-oriented interest rate objectives has helped the central banks to reduce the repercussions arising from changes in their policy stance. In addition, open market operations permit central banks to exercise considerable discretion in the day-to-day management of reserves. While relying on market forces to determine interest rates, central banks can intervene at select times to influence the range within which rates move. Furthermore, the wide variety of available domestic money market instruments (whose development was greatly encouraged by monetary authorities in most countries) allows the banks to construct intervention strategies that span the money market term structure. In practice, central banks continue to severely limit the range in which short-term interest rates fluctuate. By finetuning their market operations, usually on a daily basis, these central banks alter reserves to accommodate variations in reserve demand. This active effort to moderate even transitory interest rate fluctuations underscores central banks' desire to communicate their policy intentions clearly to market participants. In nearly all the countries under review, the stance of monetary policy is signaled through interest rates. Market interest rates respond to developments other than policy changes, however, and movements unrelated to policy must be filtered out before policy inferences can be drawn. By sharply limiting interest rate variations daily, central banks ensure that market participants can clearly identify interest rate targets and quickly ascertain changes in the monetary policy stance. To implement an interest-rate-based operating policy through periodic open market operations, central banks must be able to predict the demand for bank reserves over some relevant -6- Kasman horizon. Banks need reserves to meet reserve requirements and to make interbank payments. Central banks have considerable influence over reserve demand through their role in setting reserve requirements and interbank clearing rules. Specific rules (lagged reserve accounting, reserve averaging, and carryover provisions) and payment systems practices (timing of payments, overdraft provisions) have been designed, in part, to strengthen and stabilize the short-term demand for bank reserves. In general, the stability of reserve demand over a maintenance period has been a central element underlying central bank implementation procedures* In the past, many central banks actively managed reserve demand by changing reserve requirements and applying other administrative controls to bank behavior. These practices have greatly diminished in recent years reflecting, in part, the general trend towards market-based policy strategies. At the same time, all six central banks have reduced reserve requirement ratios over the past decade in an attempt to lighten the burden they place on banks. In some countries the relaxation of restrictions on banks' reserve holdings has led to greater variability in reserve demand, compelling authorities to adjust their reserve management procedures. Although this overview of the key features of central bank implementation strategies suggests broad similarities across countries, the specific techniques employed by individual central banks to implement monetary policy vary greatly. Central bank market operations span a wide spectrum of assets and maturities; the timing of operations and the frequency with which they are conducted also differ. Significant differences can be seen as well in the conditions determining access to central bank credit, the regulations setting required reserve levels, and the length of time granted depository institutions to meet their obligations. In many cases, these differences are institutional in nature, reflecting the particular environments in which central banks operate. For example, in conducting open market operations, central banks must depend on the markets available to them. Where active secondary security markets are not developed, central banks may need to make special arrangements for implementing their reserve management policies. The remainder of this section compares monetary policy implementation techniques across the six countries. By examining the particular institutional environment in which each central -7- Kasman bank operates and by observing the interaction of the specific instruments central banks employ — open market operations, central bank lending policy, and reserve requirements -- one can identify meaningful differences between Federal Reserve and foreign central bank operating procedures. Operating Objectives and Procedures All six central banks gear their short-term reserve management activities toward influencing interest rates, but specific interest rate strategies differ from bank to bank. The Federal Reserve in the United States limits its activities to influencing overnight interbank rates (the federal funds rate), allowing market forces to determine the transmission of policy to other financial markets. The Swiss National Bank also acts to smooth daily fluctuations in overnight interbank rates, but it is unique among these central banks in setting no explicit interest rate operating objective. Although the four other central banks also actively intervene to smooth fluctuations in overnight rates, they generally seek to influence money market rates of longer maturities as well. In Japan, overnight interbank rates remain the primary operating objective of the central bank, while in Canada, Germany, and the United Kingdom, rates of longer maturity, up to three months in some cases, are employed as the primary operating objective. A summary of the interest rates important to the banks' policy implementation is presented in Table 1. The primary interest rate operating objective for each country is highlighted. Of the central banks considered, the Bank of England (BOE) is probably most active in its daily reserve management activities. Operating in an environment in which reserve requirements are low and banks each day try to maintain a specific daily level of operational balances at the BOE, the Bank has developed a strategy of frequent intraday interventions in money markets to achieve its interest rate objectives.8 Each morning at 9:45 a.m. the BOE announces its estimate of the net reserve position of the banking system for the day. Based largely on expected government transactions and the BOE's 8. To assist the BOE in its daily forecast of the reserve position of the banking system, each clearing bank is obliged to specify the size of reserve balances that it will try to maintain daily. -8- Kasman maturing stock of short-term bills, these estimates signal the amount of reserves that the BOE anticipates must be supplied to bring actual balances of clearing banks to the levels the banks are expected to maintain.9 Because the bulk of the BOE's assets are in short-term bills (commercial or Treasury) that mature in less than three months and that do not roll over automatically, the banking system will usually be projected to have a "cash shortage" at current interest rates. To meet this shortage, discount houses, which serve as intermediaries between the BOE and private banks, are invited to offer bills to the Bank for purchase, indicating the price at which they are willing to sell.10 The BOE buys bills to meet the estimated shortage in four maturity bands: zero to fourteen days, fifteen to thirty-three days, thirty-four to sixtythree days and sixty-four to ninety-one days. It chooses the best prices offered but holds unchanged the minimum dealing rate (stop rate) on Band 1 bills maturing in up to fourteen days. As many as three rounds of these operations may take place in a day, enabling the BOE to respond to changing intraday market conditions. If late-day imbalances arise, they are met through credit facilities available to discount houses. By purchasing bills across bands (maturities), the BOE attempts to extend its influence over interest rates throughout the money market. Variations in the amount of bills purchased in Band 4 (sixty-nine to ninety-one days), for example, tend to have a strong influence on three-month Treasury bill rates. The BOE also has the option of offering repurchase agreements to discount houses on its own terms if it does not wish to validate the rates being offered. Mindful of this option, the discount houses will generally offer prices embodying their expectation of the BOE's desired rate objectives. The stop rate changes infrequently. Movements in this rate signal a shift in BOE policy and are usually reflected immediately 9. The government holds most of its balances with the BOE. Because its daily transactions with the rest of the economy are large and fluctuate widely, the BOE's forecast of net government flows is both the key component of this estimate and the greatest source of uncertainty. 10. For more detailed information on the role of discount houses in the U.K. financial system and the BOE's money market operations more generally, see "Bank of England Operations in the Sterling Money Market", Bank of England Quarterly, October 1988. -9- Kasman throughout the interbank market and in commercial bank base lending rates (Chart 1 ) . On occasion, the BOE will send a strong signal of its intention to shift policy by choosing not to accommodate a shortage in reserve needs during the day, thereby obliging discount houses to borrow from the BOE at terms determined by the Bank. Since the BOE has the flexibility to set this lending rate either above or below current stop rates, it can use this procedure to signal a tightening or an easing in policy. Japanese monetary authorities followed a similar strategy of tight control over the key intervention rate until the early 1980s. Combining reserve management operations with administrative control over interbank market participants, the Bank of Japan (BOJ) was able to stabilize the call-money overnight interbank interest rate at the level desired for long periods. As part of a broader reform of financial markets over the past decade, the BOJ has actively promoted integration of the interbank with other financial markets and encouraged greater flexibility of interbank interest rates, particularly on an intraday basis.11 The overnight call rate remains the BOJ's key operating objective, and although it is subject to greater influence from market forces than in the past, the BOJ still actively strives to limit its fluctuations around the targeted level (Chart 2 ) . The BOJ implements this policy through a variety of market operations, primarily transactions in commercial bills, and through its daily management of discount window credit. Control over the "reserve progress ratio," which measures reserves accumulated by banks relative to those required within a maintenance period, is a key element of this policy. Upward pressure on interest rates is effected by supplying fewer reserves than are necessary for the reserve progress ratio to rise at an average pace. Banks have considerable leeway in managing their reserve positions because the reserve maintenance period is a full month in Japan. Nevertheless, changes in the reserve progress ratio clearly convey the BOJ intentions concerning future interest rates and, as a result, usually lead to a quick response in interbank interest rates. 11. For a detailed analysis of the evolution of Bank of Japan policy and references to the literature on financial market liberalization in Japan, see Bruce Kasman and Anthony P. Rodrigues "Financial Liberalization and Monetary Control in Japan" this Quarterly Review (Autumn 1991). -10- Kasman The evolution of BOJ policy over the past decade reflects a movement towards procedures long practiced by the Federal Reserve System. Indeed, the two central bank implementation strategies appear quite similar in their basic characteristics — an overnight interbank market operating objective, the use of market operations and discretionary central bank lending facilities as policy instruments, and a focus on reserve management over a maintenance period. Still, important differences remain between the operating strategies of the Bank of Japan and the Federal Reserve. While the Federal Reserve conducts most of its daily operations in the repurchase market for government securities, the BOJ relies on a variety of private market instruments, including commercial bills, commercial paper, and certificates of deposit. In part, the BOJ's reserve management activities reflect the limited development of a single short-term government securities market in Japan. However, the BOJ has also employed operations in different instruments to exert direct influence on money market interest rates. Up until 1988 interbank and other open markets were not fully integrated, and the BOJ intervened actively in longer term money markets, primarily to influence the three-month certificate of deposit rate. Following a period in 1987 and 1988 in which open market rates moved well above comparable rates in the interbank market, the BOJ implemented a series of reforms to facilitate arbitrage across short-term money markets.12 Since that time the BOJ has generally limited its efforts to influence direct influence over interest rates in the interbank market to instruments of seven days' maturity or less. Market operations' in longer term money market instruments are now primarily designed to offset seasonal fluctuations in reserve demand. The administration of discount window lending also differs considerably in the two countries. In the United States, banks initiate the decision to borrow at the Federal Reserve's discount window, and borrowing is rationed through a set of administrative guidelines. In Japan, the BOJ decides on the level of bank borrowing and the length of loans (a factor that determines the 12. For a detailed discussion of money market reforms implemented since 1988, see Japan's Short-Term Money Market and Issues, Ministry of Finance and Bank of Japan, Money Market Study Group, August 1991. -11- Kasman effective cost of a loan). In administering the discount window lending, the BOJ actively manages loan provision on a daily basis to respond to intraday fluctuations in reserve positions. The BOJ is unique among the central banks surveyed in employing lending as a discretionary instrument of daily reserve management. The institutional environment in which the Swiss National Bank (SNB) operates has undergone considerable change in recent years. From 1980 through 1988 the SNB guided its policy largely with short-term bank reserve targets. Although interbank interest rates fluctuated widely on a daily basis, the SNB was reasonably successful in achieving its primary policy objective of maintaining low rates of inflation.13 In 1988, the combined effects of implementing an electronic payment system for settling interbank cash balances (1987) and introducing new liquidity rules (January 1988) led to a sharp decline in reserve deposits held at the Bank (Chart 3). 1 4 The difficulties faced by the SNB in predicting the size of this decline led to an inopportune expansionary monetary policy in early 1988. In response, the SNB shifted its operating objectives away from reserves toward short-term interest rates and exchange rates.15 Although the SNB has gradually moved back towards an implementation strategy based on operational targets for bank reserves, it has continued to emphasi'ze interest rates in its daily operating procedures. Each quarter the SNB signals its short-term policy intentions by announcing a forecast of the level of the monetary 13. See Ben Bernanke and Frederic Mishkin ("Central Bank Behavior and the Strategy of Monetary Policy: Observations from Six Industrial Countries," unpublished paper) for a recent assessment of Swiss monetary policy in relation to other central bank practices over the past two decades. 14. The new liquidity rules lowered required reserves and shifted the maintenance period from the end of the month to a month average. 15. See Organization for Economic Cooperation and Development, OECD Economic Survey-Switzerland (Paris, 1989) for a discussion of Swiss monetary policy following these institutional changes. -12- Kasman base in the subsequent quarter.16 Incorporated in this forecast is an unannounced operational target for the level of bank reserves held at the SNB. Although this target serves as a guide to policy operations over each month and each quarter, authorities have considerable discretion in deciding on their day-to-day activities. In implementing daily policy, the Bank largely seeks to smooth fluctuations in overnight interbank rates. Nonetheless, the interest rate policy of the SNB differs significantly from that of the other central banks under review. No operational targets are set for the level of interest rates, and the SNB does not employ.interest rates to signal its stance to market participants. The institutional changes that took place in Switzerland in the late 1980s have not led to substantial changes in the implementation procedures employed by the SNB. As before, market operations are generally conducted once each morning through foreign currency operations. These transactions, in the form of U.S. dollar-Swiss franc swaps, are conducted at rates close to those prevailing in Euromarkets and extend up to one year in maturity. Earlier SNB restrictions, which placed limits on end-ofmonth Lombard lending and required banks to give advance notification of their credit needs, were removed when reserve requirements were reduced in 1988.17 Nevertheless, in 1989 the Bank floated the Lombard rate 200 basis points above market rates, a move that has substantially limited recourse to this facility. In Germany, interest rates on security repurchase agreements of one- to two-month maturities are the primary 16. The forecasts are designed to be consistent with mediumrun growth targets for the monetary base. Since 1990, these medium-run targets have been defined as annual growth rates to be achieved over a period of three to five years. The targets thus give the SNB considerable flexibility in determining its quarterly forecasts. 17. Before January 1988, banks' reserve requirements were monitored only on the last day of a month. Banks' demand for reserves consequently soared at this time. With access to Lombard lending limited by these restrictions, short-term interest rates often rose very steeply at month's end. -13- Kasman operating objective of the Bundesbank.18 These rates are determined at periodic tenders typically conducted once a week. The Bundesbank normally determines the amount of repurchase agreements offered at a tender by assessing market demand for reserves, and it chooses the best prices available. On occasion, it will fix the price (interest rate) at a tender to send a clear signal of its policy intentions to markets.19 Of the central banks considered, the Bundesbank is probably the least active in its daily reserve management activities. Repurchase agreement tenders generally provide the liquidity needed each day. Occasional "supportive" operations are undertaken to influence the day-to-day money rate through a number of reversible fine-tuning measures. Short-term interest rate smoothing, however, is largely obtained through means other than market operations, a system that reflects the limited development of domestic money markets in Germany. Specifically, official rate facilities on Lombard loans and the Bundesbank's Treasury bill selling rate bound the range within which money market rates can fluctuate (Chart 4 ) . In addition, high reserve requirement ratios and long (one-month) maintenance periods provide banks with considerable flexibility to arbitrage away transitory shocks to their reserve positions. For the Bank of Canada (BOC), the three-month Treasury bill tender rate is the primary operating objective. The BOC participates in the weekly auction and buys and sells bills in the market from time to time, both on an outright and on a buy-back basis. But the BOC implements policy mainly through daily transfers of government demand deposits between the BOC and private banks.20 These transfers are decided late in the day, by which time the BOC has information on government transactions and other payment items that might affect bank reserves. Thus, the 18. For a recent discussion of Bundesbank operating procedures, see Andre Bartholomae, "Some Operational and Instrumental Aspects of Monetary Targeting in Germany," Deutsche Bundesbank, unpublished paper, 1991. 19. For example, the Bundesbank employed "volume tenders" in which it set interest rates for several months following the October 1987 stock market crash. 20. A detailed description of these operations is found in Kevin Clinton, "Bank of Canada Cash Management: The Main Technique for Implementing Monetary Policy", Bank of Canada Review, January 1991. -14- Kasman BOC is able to determine end-of-day reserve positions with unusual precision, particularly because these "drawdowns" or "redeposits" of government balances occur too late for banks to make further adjustments to their balance sheets. These transfers have a direct effect on overnight rates in the call and.interbank markets. Daily reserve management activities are geared, however, toward maintaining market conditions consistent with the BOC's weekly Treasury bill rate objective (Chart 5 ) . Kay Instruments of Rosary* Management Intervention tools vary widely across the central banks surveyed. In part, these instruments reflect the differing financial environments facing authorities in the six countries. The choice of instruments is, however, also related to specific objectives of reserve management and the means chosen by the authorities to signal their policy intentions to financial market participants. A summary of the market operations employed by the six central banks is presented in Table 2. The U.S. Federal Reserve operates mostly in the secondary market for government securities. The prototypical open market operation, the outright purchase or sale of government securities in the secondary market, has long been the major instrument for providing permanent bank reserves in the United States. The breadth and depth of this market allow the Federal Reserve to add or drain large amounts of reserves without significantly distorting yield structures. Although outright purchases of securities provide the primary source of secular reserve creation, the Federal Reserve typically conducts less than ten outright purchases and sales in the market each year.21 On a daily basis, policy is implemented primarily through repurchase agreements (which add reserves) or matched sale-purchase agreements (which drain reserves). These reversed security transactions involve lower transactions costs than outright transactions and provide a much more flexible instrument for the temporary adjustment of reserve positions. They are conducted through a large existing private market and may range up to fifteen days in maturity, although they usually mature in one or a few days. Although most of these transactions are 21. The Federal Reserve does take advantage of purchase or sale orders of foreign official accounts when these are consistent with reserve objectives. -15- Kasman designed to smooth temporary fluctuations in reserve markets, they are also employed by the Federal Reserve to implement a change in its policy stance. In Japan, Canada, and the United Kingdom, as in the United States, outright purchases of securities are the main asset counterpart to the expansion in the monetary base over time. In Japan, the purchase of ten-year government bonds meets the secular demand for reserves but is not important in short-term reserve management. The BOJ conducts a variety of other operations to affect reserve positions on a temporary basis. Outright and reversed .transactions in commercial bills and other money market instruments are designed to offset seasonal and other short-term fluctuations in reserve demand. The discount window lending activities remain the primary tool to smooth unexpected day-to-day fluctuations in reserve positions. Canadian monetary authorities also employ a variety of instruments to achieve policy objectives. The BOC's weekly participation in the three-month Treasury bill tender and its purchases of long-term government bonds at issue are the principal asset counterparts of money base increases in Canada. On a dayto-day basis, the BOC's drawdown/redeposit mechanism, described earlier, is its primary instrument of reserve management. The distribution of drawdowns and redeposits among clearing banks is determined at twice-monthly auctions where banks bid competitively for allocation ratios of government demand deposits. Supplementing this mechanism are other market operations, including outright purchases of short-term government securities and repurchase agreements. All open market operations are, however, routinely neutralized by the BOC as part of its drawdown/redeposit activities. As a result, open market operations are geared toward directly influencing particular money market interest rates. In the United Kingdom, BOE assets are held primarily in the form of short-term eligible bills. The BOE routinely purchases bills to roll over its maturing portfolio and to achieve its short-term reserve management objectives.22 22. Eligible bills include Treasury bills and commercial bills carrying two established names, usually those of a British bank and a discount house. The BOE will buy or sell bills of up to three months' maturity and does conduct some reversed security transactions. -16- Kasman As noted earlier, BOE operations are designed to relieve daily money market shortages through the outright purchase of bills from discount houses. Although it typically maintains a fixed stop rate on Band 1 bills, the BOE generally does not relieve the entire shortage through Band 1 bill purchases. It conducts bill operations in maturities as long as three months, designing these operations to exert influence on rates throughout the money market term structure. In addition, the BOE can refuse to relieve shortages through bill purchases if it is unhappy with the rates being offered. In these circumstances, the BOE can offer repurchase agreements on its own terms or invite discount houses to use their borrowing facilities at 2:30 p.m. at a rate set at the BOE's discretion.23 Neither the Bundesbank nor the SNB holds significant portfolios of securities because well-developed short-term money markets do not exist outside the interbank market in Germany and Switzerland. In this environment, the Bundesbank uses central bank lending (mainly bills rediscounted) and bond repurchase operations as the major vehicles to augment the monetary base. The Bundesbank has established special provisions for reversed security transactions with banks; these transactions serve as the Bank's primary instrument of short-term reserve management. The Bundesbank conducts periodic tenders (usually weekly) for one- to two-month repurchase agreements. These repurchase agreements consist of a secular component and a component that makes temporary adjustments to reserve positions. Repurchase agreements have steadily increased as a share of Bundesbank assets since the mid-1980s, gradually supplanting discount window lending as the principal asset counterpart of the money base. Other instruments, such as foreign exchange swaps and the transfer of government deposits from the Bundesbank to banks, are employed when daily adjustments in reserve positions are deemed necessary.24 23. The 2:30 borrowing differs from normal day-to-day late assistance in that the interest rates on loans are published and the amounts borrowed do not count against discount houses' borrowing facilities. 24. Foreign Exchange swaps are usually employed to neutralize an expansion in reserves resulting from international capital inflows. Transfers of government deposits between the Bundesbank and private banks are generally used to offset temporary reserve shortages associated with tax payments. -17- Kasman In Switzerland, the domestic securities market is extremely narrow. An active interbank swap market for major foreign currencies does exist, however, and the SNB employ's currency swaps as the primary instrument of both permanent and temporary reserve operations. Conducted daily in the form of U.S. dollar-Swiss franc swaps with a small number of banks, these operations currently provide over 90 percent of the reserve creation for Swiss banks. Since the dollars purchased by the SNB in these transactions are covered forward, these transactions can be viewed equivalent to temporary operations in domestic securities. Because swaps are settled with a two-day lag, the SNB supplements these activities with same-day shifts of government deposits between its books and those of private banks. Central Bank Credit Facilities The monetary authorities in all six countries considered offer banks a facility for obtaining credit. The market operations described above, however, have largely replaced central bank credit as the major tool for short-term reserve management in these countries. At present, most central bank lending facilities are designed to meet unforeseen and temporary end-of-day liquidity shortages or to provide assistance for institutions in times of stress. Nonetheless, the role of lending in the six central banks' implementation strategies varies. A summary of key characteristics of central bank lending facilities is presented in Table 3. In four of the countries considered (Germany, Japan, the United States, and Switzerland), a collateralized credit facility is made available to banks at below-market interest rates. In Germany, Japan, and Switzerland, discount window lending, determined by quotas, provides an ongoing source of subsidized funds to meet a portion of secular reserve demand. The Bundesbank's facility is particularly large, currently accounting for about one-quarter of total central bank assets (Table 4 ) . The large volume of subsidized discount window lending in Germany is designed, in part, to offset the costs to banks of high levels of required reserves. Because German and Swiss banks fully use their quotas most of the time, discount window lending does not accommodate banks' -18- Kasman unanticipated reserve needs in these countries. Both the Bundesbank and the SNB provide an additional line of credit at a penal rate to meet unexpected short-term liquidity needs.25 These facilities, called Lombard loans, effectively cap interest rate increases for short periods. Swiss Lombard rates float daily at two percentage points above the average of the previous two days' interbank call money rates. German Lombard rates, in contrast, are fixed by the Bundesbank and in recent years have generally remained no more than 100 basis points above the repurchase agreement rate. Lombard lending by the Bundesbank has soared for brief periods on several occasions in recent years. These surges in lending reflect, in addition to market-related liquidity developments, a strategy of tightening policy: money market rates are increased first; once market pressures build, these increases are validated in official rates.26 In the other countries reviewed, the central bank has greater freedom to decide the terms on which lending is made available. In the United States, the Federal Reserve generally sets the discount rate below short-term market rates and rations access through administrative guidelines. Lending is designed to provide for unexpected liquidity needs, particularly at the end of reserve maintenance periods. For institutions that use the window frequently, however, future access is reduced, raising the implicit cost of borrowing. Furthermore, worries about potential adverse market reactions to discount window borrowing have developed in recent years as bank failures and earnings stress have risen. The use of the discount window has, consequently, been relatively limited. Of the countries under review, only Japan makes lending an important instrument in short-term reserve management. Discount window lending makes up a substantial share of BOJ assets 25. Both central banks impose quotas on access to Lombard facilities, but the quotas rarely present an effective constraint on borrowing. 26. The maturity of Lombard loans is determined by the remaining maturity of securities rediscounted. Generally the Bundesbank grants such loans with the expectation that borrowing should be repaid the following day. Nonetheless, there exists some incentive to borrow heavily through Lombard loans when repurchase interest rates are expected to increase above Lombard rates at the subsequent weekly repo tender. -19- Kasman (currently over 10 percent), and the Bank actively manages its lending policies on a daily basis. The BOJ can either increase or call discount window loans at its discretion, and typically uses this instrument to smooth daily fluctuations in bank reserve positions. In addition, with its "plus-one-day" pricing of loans, the BOJ's effective lending rate exceeds the discount rate and can become penal for very short-term loans.27 Discount window lending thus gives the BOJ a highly flexible instrument for influencing daily conditions in interbank markets. England's central bank also has discretion in providing credit. In its transactions with discount houses the BOE can decide whether to provide credit and what the price of that credit will be. Funds are made available for "late assistance" to meet interbank clearing needs, but the terms of this borrowing are determined by the BOE and are not disclosed publicly. Generally funds are lent at or above market rates, in a way that permits the discount house to predict the cost accurately. As noted earlier, the BOE occasionally uses its lending policies to signal changes in its policy stance, allowing discount houses to borrow at a publicly announced rate after it has refrained from accommodating reserve demand earlier in the day. The central bank lending rate of the BOC (the Bank Rate) is adjusted weekly and set 1/4 percentage point above the previous Thursday's three-month Treasury bill tender. Until recently, banks were guaranteed recourse to this facility only once during a reserve maintenance period. The cost and availability of further borrowing were subject to the discretion of the BOC. Funds were provided, but at a rising cost for repeated use. These restrictions on access to BOC credit- were removed in November 1991. Banks can now borrow freely at the Bank Rate either as overnight overdrafts or to meet reserve deficiencies, a 27. The interest charged on discount window calculated on the period of the loan (using the rate) plus one day. Thus, the effective rate of the BOJ reduces the length of time for which it lend. -20- loans is official discount interest rises as is willing to Kasman change seen as a necessary prelude to the phased elimination of reserve requirements that began in June 1992.28 In addition to providing credit to meet short-term liquidity needs, most countries also offer a facility to absorb excess reserves so that short-term downward pressures on interest rates will be limited. In Japan, the BOJ has the option of withdrawing outstanding loans at will during banking hours. The Bundesbank's Treasury bill selling rate functions as an effective floor on call money rates in Germany, and in Canada, matched or outright sales of Treasury bills serve a similar purpose. In the United Kingdom, discount houses can offer to purchase securities from the BOE in the afternoon if surpluses emerge. Reserve Requirements Like central bank lending, required reserve ratios have diminished sharply in recent years. Required reserve ratios in all these countries stand well below their levels of the early 1980s; in some countries, requirements no longer effectively constrain bank behavior. In addition, the once common practice of altering reserve requirements to adjust the monetary policy stance has largely been discontinued. Nonetheless, most central banks still view reserve requirements as an important part of their implementation procedures. Requirements are seen as strengthening and stabilizing the short-run demand for reserves, thus enhancing central bank control over interest rates.. A summary of important characteristics of reserve requirement regulations is presented in Table 5. Required reserves in all six countries under review are determined by ratios linked to categories of bank liabilities.29 In the United States and, until recently, in Canada, requirements have primarily been imposed on transactions deposits, a practice 28. Under the regulations in place since June 1992 a bank with a cumulative deficiency at the end of a reserve maintenance period may pay a fee, charged at the Bank rate in lieu of taking an endof-period advance. In practice, banks have adopted the fee option so that end-of-period advances no longer appear on the BOC balance sheet. 29. In June 1992, Canada removed required reserve ratios as part of its phased elimination of reserve requirements. -21- Kasman that reflects earlier attempts to use reserve requirements to facilitate the targeting of Ml through operating objectives for bank reserves. Elsewhere, requirements are more broadly based. In the United Kingdom, Japan, and Switzerland, requirements are roughly similar across types of eligible liabilities. In all these countries, the period in which liabilities are incurred (the accounting period) ends before the period in which required reserves are held (the maintenance period). These lagged or semilagged accounting mechanisms are operationally convenient and, where reserve requirements are binding, provide central banks with a relatively good estimate of reserve demand within a maintenance period. For all six central banks except the BOE, reserve projections at maintenance period horizons are a key element in determining policy operations.30 Although lagged reserve requirements predetermine the demand for reserves, they can also severely limit the interest sensitivity of reserve demand, particularly at the end of maintenance periods. Unforeseen shifts in either the demand for or the supply of reserves have often led to large fluctuations in interbank rates at the end of a maintenance period. To provide greater flexibility in reserve management, particularly in the early stages of a maintenance period, nearly all of these central banks allow required reserves to be met by average reserve holdings over a maintenance period.31 Reserve averaging gives value to banks' excess reserve positions by enabling the banks to maintain offsetting deficiencies during other days within the period. As a result, banks have an incentive to arbitrage away the interest rate effects of temporary reserve shocks. Through this mechanism, required deposits at the central bank can function 30. As noted earlier, clearing banks in the United Kingdom provide the BOE with an estimate of the operational balances they wish to hold each day. The BOE uses these estimates as a guide in determining daily security operations. 31. Reserve averaging extends over one month in Germany, Japan, and Switzerland, and over two weeks in the United States. In Canada, reserve averaging extended over two half-month periods until June 1992, when it was extended to one month. -22- Kasman as an important aid to central banks in promoting interest rate stability.32 The extent to which bank reserves actually serve as a buffer stock is related to the level of reserve balances held at the central bank. Because overnight overdrafts are restricted in Switzerland, Japan, and Germany, and penalized in the United States, Canada, and the United Kingdom, the cost of running reserve deficiencies rises substantially when average reserve balances are low. In the United States and Canada particularly, concerns have arisen about the banking system's reduced ability to absorb reserve imbalances at low reserve levels. Reserve deposits held at the central banks of both countries have fallen sharply in recent years as a result of a secular increase in demand for vault cash to satisfy reserve requirements and, in the United States, a reduction in reserve requirements (Table 6). 3 3 Reserve management strategy in the United States traditionally focused on the two-week average reserve levels held by banks over a maintenance period. Since the cut in reserve requirements in December 1990, however, the open market desk encountered increasing conflicts between this strategy and daily federal funds market conditions. Many banks have become less tolerant of excess reserve positions early in the maintenance period, a reaction that has often led to significant late-day downward pressure in federal funds rates. At the same time, the funds rate in the morning can be a misleading guide to reserve market conditions as banks sometimes hold on to reserves early in the day to guard against inadvertent overdrafts. When faced with these conflicts in conducting its operations, the Desk has chosen to pay greater attention to daily trading.conditions in the federal funds markets to prevent misleading signals from being sent to markets.34 32. A provision for the carryover of a portion of reserve surpluses (or shortages) allows for some additional flexibility in managing reserves across maintenance periods in the United States. 33. In both countries, holdings of vault cash over previous maintenance periods satisfy current reserve requirements. Increased demand for yault cash thus lowers required deposits even when reserve requirements are unchanged. 34. See "Monetary Policy and Open Market Operations1* for further details. -23- Kasman In two countries, the United Kingdom and Switzerland, reserve requirements place no effective constraint on bank behavior. In the United Kingdom, banks must place small nonliquid deposits at the Bank of England for six months at a time. This requirement provides the BOE with operating income but is not intended to play a role in the BOE's monetary policy operating strategy. Since effective requirements are lacking, demand for reserves (operational deposits) is determined entirely by daily clearing needs. In this environment, the BOE has developed an operating strategy involving a number of daily market operations to interest fluctuations and other intraday developments. In addition, banks' uncertainty over their end-of-day clearing needs is eased by the availability of BOE late-day lending facilities to discount houses. BOE policies serve to stabilize reserve demand and encourage banks to economize on reserve holdings (Table 6 ) . Since the decline in reserve requirements in Switzerland in 1988, the SNB has placed greater emphasis on smoothing daily fluctuations in interest rates in its daily activities. In addition, central bank lending facilities in the form of Lombard loans are available to banks without restriction to meet unexpected liquidity shortfalls. Nonetheless, the SNB is much less accommodative than other central banks in its approach to offsetting temporary reserve disturbances, prohibiting overnight overdrafts and setting a large spread (200 basis points) between market and Lombard lending rates. In this environment, Swiss banks have chosen to hold substantial reserve deposits in excess of those required by regulations. RELEVANCE POR FEDERAL RESERVE OPERATING PROCEDURES The varied institutional and political environments facing these central banks make it difficult to assess whether practices followed in any one country would be useful to another. Nonetheless, the comparison of operating procedures presented above does provide interesting insights, some of which may be relevant to U.S. policy makers. The similarities in operating strategy among these central banks dominate any existing differences. All six banks currently gear their daily policies toward influencing money market interest rates; all except the SNB use short-term interest rates as operating objectives to guide their reserve management activities. -24- Kasman Furthermore, none of the banks aims to control interest rates rigidly. Although the tolerance for interest rate divergences from objectives differs across banks, authorities generally allow market forces to determine interest rates and intervene only to limit short-term fluctuations or to alter rates when changing economic conditions warrant. Since interest rate operating objectives are transmitted to economic activity largely through their linkage to longer term interest rates and other financial prices, central bank intervention strategies are designed to communicate information about current and future policy that strengthens this transmission. In most cases, interest rate objectives are changed in small steps to stabilize expectations across the term structure. In some countries, central banks intervene in assets of varying maturities to influence the money market term structure directly. In addition, these central banks actively seek to limit the daily volatility of targeted interest rates in order to reduce uncertainty about the stance of policy. In some countries (Germany, the United Kingdom) intervention rates under the tight control of the central bank send a precise signal of central bank intentions. Elsewhere, although some interpretation of money market interest rate movements is necessary, the central banks stabilize their targeted rates sufficiently so that the basic thrust of their policies is clear. Over the past decade, foreign central banks have increased the role of open market operations as a reserve management instrument, moving toward an approach long followed by the Federal Reserve in the United States. At present, each of the central banks reviewed employs some form of open market operation as an instrument for controlling reserves. Some foreign central banks conduct their operations through special arrangements with banks or other counterparties. But where these arrangements exist, they generally reflect the limited development of secondary security markets. More meaningful differences among the six central banks emerge in the functioning of their credit facilities. To be sure, the monetary authorities in all six countries extend credit to banks with temporary clearing imbalances and to banks in financial stress. The foreign central banks, however, differ from U.S. practice by moving away from administrative controls on credit allocation. -25- Kasman In three countries -- Germany, Switzerland, and Canada -banks are able to access an open-ended line of credit for temporary liquidity needs at their discretion. Borrowing rates are set above the prevailing market rates and, in Switzerland and Canada, rates adjust automatically to market rates. In Japan and the United Kingdom, access to the discount window remains at the discretion of the central bank. In practice, however, discount houses in the United Kingdom can count on the central bank to meet temporary liquidity needs at rates close to the Bank of England's prevailing intervention rates. These facilities provide foreign central banks with a flexible instrument to contain interest rate pressures, particularly late in a trading day when other intervention instruments are unavailable. In addition, each of these foreign central banks offers a facility to absorb late-day reserve excesses and thereby moderate downward interest rate pressures. The Federal Reserve's discount mechanism has considerably less value as a device for smoothing interest rates. U.S. discount window lending is provided at subsidized rates and in accordance with administrative discretion. Partly because of this subsidy, the Fed discourages frequent use of the window. In recent years, banks have shied away from approaching the window, fearing that the markets will perceive them to be dependent on discount window support. The unwillingness of banks to borrow at the discount window also reduces the ability of banks to shed excess reserves through their repayment of outstanding credit. In an environment of high, binding reserve requirements, the methods employed by central banks to allocate credit might not significantly affect their.ability to limit interest rate variability. With sufficient averaging provisions in place, banks can be expected to arbitrage away the interest rate effects of transitory shocks to their reserve positions within a maintenance period. Indeed, recourse to Lombard loans in Germany, the country that has the highest reserve requirements and longest maintenur,:a period of the six countries considered, is quite small under normal market conditions.35 35. The Bundesbank estimates normal Lombard lending levels at DM 0.5 billion, a level representing less than 0.2 percent of total central bank assets. As noted earlier, Lombard lending has risen sharply during short periods in which the Bundesbank allows repurchase agreement rates to push up against Lombard rates before it tightens policy. -26- Kasman But in the United States, recent declines in reserve requirements, coupled with increased demand for vault cash, have sharply reduced reserve deposits at the Federal Reserve. In an environment where overnight overdrafts are costly, the ability of banks to take advantage of reserve averaging has- become more limited as reserve deposits decline. These developments, coinciding with the deterioration in the functioning of the discount window, may have increased the sensitivity of the federal funds rate to reserve shocks. The central banks examined here that have faced similar concerns about the effects of lower reserve requirements have tended to revise their procedures to allow for a more elastic late-day reserve supply. The BOE, operating for over a decade in an environment where banks are effectively free from reserve requirements, has developed a strategy combining the elastic provision of central bank credit for late-day reserve imbalances with frequent open market operations during the trading day. The SNB has placed greater emphasis on interest rate smoothing in daily operations since a reduction in reserve requirements in 1988. In addition, while maintaining a large spread between rates on its Lombard lending and overnight rates, the SNB has increased access to central bank lending facilities since the decline in required reserves. In Canada, restrictions on bank access to BOC credit have also recently been removed as part of the phased elimination of reserve requirements. The example of other central banks, then, raises a question: Should the Federal Reserve consider revising its operating procedures to adapt to lower reserve requirements? It could be argued that some revision enabling the Federal Reserve to supply reserves more elastically outside of the time it conducts open market operations could help limit the variability of interest rates from objectives. To resolve this issue, an assessment of federal funds rate variability and its effect on monetary policy transmission is essential. The Appendix sheds some light on the issue by presenting evidence on actual interest rate variability. The interday volatility of the federal funds rate does appear to have risen following the decline in reserve requirements in 1990. However, U.S. federal funds rate volatility remains low in comparison with the volatility observed in overnight interbank rates in other countries. More important, perhaps, the evidence -27- Kasman indicates that increased federal funds rate volatility, within the range observed, has not diminished the response of three-month money market rates to changes in interest rate objectives. Thus, these results do not suggest that the reduction in reserve requirements has weakened the effectiveness of the Federal Reserve's policy transmission mechanisms. CONCLUSION Our analysis, while far from conclusive, provides insights that may be useful in assessing monetary policy operating procedures in the United States. Like the Federal Reserve in the United States, several foreign central banks have lowered their reserve requirements in recent years. Their experience indicates that interest-rate-oriented monetary policies can be carried out in an environment of low, nonbinding reserve requirements. Central banks operating in such an environment have been able to achieve their interest rate objectives using reserve management techniques quite similar to those employed by the Federal Reserve System in the United States. Foreign central banks have, however, seen the need to develop mechanisms that provide a highly elastic supply of reserves tb restrict the intraday fluctuation of overnight interest rates. In most, countries, the authorities have designed their central bank lending facilities, with rates set at or above current market interest rates, to achieve this goal. The empirical evidence presented in this article indicates that the recent decline in reserve requirements in the United States, combined with the increased reluctance of banks to approach the discount window, has been associated with greater variability in the federal funds rates. Nevertheless, the evidence suggests that this rise in variability has not diminished the effectiveness of U.S. monetary policy operating procedures. Within its current range, the variability of the federal funds rate remains low and does not appear to have affected the linkage between federal funds and other money market rates. APPENDIX: OVERNIGHT INTEREST RATE VARIABILITY The review of central bank operating procedures presented in the text suggests that foreign central banks, in contrast to the Federal Reserve, employ their reserve management instruments, particularly lending facilities, in a way that places strict -28- H a 3 rt »< ID H - 3 iQ H 0 *1 0 H> * < *1 H - H VO P H i *< ID 0 rt o M o CA 3 * 1 1 0 3 3 * ^ h ^ l rt p i *rj » H-PUD 0 o 0 H- 3* 3 OS W3ftftM i r p i i i i(I>QHa- H H - a , a D 3* v-i. H- rt C O ID H CHID en <n D PH O< C •A Co A*<O 3 ' frp VO W H i £ p rr N rt) n M p 3 i t rt ID H M - ID H. ID < w 3 »i]rt fl) • H - O ID CA iQ *1 ID3*MTJCA P I I D M P 3r 3* g o a vo (nrt q o f t H- P TJ H- O P P (0 ID H 3 a 3* 3 n • 3 rt i t\> VO 1 K 3 H vo * - ID CA ID i Q r t n a rt H ID 0 ID P M •-» M p i-h »i 3 • U> H-*»d P < TJ rt M ID 3* (D <D o 5 0 P ID H i Ml" i H - «i (D H C 1 3 < 2 3 ID r t H- H - ID a O a 3* ID 3 ?: W p> H- £ 3M-CA« K : M h-- <J ID 3 ID 3 rt CA f t ht H - CL C ( 0 ID H - 3 P> n 2 o ID w cID W a o a hID M > 0 ID - 0 H i CA <J pi ID ID M 0> TJ P 3 (D ID ID C H g artiii H-TJ IDIDHP> H H » 0* 3 rt C ID »1 ID 3 H- rt a> a in rt P rt ID W 3* HO 3* TJ < 0 rt • ID ID CT 3* cr o *i o IQ 3*t to £ >i rtio H*< O 3'ftao» aoo»wo> M 30»OOrtH3rt0»MlD H- H3 3 rt 3 rt tQ M»< rt 3* ID a P IDID ID 3 £ »1 H-3 a p ID P 3 rt CO CA f ) Hrt rt rt Hi 3 C 3'K 0 H P r t pi H ID rt ID M- ID H- 3 3 X w 3 rtO XlD*0 H-iQ p 3 • a ID IQ *i CA O 3 !-»*< H I rt p vo a ^ h i D M V O » a C M O C L H O HiQ 3 rt. 3* ID HH-C rt h- 3 N rt M O i Q ID 3 * P >r}*< 0 H rt Q H H - 50 O ID *< W ID O h o ID C H - r t »ij n r t rt 3 p ID 3* ID ID i Q M Q i 0» W H. K ID 3 r t h | ID • • < M - 0> ID (D 3 ?:w C D i X ID tl rt h O ID ID P H h a rt ID TJ H-* M »1IQ 0 M CA HC O < vo *v p ID 3 Di 3 a S 3 3 vo 0 M fX r t 3 H i H - (D H ?JK)H $ ID O rt rt 3* 2 0 P » ' rt 0 ( n O h rt ID c TJ 3 x ID a *i P tl ID P hP r t (D < rt ID rr 3 ID rt n 0 C 3 w ID a H3 tQ • 2: 0 rt ID * 3* 0 £ ID hh H 0 3 3 ID 0» 3 H-» ID < ID H-» (fl 0 < ID M H» VO 00 00 1 vo I- 1 rt rt 3* ID (0 ID rt H- 0 3 i-» ID (0 (0 P C H- CA H ID 3 m P» 3 a 0 3 0 rt O 1 3 0> H rt o> ID O ID (A 3 *n HID I-* a ID 0> CA H M<Q 0» 3- rt ID tfl 3* 0» < ID a *< rt H- 3 rt ID H ID 10 rt H- <; H ID »1 <Q ID 0) a <> P « 0 3 <I 3* P 3 ID rt 0> rt 3 H- < ID 0> < aID H < * H0* rt 0 C 3 CL "» ID 3P H> qui o 0 3 H TJ O 3 O O M H i O H OTJOPOfDOOTJvoOTJP 3 (D<!3HH3C(!)U)h(D3 3iQ w TJ H (HQ H - g (D 3 >1 H (Dcrv c P 3 < D *O< Pd r (tHrMt prt - nH» r t rt(D o S ID h 0» <Q ID » 0) cr 0 rct H-» **> tr o> (A H(0 rt ID H H- 0» cr HH Hrt *< H(fl h ID H-» 3 rt 0) a 0> H- !-• >< * 0> o 0 3 50 HW <J ID • 0 3 rt O z Hrt 30 rt 3* ID M O 0 C 3 <; H- rt ID (A rt O O H> Hi CA ID rt H 0» ID CA *a Hrt ID rt 3ID 3ID rt 3- £ O 3 ID 0> CA C n ID 3* ID HH •< CA rt O H ID h 0* 3 O ID HCA Hi • O HCA rt O H> H- 3 rt ID H ID CA rt Hrt ID rt ID' ID CA ID H H- h H- H 0» 3 0> H H X ID rt 3 CA rt h C g ID 3 rt CA 0> <> 0 0» P> tr HH* Hrt *< »d H ID CA ID 3 rt tr H ID P> CA 3 O O rt 3O < ID H 3 H- *Q 3rt H 0 < H 0> rt ID CA 3 rt HHi HID Cb V H ID O HCA ID H 0 tr «_i. ID O rt H- < ID • tc O 5: ID »< * • O £ O 3* 0> 3 * h ID 3 CA H- O (Q ID CA <Q 3- 3 ID H < H- HiQ rt H 0> rt ID HHi HO 0> 3 rt < O h-» 0» rt H- »-* Hrt ^< P H ID 3 0 rt iQ 3- H I 3 3 O ID CA O n 0 c 3 rt ID a O CA rt o ID 3 H- rt N. 3 P» CL ID 1Q H ID ID CA rt 0» 3 O ID CA O 0» 3 3 O rt H-« rt 3- M HO 3O K C a ID 0> *0 0> O HHi ID h ID 3 » h-* rt tr a» 3 x CA rt 3 H- 3 TJ 3 fD fjj O i— a H- tr ID 3 O 0* rt H- O 3 0 Hi 3* O < ^ ID H ID H 0» »< a < 0» H rt a H a o» !-•• 3 ID *1 171 a ID rt H-« H- P H- O H- rt ID I 3* CL 0» H rt H> CA 3 < *< aM- ID o 3 »-• 0 0» 0> ^ ID TJ HO rt C h ID H 0> K h 0 0 3 CA H- •a 0* rt HID H H p> h-« Hi »i 3 ID CA ID »i rt H HID (A TJ 0 H- 3 3' < ID H- 3 ct ID h ID CA rt H 0> rt ID CA H- O S *< M 3 TJ 1 H- CA rt 0> 3 O ID 3 - »0 P 3 O H- M ID "• rt 3- M-. CA 3 ID 0> CA c !-• H* rt *< tr 0 Q 3 3 c CA a rt ID P 3 l-* ID < ID M c H a Hi rt. ID H ID O TJ H 0 a <J H- a ID o> M ID 0 0) x o o c CA H 0» rt ID c 0* H O c 3 a rt 3- H(A a TJ HiQ 3 3* 0> H O ID 3 rt H 30» 3 M 3 P 3 ?r <Q <Q < ID H ID CA ID rt P tr * CA a P rt ID CA O 3 H 3 rt P rt (D CA C 3 a ID H P TJ O h-» H- CA rt O 3 a • H XT 3O 3 rt H P H- tr P 3 X HCO O 3 a 3 ID p CO <• c H rt ID M ID (A rt H p rt ID O TJ ID *1 p rt H- 3 iQ O Hi ID MCA O 0 3 CA rt »i C a rt ID 3 O < H- O 3 < iQ ID tl 3 HIQ 3* rt CA CD O CA 3 < 3 rt ID HID P O ID O ID ID Hp rt H- t xr p. o o ID XT CA ID a o K P 3 H- O O P a* iQ ID TJ 3 P 3 HCA ID rt N H-» H« HO tr P 3 3 < ID o ID O < ID cr O 0» TJ 0 H- CA 3 CA ID H- Hi H rt H- ID H O !-• < P ID 0> MCA H iQ ID C 0 tr <_i. ID O rt H- a »0< a tr < UJ. 0» H ID •^ O H3* H CL Tl 0» H rt HO H P rt ID CA H O C 3 a P 3 ID p 3 P CL i_». c CA rt ID • •^ 3CA ID O O 3 c 3 Hi iQ ID CA rt •< H3 iQ H3 O 3 O Hi O (A 0 M c rt ID ID H-» vo 00 00 1 vo I-* T3 ID N TJ H ID CA ID 3 rt CA rt € O rt a •-9 3* ID P *1 O C 3 P Hi H- 'M CA rt ^1 rt i a P K ID 1 O O < 3 TJ P M < p 3 O ID rt 3* 0 Hi ID < P 3- H H- Hi O P P C O tr H- 3 M H* rt c rt M HID CA .»< 0 O < ID CA < 0 3 ID ID Hi H- Hrt CA o O HM Mrt HID CA H 3 MiQ 3rt M- 3 rt ID »i ID CA rt Hi £ O 3 CA H- a ID H ID a Hi H- 3 O H *i P rt ID X P < P p 3 rt ID *1 ID CA rt 3- H P a ID 50 *< • rt ID CA rt (D ^ ID H P ^ H H- P cr HM Hrt ID *1 H- 3 3 rt ID CA ID 3 3* 3* H 3 ID rt ID CA H- C 3 < ID » P Hrt ID Hrt a HCA rt C o Q HiQ TJ C rt rt 3* tr O Hi H O rt P H ID CA !-•• 3 0 ID CA M ID H-» ID 0 3 C H- cr tr 3 < a P H* < !-•• 0 ID H P ID > ID CA ID h P iQ ID O P rt H- rt tr cr ^ ID rt rt ID CA a 3* P 3 O H P *ID K ^1 0 rt ID •^ P rt a 3* cr HH« Hrt rct tr ID P Hi 3 CA P 3 < ^ HP M HiQ Hi HQi ID tr CA O a cr H. P < aID < M- ID N. p «Q ID P *< N. p iQ ID ID CL O < ID P ID P P rt tr ID 3 rt ID H ID CA rt H. P rt ID P rt ID CA CA ID Hi c !-• M CA CA ID (A CA M- 3 iQ rt 3ID CA 3 Kasman indicate the degree of intraday interest rate variability, an issue of some concern to U.S. policymakers. The evidence also points to a relationship between required reserves and overnight interest rate variability. In the United.* Kingdom and Switzerland, the two countries operating with low, nonbinding reserve requirements, overnight rates are much more volatile than the rates elsewhere. In addition, in the United States and Canada, where reserve deposits held at the central bank have fallen in recent years, the decline in reserves has been accompanied by rising interest rate variability. These findings support the view that central banks face greater difficulty in stabilizing interest rates around desired levels when reserve requirements are eased. Nevertheless, increased overnight interest rate volatility, per se, need not erode the effectiveness of monetary policy, particularly if fluctuations in overnight rates are transitory and do not reduce the ability of market participants to identify the authorities' policy intentions. To assess whether overnight interest rate variability has influenced the monetary transmission mechanism, one must determine whether the overnight rate variability impacts on longer term market interest rates. Table A.2 presents regression results estimating the effect of overnight rate variability (MAD°) on the measured volatility of three-month money market rates (MAD").37 As the Table shows, overnight rate variability is not systematically related to three-month money market rate divergences in the United States. Indeed, of the countries surveyed, only Switzerland has large and statistically significant coefficient estimates for transmission. Perhaps a more important issue is whether interbank rate volatility influences the transmission of changes in central bank operating objectives to money market rates. To resolve this issue in the case of the United States, one can test whether the federal funds rate variability measure affects the response of three-month 37. In Table A.2 the volatility of interbank (MAD°) and threemonth money market rates (MAD") are measured as the absolute deviation of rates adjusted for changes in the monetary policy stance. For Switzerland, however, deviations around a thirty-day centered moving average are used. Note that the results are qualitatively unchanged by the choice of volatility measures. -30- Kasman Treasury bill rates immediately after a change in the Open Market Desk's federal funds rate objective. In the regression ARt = c + (bl + b2 MAD°t.i)Afft + ^ ARt is the change in the three-month Treasury bill rate; MAD°t_i is the average absolute deviation of the federal funds rate from the Desk's objective, measured over the preceding objective period; and Afft is the change in the Desk's federal fund objective.38 The coefficient estimate for b2 provides an indication of how variability has affected the transmission of federal funds rate changes. The regression results are presented in Table A.3. Estimates are given for the responsiveness of the three-month Treasury bill on both the day of the federal funds rate change and the five days following the change. As the Table shows, the three-month Treasury bill rates rose on average 22 basis points in response to a percentage point rise in the federal funds rate objective on the day the objective increased. This response increased to 26 basis points after five days. The variability of federal funds rates does not appear to have altered this response. In both regressions, the coefficient on variability is not significant and enters with the wrong sign. Taken together, the results suggest that federal funds rate variability, within the range observed has not altered monetary policy transmission in the United States. 38. This analysis closely follows earlier work by Timothy Cook and Thomas Khan, "The Effect of Changes in the Federal Funds Rate Target on Market Interest Rates in the 1970s," Journal Of Monetary Economics, vol. 24 (1989), pp. 331-51. -31- Kasman REFERENCES Bank for International Settlements. Financial Monetary Policy, BIS, 1984. Innovation and Bank of England. "Bank of England Operations in the Sterling Money Market", Bank of England Quarterly, October 1988. Bartholomae, Andre. "Some Operational and Instrumental Aspects of Monetary Targeting in Germany". Unpublished paper presented at XXVIII meeting of Technicians of Central Banks of the American Continent. Batten, Dallas S. and Michael Blackwell et.al. "The Conduct of Monetary Policy in the Major Industrial Countries". IMF Occasional Paper. No. 70, July 1990. Bernanke, Ben and Frederic Mishkin. "Central Bank Behavior and the Strategy of Monetary Policy: Observations from Six Industrialized Countries". Mimeo. Clinton, Kevin. "Bank of Canada Cash Management: Technique for Implementing Monetary Policy". Review, January 1991. The Main Bank of Canada Cook, Timothy and Thomas Khan. "The Effect of Changes in the Federal Funds Rate Target on Market Interest Rates in the 197 0s." Journal of Monetary Economics, vol. 24 (1989), pp. 331-51. Dale, Spencer. "The Effect of Changes in Official UK Rates on Market Interest Rates Since 1987". Unpublished, Bank of England. Federal Reserve Bank of New York. "Monetary Policy Operations during 1990". Quarterly Review, Spring 1991. . "Monetary Policy Operations during 1991". Review, Spring 1992. Quarterly Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary Policy". Carnegie-Rochester Conference Series on Public Policy, no. 34 (1991), pp. 7-39. -32- Kasman Kasman, Bruce and Anthony Rodrigues. "Financial Reform and Monetary Control in Japan". FRBNY Quarterly Review, Autumn 1991. Kneeshaw J.T. and P. Van den Bergh. "Changes in Central Bank Money Market Operating Procedures in the 1980s". BIS Economic Papers, No. 23, January 1989. Longworth, David and Patrice Muller. "Implementation of Monetary Policy in Canada with Same-Day Settlement: Issues and Alternatives". Bank of Canada Working Meulendyke, Ann-Marie. Markets. U.S. Monetary Policy Paper, and No. 91-3. Financial Federal Reserve Bank of New York, 1989. Ministry of Finance and Bank of Japan. Japan's Short-Term Money Market and Issues. Money Market Study Group, August 1991. Organization for Economic Cooperation and Development. OECD Economic Survey-Switzerland, OECD (Paris, 1989). Poole, William. "Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Macro Model". Quarterly Journal of Economics, vol. 84 (1970), pp. 197-216. Swiss National Bank. Temperton, Paul. U.K. Annual Report Monetary various issues. Policy: The Challenge for 1990s. London, MacMillan 1991. Thornton, Daniel L. "The Borrowed Reserves Operating Procedure: Theory and Evidence". FRB St. Louis Review, January/February 1988. -33- 1. Structure of Short-Term Interest Rates Country Official Rates Overnight Interest Rates Other Key Interest Rates United States Discount rate Federal funds rate Treasury bill rate Germany Discount rate Day-to-day money rate Repurchase agreement rate (one-to twomonth) Lombard rate Three-month interbank loan rate Treasury bill selling rate Interbank call money rate Japan Certificate of deposit rate (three month) Discount rate Bill discount rate i Overnight interbank rate United Kingdom Bank of England dealing rate No posted rate Commercial bank base lending rate Three-month interbank loan rate Bank Rate Canada Money market financing rate Three-month treasury bill tender rate Ninety-day prime corporate paper rate Switzerland1 Discount rate Call money rate Three-month Euro-franc deposit rate Lombard rate Note: 1. Each central bank's primary interest rate objective appears in bold face type. The Swiss National Bank does not employ interest rate operating objectives. 1. United Kingdom: Short-Term Interest Rates Weekly Observations, Wednesdays Percent 16 Base Rate, 14 i p ^ W / i V " " ^ |f=*N \ * i . : lil ' y p a p f t s J 11 ^ Stop Rate ^w ' 12 en i CA '\ l< Overnight I LIBOR 10 I 1988 I 1989 I I I L 1990 J 1991 L 2. Japan: Short-Term Interest Rates Weekly Observations, Wednesdays Percent 9 1, . - Three-month certificate of deposit rate . # #. biP^^i wi j v • •• • » A y \ / ?••• J^\/* / 0 m Overnight call money rate J 1 1 i l l # # \ ^ •L - — i » t II i i 1988 * Values are month-end observations. Discount rate * » ^J^^l I I I Va (A --Y I I I * ••vv • / If i •I M ••••••• J ay * «L 1 i i 1 i i 1 i 1989 i i 1 1 1 1 i l l 1990 1 1 1_L i i 1 i i l.-LX-L„i. i ! 1991 iill 1992 3 * M W ^ u ? r l t n d : R e s e r v e Deposits and Interest Rat* Monthly Averages Billions of SF Percent 12 U *, ,\ 10 h 12 : 1 :•% « » : J :,'*'-""» '; . Lombard1 ? \ ' i \ /%» Rate Hio » / : *» \ ': i •-•-'•-»-•-»»••-«-....-....* iI ; \ x x Resen/e 'll Deposits 1: \ v ^ ' V'i' 8 [~ h .--ij M i CO I / / i \ ;/ ^ \ ; / 6 H" Vv i Overnight • Call Money • Rate "j H 6 /V \: \• * i A 4 H ./: (A • fV H 8 1 1 H 4 / \ "™ . ^ *"• • 1 J H 2 2 K Li i i 1987 i r„.i __L 1988 i i i l 1989 i i t i 1990 i i i l 1991 l i 1 1 1992 4. Germany: Short-Term Interest Rates Weekly Observations, Wednesdays Percent I p» 1988 1989 1990 1991 1992 Canada: Short-Term Interest Rates Weekly Average Percent i u> (A 1988 1989 1990 1991 At Thursday tender. The central bank lending rate (bank rate) is set 1/4 percentage point above this rate. 1992 2. Instruments for Reserve Management Country United States Germany Japan Primary Short-Term Reserve Management Tool Instrument Activity Repurchase agreement Government security Matched purchase and sale Government security Repurchase agreement Government security Repurchase agreement Commercial bills, government securities Discount window lending Other Operations Activity Instrument Purchase or sale Government security Purchase or sale Government security Foreign exchange Swap Purchase or sale Government security Repurchase agreement Commercial paper United Kingdom Purchase or sale Government security, commercial bills Repurchase agreement Government security Canada Drawdown/ redeposit Government deposits Purchase or sale or repurchase agreement Government security Switzerland Foreign exchange Swaps Purchase or sale Cantonal and bank bonds Drawdown/ redeposit Government deposits 3. Central Bank Lending Facilities United States 1) Credit available at below market rates Access restricted by: Q = quotas, D = administrative discretion Germany Japan Yes Yes Yes Q,D Q Q,D Canada Switzerland No No Yes Yes Yes Yes P,D* Interest rate setting: P = posted rate D = set at discretion of central bank 2) Other credit sources available United Kingdom No Yes No Access restricted by: Q = quotas, D = administrative discretion 0 = other Interest rate setting: F = Floats in relation to market rate P = Posted rate D = set at discretion of central bank 1. The Bank of Japan provides credit at the official discount rate. The Bank can add or call loans at will, however, and interest charged is calculated on the period of the loan plus one day. The effective cost of borrowing thus rises as the maturity of a loan is reduced. 2. Generally non-binding. 3. Bank of Canada advances are provided only for overdrafts to meet a deficiency of clearing balances or for an end-of-averaging period reserve deficiency. 4. Central Bank Lending as a Share of Central Bank Assets (Annual Average of End-of-Month Observations) 1985 1988 1991 United States 0.7 0.9 0.1 Japan 8.4 13.6. 12.1 29.4 22.5 25.0 United Kingdom 1.8 1.8 3.0 Canada 7.4 7.4 2.2 2.2 2.0 2.0 Switzerland 9.9 0.9 1.2 Germany ^ J"? Q> Reserve Requirement Regulations i Ui i United States Japan Germany United Kingdom Canada* Length of reserve accounting period 14 days 1 month 1 month 6 months 1 month 3 months Length of maintenance period 14 days 1 month 1 month 6 months 15 days 1 month Interval from end of accounting period to end of maintenance period 2 days 15 days 15 days 180 days 30/45 days 50 days Highest reserve ratio for demand deposits 10 1.3 12.1 0.5 10 2.51 1.2 4.95 0.5 Yes No Yes No No No Yes No Yes No No Up to 50 percent No Yes Yes No No No No Highest reserve ratio for other deposits Averaging provisions Carryover provisions Vault cash satisfies requirement 1. 0.5 w Yes Yes Yes Penalty for reserve deficiency (percentage above central bank lending rate) Interest paid on reserves Switzerland 3-5 No No Includes time deposits with a term to maturity up to three months. * As of June 1992, reserve ratios were eliminated in Canada as part of a planned phaseout of required reserves. Currently required reserves are set at a predetermined amount; this amount will decline to zero in 1994. The maintenance period has been extended to one month. Banks incurring a reserve deficiency pay a penalty calculated at the Bank rate. 6. Reserve Deposits Held at Central Banks as a Share of Total Bank Liabilities (Year Average of End-Month Observations, in percent) 1980 1985 1988 1991 United States 1.6 0.8 1.0 0.6 Japan 1.6 1,1 1.0 1.0 Germany 7.2 5.6 5.5 5.5 Switzerland 4.0 3.1 1.7 0.7 United Kingdom 0.3* 0.1 0.1 0.1 Canada 3.9 1.4 0.8 0.4 * Figure is for year-end 1981. to A.l. Overnight Interest Rate Variability (Mean Absolute Deviation of Daily Observations, in Basis Points) Deviations from Thirty-day Centered Moving Average 1988 1989 1990 1991 Average 1988-91 1988 12.3 11.9 12.3 21.1 14.4 13.0 11.8 12.8 18.5 14.0 8.7 8.5 7.1 8.4 8.2 12.5 8.5 7.4 5.8 8.6 Germany 15.7 18.2 13.6 13.4 15.2 15.8 17.4 14.5 14.8 15.6 United Kingdom 50.4 32.9 14.8 25.3 30.9 52.5 39.7 14.2 25.0 32.9 9.7 13.4 21.3 28.7 18.3 11.0 15.7 21.3 28.8 19.2 33.8 34.8 37.8 35.5 United States Japan 1 Deviations from Mean Adjusted for Changes in Policy Stance1 en i Canada Switzerland 1989 1990 1991 Average 1988-91 Note: Overnight interest rates are the effective overnight Fed funds rate (the United States), overnight call rate (Japan), day-to-day money rate (Germany), London interbank offer rate (the United Kingdom), overnight money market financing rate (Canada), and overnight call rate (Switzerland). 1. Values are average absolute deviations of overnight rates from a mean that changes along with estimated shifts in central bank interest rate operating objectives. (A A.2. The Transmission of Overnight Rate Variability to the Variability of Three-Month Money Market Rates (Based on Monthly Observations, 1988-1991) MAD",, = C + B MAD°t + Ut . B j? DW United States 0, 0.12 .12 ((4.79) 4 ,. 7 9 ) --0.16 0 ..16 (-0..95) (-0.95) -0.01 2.23 Japan 0, .04 0.04 ((0.90) 0 ,. 9 0 ) 0. . 2 2 0.22 ((0.41) 0 ,. 4 1 ) -0.01 2.34 Germany 0. . 0 5 0.05 ( 1 .. 4 6 ) 0, .25 0.25 ( 1 .. 2 8 ) -0.02 1.92 United Kingdom 0. ,14 0.14 ((7.14) 7 ,. 1 4 ) --0.01 0 ,.01 ((-.14) -. .14) -0.01 1.67 Canada 0, 0.05 .05 ((3.71) 3. 71) 0, 0.04 .04 ( 0 .. 5 8 ) (0.58) 0.10 1.90 0. . 7 0 * 0.70* ( 2 .. 0 7 ) 0.23 2.32 Switzerland1 -0.13 - 0 . .13 (-0.,79) Note: Equation is estimated using instrumental variables. Instruments include lagged MAD0 and lagged levels of interbank interest rates. Overnight interest rates are those described in Table A.l. Three month money market rates are the three-month Treasury bill rate (the United States and Canada), Gensaki rate (Japan), three-month interbank loan rate (Germany, Switzerland) and the three-month Sterling interbank deposit rate (the United Kingdom). 1. Sample covers June 1989-December 1991. * Significant at 5 percent level. A.3. I n t e r e s t Rate V o l a t i l i t y and t h e Transmission of Changes in Federal Funds Rate O b j e c t i v e s : ARt = C + (Bl + B2 MAD°t_i)Afft 1988-1991 + ji t B, DW Response of three-month Bill Rates (ARt) Day of Federal Fund Objective Change - 0 , .02 ( - 1 , .51) 0.22** (4.03) 0.06 (0.22) 51 Five Days Following Federal Fund Objective Change - 0 , .38 ( - 1 . .39) 0.26* (2.42) 0.58 (1.31) .40 1.86 7* i i * Significant at the 5 percent level ** Significant at the 1 percent level 2.25 COMMENTS Stephen A. Meyer1 The goal of these two papers is to develop an understanding of monetary pplicy operating procedures in countries other than the U.S. in the hope that we can learn something applicable to U.S. monetary policy. Potentially the most useful insights in these two papers come from examining operating procedures in those countries, such as Switzerland and the U.K., that have had low, non-binding reserve requirements. In those countries, low or nonexistent reserve requirements combine with tight restrictions on daylight overdrafts to create a situation in which required reserve balances are lower than the reserve deposits that banks need to hold to clear payments through the central bank. is moving toward such a regime. The U.S. Now that reserve requirements on transactions deposits have been lowered, we are likely to find ourselves in that situation during the early months of each year. As Anne-Marie Meulendyke's paper for this conference notes, those responsible for implementing monetary policy in the U.S. are concerned that reserve requirements now are low enough that banks' reserve deposits sometimes will be lower than the operating balances they need to clear payments. Under current operating procedures, the Desk attempts to make the supply of non-borrowed reserves roughly equal to the forecasted demand each day. If the forecast is wrong, especially near the end of a reservemaintenance period, the federal funds rate will deviate from its target. Non-binding reserve requirements, because they give rise 1. Vice President and Associate Director of Research, Federal Reserve Bank of Philadelphia, and Adjunct Professor of Finance, The Wharton School, University of Pennsylvania. Meyer to larger errors in forecasting the demand for reserves, will generate more variability in the funds rate unless U.S. operating procedures are changed to compensate. Most economists, including those in the Federal Reserve System, are not convinced that more variability in the funds rate would be harmful. But U.S. policymakers have revealed an aversion to funds rate volatility, at least in part out of concern that greater funds rate variability will reduce the Fed's ability to communicate the stance of monetary policy to the markets. Policymakers abroad share that aversion. Given policymakers' aversion to interest rate volatility, we would do well to learn whether other countries with low reserve requirements have designed operating procedures that yield less day-to-day interest rate volatility than does our current procedure. Or perhaps we can learn how to construct operating procedures that clearly communicate policymakers' intent to the markets despite volatility in short-term rates. THE PAPERS The papers by Kasman and by Morton and Wood treat operating procedures in major industrial countries, so they do overlap. Nonetheless, the papers neatly complement one another. Reading the two papers together, we learn about the evolution of operating procedures over time, including how central banks have adapted to changing financial conditions, and also about current practices for the day-to-day implementation of monetary policy. Both papers clearly indicate that operating procedures converged to a large extent during the 1980s, with central banks of all the countries examined now using a short-term interest rate as their operating instrument. Nonetheless, some important differences remain. From Bruce Kasman's paper we learn about differences in the day-to-day operating procedures used to smooth short-term interest -2- Meyer rates. Central banks rely mostly on open market operations in one or more financial assets, but non-market techniques such as direct lending to commercial banks (discount window lending, in U.S. parlance) or shifting government deposits between the central bank and commercial banks are used, too. From the paper by John Morton and Paul Wood we learn how the institutional context in which dayto-day operations take place has changed over time, and how monetary policy procedures have changed in response. While we learn a great deal from the papers, we learn less that 1 would like about the roles that the European Monetary System and international integration of money markets have played in the development of operating procedures. I suspect they have been important, but the papers do not tell us how important. As the authors note, interest rate operating procedures are appropriate under some conditions. They were adopted or revived during the 1980s by all seven of the countries examined in these two papers as the seemingly robust empirical relationship between money, interest rates, and economic activity appeared to break down under the pressure of continuing deregulation of financial firms, spreading financial innovations, increasing international mobility of financial capital, and declining costs for financial transactions. The common movement toward interest-rate operating procedures was also driven, in part, by policymakers' perceptions that reserves or money supply targeting allowed too much interest rate volatility. I do not have a comparative advantage in knowledge of the details of other countries operating procedures. Furthermore, reading some relevant literature and discussing the topic with a few participants in foreign financial markets reveals that the authors have done a generally good job of laying out those details. Thus I want to step back from the details and try to put the information presented in these two papers into perspective by -3- Meyer discussing, at least in a general way, some analytic issues. I hope to focus the material in the papers on the search for alternative operating procedures before we turn to the lessons that we might draw from the experience of countries with nonbinding reserve requirements. INTEREST RATE OPERATING PROCEDURES I turn first to a brief discussion of the objectives of interest rate operating procedures, and then to a broad-brush characterization of different ways to structure such procedures. I will focus on the implications of those structures for short-run variability in interbank interest rates. That will lead to some comments on the statistics presented in the two papers. Finally, I will offer comments on what we might adapt for use in the U.S. The Objectives of Interest-Rate Operating Procedures At the macroeconomic level, the objective of any monetary policy operating procedure is to achieve policymakers' desired outcomes for real GDP growth, inflation, or other macroeconomic variables. At the tactical level, we can identify at least four not-always compatible objectives of day-to-day operating procedures: (1) to set the short-run operating instrument at a level believed consistent with policymakers' desired outcomes for intermediate targets such as money growth or the exchange rate, or for final goal variables; (2) to smooth day-to-day variability in shortterm interest rates while nonetheless allowing the level of interest rates to move in response to "goods market" shocks; (3) to convey information about the stance of monetary policy to the markets -- sometimes clearly and sometimes not; and (4) to extract information about economic and financial shocks from the markets. Trade-offs among these objectives condition the design of interest-rate operating procedures. -4- Meyer The Design of Interest-Rate Operating Procedures. As both papers note, an interest-rate operating procedure is, in practice, a set of techniques for managing the supply of bank reserves so as to keep the supply of reserves equal to the demand at the target interest rate. I will focus on market-related procedures. The Simplest Procedure. Perhaps the simplest procedure is one in which the central bank creates a perfectly elastic supply schedule for reserves by posting an interest rate and announcing that it will supply any quantity, of reserves that banks want to obtain at that rate. A central bank could implement such a procedure either by posting a rate at which it will freely lend to banks, or by posting a yield at which it will buy and sell some short-term financial instrument. Clearly this simple procedure allows policymakers to set the operating instrument exactly at its target level. Changes in the target level are immediately observable, so this procedure provides full information about the stance of monetary policy to financial markets. Central banks in some of the countries considered in these two papers used to follow such a procedure, more or less, but no longer do so. The papers indicate that this procedure was dropped because it was perceived to slow the response of interest rates to shocks. That statement must be an argument about policymakers' willingness to change the target level of interest rates or about the need for political cover, rather than an argument that the simple fixed-interest-rate procedure does not extract information about shocks. The observed change in the quantity of reserves that results from a shock under a fixed-interest-rate procedure provides the same information about the economy as would the change in interest rates under a variable-interest-rate operating procedure, so long as we know the -5- Meyer interest elasticity of the supply and demand for reserves. Evidently the simple fixed-rate operating procedure yields too little interest rate variability. More Flexible Procedures. As an alternative, we can construct interest rate operating procedures that generate an inelastic or less-than-perfectly elastic supply of reserves over some range of interest rates, but that also keep rates from moving out of that range. Such procedures would allow shocks to immediately affect interest rates, and would also provide less information about policymakers intentions or targets by allowing some day-to-day interest rate variability. One possibility is a two-tier lending mechanism, as in Germany. The central bank posts one interest rate at which banks can borrow up to a rationed amount, usually not quite enough to satisfy their total demand for reserves at that rate, and a second, higher interest rate at which banks can borrow freely. A second possibility uses a two tier intervention or repurchase rate to produce the same result, as in France. The central bank offers to buy government securities at a yield it chooses, up to some maximum quantity that is less than needed to satisfy banks' total demand for reserves at that yield. The central bank also offers to buy securities at banks' initiative, but at a higher yield. The central bank could either buy very short-term securities outright, or buy longer-term securities through repurchase agreements. Both of these procedures would generate a supply schedule for bank reserves that is a step function. These two procedures would allow shocks to reserve demand to affect the level of interbank interest rates, at least within the range defined by the lower and upper discount or intervention rates. By choosing the spread between upper and lower rates, policymakers can control the -6- Meyer maximum volatility in the targeted rate. By allowing the targeted interest rate to vary somewhat on a day-to-day basis, these procedures can maintain some flexibility or ambiguity about the central banks' exact target. And by announcing changes in the lower and upper rates, policymakers can make clear announcements about the stance of monetary policy when they need to do so. Policymakers might feel that a vertical step in the supply schedule allows too much day-to-day variability in the targeted rate. In another variation on operating procedures, the central bank can manage the supply of reserves to get virtually any positive slope for the portion of the reserve supply schedule between the lower and upper intervention rates. By undertaking short-term repurchase agreements or foreign exchange swaps in response to forecasts of changes in the demand for reserves or to observed variations in interbank interest rates, or by shifting government deposits between commercial banks and the central bank, those responsible for implementing policy can generate an upward sloping supply schedule for bank reserves over a range of interest rates between the lower and upper lending or intervention rates. Of course the upper and lower intervention rates can be used at the margin, rather than to provide the bulk of banks' reserves. The bulk of reserve deposits can be provided through outright purchases of securities, as in the U.S., Canada, and the U.K., or through long-term repurchase agreements as in Germany, or through foreign exchange swaps as in Switzerland. Thus the central bank can manage not only the slope of the upward-sloping portion of the reserve supply schedule but also its position. These mechanisms for providing the bulk of reserves can be biased toward keeping the banking system short of reserves on average, making it likely that interbank rates will trade near the top of the range defined by the upper and lower intervention rates. Or they can be biased toward keeping the banking system flush with reserves on average, -7- Meyer so that interbank rates tend to trade near the bottom of the range. Do Flexible Operating Procedures Meet the Stated Objectives? How does such an operating procedure stack up relative to the goals 1 discussed earlier? Such a mixed strategy can provide a well defined trading range for the targeted interest rate, thus limiting interest rate volatility. The targeted rate will respond to shocks in the demand for reserves, allowing policymakers to observe such shocks. The provision of reserves can be biased so as to keep the targeted rate near any desired level within the range, allowing policymakers not only to hit a specific interest rate target on average, but also giving them flexibility to make adjustments to their target without making explicit announcements of such changes. And changes in the top or bottom of the range can provide clear signals of changes in the stance of monetary policy. Other Countries Use These Mixed Strategies. This mixed strategy of lower and upper discount or intervention rates with an upward sloping supply of reserves between them is a reasonably good characterization of operating procedures used in Germany and France. It is a less good, but still reasonable characterization of operating procedures in Canada and the U.K. In Germany the spread between the lower and upper lending rates is quite large, usually around 200 basis points. In the other countries the range is much narrower. The U.S. Does Not. U.S. operating procedures, in contrast, generate no clearly identifiable minimum or maximum for the target interest rate. There is no minimum -- other than zero -- because discount window borrowing is so sharply restricted by -8- Meyer administrative controls; there is no maximum because the Desk enters the markets at most once each day and there is no "Lombard facility." We can characterize the current U.S. operating procedure as one in which a nearly vertical supply curve for reserves is placed, daily, so that it intersects a forecast of that day's demand for reserves at the target federal funds rate. The reserves supply schedule is "nearly vertical," rather than vertical, for three reasons: (1) the Desk does shade its forecast of the demand for reserves up or down in response to movements in the funds rate; (2) dealers can and do withdraw from repurchase agreements with the Fed when market yields fall; and (3) discount window borrowing still responds a little to changes in the spread between the fed funds rate and the discount rate. With this nearly-vertical supply schedule, we sometimes see very large daily movements in the funds rate, particularly on the last day of reserve maintenance periods. COMPARING OPERATING PROCEDURES AND INTEREST-RATE VOLATILITY This discussion of operating procedures, along with the earlier discussion of the effects of non-binding reserve requirements, might lead us to expect that interbank interest rates in the U.S. would be more variable than those in other countries, except perhaps Switzerland and the U.K. That conclusion turns out to be half right, as the papers by Kasman and by Morton and Wood indicate. Interbank interest rates do seem to deviate more from their targets in Switzerland and the U.K. than in the U.S., but there is no apparent difference in the variability of U.S. interbank rates and those of the remaining countries. Three cautionary notes on the interpretation of the statistics on interest rate variability presented in the papers are in order. First, as Kasman argues, we do not want to confuse changes in interest rate targets with the variability in interest -9- Meyer rates that occurs around unchanged targets. Presumably it is only the latter kind of variability that policymakers find disturbing. As the charts presented in the two papers show, changes in target rates were quite frequent in some countries. For that reason I find it difficult to interpret the statistics presented by Morton and Wood; the average of standard deviations of daily interest rate changes around monthly means, for example, would correspond to unintended interest rate variability only if policymakers never changed interest rate targets except at the turn of the month. Second, as Kasman himself notes, his statistics for Switzerland and Japan are based on a less accurate identification of the central banks' target rates than is the case for other countries. Thus we should not be all that confident that the variability reported for Switzerland gives us an accurate measure of the effects of non-binding reserve requirements. Third, I suspect that the observed differences in interest rate variability reflect not only differences in operating procedures but also differences in policymakers' aversion to interest rate variability. Thus I am reluctant to draw strong conclusions about the effects of operating procedures on interest rate variability from the statistics presented in these two papers without knowing more about how much variability each country's policymakers find acceptable. WHAT MIGHT THE U.S. ADAPT FROM OTHER'S OPERATING PROCEDURES? I will conclude with a possibly provocative suggestion on what the Federal Reserve might adapt from other countries' operating procedures. I offer this suggestion in the hope that it will stimulate discussion and lead to wide-ranging consideration of alternatives. I should make clear that I have not worked out all necessary details, nor, I am sure, have I thought of all potential problems. -10- Meyer A Suggestion for U.S. Monetary Policy Operating Procedures My suggestion is that the Federal Reserve augment its current operating procedures by setting up either a second discount rate - a penalty discount rate modeled loosely on German practice --or by setting up a penalty-rate repurchase facility modeled loosely on French practice. Adding either of these could serve as a first step in modifying U.S. operating procedures. To set up a two tier discount rate, the Federal Reserve would establish the equivalent of a "Lombard rate" --an additional, higher discount rate at which banks could borrow freely against eligible collateral. That second rate would be higher than the target federal funds rate and would exist alongside the current subsidy discount rate. Banks would be able to borrow at the lower, subsidy discount rate only in the event of truly unforseen reserve shortfalls due to events such as computer failures or wire transfer delays. But banks would have ready access to borrowing at the higher "Lombard rate." To set up a penalty-rate repurchase facility, the Federal Reserve could announce that it stands ready to provide reserves to banks through short-term repurchase agreements arranged at banks' initiative, but at a rate that would be set above the target federal funds rate. Such a facility would require that banks have appropriate collateral; it also would require the Federal Reserve to put in place safeguards to limit counterparty risk. Potential benefits of Modifying U.S. Operating Procedures. Either of these facilities would provide a backup source of liquidity to the banking system when reserve deposits plus clearing balances fall short of balances needed for funds transfer purposes, or when the Desk underestimates the demand for reserves. Either facility could prevent spikes in the federal funds rate such as we have seen on some end-of-maintenance-period Wednesdays. -11- Meyer By establishing clear, although perhaps broad, limits on movements in the funds rate, either of these facilities might reduce policymakers' concerns that day-to-day movements in the federal funds rate could be misinterpreted as a change in monetary policy. That is particularly likely if market participants knew that significant changes in the stance of monetary policy would be signaled by changes in the penalty discount rate or repurchase rate, and perhaps also by changes in the subsidy discount rate. Seemingly paradoxically, a change in operating procedures that would prevent large movements in the federal funds rate could actually allow more day-to-day volatility by reducing markets' reliance on changes in the funds rate as an indicator of the stance of monetary policy. In addition, the existence of a liquidity safety valve would reduce the Desk's need to match the supply of reserves to the predicted demand each day. The Desk might well be more able to focus on the reserve need for the maintenance period as a whole, and thus be free to conduct fewer daily open market operations aimed at smoothing the funds rate. Finally, to the extent that the proposed changes allow greater day-to-day variability in the funds rate, they will enable market forces to move the average level of the funds rate more readily than is the case today. Those movements, in turn, might allow policymakers greater flexibility in making a series of small changes in their federal funds rate target, at least within the band defined by the subsidy discount rate and the penalty rate. -12- MONETARY TRANSMISSION CHANNELS IN MAJOR FOREIGN INDUSTRIAL COUNTRIES Robert B. Kahn and Linda S. Kole The past two decades have witnessed far reaching transformations of financial markets in major foreign industrial countries. The process of financial liberalization that has taken place abroad in many ways parallels changes that have occurred in the United States. Although the specific circumstances of individual countries have differed, there have been important common elements, including the introduction of new financial assets and markets, fuller integration of domestic and international financial markets, greater reliance on market-determined interest rates, and significant structural change in banking systems. The changes in financial markets that occurred in most of the major industrial countries in part reflected the response to a common set of global economic forces. These changes, in turn, had an impact on the conduct of monetary policy and how monetary policy actions fed through to the real economy. This paper characterizes the main financial channels through which monetary policy affects real economic activity in major industrial countries, and analyzes whether and how these transmission channels have changed during the past two decades. Because of the broad nature of the question, we have chosen to limit our country coverage. Our primary focus in this paper is on Japan, Germany, and the United Kingdom, three countries that have had a wide range of diverse experiences with financial deregulation and monetary control. We find evidence that in all three of these countries, wealth is crucial in the determination of money demand, the first link in the monetary transmission channel. Further the demand for broad money seems to have become more portfolio 1. Linda S. Kole is on the staff of the International Finance Division of the Federal Reserve Board. Robert B. Kahn, formerly a member of the staff of the International Finance Division of the Board, is now on the staff of the International Monetary Fund. We thank Hali Edison, Neil Ericsson, Mike Gavin, Craig Hakkio, Dale Henderson, Karen Johnson, Steve Kamin, Eric Leeper, Eileen Mauskopf, and Larry Promisel for their comments. Peter Fishman, John Maluccio, and Tina Sun provided exemplary research assistance. - 1- Kahn and Kole based and interest sensitive in the past decade. We also find evidence that different monetary variables affect real economic activity differently across countries, but that the role of the term structure and the exchange rate seems to be increasing in the transmission channel. The next section outlines reasons why monetary policy transmission mechanisms have changed in the past two decades. The third section then traces out a basic Keynesian model to illustrate the importance of asset prices in the monetary transmission process. Section 4 analyzes the relationship between money, interest rates, income, and wealth by estimating money demand functions and considering how they have changed over time. Section 5 then examines the links between various instruments of monetary policy and real variables. Finally, in the last section we summarize our results and explore possible extensions for future work. MONETARY POLICY TRANSMISSION CHANNELS, 1970-1990 The reasons why monetary transmission mechanisms may have changed in the past two decades are well known. They include: Financial liberalization and innovation. In the early 1970s, many foreign industrial countries had relatively underdeveloped financial systems that limited the channels through which monetary policy influenced economic activity. Subsequent changes in the transmission of monetary policy in these countries generally reflected governments' attempts to modernize and integrate their financial markets in a changing world economic environment. In Japan, financial markets were tightly controlled and segmented in the early 1970s. Major bank deposit rates were regulated while lending rates were closely linked to the Bank of Japan's discount rate. Bond markets were small and underdeveloped, and international transactions tightly controlled. Thus, credit was effectively rationed, and bank lending was the main channel through which monetary policy influenced the economy. Japan liberalized its capital markets later than many other industrialized countries. 1980s, deregulation gradually gained momentum. In the early Controls on domestic interest rates and on external capital flows were dismantled, and new financial products proliferated. There was also substantial liberalization in the United Kingdom over the course of the 1970s and 1980s. - 2 - Domestic financial markets were Kahn and Kole quite segmented (e.g. between the clearing banks and the building societies) in the early 1970s, restricting competition for both loans and deposits. In the 1970s, monetary policy mainly operated through restrictive guidance of bank and building society lending. Between 1973 and 1979, the "corset", in essence a tax on the expansion of U.K. banks' liabilities (and thus credit) was intermittently activated, although eventually it became ineffectual. When the Conservatives came to power in 1979, the new government recognized that the controls were increasingly ineffective and initiated major financial market reforms. Most of the restrictions on lending and exchange controls were abolished in 1979. The corset was scrapped in 1980, and reserve requirements were abolished in 1981. A series of other regulatory changes followed that culminated in the collapse of the building societies' cartel in 1983 and the Big Bang in 1986. Deregulation progressively put banks and building societies on a more even footing, broke down the separation of the mortgage market from other forms of personal credit, and freed consumer credit. In contrast, German domestic financial market liberalization was already advanced by 1970, even though markets remain cartelized and underdeveloped. All interest rate controls were abolished in 1967. Exchange controls were removed in the 1950s. Such an environment might be expected to be one with particularly strong international linkages, and international financial innovation has made a mark on German capital markets, although relatively few new financial instruments have been introduced there. One special aspect of German financial markets has been their proximity to Luxembourg, a financial center that is considerably more developed than German markets. During the past two decades, Luxembourg has often served as an outlet for German portfolio shifts, providing Germans with the full array of instruments (offered by German banks) that their own market lacks. Without Luxembourg, German financial innovation would undoubtedly have proceeded at a faster pace. With the approach of a fully integrated European financial market, the process should accelerate. It is often noted that one factor influencing liberalization of financial markets during the 1970s and 1980s was the sharp increase in industrial countries' interest rates and inflation rates after 1973. - 3 - Kahn and Kole Chart 1 depicts short-term interest rates and consumer price inflation from 1973 to the present in major foreign industrial economies. The increase in nominal and real interest rates created incentives for asset holders to reduce their balances bearing below-market interest rates. At the same time, the growth in government deficits in a number of countries contributed to the deepening of debt markets. Japan, France, and Sweden are countries where changes in heavily regulated bank systems were stimulated, by the development of the bond market under the pressure of 2 rising government budget deficits. Clearly other factors were also at work. Structural change in the banking system has contributed to a changing role for banks. And technological change, particularly in the advance of computer technology, altered the environment of financial markets, decreasing costs and increasing the speed with which financial transactions were transmitted between markets. Most of the countries we examined responded to these changes by shifting monetary policy away from the use of credit constraints toward a greater reliance on market-determined interest rates. With the rapid development of financial markets, monetary aggregates became less stable and less subject to control by the monetary authorities. In general governments began to deemphasize monetary aggregates, by changing the method of targeting while paying more attention to short-term interest rates and their impact on the economy. For example, as the velocity of the broad aggregate £M3 became more unstable, the response of the U.K. government was to shift from targeting £M3 towards a narrower aggregate, MO. In the second half of the 1980s, British monetary policy became more interest-rate oriented, and at times, was used to pursue exchange rate targets as well. In contrast, when German CBM became unstable and its targets were overshot in part as a result of volatility in currency flows, the Bundesbank shifted its focus to M3, an aggregate that puts less weight on currency. The Bundesbank was one of the first central banks to adopt (both informally and formally) monetary targets after the breakdown of the Bretton Woods system. Monetary policy relied heavily on interest 2. For a comprehensive discussion of these issues, see Germany and Morton (1985). - 4- Kahn and Kole rates, supplemented by window guidance. Despite the uncertainties surrounding the unification of western and eastern Germany, the Bundesbank continues to put more weight on targeting broad money than do central banks in most other major industrialized countries. Although the last two decades saw a gradual move towards more active use of short-term interest rates as instruments of monetary policy, evidence is generally inconclusive as to whether foreign authorities have acquired greater influence over long-term interest rates. Some have argued that financial innovation implies that short- term and long-term rates move more closely together than previously. However, as both government and corporate bond markets deepened, the ability of monetary authorities to influence long-term rates may have been reduced. Because long-term rates are more important in the determination of real economic variables, the nature of the relationship between monetary policy and the term structure is crucial. Financial liberalization has had conflicting effects on the interest sensitivity of aggregate demand. In Europe and Japan, there have been significant structural changes that reduced liquidity constraints on consumers and rationing of loans by banks. To the extent that there is now less disintermediation from the banking system associated with a monetary contraction, interest rate changes will be measured to have a smaller impact on economic activity than previously. On the other hand, as discussed further below, individuals now carry more debt and their cash flow is more vulnerable to interest rate changes. In addition, households now have more assets whose return can be squeezed. In fact, one of the general conclusions that we suggest below is that wealth channels for the transmission of monetary policy now matter more. Greater international openness and integration. Liberalization of domestic financial markets had its counterpart in greater economic integration of trade and financial relationships across countries. Part of this transformation reflected the continued liberalization of trade on a multilateral basis. Table 1 illustrates that, throughout the industrial countries, merchandise trade as a share of GDP/GNP rose between 1971-75 and 1987-91. Indexed by this measure, Germany, the United Kingdom, Canada and France appear to have been quite open by the late 1980s. In contrast, the external sector in Japan remains a smaller - 5 - Kahn and Kole part of aggregate demand than is widely realized. Exchange rate considerations have figured prominently in Germany in part because of the relatively large size of the export sector (and-because the Bundesbank has a statutory mandate to "safeguard the currency"). The greater openness of trade flows reflected a revolution in technology, which contributed to greater international integration of financial systems. Greater openness and financial market integration meant that governments increasingly had to respond to global economic disturbances. Spillover effects of one country's change in monetary policy could no longer be ignored. The growth of the tradable goods sector of these economies supports the common contention that monetary policy now has a greater impact on economic activity through the exchange rate channel than previously was the case. The usual argument is that a monetary-policy induced change in interest rates at home causes nominal interest differentials to change, leading to movements in nominal exchange rates. With the growth in the size of tradable goods sectors, movements in exchange rates will then have greater effects on aggregate economic activity. Mauskopf (1990) finds some evidence for an increase in the influence of the exchange rate on the volume of U.S. exports and imports using the Board's MPS model. However, there are a number of reasons why we may not be able to observe this change clearly in the data. First, there is some evidence that long-term rates in the major industrial economies have moved more closely together in recent years than in the past. Whether this reflects greater financial integration or governments' policy responses to similar economic disturbances, the implication is that an interest rate innovation in one country may on average be associated with smaller movements in interest rate differentials and exchange rates than in the past. Further, changes in the passthrough of exchange rates to import prices (e.g., Hooper and Mann, 1989) will also affect the impact of monetary policy monetary policy through this channel. The large movements in exchange rates of the major foreign industrial countries (especially against the dollar) during the 1980s also affected the monetary policy reaction function in some countries. The most obvious example was the creation of the EMS in 1979 and the gradual convergence of interest rates in member countries during the - 6- Kahn and Kole 1980s, which for all practical purposes devoted the conduct of monetary policy in France, Belgium, and the Netherlands to the goal of exchange rate stability against the mark. The progressive decline in the frequency of realignments in the EMS along with the move towards fuller EMU gradually enhanced the ERM's credibility and to some extent has made 3 monetary targeting in separate nations obsolete. Germany, arguably the most zealous G7 nation in terms of monetary targeting, occasionally subordinated its monetary objectives to external goals. For instance, after meeting its CBM targets each year between 1979 and 1985, the Bundesbank significantly overshot its target in 1986, when the mark appreciated 35 percent against the dollar and the current account reached 4 a record level. Japan also tolerated faster money growth at times due to exchange rate considerations. have been inappropriate. Occasionally, exchange rate targets For example, it is by now well recognized that the British experiment with shadowing the ERM at a rate of DM3 between 1987 and 1988 led to overly expansionary monetary growth, a boom, and excessive inflation. The rise in consumer and business indebtedness. During the 1980s, the deregulation of financial markets and the provision of new financial instruments in many industrialized countries allowed both households and firms to accumulate more debt than ever before. The removal of liquidity constraints for less well-off households and small firms and the wider selection of financing methods available to the private sector changed monetary transmission mechanisms in the following ways. Lower liquidity constraints allowed more consumers to smooth consumption and enabled them to react more to changes in permanent rather than current 3. It is interesting to note however that most of the ERM countries continue to target some aggregate, despite the essentially fixed nature of their exchange rates. One justification for the seeming inconsistency between exchange rate and monetary targets is that the anchor of the system, once the mark, has become less clear as Germany has had to adjust to the inflationary pressures of unification. 4. Note that domestic considerations were not in conflict with a monetary expansion, as consumer price inflation was negative for the first time since the 1950s. 5. In some countries a change in tax incentives contributed to the buildup of debt as well. - 7 - Kahn and Kole income. Deregulation lowered the cost of financial intermediation, particularly for households, causing a shift towards borrowing. The larger proportion of households in debt meant that the private sector's sensitivity to interest rates might increase, thereby making monetary policy more potent. For example, empirical work by Dicks (1990) found that the interest elasticity of U.K. consumption increased during the 1980s, increasing the leverage of monetary policy. Table 2 shows that the increase in debt of the household sector since the 1970s has been most dramatic in Japan and the United Kingdom. In both these countries, it could be argued that the conjunction of easy monetary policy and the rapid accumulation of mortgage debt fueled a real estate boom in the latter part of the 1980s. While land and equity prices were rising, the increase in consumers' liabilities seemed to be matched by the increase in their gross wealth. However, when tighter monetary policy finally burst the asset price bubbles in late 1989, both firms and households were left with an excessive stock of outstanding debt. After two years of recession, U.K. spending is still quite weak as firms and households continue to adjust their balance sheets. Debt overhang may have made U.K. consumers less responsive to monetary easing than previously was the case. For similar reasons growth in Japanese spending is now expected to be sluggish for some time. Note that these episodes have meant that asset prices, including those of tangible assets such as land, may be playing more of a role in the monetary transmission mechanism than previously. The increase in household indebtedness has been more moderate in continental European countries such as Germany and France. The comparatively lower increase in the debt burden may mean that interest rate sensitivity has not changed as much in these countries. Also these countries did not experience the rapid asset price inflation seen in Japan and the United Kingdom in the late 1980s, so monetary policy may have worked less through the asset price channel. 6. Bayoumi (1990,199?) empirically documented this in his work on financial deregulation and consumption in the United Kingdom. He estimated that during the 1980s, the U.K. personal savings rate fell by 2-1/4 percent as a result of financial deregulation. - 8- Kahn and Kole The cycle of debt growth and asset price inflation may have changed monetary reaction functions as well. In both the United Kingdom and Japan, the asset price inflation preceded periods of general consumer price inflation, giving monetary authorities an early warning that 7 tightening would be a good idea. The responsiveness of firms to interest rates also was altered by the rise in business indebtedness, which again has been greatest in Japan and the United Kingdom as shown in the middle panel of table 2. Financial market deregulation expanded the access to credit and lowered the cost of borrowing for non-financial firms. With the development of corporate bond markets and innovations such as securitization, businesses have resorted increasingly to non-bank financing. However, as discussed with respect to the United States (in Bernanke and Blinder, for example) there are good reasons to believe that bank credit remains an important channel for monetary policy. The switch to a less active stabilization policy in a lower inflation environment. In the 1980s, there appears to have been increased recognition on the part of foreign industrial governments that there are limits to what can be achieved through countercyclical fiscal policy and that monetary policy should be devoted more seriously to achieving price stability. As shown in the bottom panel of table 2, public sector debt as a share of GDP leveled off or declined in a number of countries between 1985 and 1990, although it probably has risen recently. Further, to some extent the drop in inflation and inflation expectations throughout the industrial world during the course of the 1980s can be credited to effective monetary policy. Lower inflation should have contributed to a decline in the velocity of various monetary aggregates and increased economic efficiency. A CONVENTIONAL KEYNESIAN MODEL WITH WEALTH AND EXCHANGE RATE DYNAMICS In light of our hypothesis that asset prices and wealth have become more important in the transmission of monetary policy, this section develops a 7. The Bank of Japan explicitly tightened policy in 1989 to quell asset price inflation. It appears that the U.K. government did not adjust monetary policy quickly enough in response to the asset price signal, leading to the boom and subsequent bust that occurred. - 9- Kahn and Kole model of the open economy in which wealth plays a key role along with income in the determination of aggregate demand and prices. The model is Keynesian; the price level is sticky and output is demand determined in the short run. Wealth, modelled for simplicity as the value of the stock market, affects output directly through its effect on aggregate demand and is also a factor in determining money demand. Output, through both profitability and interest rate channels, in turn is a key determinant of asset prices. The model is a variant on Gavin (1989) who built on the closed economy analysis of Blanchard (1981). Table 3 presents the equations of the model. The model is formulated so that all parameters in the structural equations are positive and all variables other than interest rates are in logs. In equations 1-2, real aggregate demand, D, depends on the real value of the stock market, q, real output, Y, the real exchange rate, E, and fiscal policy, G. Output then, adjusts to the difference between aggregate demand and output, where the coefficient d is a measure of the speed of g adjustment of Y to D. Equation 3 can be interpreted as a liquidity formulation for money demand, solved for the nominal interest rate i, where the money supply is exogenously given. wealth affect money demand. Money, income, prices, and Alternatively, this equation could be thought of as a policy reaction function of the monetary authority. In this case, a positive coefficient on wealth would suggest that monetary authorities respond to asset price inflation by tightening policy. For simplicity, equation 4 defines the real interest rate as the nominal rate less the rate of change of home goods' prices; the real exchange rate does not enter. We also have assumed that perfect foresight prevails. The dynamics of the price process are given in equations 5. The price adjustment parameter is simplistic, and abstracts from Phillips curve or staggered wage contracting considerations, but still allows for significant interactions in the results described below. The steady- state price level (equation 6) equilibrates the money market at the steady-state interest rate and the equilibrium level of output. Asset market equilibrium is given in equations 7-9. In equation 7, the expected real return on a share of the stock market consists of 8. One could allow d to equal 1. equilibrium value. in that case, Y jumps to its new - 10 - Kahn and Kole both capital gains, q/q, and the rate of profit, Z/q. equal the real return on domestic bonds. parity condition, where r exogenously given. This return must Equation 8 is the open interest is the foreign real interest rate and is Equation 9 specifies real profits as a function of real output and equation 10 gives the relationship between the long-run real interest rate and the short-run real interest rate. Equations 1-9 can be solved for the steady state values of Y, q, E, and P, and the two negative roots that describe the adjustment of the economy after a shock. Chart 2 plots possible dynamic paths of these variables after a monetary shock occurs. Consider an unanticipated monetary expansion at time 0, with the economy initially at a steady state. The new steady state is identical to the old, except that the price level is increased by the amount of the monetary injection. As shown in chart 2, the simple price adjustment mechanism in this model ensures that prices adjust monotonically to their new equilibrium level. The dynamic responses of output, interest rates, and the real exchange rate are more complicated. Output is "humpbacked", rising from the initial.steady state to a point and then falling toward the unchanged steady state. Stock market wealth jumps up at the time of the monetary expansion to equilibrate expected stock market returns with returns in the bond markets, then gradually falls as interest rates and income adjust to the shock. In the left middle panel, the normal pattern of short- and long-term interest rates following the monetary expansion is shown. Interest rates initially fall following the monetary shock, then rise as output expands. Rates eventually rise above their baseline level before beginning to fall again as output returns to baseline. However, the introduction of stock market prices into the model creates the possibility that long-term interest rates could rise following the monetary expansion. Short-term interest rates always fall on impact of the monetary impulse, but rise as activity responds. If the monetary expansion has a strong effect on the stock market, and the increase in wealth in turn has a strong effect on activity, short term rates may respond quickly and soon rise above baseline. In this case, the long- term interest rate as the weighted average of current and future shortterm rates might rise immediately, implying a stronger movement in the term structure. - 11 - Kahn and Kole Chart 2 also shows possible paths for the real exchange rate. The left and center bottom panels show the usual depreciation of the real exchange rate following a monetary expansion. As discussed by Gavin, the possibility exists in this model for the exchange rate to exhibit perverse behavior and appreciate following the monetary expansion. The logic is similar to the reason that long-term rates can rise following the innovation to money: If the stock market response is strong enough, the actual and expected increase in interest rates causes the exchange rate to appreciate. This model highlights the importance of wealth in the monetary transmission channel. MONEY DEMAND IN JAPAN, GERMANY, AND THE UNITED KINGDOM Although wealth plays a crucial role in the monetary transmission process, most of the vast literature on money demand in industrialized countries excludes this variable from the analysis. Especially when one considers the demand for broad aggregates, that are less oriented towards making transactions and have more of a role in portfolio management, the omission of some measure of wealth is likely to seriously bias one's results. M2+CDs, the aggregate most closely monitored by the Bank of Japan, and M3, the aggregate targeted by the Bundesbank, are both likely to be determined primarily by portfolio flows rather than fluctuations in the transactions demand for money. The same is true of British M4, which we chose to analyze here because of its closeness in definition to £M3, 9 the aggregate targeted over most of the period. Corker (1989) noted the importance of including Japanese wealth in order to achieve a stable equation for the evolution of real M2+CDs and Hall, Henry, and Wilcox (1992) found personal sector wealth to be significant in the long-run determination of U.K. M4. In the spirit of these studies, we estimated money demand functions for broad monetary aggregates that allow wealth to play a role, both in short-run fluctuations of money demand and in a long-run (error-correction mechanism) context. 9. We chose not to model British MO, the aggregate targeted by the British government for the past few years. This aggregate, which includes notes, coin,, and banks' operational deposits with the Bank of England, is extremely narrow and more of a coincident indicator of economic activity than a indicator of the government's monetary policy. - 12 - Kahn and Kole As we mentioned above, Japanese financial markets were deregulated later than those in Germany and the United Kingdom, and the proce'ss of removing interest rate regulations is still ongoing. As a consequence, among the three countries considered, Japan is the most likely candidate for instability in the demand for money. This is indeed what we find. In part, the problem in estimating a stable money demand function for the entire period between 1973 and 1991 is that different interest rates were probably relevant at different periods. the relevant interest rates: The top panel of Chart 3 shows the Gensaki rate was the most market- related rate in the 1970s whereas the own rate was very sticky and unrelated to market interest rates. Both the own rate and the postal savings rate became more, important over the course of the sample, while the Gensaki rate became progressively less significant. The chart also shows that the opportunity cost of M2+CDs, the gap between either outside rate and the own rate continued to narrow until quite recently. Using quarterly data, we regressed the change in the log of M2+CDs deflated by the log GNP deflator, A(m-p), the change in log GNP (Ay), the rate of GNP price inflation (Ap), the growth in the log of real wealth, A(w-p), and the change in various interest rate measures: an own rate, i , the postal savings rates i , the Gensaki rate, and the long rate, i as well as the error correction terms. We started with a specification that included 4 lags of each of the first-differenced variables, and gradually eliminated those variables that were not significant at the 10 percent level. The wealth variable used was an updated version of that used by Corker (1989), the total stock of assets held by the personal and corporate sector. We originally started with error correction mechanisms similar to Corker's, (m-w) -and (m-p-y) -, but found we were almost able to reject the implied restrictions, so we allowed the long run relationship between real money, real wealth and real output to be estimated freely. 10. We followed Corker (1989) in the construction of the own rate on M2+CDs as a weighted average of the interest rate on 3-month CDs and the average rate on savings deposits. 11. Unfortunately, this measure of wealth only includes financial wealth, not tangible wealth such as real estate. - 13 - Kahn and Kole The first column of table 4 presents the results. Note that the wealth variable, w-p, is quite significant, both in terms of its rate of change and in its lagged level. The rate of inflation, Ap, and its first lag, are strongly significant, and we were able to reject the hypothesis that the relevant demand is for nominal rather than real money. The rate of change of GDP, Ay, and its lags, did not add to the explanatory power of the equation, but y ., is significant. Both the postal savings rate, a rate on assets outside M2+CDs, and the own rate, a weighted-average rate on assets inside M2+CDs, have estimated coefficients that are significant and of the correct signs. Assuming price stability, the long-run money demand function implied by the error correction variables is: m-p - .43y +.48(w-p) + 6.28i° - 5.01iPS We then estimated the equation over two subperiods; 1973-1982 and 1983-1991. We broke the sample after 1982, before financial deregulation and innovation really started to take off in Japan.12 The Fisher test indicates that one can reject the hypothesis that the demand for M2+CDs remained stable between these two periods at the 1 percent level of significance. Note that the dynamics have changed considerably between the two periods, and that during the first period this specification leads the errors to be serially correlated. There is some evidence that the lags were longer in the first period than in the second, perhaps indicating that money holders have become quicker to adjust their balances in response to changes in wealth, interest rates, and other explanatory variables. Turning to the estimated coefficients for the two subperiods, several points are worth noting. First, both the short-run and long-run interest rate elasticities are estimated with the wrong sign during the 12. Kasmin and Rodriques (1992) chose to break their sample in 1984. 13. We were able to find a stable money demand function between the 2 periods when we left many lags in the equation. These lags were more significant during the first period. Their addition however, did not improve the fit of the equation in the overall period, so they were discarded. - 14 - Kahn and Kole earlier period; with the change in the own rate significant, whereas in the later period all the estimated interest rate elasticities have the correct sign. Second, although the lagged wealth variable is significant in both periods, the long-run coefficients indicate that the proportion of an increase in wealth allocated to M2+CDs fell from 60 percent in the early period to about 50 percent in the later period, which seems consistent with the wider diversity of assets available outside the aggregate. Finally the higher significance of the lagged level of income in the earlier period is evidence that the transactions motive for holding money has become less important over time. In contrast to the Japanese case, due to the lack of financial deregulation and the consistency of monetary policy, German money demand before 1990 was, a priori, the most likely to be stable among the countries we considered. The middle panel of chart 4 shows the interest rates relevant for German money demand. It is immediately apparent that they move together during the entire sample. Unlike Japanese savings instruments, German time deposits have had market-related interest rates for some time, reducing the incentive for financial innovations such as those which change the nature of the relationship between Japanese interest rates during the sample period. We used the same estimation technique employed for Japanese money demand to determine the proper lag lengths of the included variables. Table 5 presents the results for the overall period and two subperiods: 1970-79 and 1980-89. the EMS. The split in the sample is close to the creation of We ended the sample in the fourth quarter of 1989 because of the possibility that the fall of the Berlin wall induced instability in the demand for M3. The rate on public bonds was found to be more important than short-term rates as an opportunity cost of M3. This may be due to the fact that M3 contains assets of maturities of up to 4 years, so that the marginal investor might substitute into a longer-term asset such as public bonds. Although some studies of German money demand (see for example, von Hagen and Neumann (1988)) have included foreign interest rates and/or the exchange rate, (spot and forward), Frowen and Schlomann (1992) found them not to be very important for M3. The rate on public bonds is more important in the 1980s than it is in the 1970s, as is the - 15 - Kahn and Kole own rate. Note that the coefficients on own rate and the public bonds rate are of equal and opposite sign in the later sample, indicating that an opportunity cost specification using the interest rate differential may be best for the period of the 1980s. The wealth variable for Germany was constructed from flow-of-funds data of the household sector. Since data on households' net assets are available only on a biannual basis, we interpolated resulting in a smooth version of quarterly wealth. (This measurement error associated with interpolation and constructing stock data from flow data may account for the lower significance of wealth in German money demand, at least in comparison to Japan and the United Kingdom.) However, its significance increases between the two periods, indicating that portfolio management plays an increasing role in the determination of the demand for M3. The long-run solution implied by the error correction coefficients is: m-p - .19y + .54(w-p) + 3.75i° - 5.41iPB Finally, it is evident that the transactions demand for money has become less important as the lagged level of real GDP has become progressively less significant. The fit for the earlier period improves slightly if one adds in the change in GDP. None of these changes are statistically significant, however, as is indicated by the fact that the Fisher test does not reject the hypothesis that the two subsamples can be pooled. As we expected, and as others have found, it does not take much to arrive at a stable demand for German money in the pre-unification period. It is difficult to say much about how the monetary transmission mechanisms have changed since the unification of east and west Germany. The lack of a sufficiently long data set post-unification rules out econometric analysis of the period. Nonetheless, a few preliminary judgments can be ventured. First, German money demand has become less certain following the 1990 currency conversion of Osmarks for Deutchmarks that added eastern German demand, and at times movements in interest rates and money growth have sent conflicting signals regarding the stance - 16 - Kahn and Kole of policy. In this environment, the Bundesbank has recognized that monetary policy changes will be at best a blunt instrument for influencing prices and activity. Second, a case can be made that aggregate demand in Germany currently is less interest sensitive than before unification. Specifically, a significant portion of German domestic demand is associated with the surge in investment, and specifically construction, in the eastern portion of Germany. Much of this investment appears to be limited by capacity constraints, suggesting that changes in interest rates may have a limited effect on these flows in the near term. In addition, a substantial portion of this investment is publicly subsidized, and thus less responsive to monetary policy. Further, to the extent that investment in eastern Germany is subject to substantial systemic risks associated with the region's transition to a market economy, nominal interest rates will be a smaller part of the total perceived cost of investing than in western Germany. Another argument that has been made is that with the completion of the internal market in the European community, barriers to entry into German banking markets have been reduced. New entrants, actual or potential, are expected to enhance competition in these markets and should lead to more competitive pricing of deposits. This may lead to less disintermediation associated with changes in Bundesbank policy, meaning that interest rate changes may have less effect on activity through the bank lending channel than previously. Table 6 presents the results of estimating the demand for real U.K. M4. Here we adjusted the M4 data to account for a break in the fourth quarter of 1981 when the monetary sector replaced the banking sector and many new institutions such as the Trustee Savings Banks were for the first time included in the broad monetary aggregates. We tried a variety of interest rates including long-term rates, local-authority rates, and deposit rates, and found that over the entire sample, the rate on 3-month Treasury bills fit the best. From the bottom panel of chart 14. Some have argued that strong eastern German money demand suggests a combination of a high income elasticity and strong income growth due to large transfer payments from the government. - 17 - Kahn and Kole 4, it appears that the opportunity cost of M4 has not varied substantially over the sample. No interest rates seemed to matter in the- rate of change form, and it was difficult to find specifications with the rate of change of GDP entering the equation significantly and with the correct sign. The change in real wealth (and its lags) also had nothing to add to the explanatory power of the other variables. Wealth is total wealth of the personal sector, including tangible assets. We also tried financial wealth, and it fit, but not as well, especially in the latter period. The long-run money demand function implied by the error-correction terms is: m-p - -.51y + 1.58(w-p) + 5.4i° - 5.84iTB The estimated long-run elasticity of real M4 demand with respect to GDP is about -1/2. The equation can be rewritten in the following way: m.p-y - 1.58(w-p-y) + .07y + 5.4i° - 5.84iTB The inverse velocity of M4 is positively related to the wealth/GDP ratio and negatively related to the opportunity cost of money. We chose as a sample breakpoint the end of exchange controls, and the beginning of the Thatcher government's reorientation of monetary policy from restrictive lending towards more market and monetarist goals. The equation passes the Fisher test for stability between the two periods at the 5 percent level of significance. It seems intuitively plausible that demand for U.K. M4 would be less stable than that for German M3, but more stable than the demand for Japanese M2+CDs. However, the dynamics seem to have changed between the two periods, as do the signs and significance of some of the coefficients. The estimated coefficient on the growth of GDP (two quarters earlier) switches from positive to negative between the two periods. It is interesting to note that the error correction mechanism implies the 15. Note that this result does not disappear if current and once-lagged GDP are included in the equation. The sum of the estimated coefficients still adds up to a negative number for the second period. - 18 - Kahn and Kole opposite change in sign between the two periods. The long-run elasticity of real M4 demand with respect to real GDP changes from an insignificant negative number to a significant positive 1.5. . (Note that the negative long run income elasticity estimated for the full period seems to be coming from the 1970s rather than the 1980s.) The implication is that between the two periods the relationship between transitory changes in income and real M4 changed, but income became more important as a longrun determinant of money demand. As in Japan, the long-run wealth elasticity declines between the two periods, from 1.4 in the 1970s subperiod to .5 in the latter period. It appears that in the earlier period, investors were moving into M4 on balance, while in the later period they were diversifying into assets outside M4. The estimated coefficients for 3-month Treasury bill rate and the own rate decline in significance between the first and second period. This could be a result of other interest rates being more relevant. There is some indication that long-term rates seem to become more important between the two periods. This may be an indication that investors became more sophisticated as financial markets deepened. To summarize our results thus far, we find evidence that wealth is a significant determinant of money demand in Japan, Germany, and the United Kingdom, and may have become more important than in the past. In all three countries, there seems to have been a shift from holding money for transactions purposes to holding it as part of a portfolio between the 1970s and 1980s. We find that the demand for broad money remained fairly stable in both Germany and the United Kingdom during the past two decades. However, in Japan, where financial deregulation was more gradual and is still ongoing, we find more instability. THE RELATIONSHIP BETWEEN MONETARY AND REAL VARIABLES. What is the relationship between monetary variables, be they monetary aggregates, interest rates, or exchange rates, and the real economy? This is an age-old question that has been analyzed most extensively for the United States without resulting in anything close to a consensus. surveyed the evidence for Japan, Germany, and the United Kingdom. Some of the changes in the transmission mechanism discussed above could be - 19 - We Kahn and Kole expected to change the reduced form relationship between monetary variables and various activity variables. For example, if consumers have become more interest sensitive because of lower liquidity constraints, interest rates may be more related to real spending. Estimation of the structural model laid out in section 3 is beyond the scope of this paper, but in this section we take a first pass at analyzing the relationship between monetary and real variables by looking at reduced-form results. Specifically, we ran a battery of regressions to assess the predictive power of monetary variables in the determination of real economic activity, over the 1973-91 period and two subperiods for Japan, Germany, and the United Kingdom.' For the activity variables for which monthly data are available, industrial production (IP), retail sales (RS), the trade balance (TB), and the unemployment rate (U), we included six lags of the relevant variable, a trend, six lags of various price indices to proxy for supply shocks (and to stack the deck against us), and six lags of the three monetary indicators shown as column heading in tables 7-9. The monetary indicators chosen were the major aggregates (in most cases the targeted ones), the 3-month -interbank rate, and the term spread between a long-term rate and the 3-month rate. The latter variable has been shown by Stock and Watson (1989) and others to have remarkable explanatory power for U.S. output. We also included the exchange rate as a monetary indicator, when the activity variable in question was an external balance. Quarterly regressions were run with the following variables: real GNP or GDP, real consumption expenditures (C), real gross fixed investment (I), and real net exports (NX). These regressions included four lags of the activity variable, four lags of the monetary variable (on an average quarterly basis), and four lags of the GNP/GDP deflator, the CPI, the PPI, and the world commodity price index, respectively. Because we were primarily interested in the cumulative impact of a given money or interest rate innovation, we tested whether the addition of the lags of the relevant monetary indicator variable summed to a 16. Our equations for industrial output are similar to those estimated for the United States by Stock and Watson (1989). - 20 - Kahn and Kole quantity significantly different from zero. When we find the sum of the coefficients on a monetary variable to be significant, we can say that that variable helps predict the real variable. Table 7 presents the marginal significance levels for rejecting the hypothesis that the estimated coefficients on monetary indicator variables (in the columns) sum to zero in regressions where the activity variables (in the rows) are the dependent variables. A low marginal significance level means that it is easy to reject the null hypothesis that the 6 lags of a monetary variable sum to zero, i.e. low significance levels are associated with monetary variables adding to the predictability of real variables. For Japan, it is evident that of the monetary variables considered, M2+CDs has the strongest relationship with real activity variables over the entire period in question. In addition, the additional explanatory power of M2+CDs (as well as bank credit) rose between the two periods, at least for retail sales, consumption, industrial production, and GNP. The 3-month rate is only a good additional predictor for industrial production and real GNP for the whole period, although it seems to be an important variable for investment in both of the subperiods, and for retail sales and net exports in the later period. The term spread is a good predictor of GNP, consumption and the trade balance, but only for investment in the earlier period. The fact that it increases in significance for several activity variables in the later period could mean that the sensitivity of activity variables (excluding investment) may have increased between the periods. The amount of financial deregulation and innovation in Japan during the past decade may have made the term structure relevant for the first time. As financial markets deepen, interest rates and term spreads may transmit a clearer signal of both monetary and real disturbances. For Germany, the 3-month rate is the best indicator for the entire period, although M3 is a good indicator of industrial production, unemployment, GDP, and investment. The term spread is a good indicator for the trade balance, retail sales, GDP, and consumption. In the full 17. Note that this is a weaker test than Granger causality in that it does not require that each individual coefficient equal zero, but only that they sum to zero. However, we wanted to rule out cases where monetary variables entered significantly, but with equal and alternating signs. - 21 - Kahn and Kole sample, the regressions yield coefficients with the correct signs. One interesting finding is that money is negatively (and 3-month interest rates are positively) related to real net exports and the nominal trade balance. Not surprisingly, real net exports are also positively related to the DM/$ exchange rate especially in the case where the regressions are specified in terms of rates of change. Comparing the two subperiods, it looks as if M3 and the exchange rate have somewhat more significance for real net exports in the later period, while the 3-month rate and the term structure have more significance for several real variables in the earlier period. It appears that interest rates were more important in the earlier period. Retail sales were much more related to monetary variables, including bank credit, in the earlier period indicating that they may have played more of a role in the monetary transmission mechanism. Overall, there are less differences to point to than in the case of Japan. In the United Kingdom, interest rates are more strongly related to real activity variables than the two monetary aggregates we considered. The best overall indicator for the entire period seems to be the 3-month rate, although the term-spread is quite significant as a predictor of IP, GDP, investment and net exports. There is little evidence that consumption became more interest-sensitive over the two periods, and unfortunately, we only have retail sales data on the latter half. There is evidence that industrial production became more related to the 3-month rate, the term spread, and M4 while becoming less related to MO. The short rate and the term spread also have more impact on GDP and investment in the 1980s, i.e. the marginal significance levels decline between the two periods. We also repeat these tests with first-differenced variables as well. These results were not as strong (not surprisingly), but there were a few points worth noting. In the case of Japan, the growth of M2+CDs is the best additional predictor of real economic growth, especially in the later period. There is again evidence that the term spread increased in importance between the two periods. The rate of change of the dollar exchange rate (the variable most likely to be nonstationary) has considerable significance for the rate of change of net exports for the full period and the early period, but not for the latter. - 22 - Kahn and Kole In contrast, the rate of DM depreciation against the dollar has more significance for German real net exports in the later period than in the earlier one, even though it its marginal significance level for the full sample is .0002. The 3-month rates still stands out as the best additional predictor for German economic activity. Finally, the result for the U.K. rate of change regressions were the worst of the three countries. The change in the dollar exchange rate did not add explanatory power in regressions involving the the trade balance or net exports, but both the change in the 3-month rate and the term spread were significant. Next we considered the role of bank credit for each of the countries and the monthly regressions in tables 7-9. To summarize, bank credit did not play much of a role in predicting German activity variables, except retail sales. The marginal significance of bank credit was .0001 for retail sales in the earlier period. Bank credit appeared to matter for all the activity variables in Japan, primarily in the later period. For the United Kingdom, available data only allowed us to conduct the tests on the later period, and we found that bank credit did not have significant explanatory power for real activity in the 1980s. To get at the issue of the change in international linkages, we decided to test whether a foreign monetary policy variable could add explanatory power to domestic monetary variables in Japan, Germany, or the United Kingdom. We reran all the regressions included in tables 7-9 adding 6 lags of the U.S. term spread to the monthly equations and 4 lags of the U.S. term spread to the quarterly Equations and tested whether the U.S. term spread had additional predictive power for foreign real economic variables once their own monetary variables were accounted for. For Japan, we found that the U.S. term structure was most significant as an additional explanatory variable for retail sales, the trade balance, the rate of change in unemployment, and GDP. In Germany, the U.S. term spread had additional explanatory power for GDP, consumption, net exports, and retail sales. The U.S. term spread is most significant in the U.K. regressions involving the trade balance, net exports, GDP, and investment. One result we found across countries was that the U.S. term spread seemed to have more significance in the earlier period than in the later - 23 - Kahn and Kole period. This could be evidence that international interest rates were less integrated in the earlier period, so that the U.S. term structure conveyed information lacking in domestic monetary variables. Another finding of note was that for most of the regressions involving net exports or the trade balance, the significance of the U.S. term spread substantially diminished when the exchange rate was chosen as the monetary variable, indicating that the U.S. term spread may be proxying for the exchange rate. In summary, the U.S. term spread had some additional explanatory power for foreign real variables, but nowhere near the power is seems to have for the U.S. real economy. CONCLUDING REMARKS This paper attempts to characterize the ways in which monetary transmission mechanisms abroad have changed in the past decades. We find that despite a major transformation in global financial markets, the demand for German M3 (pre-unification) and British M4 seems to be stable during the past two decades. Only in Japan, where financial deregulation was later and is still ongoing, do we fail to find a stable demand for broad money. We find some evidence that financial deregulation and innovation may have changed the interest rates relevant for money demand, and that the elasticity of money demand with respect to the opportunity cost of holding money has increased. There also seems to have been a shift from holding money for transactions purposes in the 1970s to holding it for portfolio motives in the 1980s. We find that wealth plays an important role in the determination of money demand in each of the countries we considered. While this paper does not take the further step of empirically measuring the importance of wealth channels in the transmission of monetary policy to the real economy, we believe that more work on this area is crucial. Furthermore, our analysis suggests that a greater role for asset prices as indicators, instruments, or targets of central bank operating procedures. The reduced form nature of our tests do not allow us to draw many conclusions about the transmission of monetary shocks to the real economy. However, we do find evidence that corroborates the view that the interest rate sensitivity of spending (especially that of consumption) has increased in the past decade. This is not surprising - 24 - Kahn and Kole given the wider availability of consumer and mortgage credit for households and the development of new financing methods for firms*. Further investigation into the impact of the debt buildup of the late 1980s on the monetary transmission mechanisms in Japan and the United Kingdom is clearly warranted. We also find some evidence that the exchange rate is an important channel for monetary policy. In addition, the term structure of interest rates seems to be gaining significance as a predictor of real economic activity in Japan and the United Kingdom. However, the term spread is not found to be nearly as significant in Japan and Germany as it has been found to be as a predictor of U.S. real output, which raises the question of why such a difference has emerged. Short of estimating a structural model across these foreign countries, there are several extensions to this paper that would be worthwhile. Estimating monetary reaction functions and attempting to pinpoint how they have changed over the past several years could give us valuable information on how money supply processes have evolved. - 25 - Kahn and Kole APPENDIX: Major Events or Policy Changes Affecting the Monetary Transmission Mechanism 01 1973 The Bretton Woods system of fixed exchange rates collapsed. 1979 The exchange rate mechanism of the EMS began operation in March. 1985 G7 Finance ministers met at the Plaza Hotel in New York in September and agreed on the Plaza Accord, that the dollar (which peaked in February 1985) was fundamentally overvalued and that a further fall in its value was warranted. 1987 G6 Finance ministers met at the Louvre in February and decided to manage major exchange rates within unannounced target zones. 1989 At the Madrid summit, EC governments agreed to start Stage One of European Economic and Monetary Union (EMU) on July 1, 1990. 1991 At the Maastricht summit, EC governments agree to start Stage Two of EMU in 1994, and to begin full monetary union sometime between January 1997 and the end of 1999. Japan 1974 Following the oil price shock, inflation rose sharply and the government began to run substantial deficits that contributed to growth in the government securities (Gensaki) market. Also, interest rates on foreign currency deposits were liberalized. 1975 Bank of Japan announced the "Money supply policy" in July acknowledging a shift in emphasis to focus on M2. 1978 Bank of Japan began to present a public forecast of the money supply. At the beginning of each quarter, a forecast (not an official target) - 26 - Kahn and Kole for the year-on-year growth rate of the monetary aggregate (initially M2) during the quarter was announced. The government began to issue, by public tender, 2-4 year medium-term bonds. 1979 Bank of Japan shifted to announcing target for M2+CDs. Banks were allowed to issue Certificates of Deposit (CDs) worth at least -P500 million. 1980 "Chuki-Kokusai" funds (similar to money-market mutual funds) were introduced by securities firms. "Foreign Exchange and Foreign Trade Control Law" led to significant relaxation of capital controls. 1985 Continued financial market reform occurred such as the introduction of money market certificates (MMC) for banks and credit associations, with minimum denominations (-50 million) and maturity 1-6 months. Minimum denomination and maturity of CDs were shortened, and the ceiling on issue size was raised. Interest rates on deposits of -1 billion or more with maturity of 3-24 months were deregulated. 1986 Interest rates on deposits of -300 million or more were decontrolled. Further liberalization on ceilings, size, and maturity of CD issues took place. Germany 1972 The annual report of the Council of Economic Experts advocated the control of the money stock to combat inflation. 1973 The Bundesbank discarded its previous monetary indicator, "free liquid reserves" and replaced it with "central bank money" (CBM). In order to maximize control over CBM, it reduced to "practically zero" the formerly generous reserves that allowed banks easy access to CBM. The mark began to float. - 27 - Kahn and Kole 1974 The Bundesbank announced a target for CBM for the first time in December. The growth of CBM was to be held to 8 percent between December 1974 and December 1975. 1978 As a result of a large appreciation of the mark and a substantial overshooting of its 8 percent CBM target for 1978, the Bundesbank announced that it would adopt a target range of 6-9 percent between the fourth quarters of 1978 and 1979. 1981 The Bundesbank made Lombard credit available only under a Special Lombard Facility at a cost higher than the ordinary Lombard rate. 1985 In January, the Bundesbank raised the Lombard rate to a level that applied in its temporary security operations and began to use market operations to supply reserves more liberally. 1986 Reserve requirements were restructured to ensure that they covered liabilities in the form of bearer securities at up to two years, when German banks were authorized to issue CDs. 1987 Minimum reserve ratios were increased for the first time since 1979, to offset the effect on bank reserves of large Bundesbank purchases of foreign exchange. After the substantially exceeding its CBM target, the Bundesbank decided to switch to an M3 target in December to decrease the influence of notes and coins on the aggregate. 1988 A withholding tax on capital incomes to take effect the beginning of 1989 was passed and reportedly caused substantial capital outflows. 1989 The withholding tax was withdrawn. The Berlin wall fell in November. 1990 Monetary union, including the swaps of Osmarks for DM occurred in July and was followed by full unification of east and west Germany in October. - 28 - Kahn and Kole The United Kingdom 1972 Sterling floated in June. 1973 Supplementary Special Deposits scheme (better known as the "corset) was introduced, under which banks were required to place supplementary special deposits with the Bank of England if their interest bearing eligible liabilities grew faster than a specified rate (8 percent in the first 6 months). This scheme was suspended and reactivated intermittently throughout the rest of the decade. 1979 The Conservative party came to power in May and essentially ended restrictive guidance on building society lending. In its first budget, the new government announced relaxation of exchange controls, continuation of the "corset", and a £M3 target range of 7-11 percent from mid-June 1979. Starting with the 1978-79 fiscal year it was announced that targets would be rebased every six months. 1980 The green paper on Monetary Control was published in March, and in the medium term financial strategy (MTFS), the government announced a gradual reduction in money supply growth. The 1980-81 target for £M3 was to be 7-11 percent, and the range was to decrease by 1 percent each subsequent fiscal year to 4-8 percent in 1983-84. The "corset" was discontinued in June. In November, the government published a note on Methods of Monetary Control, in which it advocated phasing out the reserve assets ratio, under which banks had to hold at least 12.5 percent of their deposits in a specified range of liquid assets, and considering a cash ratio instead. The Bank of England was to change its money market intervention to emphasize open market operations rather than discount window lending, to try and keep very short-term interest rates within an unpublished band, and to gear daily operations primarily towards offsetting cash flows between the Bank and the money market. 1981 In the MTFS, a new target range for £M3 was set at 6-10 percent for the 14 months starting in mid-February 1981. It was announced that the minimum reserve assets ratio would be abolished, and that all banks would be required to hold non-operational non-interest bearing deposits with the Bank of England. The new arrangements for monetary control took effect in August. 1983 The conservative government of Margaret Thatcher was reelected. Nigel Lawson, new Chancellor of the Exchequer, reviewed monetary policy and - 29 - Kahn and Kole stated that narrow measures of money were linked more closely to inflation. The building societies' cartel collapsed in October. 1984 A target range for MO was set for the first time, at 4-8 percent for the 14-month period beginning in mid-February 1984. The target range for £M3 was set at 6-10 percent. 1985 Lawson suspended the 1985-86 £M3 target of 5-9 percent, because it had been set too low. Overfunding (the policy of issuing more government debt than required by the budget deficit which had started in 1981) was dropped and a full-funding policy was adopted in order to ease persistent shortages in the money market. 1986 The Building Societies Act was passed and mortgage lending guidance was withdrawn. In October, financial reforms known as the "Big Bang" eased entry into dealing on the stock exchange and in the gilts market. A new target range for £M3 was set at 11-15 percent for the 1986-87 fiscal year, much higher than the 4-8 percent target range set out in the previous MTFS. The target range for MO was set at 2-6 percent. 1987 Target ranges for broad monetary aggregates were abandoned. The MO target was set at 2-6 percent. The U.K. authorities begin to "shadow the ERM", by attempting to keep the value of the pound below DM3. 1988 Sterling was allowed to appreciate above DM3 beginning in March. the MTFS, M4 replaced £M3 as the main measure of broad money. In 1989 At the Madrid summit in June, Britain committed itself to becoming a member of the ERM by the end of Stage One of EMU. 1990 In October, Britain joined the ERM at a central rate of DM2.95 and with 6 percent bands. - 30 - Kahn and Kole REFERENCES Bayoumi, Tamim (1990), "Financial Innovation and Consumption in the United Kingdom", IMF Working Paper WP/90/95, forthcoming in Review of Economics and Statistics. , Tamim (1992), "Financial Deregulation and Household Saving", mimeo. Bernanke, Ben, and Alan Blinder (1990), "The Federal Funds Rate and the Channels of Monetary Transmission." National Bureau of Economic Research Working Paper no. 3487, October. Bernanke, Ben and Frederic Mishkin (1992), "Central Bank Behavior and the Strategy of Monetary Policy: Observations from Six Industrial Countries", mimeo. Blanchard, Olivier J. (1981), "Output, the Stock Market, and Interest Rates," American Economic Review, vol. 71, no. 1, pp. 132-143, March. Boughton, James M. (1992), "International Comparisons of Money Demand, IMF Working Paper no. WP/92/7. Brayton, Flint, and Jaime Marquez (1990), "The Behavior of Monetary Sectors and Monetary Policy," in Financial Sectors in Open Economies: Empirical Analysis and Policy Issues, P. Hooper, K. H. Johnson, D. L. Kohn, D. E. Lindsey, R. D. Porter and R. Tryon (eds.), Board of Governors of the Federal Reserve System, Washington, D.C., pp. 365-393. Bryant, Ralph C. (1990), "Model Representations of Japanese Monetary Policy," mimeo. Corker, Robert (1989), "Wealth, Financial Liberalization and the Demand for Money in Japan", IMF Working Paper WP/89/85. - 31 - Kahn and Kole Dicks, M.J. (1990), "Interest Elasticity of Consumers' Expenditure", in S. G. B. Henry and K. D. Patterson (eds.) Economic Modelling the Bank of England, Chapman and Hall, London, at pp. 73-106. Estrella, Arturo and Gikas A. Hardouvelis (1991), "The Term Structure as a Predictor of Real Economic Activity", Journal of Finance, vol. 46, pp. 555-76. Frowen, Stephen F. and Heinrich Schlomann (1992), "Financial Innovations and the Stability of the Demand for Money in Germany since 1974", in Monetary Policy Countries: and Financial Innovations in Five Industrial The UK, the USA, West Germany, France and Japan, S. F. Frowen and D. Kath (eds.), St. Martin's Press, New York, pp. 59-81. Fukui, Toshihiko (1992), "The Recent Development of the Short-Term Money Market in Japan and Changes in the Techniques and Procedures of Monetary Control Used by the Bank of Japan," mimeo Gavin, Michael K. (1989), "The Stock Market and Exchange Rate Dynamics" in Journal of International Money and Finance, vol. 8, pp.181200. Hess, Gregory D. and Richard D. Porter (19.92), "Comparing Interest-Rate Spreads and Money Growth as Predictors of Output Growth: Granger Causality in the Sense Granger Intended", mimeo. Germany, J. David, and John E. Morton (1985), "Financial Innovation and Deregulation in Foreign Industrial Countries", Federal Reserve Bulletin, pp. 743-753. Hall, S. G., S. G. B. Henry and J. B. Wilcox (1990) "The Long-run Determination of the UK Monetary Aggregates" in S. G. B. Henry and K. D. Patterson (eds.) Economic Modelling England, Chapman and Hall, London, pp. 127-166. - 32 - at the Bank of Kahn and Kole Hooper, Peter and Catherine L. Mann (1989), "Exchange Rate Passthrough in the 1980s: The case of U.S. Imports of Manufactures", Papers on Economic Activity, Brookings no. 1, 297-329. Jappelli, Tullio and Marco Pagano (1989), "Consumption and Capital Market Imperfections: Economic Review, An International Comparison" in American vol. 79, no. 5, December, pp. 1088-1105. Kasman, Bruce, and Anthony Rodrigues (1991), "Financial Liberalization and Monetary Control in Japan." York Quarterly Review, Federal Reserve Bank of New Autumn, pp. 28-60. Kole, Linda S. and Michael P. Leahy (1991), "The Usefulness of P* Measures for Japan and Germany", International Finance Discussion Paper No. 414. Shigehara, Kimiharu (1992), "Current Monetary Policy Issues in Japan," mimeo. Stock, James H. and Mark W. Watson (1989), "New indexes of coincident and leading indicators", NBER Macroeconomics Annual 1989, pp. 351-93. Tamura, Tatsuya (1992), "Monetary Control in Japan," in Monetary Policy and Financial Innovations in Five Industrial Countries: The UK, the USA, West Germany, France and Japan, S. F. Frowen and D. Kath (eds.), St. Martin's Press, New York, pp. 101-119. Temperton, Paul (1991), UK Monetary Policy: St. Martin's Press, New York. The Challenge for the 1990s, Topping, S. L. and S. L. Bishop (1989), "Breaks in Monetary Series", Bank of England Discussion Paper, Technical Series No. 23. Ueda, Kazuo (1990), "Financial Deregulation and the Demand for Money in Japan" in Financial Sectors in Open Economies: Analysis Issues, Edited by P. Hooper, K. H. Johnson, and Policy - 33 - Empirical Kahn and Kole D. L. Kohn, D. E. Lindsey, R. D. Porter and R. Tryon, Board of Governors of the Federal Reserve System, pp. 175-199. Von Hagen, Jurgen and Manfred J.M. Neumann (1988), "Instability versus Dynamics: A Study in West German Demand for Money", Journal of Macroeconomics, vol. 10, no. 3, pp. 327-49. West, Kenneth D. (1991), "An Aggregate Demand - Aggregate Supply Analysis of Japanese Monetary Policy, 1973-1990," National Bureau of Economic Research Working Paper no. 3823, August. - 34 - Kahn and Kole TABLE 1 Measures of Openness (Annual Average, in Percent) Exports GDP Exports + Imports GDP 1971-75 1987-91 197J-75 1987-91 Japan 24 33 10 17 Germany 50 .74 25 38 United Kingdom 47 67 23 32 France 38 51 18 24 Italy 10 12 4 6 Canada 40 58 21 28 United States 14 21 6 10 - 35 - Kahn and Kole TABLE 2 Gross Debt as Percent of GNP/GDP 1960-70 1970-80 1985 1990 47.4 57.1 51.4 59.0 50.2 38.1 6.9 55.5 63.2 52.6 73.8 71.7 63.2 47.1 9.8+ 66.9 Households Japan Germany United Kingdom United States Canada France Italy Sweden 20.3 34.7 31.0* 48.0 43.9 34.0* n.a. 46.0* 34.2 43.2 33.6 50.3 53.2 38.1 6.5 51.1 Non-Financial Firms Japan Germany United Kingdom United States Canada France Italy Sweden 85.6 54.6 44.3* 35.6 62.0 63.6* n.a. n.a. 92.1 65.1 46.7 35.6 65.8 62.0 48.9 59.9 100.6 72.9 46.5 40.7 69.6 60.1 57.2 68.1 134.7 74.4 78.7 48.9 76.4 68.9 60.9 99.6 Public Sector Japan Germany United Kingdom United States Canada France Italy Sweden 22.5 20.3* 87.1* 56. 94. 45. 36, n.a. 45.8 24.5 64.9 44.9 82.8 36.4 56.6 55.0 * Data for only part of period + 1989 Source: BIS - 36 - 88.9 41.4 58.6 56.6 106.8 42.5 84.7 80.1 75.7 43.3 45.3 64.1 108.9 44.2 102.5 55.7 Kahn and Kole TABLE 3 The Model (1) D - a1q + a2Y + a^E + D - value of stock market (2) Y - d(D-Y) (3) aggregate demand i - gY - h(M-P) + jq Y — real output (4) r - i - P E - real exchange rate (5) P - -f(P-P) G - government spending (6) P - M + (r - gY)/h i - nominal interest rate (7) q/q + Z/q - r M - money stock (8) r - r P - price level (P in equilibrium) (9) Z - bj+ b 2 Y r - real interest rate (in terms of home goods) (10) R - r + R/R Z - profits r - foreign real interest rate R - long-term real interest rate X - dx dt X - equilibrium value of X + E - 37 Kahn and Kole TABLE 4: Estimates of Japanese Demand for Real M2+CDs, A(m-p) 1973:IV 1982:IV 1973:IV 1991:11 Variable 1983:1 1991:11 .464 (.58) .065 (.06) .571** (5.07) .328** (2.86) .365* (2.04) A(w-p) t .150** (3.44) .269** (3.92) .196** (3.00) A(w-p)tl .075* (2.04) .001 (1.59) .073 (1.53) -.860** (-10.13) -.552** (-7.32) -.952** (-5!l8) .530** (3.44) .089 (-.72) .462 (1.54) 1.196* (2.10) -1.292* (-2.33) 2.193* (2.20) -.988* (-2.11) .486 (1.20) -1.277 (-1.24) -.253** (-4.84) -.413** (-5.01) -.311* (-2.28) Intercept 1.005* (2.31) A(m-p) t _ 1 Ap, Ap t-1 "°t A -PS Al t m t-r p t-i yt-i w t-r pt-i .0 t-1 L .PS 1 t-l Regression .103* (2.16) .182** (2.98) .179 (1.54) .121** (4.88) .252** (3.20) .157** (2.72) 1.588** (4.80) -1.271 (-1.79) 1.107 (.86) -1.268** r-4.811 .651 (1.31N -.907 (-.75) statistics R2 .895 Standard error .00405 Sample size 71 2 Serial correlation (X ) lrst order .070 lrst-4th order 2.843 .973 .00234 37 9.493** 12.271* Fisher test F 13.45 - 4 ' 6 2 ** T statistics are in parentheses •^Significant at the 5 percent level **Significant at the 1 percent level - 38 - .850 .00362 34 2.545 6.547 Kahn and Kole TABLE 5: Estimates of German Demand for Real M3, A(m-p) 1970:11 1989:1V Variable 1970:11 1979:IV -1.610 (-1.49) .331 (.88) Intercept 1980:1 1989:IV .381 (.94) A(w-p) t .456** (4.02) .367* (2.19) .708** (4.04) A w .159 (1.69) .244 (1.59) -.060 (-.44) Ay t .050 (.63) .289 (1.97) -.073 (-.73) Ap t -.481** (-3.15) -.509* (-2.60) -.153 (-.47) m -.099 (-1.66) -.327* (-2.31) -.105 (-1.25) ( -P>t-1 t - r Pt-i yt-i w t - r Pt-i .019 (.31) .349 (1.95) -.026 (-.40) .054 (1.76) .077 (1.35) .102 (1.91) .o t-1 .371* (2.04) .PB -.536** (-4.03) L 1 t-1 Regression .072 (.23) -.484* (-2.41) .586* (2.26) -.588** (-3.34) statistics R2 .594 Standard error .00620 Sample size Serial correlation (X ) lrst order .044 lrst-4th order 1.776 lrst-8th order 3.307 .601 .00706 39 .184 7.865 10.094 Fisher test F 10,59 "X - 2 3 T statistics are in parentheses ^Significant at the 5 percent level ^^Significant at the 1 percent level - 39 - .554 .00499 40 1.586 4.333 8.875 Kahn and Kole TABLE 6: Estimates of U.K. Demand for Real M 4 , A(m-p) 162 (-' 36) -.147 (-1.08) Intercept A(m-p)tl A 1970:1 1979: III 1970:1 1991:11 Variable ^t-2 278 .337** (4.01) a. 78) .182** (2.74) .265* {2. .45) -.944** (-13.67) (-s'• 79) .972** 1979:IV 1991• ; H ,. -2. 583** (-2.94) 057 (• 36) -2..87* (-2..08) -1..116** (-8 • 26) Ap t-1 .315** (3.00) (1 .41) .091 ( • 52) Ap t-2 .190** (2.66) .128 (1 .21) . .314 (-1 • 97) -.055** (-3.44) .058 (-1 .67) . .187** (-4 .68) m t-rpt-i -.028 (-1.45) 't-1 w L .285 .019 (- • 72) .282* (2 .52) t-rpt-i .087** (4.71) .082* .094* (2 • 15) (2 • 53) t-i .297* (2.49) (1 • 61) .TB L t-1 Regression .696 -.321** (-3^80) .553** (-3 • 14) .087 ( • 42) - .013 (- • 10) statistics R< .881 Standard error .00663 Sample size 86 Serial correlation <*2> lrst order .920 Irst-•4th order 2.329 .906 .00726 39 2.437 3.957 Fisher test 11,64 T statistics are in parentheses *Significant at the 5 percent level **Significant at the 1 percent level - 40 - - 1.88 .805 .00531 47 2.065 10.802* Kahn and Kole TABLE 7: Marginal Significance Levels of Monetary Indicators for Forecasting Alternative Measures of Economic Activity: Japan Activity Variables M2+CDs 3-month Rate Term Soread Dollar Exchange Rate «^i^___^»~»a~ 1973:111-1992:1 .014 .086 .168 2) Retail Sales (RS) .00003 .020 3) Trade Balance (TB) .144 .302 .771 .064 .081 4) Unemployment (U) .015 .303 .115 -- 5) GNP .001 .023 -- 6) Consumption (C) .005 .365 .304 .059 -- 7) .022 .395 .759 -- .286 .264 .273 .021 1) IP .135 .380 .432 2) RS .677 .189 .329 -- 3) TB .050 .875 .737 .499 4) U .930 .959 .930 -- 5) GNP .253 .187 .604 -- 6) C .149 .298 .962 -- 7) I .912 .191 .001 -- .420 .003 .594 1) IP .033 .678 .069 2) RS .007 .005 .006 -- 3) TB .005 .972 .003 .771 4) U .095 .022 .0003 .660 .144 .232 .002 -- 1) Industrial Production (IP) Investment (I) 8) Net Exports (NX) 1973:111-1982:IV 8) NX -- .001 1983:1-1992:1 5) GNP 6) C 7) I 8) NX .522 .302 .491 .015 .029 -- .376 .200 -- .484 .040 -- For each activity variable, entries across the rows are the marginal significance levels for the F-test that the coefficients on 6 lags of the monetary indicators (columns) sum to zero in an unrestricted OLS prediction equation that also included a constant, trend, 6 lags of the forecast variable, and 6 lags of a price variable. The price variables used were the PPI in regressions including IP or U, the CPI in regressions with RS, and a world commodity price index in regressions with TB. Data are monthly -and all variables but interest rates, the trade balance, and net exports are log levels. The term spread is the 10-year rate less the 3-month rate. Quarterly data include real GNP, consumption, investment, and net exports and are in 1985 yen. - 41 - Kahn and Kole TABLE 8: Marginal Significance Levels of Monetary Indicators for Forecasting Alternative Measures of Economic Activity: Germany Dollar •Exchange Rate 3-Month Rate Term Spread .080 .005 .405 .006 .039 --.164 .442 -- .010 .006 .004 .013 .681 .001 .341 .00002 .016 .008 .002 .106 --- .103 -.007 1) IP 2) RS 3) TB .779 .770 .995 -- .018 .0004 .006 -- .585 .720 .077 .542 4) U 5) GDP 6) C .131* .072 .035* .071 --- .949 .00008* .460 .041 .509 -- 7) I 8) NX 1981:1- 1989:IV .611 .044 .517 -- .647 .375 .019 .231 .084 .018 .514 .623 .833 .415* .404 .667 .429 .605 .219* .189 .475 .153* .064 --.209 --- .690 .612 .555 -- .530 .272 .588 -- .021 .276 .299 .019 Activity Variables Ml L973:Ii:I-1989:IV 1) Industrial Production (IP) 2) Retail Sales (RS) 3) Trade Balance (TB) 4) Unemployment (U) 5) GDP 6) Consumption (C) 7) Investment (I) 8) Net Exports (NX) .036 .035 .913 .056 .048 1973:III-1980:IV 1) IP 2) RS 3) TB 4) U 5) GDP 6) C 7) I 8) NX For each activity variable, entries across the rows are the marginal significance levels for the F-test that the coefficients on 6 lags of the monetary indicators (columns) from an sum to zero in an unrestricted OLS prediction equation that also included a constant, trend, 6 lags of the forecast variable, and 6 lags of the PPI. Data are monthly and all variables but interest rates, trade balance, and net exports are log levels. The term spread is the rate on 5-7 year public bonds less the 3-month rate. Quarterly data include real GDP, consumption, investment, and net exports and are in 1985 DM. * 12 lags used for unemployment, PPI, and monetary variable. - 42 - Kahn and Kole TABLE 9: Marginal Significance Levels of Monetary Indicators for Forecasting Alternative Measures of Economic Activity: United Kingdom Activity Variables 1974:1-1991:11 1) IP MO M4 .186 .023 016 3-Month Rate Term Spread Dollar Exchange Rate .0002 .003 -- .065 291 -- .028 .149 .105 5) GDP 6) C 7) I 658 .049 .042 .053 182 307 .130 .750 .074 8) NX 109 .145 .299 .372 .011 .051 .053 1974:1-1979:111 1) IP .015 .660 641 .796 .875 -- .022 .545 .747 .016 .687 .070 .886 .021 -.191 .065 .317 .071 .961 .434 .628 .056 .328 .252 .466 .197 .092 .640 .123 .336 .046 .0001 .009 .0008 .209 .016 .136 .336 .009 .407 .768 .546 .360 .013 .603 .256 .910 .530 .669 .190 .561 .131 .424 .016 .842 .213 .411 .442 .545 .018 .063 2) RS 3) TB 4) U 2) 3) 4) 5) 6) 7) RS TB U GDP C I 8) NX 1979:IV -1991:111 1) IP 2) RS 3) TB 4) U 5) GDP 6) C 7) I 8) NX .068 .101 .131 .751 For each activity variable, entries across the rows are the marginal significance levels for the F-test that the coefficients on 6 lags of the monetary indicators (columns) from an sum to zero in an unrestricted OLS prediction equation that also included a constant, trend, 6 lags of the forecast variable, and 6 lags of the PPI. Data are monthly and all variables but interest rates, trade balance, and net exports are log levels. The term spread is the 10-year rate less the 3-month rate. Quarterly data include real GDP, consumption, investment, and net exports and are in 1985 pounds. - 43 - Kahn and Kole Cham INTEREST RATES AND INFLATION IN SELECTED G-10 COUNTRIES UNITED STATES JAPAN Percent Percent 25 25 •3-month Interbank Rate — — - . - 12-month Rate of CPI Inflation 20 L * 15 10 * X.Wv ~r**\ * ' • % , / I M 1970 M M 1975 I I I I M 1980 M ll 1985 GERMANY I I I M 1990 I 1I I I I I I I I I I I M I 1I I I I 1I 5 1970 1975 1980 -985 FRANCE Percent 1990 Percent 20 ' I 16 I— 16 12 f— 12 I I I I I 1 I I I I I I I t I t I I I I 1970 1975 1980 1985 UNITED KINGDOM I M M I I I I I I I I I I I I I I I I I 1990 1970 "™^™l A L^ 1975 CANADA Percent r-" 20 1980 1985 1990 Percent 30 25 ^ J 25 11 20 1 • U- i 1 • —| 20 15 J s& A t It XI \ 15 if 1 h 10 y ¥-V v-\- ''vl \J £1 1 1J_U 1 _LU M l I 1 1 1 1 1 1 1 1 L-LL 1970 1975 1980 1985 10 5 0 1990 - 44 1 M 1970 IIIIIIIIIIIM IIIII I 1975 1980 1985 1990 Kahn and Kole Chart 2: Model Dynamics An Unanticipated Monetary Expansion Time To Time To Time Interest Rates r.R To Time Case a. "baa news" (normal case) Case D: "good news" Possible Exchange Rate Paths To Time Case a* norma! case initial deoreciation. then graoual return to baseline To Time Case 0: initial depreciation, then overshoot - (Ef = depreciation) 45 - To Casec: initial appreciation Time Kahn and Kole Chart 3 Market Interest Rates and Own Rates of Return On Broad Money 14 Japan T 12 1 H 6 l 4 i 2-r w ' M2 + CDS 1976 1979 1982 1985 1988 1991 1988 1991 Germany T979 i^IT United Kingdom 20i q 973 1982 197T 1979" 1982 - U(y - 1985 MONETARY TRANSMISSION CHANNELS IN MAJOR FOREIGN INDUSTRIAL COUNTRIES: A COMMENT Craig S. Hakkio In "Monetary Transmission Channels in Major Foreign Industrial Countries," Robert Kahn and Linda Kole address "whether and how these [monetary] transmission channels have changed during the past two decades." In studying the U.S. economy, Friedman (1989, p. 96) finds that changes in the U.S. economy have "led to major changes in standard reduced-form relationships of the kind that often stand behind quantitative analysis of monetary policy at either formal or informal levels." Kahn and Kole have the more difficult challenge of finding whether the transmission channels have changed in Japan, Germany, and the United Kingdom. In the first part of their paper, Kahn and Kole discuss why the channels of monetary policy may have changed. They then present a stylized Keynesian model that supposedly "captures the basic relationships that [the authors] hope to capture in the empirical work." However, the changes that occurred during the last two decades are probably more complex than what can be captured in a simple Keynesian model. In these comments, I will not discuss either of the first two sections. Instead, I will discuss the empirical results--presented in sections 4 and 5. The evidence presented in "Monetary Transmission Channels in Major Foreign Industrial Countries" suggests that the 1. Craig S. Hakkio is an assistant vice president and economist at the Federal Reserve Bank of Kansas City. He thanks Sean Becketti for comments on an earlier draft. Hakkio transmission channel may, or may not, have changed. The authors estimate money demand functions for Japan, Germany, and the U.K. for the whole period and two subperiods. They find that money demand functions are stable in Germany and the United Kingdom, but unstable in Japan. The authors then estimate reduced form equations for 8 activity variables, using 3 measures of monetary policy, for the whole period and two subperiods. They "find evidence that corroborates the view that the interest rate sensitivity of spending has increased over the past decade." In these comments, I will first make some specific comments on the paper by Kahn and Kole. Then, I will extend their results by looking at whether financial markets have become more integrated. Finally, I will test whether their results are sensitive to specification problems. SOME SPECIFIC COMMENTS ON THE KAHN-KOLE PAPER The authors1 finding that "the interest sensitivity of spending (especially that of consumption) has increased in the past decade" is surprising. (1989, p. 30) states: In studying the U.S. economy, George Kahn "Empirical evidence suggests a reduction in the economy's overall interest sensitivity. This reduction in interest sensitivity is not spread equally across all sectors of the economy, however." In particular, Kahn finds that consumption is less interest sensitive, not more interest sensitive as found by Kahn-Kole. -2- Benjamin Friedman (1989, p. 97) Hakkio reports similar findings. In discussing their results, Kahn-Kole seem to argue that a smaller marginal significance level means that the effectiveness of monetary policy is larger.2 This is not necessarily true. The effectiveness of monetary policy depends on the size of the coefficient, in addition to its significance. And the marginal significance level says nothing about the size of the effect; rather it says something about the size of the coefficient relative to its standard error. If the coefficient falls and the standard error falls more, the marginal significance level will rise even though monetary policy has become less effective. HAVE MARKETS BECOME MORE INTEGRATED? Kahn and Kole argue that the transmission channels of monetary policy have changed due to financial liberalization and greater international openness. Since financial liberalization often took the form of opening financial markets to international competition, I will consider these two explanations as one. With more integrated financial markets, interest rates are determined in a single world capital market. Therefore, we would expect German interest rate changes to be highly correlated with U.S., Japanese, and U.K. interest rate changes. In addition, we would 2. The authors recognize this problem when they state: "The fact that it increases in significance for several activity variables in the later period could mean that the sensitivity of activity variables (excluding investment) may have increased between the periods." -3- Hakkio expect monetary policy would be less able to influence interest rates. Therefore, instead of determining whether money demand functions have shifted, or the sum of lag coefficients have changed, we can look directly at whether changes in interest rates have become more or less correlated over time. If markets are more integrated, then we would expect interest rate changes to be more highly correlated. To test this hypothesis, I collected daily interbank bid rates from FAME for Japan, Germany, the U.K., and the United States. The interbank rate is a short-term interest rate. Since the timing may be important, the June 24 interest rate quote is at 10:00 am (local time) in Germany and the U.K., and at closing (local time) in Japan; in the United' States, the interest rate is the effective federal funds rate. I then calculated the Spearman rank correlation coefficient between changes in interest rates and between the level of interest rates. The Spearman rank correlation coefficient is a robust measure of association between two variables; it is simply the correlation between the ranks, as opposed to between the actual values. I did not include exchange rates in calculating a covered interest rate because the results would be dominated by the exchange rate. Table 1 gives the Spearman rank correlation coefficients. The top half of the table gives the correlation coefficient for the first difference ir* short-term interest rates, while the -4- Hakkio bottom half gives the correlation coefficient for the level of short-term interest rates. The first number in the cell is for the whole period, while the other two numbers are for the 2 subperiods. The subperiods are chosen to match those used in the Kahn-Kole paper. The breakponts are: Germany, December 31, 1979; Japan, December 31, 1982; and the United Kingdom, September 28, 1979. The table shows that the rank correlation is about zero for the first difference and between 1/3 and 3/4 for the level. For example, the correlation between German and Japanese interest rate changes, for the whole sample, is -0.001; the correlation between the level of German and Japanese interest rates is 0.77. There is little evidence that the correlations are bigger in the second subperiod. For example, for the first differences, 3 correlations become bigger in absolute value, 2 becomes smaller, and 4 remain the same. become smaller. For the levels, 5 become bigger and 4 Of course, without standard errors we cannot determine whether the changes are significant. 3. The marginal significance levels for the correlation coefficients equal 0.00 for the correlation of the levels, and are generally greater than 0.30 for the first differences. The only exceptions for the first differences are: Germany and Japan in the first subperiod (correlation = 0.07, msl = 0.10); Germany and the U.S. in the whole period (correlation = 0.03, msl = 0.09), and in the second subperiod (correlation = 0.03, msl = 0.08); the U.K. and the U.S. in the whole period (correlation = 0.05, msl = 0.00), in the second subperiod (correlation = 0.05, msl = 0.01). -5- Hakkio Another way to test for changes in the extent of financial market integration is to calculate whether big changes in shortterm interest rates are independent. Define "big" to mean a change in the upper or lower 5 percent tail of the distribution. With this definition, 10 percent of interest rate changes are "big." Table 2 shows the results for changes in U.S. and German interest rates. Table 2 is a two-way classification of interest rate changes. The table shows that of 3978 observations, 56 (or 1.4 percent) had a big change in U.S. interest rates and a big change in German interest rates. Given the definition of "big," we would expect 1 percent of the observations to fall in the BIGBIG cell if big changes in U.S. interest rates were independent 4 of big changes m German interest rates. is a test of independence. Fisher's exact test According to the table, big changes are correlated with big changes. Table 3 reports similar results for all countries and for 2 subperiods. The periods were chosen to match those used in the Kahn-Kole paper. The table reports the marginal significance level of Fisher's exact test statistic for independence. A small marginal significance level means you can reject the hypothesis that big changes are independent of big changes; less precisely, 4. Actually, we would expect 0.95 percent of the observations to fall in the BIG-BIG cell. Since there are missing observations for each variable and the table is constructed for non-missing observations for both variables, the BIG row and column sums do not equal 10 percent. As a result, if the changes are independent, we expect to find 0.107*0.089 = 0.0095 =0.95 percent of the observations in the BIG-BIG cell. -6- Hakkio a small marginal significance level means that big changes are correlated with big changes. The results in Table 3 suggest that financial markets did become more correlated in the second subperiod. In most cases, the marginal significance level fell in the second subperiod. The two exceptions were (1) Germany and the U.S., where the marginal significance level rose from 0.00 to 0.01, and (2) the U.K. and Japan, where the marginal significance level rose from 0.46 to 0.49. Also, in many cases the marginal significance levels are less than 10 percent in the second subperiod. For example, we can reject the hypothesis that big changes in German interest rates are independent of big changes in Japanese and U.S. interest rates. To summarize, the results in this section complement the results in the Kahn-Kole paper. There is some weak evidence that financial markets have become more integrated. As a result, we would expect that monetary policy transmission channels would change. However, since the evidence on greater integration is weak, the change in transmission channels is probably also weak. WHY ARE THE RESULTS WEAK? The inconclusive or weak results could truely reflect little or no change in the transmission channels, or they could reflect statistical problems. If we can minimize the chance of statistical problems, then we can be more confident that the -7- Hakkio results really reflect little or no change in the transmission channels. Therefore, this section looks at some potential statistical problems. Are the Results Sensitive to Outliers? The presence of outliers could produce the inconclusive results reported by Kahn-Kole. To check for outliers, I first estimate a reduced form equation for industrial production. The general form of the equation is similar to that used by the authors: K log (ind prod) t - <x0 + a1TIMEt + ]£ P i (monetary policy) t_i i-l X + ]T Yilog( ind prod) t_± + et 2-1 P -EPi 1-1 Monetary policy is measured by a monetary aggregate and by shortterm interest rates. The lag length, K, is determined from Akaike's Information Criterion and Amemiya's Prediction Criterion. I search for influential observations in this regression in several ways. An influential observation, or a small influential subset of data, is one which "can have a disproportionate influence on the estimated parameters or predictions" of a regression equation 5. Generally, 2 to 4 lags were required, fewer than used by the authors. -8- Hakkio (Belsley, Kuh, and Welsch (1980), p. 6). An observation may have a big influence on the fitted values of a regression, on the variance-covariance matrix of the coefficients, or on the sum of lag coefficients. Accordingly, three statistics are used in this paper to detect influential observations.6 The first, Cook's distance, measures the influence of the t-th observation on the fitted values from a regression. The second, COVRATIO (Belsley, Kuh, and Welsch), measures the influence of the t-th observation on the variance-covariance matrix of the coefficients. The last, a variant of DFBETA (Belsley, Kuh, and Welsch), measures the influence of the t-th observation on the sum of the lag coefficients of the regression. each statistic. Critical values are given for In the results presented below, I focus on only the largest value of the statistic (which is also greater than the critical value). To find an influential observation, the regression is estimated with all observations and with all but the t-th observation. Then, a normalized difference in some statistic is calculated with and without the t-th observation. Finally, the normalized difference is compared to a critical value. A large normalized difference means the observation is influential. As an example, consider the variant of the DFBETA statistic. 6. See Chatterjee and Hadi for a discussion of influential observations in linear regressions. They state that these three measures "seem sufficient for detecting influential observations" (p. 387). -9- Hakkio I calculated a time series of sum of lag coefficients obtained from omitting one observation at a time. More specifically, DFBETA « (j8 - $t)/aR (t) ' w h e r e & i s t h e sum of the la 9 coefficients, j8. is the sum of the lag coefficients obtained from omitting the t-th observation, and a*. . is the standard error of the sum of lag coefficients. In other words, DFBETA is like a t-statistic: it equals the difference in coefficient estimates divided by the standard error. If results are sensitive to an outlier at observation t, then DFBETA will be "large." Table 4 shows the dates of the influential observations in a reduced form with monetary policy measured as money and shortterm interest rates. Each cell in the table gives the date of the most influential observation, the fraction of observations that are influential (the number to the left of / ) , and the size of the largest statistic relative relative to the critical value (the number to the right of / ) . As expected, the different statistics find different influential observations. The fraction of influential observations ranges from 1 percent to 10 percent; and the size of the largest statistic ranged from 1.3 times the critical value to 31.8 times the critical value. Unfortunately, no simple conclusions can be drawn from the table. While no simple conclusions can be drawn, it is clear 7. See Belsley, Kuh, and Welsch (1980) for an extended discussion of the DFBETA statistic. -10- Hakkio that with so many influential observations, some of which are very "large," the results may be due to influential observations. Influential observations can be due to improperly recorded data. Alternatively, they can be legitimately extreme observations that contain valuable information about the parameter estimates. However, even in this situation it is important to identify the observations and determine the extent to which they are responsible for the results. That is, we want to know whether the results are due to this one observation, rather than the entire dataset. Unfortunately, such an analysis is beyond the scope of these comments. Are the results sensitive to choice of subperiods? The results could also be inconclusive because the authors split the sample at the wrong place. As a result, estimating a reduced form equation over two subperiods may miss the change. In addition, while it may be reasonable to think the change occurred at a single point in time, it is probably more likely that there have been several changes or that the changes occur gradually over time. If the effectiveness of monetary policy is changing gradually or has changed more than once, then estimating a reduced form over two sample periods may again miss the change. To allow the change in monetary policy to occur over time, I estimated a series of rolling regressions. I estimate the same reduced form as in the previous section. Monetary policy is -11- Hakkio measured as either a monetary aggregate or the short term interest rate. The sample period is a fixed 20 percent of the observations. Charts 1-3 plot the sum of the lag coefficients with a 2 standard deviation confidence band. The top panel is the sum of money lag coefficients and the bottom panel is the sum of short-term interest rate lag coefficients. In Japan, the sum of money lag coefficients changed from negative in the first part of the sample period to positive in the second part. Furthermore, the standard errors have become smaller over time. The sum of interest rate lag coefficients has changed over time, but there is no pattern. Except for two episodes, the sum was about 0 for most of the 1980s; the sum then turned positive in the 1990s. There is little evidence of a change in the effectiveness of German monetary policy. The sum of both money and interest rate lag coefficients has fluctuated around 0 for most of the sample period. In the United Kingdom, monetary policy seems to have become less effective. The sum of money coefficients was positive in the early part of the sample, and has been zero since then. However, the sum of interest rate coefficients has fluctuated around zero for most of the sample period. To summarize, while the sum of lag coefficients have changed over time, the changes may not have been significant. Therefore, the weak results reported are probably not due to the authors1 -12- Hakkio choice of subperiods. CONCLUSIONS The authors present a comprehensive study of changes in the monetary policy transmission mechanism in Japan, Germany, and the United Kingdom. change. The find weak evidence that there has been a In looking at different data and techniques, I have confirmed their results. -13- Hakkio REFERENCES Belsey, David A., Edwin Kuh, and Roy E. Welsch, Regression Diagnostics: Identifying Influential Data and Sources of Collinearity, John Wiley & Sons, New York, 1980. Chatterjee, Samprit and Ali S. Hadi, "Influential Observations," Statistical Science. August 1986, pp. 379 - 392. Friedman, Benjamin, "Changing Effects of Monetary Policy on Real Economic Activity," in Monetary Policy Issues in the 1990s, a symposium sponsored by the Federal Reserve Bank of Kansas City, 1989. Kahn, George A. "The Changing Interest Sensitivity of the U.S. Economy," Economic Review, Federal Reserve Bank of Kansas City, November 1989, pp. 13 - 34. -14- Hakkio l. Correlation of Short-term Interest Rates (Spearman rank correlation matrix of first differences and levies) Germany Japan United Kingdom U.S. First difference of short-term interest rates Germany -0.00 0.01 0.03 -0.01 0.00 -0.02 0.03 -0.02 0.06 0.01 -0.01 0.01 0.01 0.01 -0.01 * -0.00 | || * Japan -0.01 0.01 United Kingdom 0.01 0.05 0.01 * | 0.01 0.02 0.05 0.05 0.03 0.00 Levels of s hort-term interest rates 1 Germany | || 0.77 0.61 0.47 0.74 0.73 0.42 0.59 0.13 0.42 0.63 0.43 0.73 0.49 0.61 0.12 • 0.77 * Japan 0.72 0.66 United Kingdom 0.61 0.45 0.63 * 1 0.34 0.33 0.10 P_._5_9.__ 0___6_4. 0.40 1 Note: Each number is a rank correlation coefficient. The numbers are for different subsamples: row 1 whole subsample row 2 first subsample, determined by row variable row 3 second subsample, determined by row variable The German/Japan correlation does not equal the Japan/German correlation in rows 2 and 3 because the breakpoints for the subsamples are different. -15- 2. Normal and Big Changes in German and U.S. Short-term Interest Rates Change in U.S. interest rates Change ! row sums normal change BIG change 3253 300 3553 81.8 % 7.5 % 89.3 % 369 56 425 9.3 % 1.4 % 10.7 % 3662 356 3978 91.0 % 8.9 % in German interest normal change rate BIG change column siims ; Test of independence of row and column variables: Fisher's marginal significance level = 0.002 100 % Hakkio 3. Normal and Big Changes in Short-term Interest Rates Two-way Table of Frequency Counts Germany Germany Japan U.K. U.S. 0.00 0.26 0.00 0.50 0.00 0.67 0.26 0.00 0.01 0.69 0.37 1/1/80 * Japan 1/3/83 U.K. 10/1/79 0.00 0.29 0.00 * 0.26 0.77 0.66 0.17 0.46 0.49 0.11 0.10 1.00 0.06 0.03 * 0.41 0.08 Note: A big change is defined as a change in the top or bottom 5 percent of changes. Therefore, 10 percent of the changes are big. The number in2the cell is the marginal significance level of the Pearson x test of independence between the row and column variable. Each cell has 3 numbers, corresponding to different sample periods: row 1 whole subsample row 2 first subsample, determined by row variable row 3 second subsample, determined by row variable Note, the German/Japan pair does not equal the Japan/German pair in rows 2 and 3 because the breakpoints for the subsamples are different. The first observation of the second subsample is given below the country name; the date corresponds to the breakpoint used in Kahn-Kole. -17- Hakkio 4. Detecting Influential Observations in a Reduced Form of Industrial Production Germany Japan United Kingdom Feb 1991 May 1991 June 1985 5% / 20.4 7% / 2.8 10% / 3.2 July 1984 Feb 1976 Jan 1974 4% / 15.6 6% / 6.0 6% / 12.3 March 1991 June 1989 May 1990 6% / 31.8 11% / 4.3 7% / 5.3 May 1981 June 1975 Sept 1973 8% / 9.5 10% / 8.0 6% / 11.6 June 1984 Oct 1982 Sept 1985 4% / 2.4 3% / 1.3 5% / 2.1 April 1979 April 1989 Jan 1974 1% ../ 1.6 3% /._1=_8 3% / 4.7 Statistic: moneyCook•s distance interest • rates money COVRATIO interest rates money DFBETA interest rates Note: j The first number in each cell is the date of the largest statistic. The pair of numbers on the second line give the fraction of observations greater than the critical value (the number to the left of /) and the size of the statistic relative to the critical value (the number to the right of / ) . -18- Chart 1 Sum of Lag Coefficients - Japan 15 15 M2 H 10 10 H5 -5 -10 -10 I I I I l l 1878-1 1979-1 196*1 1991*1 1912:1 1993*1 -15 1994*1 1995:1 199*1 l l I l l I 1997:1 1999*1 1999:1 1990:1 1991*1 1992:1 -15 Year 0.06 0.06 Interbank Interest Rate 0.04 h 0.04 0.02 h 0.02 -0.02 -0.02 -0.04 -0.04 -0 06 -0.06 Chart 2 Sum of Lag Coefficients - Germany M3 H2 -2 h J 197T1 tfTKI L HT«1 Itaaci l I 1WV.1 IMtl l 1W*1 i l l 1«MC1 1«Mct 1MK1 t i 1 I I I 1WK1 1WT1 IMMC1 tMVI t«t£1 Year 0.04 0.04 Interbank interest Rate 0 02 H 0.02 -0.02 -0.02 J -0.04 tITM t»7fl t«71-1 L ItN-l J IM1 t tMBM INTI J L 10««'1 IMf:f Year 1MT1 tttT"! I J9W1 I tNTf ' ItNtl ' ' IHt 1 t»S2:f -0.04 Chart 3 Sum of Lag Coefficients - The United Kingdom 15 15 10 H 10 -5 -5 -10 -10 t ! -15 i l l J I L ' I ! -15 1977:1 1978:1 1979:1 1980:1 1981:1 1982:1 1983:1 1984:1 1985:1 1986:1 1987:1 1988:1 1989:1 1990:1 1991:1 1992:1 Year .015 0.015 Interbank interest Rate 0.01 H 0.01 h t \ i .005 hh t H 0.005 A A */ J A V /w / / x \ \ r v t 005 VY v * hh v \/ ' --V/ H J » / 1 r l .015 V h -J I L_ 1 1 ! 1 1 1 1 1- -L ' 1 /V/' \ /' V 0.01 i 1 . *' / ' V ' H -0 01 ' ! I -0.015 1977:1 1978:1 1979:1 1980:1 1981:1 1982:1 1983:1 1984:1 1985:1 1986:1 19871 1988:1 1989:1 1990:1 1991:1 1992:1 Year -0.005 ANOTHER HOLE IN THE OZONE LAYER: CHANGES IN FOMC OPERATING PROCEDURE AND THE TERN STRUCTURE William Roberds, David Runkle, and Charles H. Whiteman1 To economists schooled in the Walrasian tradition, there could be no more enigmatic ritual than the practice of central banking. The classical models of this tradition show how Pareto-optimal allocations can be realized in competitive equilibrium, in which prices reflect the fundamentals of tastes, technology, and endowments. In response to changes in these fundamentals, prices must also change if competitive allocations are to remain optimal. By contrast, since 1914 the unifying theme of real-world monetary policy has been the elimination of shortrun movements in short-term interest rates, typically via open market operations in government securities. The obvious implication of this practice, which has become known as interest-rate "smoothing," is that central banks (and their sponsoring governments) find such fluctuations in the time price of money to be inherently undesirable. This apparent contradiction between high theory and everyday practice has hardly gone unnoticed by the economics profession. In fact, this contradiction has formed one of the traditional jumping-off points for much of the monetarist and neoclassical criticism of the policies of the Fed and other central banks. Yet the constant criticism of interest-rate "smoothing" from this quarter seems to have had almost no effect on the practice of monetary policy. A recent survey of operating procedures in five major industrialized countries (Batten et al., 1990) found that short-term control of interest rates was tightened in virtually all of these countries during the 1980s. Perhaps in response to the continued popularity of interest-rate smoothing, a number of papers have appeared in the macroeconomics literature, in which economists working in the Walrasian tradition*have taken a more benign view of interest-rate smoothing. These papers run the gamut from Sargent and Wallace (1982), which presents a model where 1. William Roberds is on the staff of the Federal Reserve Bank of Atlanta. David Runkle is on the staff of the Federal Reserve Bank of Minneapolis and on the faculty at the University of Minnesota. Charles H. Whiteman is on the faculty at the University of Iowa. We thank David Stowe for help in data collection. This paper was written while Whiteman was a visiting scholar at the Federal Reserve Bank of Atlanta; it reflects only the views of the authors, and not the Federal Reserve Banks of Atlanta and Minneapolis or the Federal Reserve System. Roberds, Runkle, and Whiteman the best monetary policy is an interest rate peg, to Poole (1991), which suggests that the practice of interest-rate smoothing could in some instances serve as a potentially useful method for communicating a central bank's intentions to the public. While these papers have offered a number of insightful explanations for the smoothing phenomenon, it is fair to say that none of the explanations has been widely accepted within the economics profession. Instead, various factions within the profession have supported disparate views of smoothing, reflecting the more general professional quandary over the proper role for money in macroeconomic models.2 At the same time, there has been some acceptance on the part of the Federal Reserve System of the idea that it is possible to be too aggressive in the smoothing of short-term interest rates. In particular, the operating procedure of the late 1970s, which was almost completely focused on smoothing the Federal Funds rate, is typically viewed as a mistake. In an article in the Federal Reserve Bulletin, Heller (1988, p.425) notes that the emphasis on the funds rate contributed to a loss of control over the growth of the monetary aggregates. Essentially identical sentiments are voiced in a later issue of the Bulletin, in an article by Donald Kohn (1990, pp. 4-5). In a New York Fed-sponsored survey of various Fed approaches to open market operations, Meulendyke (1990) observes that during the late 1970s, "[Fed funds] rate moves during the week were so limited that they provided little or no information about reserve availability or market forces." This combination of a desire on the part of policymakers to avoid past mistakes, together with the ongoing professional impasse over the conduct of monetary policy appears to have led to an "eclectic" or "compromise" approach to the day-to-day conduct of open market operations. The essence of this approach is perhaps best summarized in the survey of Batten et al. (1990, pp.30-32). Describing the general approach adopted during the 1980s, Batten et al. note that "operating procedures in [the U.S. and other major industrialized countries] generally allow short-term rates to be primarily marketdetermined, while at the same time, permit monetary authorities to limit the range within which these rates fluctuate. Each monetary authority sees the need for interest rates to adjust expeditiously to reflect new economic developments but also recognizes the importance of maintaining 2. See Goodfriend (1991) for a survey. -2- Roberdsf Runkle, and Whiteman some discretionary control over interest rate movements to avoid excessive volatility." [pp. 30-31] An inherent limitation of this approach is that, lacking any real theoretical guidance, it provides no specific definition as to what level of interest rate volatility is "excessive." Without some specific criteria that define a well-functioning credit market, the "avoidance of excessive volatility" in short-term interest rates cannot be construed as a meaningful objective for monetary policy. One reasonably objective and recently popular metric for evaluating the impact of interest-rate smoothing on the bond markets has been to compare the information content of the term structure, particularly at horizons of less than a year, across periods of time associated with different regimes for monetary policy. The intuition behind the use of the term structure for this purpose is fairly simple. If interest rates are to "adjust expeditiously to reflect new economic developments," then the spreads between long and short rates should contain some useful information about the future course of interest rates. This is because interest rates represent intertemporal prices, prices one would expect to be affected by news about the likely future course of the economy. One of the most widely cited papers in this area is by Mankiw and Miron (1986), who consider the performance of the short end (less than one year) of the term structure over various periods ranging from 1890 to 1979. They find that the term structure was more informative prior to the founding of the Fed. They hypothesize that this result is due to interest-rate smoothing activities on the part of the Fed after 1914. The salient claim of their paper is that there exists a tradeoff between the desire to smooth interest rates on the one hand, and the informativeness of the term structure on the other. Other papers in this tradition include Cook and Hahn (1990), Hardouvelis (1988), Mankiw, Miron, and Weil (1987), and Simon (1990). In what follows, we seek to apply the term structure yardstick to the Fed operating procedure that has been in place since early 1984, technically known as borrowed-reserves targeting with contemporaneous reserve accounting. Specifically, we are interested in the ability of the term structure in the Fed funds market to predict subsequent moves in Fed funds rates, at horizons ranging from one to six months. We also try to measure the information content of the spreads between Fed funds rates and closely related rates on Treasury bills and repurchase agreements (repos). We are especially interested in comparing the term structure during the current operating procedure to the term structure -3- Roberdsf Runkle, and Whiteman under other recent operating procedures* Our results should be of interest to policymakers, given the widespread acceptance of the idea that successful monetary policy should not incorporate interest-rate smoothing to the same extent as was the case during the late 1970s. Our results should also be of interest to monetary theorists, as the set of stylized facts presented below presents a challenge to any theory that would attempt to explain the interaction between a central bank's open market operations and the information contained in the term structure. Our study differs from previous studies in this area primarily in that we make use of daily data on yields for Fed funds and related markets. Previous studies that have attempted to measure the impact of interest-rate smoothing on the term structure have made use of weekly or lower frequency data. Since a major emphasis of the Fed's open market operations has traditionally been the smoothing of day-to-day interest rate changes, the use of daily data is necessary to fully capture the dynamics of the yield curve. INSTITUTIONAL BACKGROUND Although the smoothing of short-term interest rates has always been an important component of Federal Reserve policy, this practice reached a new stage during the 1970s. The development of the overnight market for bank reserves, popularly known as "Fed funds" provided the Fed with an efficient vehicle for large, frequent, short-lived interventions in this market.3 Particularly during the latter half of the 1970s, short-run Fed policy came to focus almost exclusively on the funds rate target. From October 1979 through October 1982, nonborrowed reserves (bank reserves not borrowed from the Fed) replaced the funds rate as the official short-run operating target. This change in operating targets was accompanied by a marked increase in the volatility of the funds rate. As is discussed in further detail below, the standard deviation of daily changes in the funds rate increased roughly threefold. Despite this degree of volatility, however, it is doubtful that fluctuations in the funds rate were completely ignored during this time period. A recent study by Cook (1989), found that despite the nominal adoption of the nonborrowed reserves target, two-thirds of the variation in the funds rate during the October 1979-October 1982 period can be directly attributed to policy actions on the part of the Fed; that is, these 3. Goodfriend and Whelpley (1986) present a useful historical summary of the Federal funds market. -4- Roberds, Runkle, and Whiteman movements in the funds rate were not necessary to meet the nonborrowed reserves target. In October 1982, the Fed's short-term operating target was changed from nonborrowed reserves to borrowed reserves. Despite the continued nominal use of a reserves target, this change has been widely perceived (e.g., by Friedman, 1988) as a retreat towards the funds ratetargeting of the 1970s. Statistical comparisons of the two periods are somewhat problematic due to a change in reserve accounting that was instituted by the Fed in early 1984. Under the pre-1984 accounting procedures (commonly referred to as "lagged reserves accounting") required reserves were computed over a week-long computation period, and had to be maintained with a two-week lag. Under the post-1983 accounting procedures (commonly known as "contemporaneous reserves accounting"), required reserves are computed over a two-week period, and must be maintained with a two-day lag. A characteristic feature of the new accounting procedure has been the introduction of an occasional spike in the overnight funds rate on alternate Wednesdays, i.e., the last day of the two-week reserve maintenance period.4 STATISTICAL FINDINGS Data and Suaaary Statistics In what follows, we use daily data on Fed funds, repo, and T-bill rates at maturities of one day, 30 days, 90 days, and 180 days to examine the predictions of yield spread about future movements in short term interest rates.5 The sample starts in the fall of 1974 and runs until the summer of 1991. Within that sample, we examine three of the different operating regimes:6 the Fed-funds targeting regime (using a sample from January 2, 1975 to October 3, 1979); the nonborrowed 4. For a more detailed description of the mechanics and implications of the change in reserve accounting procedures, see Goodfriend (1984). Additional details concerning recent approaches to open market policy can be found in Meulendyke (1989,1990) and Heller (1988). 5. Repurchase agreements, or repos, are short-term loans collateralized by a fixed-income security. For more information on repos, see Lumpkin (1986) or Stigum (1989). One problem in comparing repo rates is that different rates can be quoted for repos using different types of collateral. To minimize this problem, we look at data specifically for repos that are collateralized by Treasury securities. 6. We exclude the borrowed reserves targeting-lagged reserves accounting regime (October 1982-January 1984) because there are too few observation to conduct meaningful inference about the term structure of interest rates. -5- Roberds, Runkle, and Whiteman reserves targeting regime (using a sample from October 11, 1979 to October 6, 1982); and the present regime-borrowed reserves targeting with contemporaneous reserve accounting (using a sample from February 2, 1984 to July 24, 1991.) The overnight Federal funds rate we use is the effective Federal funds rate computed by the Federal Reserve Board, which is a transaction-weighted average. All other data for Federal funds rates and repurchase agreement rates represent the daily closing quotes from the Bank of America at 5:00 p.m. Eastern Time. The repurchase agreement quotes are for transactions collateralized by Treasury securities. Since both Fed funds and repo rates are originally stated on a 360-day basis, they are all converted to bond-equivalent yields for comparison with other data. Data for one-month, three-month, and six-month7 Treasury bill rates come from the Federal Reserve Board. These data are stated as discounts for an average of bid quotations for the most recent issue, and are also converted to bond-equivalent yields. A brief description of our dataset is presented in Table 1. Summary statistics for daily changes in the various interest rates during the four operating regimes are summarized in Tables 2 and 3, and Figures 1-5.8 The results in Table 2 reveal four characteristics of fluctuations in the rates. First, the sample standard deviations of daily changes in the rates document the well-known increased volatility in interest rates across all maturities during the nonborrowed reserves targeting period. For example, the standard deviation of the daily change in the effective Federal funds rate FFEY increased from 0.301 (30.1 basis points) during the funds targeting period to 0.823 during the nonborrowed reserves period. The volatility dropped markedly after 1982, to 0.316 in the borrowed reserves-lagged accounting period, and 0.378 during the most recent contemporaneous accounting period. Second, the higher-order moments summarized in the skewness (Sk) and kurtosis (Ku) measures are consistent with the view that while volatility increased during the nonborrowed reserves period, outliers were less important. With few exceptions, Sk and Ku are smaller during 7. Since bills are auctioned only once weekly, their maturities fluctuate on a periodic basis. For example, a "91-Day" T-bill typically has a maturity of 91 days on its issue date (Thursday), a 90-day maturity on the following day, etc. 8. The complete set of tables of summary statistics (112 pages in 16.67 pitch type) is available from the first author for a nominal fee to cover reproduction and postage. -6- Roberds, Runkle, and Whiteman 1979-82 than in the funds-rate targeting regime or the borrowed-reserve contemporaneous-reserve-accounting regime* The similarity between the higher moment measures pre-1979 and post-1984 provides further evidence that these periods were alike, and highlights the difference between the practice of permitting only infrequent changes in the Federal funds target during 1975-79 and post-1984 and the "practice" of permitting the Funds "target" to change daily during 1979-82. Third, except for 1979-82, there is a tendency for the volatility in daily changes to fall with the maturity of the underlying contract. For example, in the 1984-91 period, the standard deviations of overnight, thirty-, sixty-, ninety-, and one hundred eighty-day Federal funds rates were 0.378, 0.151, 0.136, 0.151, and 0.16. During 1979-82, rates on daily contracts were more volatile than those on longer contracts, but otherwise the volatility-maturity relationship seems absent. Fourth, the Federal funds rate tends to be more volatile than the repo rate at each maturity. An exception to this is the 1979-82 period, when the two rates were about equally volatile. The composite statistics of Table 2 conceal large interday differences displayed in Tables 3 and 4. Table 3 presents statistics by day of the Federal Reserve maintenance period for overnight Federal funds, and shows characteristics shared by the other overnight rates and to a lesser extent by rates on weekly contracts; Table 4 presents statistics for 3-month T-bills, and shows the characteristic pattern of other medium-term rates. Table 3 displays the striking effects on short-term rates of reserve requirements. Even during 1979-82, volatility in daily changes is noticeably higher on the first and last days of the settlement period. Before 1979 and after 1982, the differences are quite large: pre-1979, volatility increases by a more than factor of four from Monday (day 1 of the five-day settlement period) to settlement Wednesday. From 1988 on, volatility increases by a factor of greater than six between nonsettlement Wednesday (day 5) and settlement Wednesday (day 10) and the following Thursday (day 1 ) . Table 3 also hints that the large movements on Wednesdays and Thursdays tend to be offsetting. Skewness switches sign between these days, as does the midpoint between the minimum and maximum values. The fat tails and otherwise odd features of the distribution of daily changes is apparent in Figures 1-5. Figure 1 displays the distributions of daily changes for the four operating regimes, and illustrates clearly the leptokurtic nature of the distribution of -7- Roberds, Runkle, and Whiteman changes during the nonborrowed reserves period. The remaining figures illustrate the distributions by selected days of the settlement period, and indicate the generally fatter tails which occur on settlement day. The results in Table 4 indicate that the interday pattern of daily rates does not extend to rates for longer horizons. Volatilities across days are quite similar, and the distributions more nearly symmetric. Furthermore, the results differ little across operating regimes—a characteristic not shared by the term structure restrictions investigated below. Tests of the Tern Structure Restrictions The most direct method for testing the implications of the expectations theory of the term structure is the so-called VAR (vector autoregressive) approach.9 While the results of the VAR tests are less easily interpreted than those of the other tests presented below, they do provide summary measures of the overall validity of the expectations model of the term structure over the various policy regimes. To describe these tests, let R^ denote a longer-term, n-period rate of interest, and let RVB denote a shorter, m-period rate of interest, where a divides n. The risk-adjusted expectations hypothesis then states that the n-period interest rate at time t, R^ is the average of the current m-period interest rate R^, and current expectations about future m-period rates, plus a time-invariant risk premium; i.e., (1) R^ « (l/k)^:i tRt+H^+c, k = n/m, where tRt+krin is the expectation at time t of the m-period interest rate starting in period t+k. In the textbook case, the {R^} and {R^} processes are jointly Gaussian and covariance stationary.10 Then it is a straightforward, though some what tedious exercise to apply the standard techniques of rational expectations to derive the implications of equation (1) on the fundaunental moving average representation for the U\», Km)} process. 9. The approach is discussed in more detail in Campbell and Shiller (1987,1991) and Hodrick (1991). 10. These processes must satisfy other technical requirements in order to apply standard rational expectations methodology. See Hansen and Sargent (1991) for a thorough discussion of the econometric issues associated with tests of the expectations model of the term structure. -8- Roberds, Runkle, and Whiteman We adopt a number of modifications on this basic strategy, following the approach taken by Campbell and Shiller (1987). First, equation <1) is approximated by assuming that n/m is large, and taking an infinite horizon counterpart, i.e., (2) R ^ « (1-5) ST.* 6> ^ ^ -he where 6 is a discount factor in (0,1). This modification allows the restrictions imposed by the expectations model to be expressed in a linear form, which reduces computational complexity* Second, in this application we assume ««1 day, that is, the short rate is taken to be an overnight rate. Third, due to evidence in favor of differencestationarity of the various interest rate processes, it is advantageous to rewrite (2) as (3) R^ - R ^ H S*-> = Zmiml 6l (tRt+nri4--Rt+«(M),») +C Equation (3) states that the spread between the long rate and the short rate, S ^ must equal a discounted sum of expected future changes in the short rate. The fourth and final modification is to assume that the bivariate, stationary process (R^-R^, Ru-Rt-u) h a B a V A R representation. Under these modifications, the implications of the expectations model of the term structure can be shown to be equivalent to a set of linear restrictions on the coefficients of the VAR for (R^-R^, RU-RM.I). 11 Representative results for these tests are shown in Table 5. In these applications, the short interest rate was taken to be the effective overnight Fed funds rate, and the long rate was taken to be the 3-month Treasury bill rate. A VAR model was fit to daily observations on first differences in the overnight funds rate and the spread between the T-bill rate and the funds rate. Missing observations were filled in by repeating the previous day's values. Standard tests for lag length revealed that 21 lags were sufficient to capture the model dynamics after October 1979. For the funds rate targeting period, however, these tests were somewhat ambiguous. Hence, for this period, Table 4 presents results for a 42-lag VAR system as well as for a 21-lag system. 11. See Campbell and Shiller (1987, pp. 1066-1068). Following Campbell and Shiller, the expectations-model restrictions are tested using a Wald test. The variance-covariance matrix of the coefficients is the heteroskedasticity-consistent estimator suggested by Hansen (1982). -9- Roberds, Runkle, and Whiteman The results in Table 5 show that the expectations model can be rejected at arbitrary significance levels for the funds rate targeting period. The expectations model cannot be rejected on the basis of the data from the 1979-82 period, although the smaller sample size associated with this period makes this finding somewhat less informative than might be the case otherwise. The test results for the post-1984 sample fall in an intermediate range: the expectations model can be rejected at the 5% but not at the 1% level of significance* The VAR results accord with other studies of the short end of the term structure that report subsample results for different Fed operating procedures, e.g., Simon (1990) and Hardouvelis (1988). They indicate that the expectations model can be taken literally only over the relatively short subsample associated with nonborrowed reserves targeting, that it is unlikely that much information can be recovered from the short end of the yield curve during the late 1970s, and that some information may be present in the yield curve after 1984. Information in Spreads Although the testable implications of the expectations theory of the term structure are rejected for two of the subsamples by the using a VAR model, that rejection does not necessarily mean that there is no information in the term structure. The expectations theory implies that current spreads between interest rates at different maturities predict future interest rate changes. This implication of the expectations theory warrants separate examination. Campbell and Shiller (1991) show how under the expectations hypothesis, yield spreads can be used to predict changes in both shortand long-term interest rates. To use the hypothesis to predict short rates, follow the approach of the previous section by subtracting R ^ from both sides of (1) and reverse sides, giving (4) (l/k)2j:i tRt+ni,m - R ^ = Run-R^n + c, k = n/m. The right-hand side of (4) is just the current spread between n- and •period interest rates. Equation (2) thus suggests that the difference between the average expected m-period rate and the current m-period rate is equal to the current spread between n- and m-period rates plus a risk premium. Equation (4) can be tested by regressing the realized difference between average m-period rate and the current m-period rate, (l/k)2^;i Rt+mMn - R^B S St(n-m)" on the current spread, R^-R^ = SfK The expecta- -10- Roberds, Runkle, and Whiteman tions theory implies that the coefficient on the current spread should be unity. Thus, the current spread should be a good predictor of the future average change in short-term rates. Campbell and Shiller also consider the implications of equation (1) for changes in future long-term rates. They note that (5) sj*»> s (m/(n-m))S?^> * ^^-K^. This implication of the expectations hypothesis can be tested by regressing the realized value Rt+BMMIrRtta on s ^ . The expectations theory predicts that the estimated coefficient on s ^ will be unity. Thus, a known multiple of the current spread should be a good predictor of the future change in long-term rates. Campbell and Shiller test both (4) and (5) using Mcculloch's (1990) monthly data on U.S. Treasury bill, note, and bond prices from 1952:1 to 1987:2. Their analysis is especially complete: they look at all possible combinations of short and long maturities that are multiples of each other from one month to ten years. By conducting such an exhaustive analysis, Campbell and Shiller are able to pinpoint those maturity combinations for which the expectations theory of the term structure works well, as well as those combinations for which it works poorly. One of the Campbell-Shiller findings is that for any two maturities, n and a, equation (5) performs abysmally. That is, the current spread between n- and a-period rates has no power in predicting the difference between the (n-a)-period rate a-periods from now and the current n-period rate. In fact, equation (5) performs so poorly that the coefficient on s ^ is usually negative, while the expectations theory predicts a value of one for that coefficient. The Campbell-Shiller estimates of equation (4) are somewhat more promising for the expectations theory. For maturities beyond three or four years, they cannot reject the hypothesis that the coefficient on This means that the current spread between n- and S(iwm) ^g unity. m-period rates predicts how the average m-period rate will change over the next n-periods.12 But for shorter maturities, especially those below one year, Campbell and Shiller's tests reject the hypothesis that the coefficient on S ^ is unity. Their results are consistent with 12. Or to be more precise, it says how the average a-period interest rate every a periods from the current period to the n-ath period will change from the current a-period interest rate. -11- Roberds, Runkle, and Whiteman earlier research by Shiller, Campbell, and Schoenholtz (1983), Fama (1984), and Mishkin (1988). Although the Campbell-Shiller results are useful for analyzing the predictive power of the yield spread over the entire post-Treasuryaccord period, they do not tell us much about how different Federal Reserve operating procedures have affected the term structure. There are two ways in which we believe the Campbell-Shiller results must be extended to understand the effect of different operating procedures. First, we must examine the predictive power of yield spreads during each different operating regime, since the amount of information contained in the spread could differ greatly across the different regimes. Second, we must examine the predictive power of yield spreads for each different day of the maintenance period for the different operating regimes, since operating procedures and volatilities vary greatly by day of the maintenance period. Since the Campbell-Shiller results show that there is almost no hope for equation (5), we concentrate our efforts on equation (4). We want to see whether differences in operating procedures can explain the Campbell-Shiller finding that average future short-term interest rates do not change as much from the current short term rate as the current yield spread predicts that they will. To do this, we estimated ex post versions of equation (4) using data consolidated according to operating regime, as well as broken out by day of the settlement period within each regime.13 Results for term Fed funds rates under the non-borrowed and borrowed reserves operating regimes are presented in Table 6. In the table, three characteristics of the term structure emerge. First, under the current operating regime, the short end of the term structure displays the characteristic pattern found by Campbell-Shiller for intrayear rates: the bias of the term structure forecast increases with the maturity of both the long and short term rates. Second, under the nonborrowed reserves targeting regime, the short end of the term structure was substantially more informative about movements in future 13. Because we use daily data, the errors in our term-structure regressions are serially correlated, for reasons noted by Hansen and Hodrick (1980). We correct for both serial correlation and conditional heteroskedasticity using the methods suggested by Hansen (1982). Missing observations are dealt with in the following manner. We repeat missing observations in order to calculate the forward averages on the LHS of (4). However, these repeated observations are not used to calculate the regression results in Tables 6-10. This procedure de facto extends the maturity of pre-holiday short rates by one day. -12- Roberds, Runkle, and Whiteman short rates. With the exception of the overnight—30-day connection, each of the slope coefficients in the nonborrowed reserves table is within one standard error of unity, the value predicted by the expectations hypothesis. Third, the fraction of the variation in the spread between average future short rates and the current short rate which can be explained by the current long-short spread is much lower at the shortest end of the term structure during the nonborrowed reserves period. For example, the R2 in the spread regression using the overnight—60-day spread was 0.45 after 1984, but only 0*16 between 1979 and 1982. However, this deterioration in quality of fit does not characterize the longer end of the term structure—primarily because so little of the variation is explained even in the best cases. For the borrowed-reserves/contemporaneous-reserves accounting regime, the results for estimating (4) using data on repo rates are very similar to those using Fed funds data, as can be seen in Table 7. There are only two important differences between the repo results and those for Fed funds: First, the slope coefficients are somewhat higher using repo data if the long-rate maturity is 90 days or less. Second, the slope coefficients are actually negative if the long-rate maturity is 180 days and the short-rate maturity is 30 days or more. For the results for estimating (4) using repo data are very different from those using Fed funds data for the nonborrowed-reserve targeting regime, as can be seen in Table 7. As noted above, the slope coefficients in the Fed funds regressions were all within one standard error of unity, except with a short-rate maturity of one day and a longrate maturity of 30 days. With the repo data, only two of the slope coefficients are within two standard errors of unity. The relatively poor performance of the term-structure regressions on repo data in the nonborrowed-reserve targeting period may well be explained by institutional factors affecting the repo market during this time. Before the fall of 1982, the repo market was quite immature and contained many legal uncertainties. One indication of the repo market's immaturity is the fact that most trades during this period were done at 1/4 percentage point increments. Fed funds were trading on 1/16 or finer percentage point increments. Because Fed funds rates had a higher resolution than repo rates, the term structure of repo rates could not contain as much information as the term structure of Fed funds. Other institutional issues besides market immaturity may also explain the poor performance of these regressions. Until 1982, courts did not decide who actually owned the pledged securities if a broker -13- Roberds, Runkle, and Whiteman went bankrupt. Also, pricing by dealers was not uniform until October 1982, when the New York Fed required repo pricing to be based on accrued interest on the pledged securities. This ruling came after abuses of alternative pricing mechanisms lead to nearly $300 million in losses to Chase Manhattan when Drysdale Government Securities collapsed. Since term Fed funds data are not available for the funds-rate targeting period, Table 8 replicates some of the results in Table 6, using observations on the overnight funds rate and the 1, 3, and 6-month T-bill rates. The Table 8 results for the nonborrowed and borrowed reserves regimes are generally not as favorable to the expectations hypothesis as the analogous results in Table 6. We speculate that this deterioration in the fit of the expectations model is driven by the existence of a secondary market for T-bills that does not exist for term Fed funds. The existence of a secondary market implies that the price of T-bills should reflect the value of this "put-option" feature. It also seems likely that the value of this feature of T-bills would incorporate factors other than the conditional first moment of the short-term interest rate. The Table 8 results also differ from the analogous figures in Table 6 in terms of the patterns displayed by some of the statistics over the various maturities. For both the borrowed and nonborrowed reserves subsamples, the bias of the term structure is less at 91 days than at 28 or 182 days. However, the two subsamples still differ substantially in terms of the bias and fit of the term structure equations. The results for the nonborrowed reserves subsample still dominate those for the borrowed reserves subsample in terms of bias, i.e., the slope coefficients are closer to unity. At a horizon of 182 days, the R2 statistics are still larger for the nonborrowed reserves period. At the shorter horizons, the R2 statistics are roughly the same for both periods. Results for the funds-rate targeting period are also shown in Table 8. For the funds-rate targeting regime, the information content of interest-rate spreads appears to be uniformly low, as evidenced by the very low R2's obtained for equation (4). The slope coefficients are also generally quite small, and in most cases are within a standard error of zero. The exception is the slope coefficient on the 6-month/ 3-month T-bill spread, which is greater than the analogous estimates for the borrowed- and nonborrowed-reserves regimes, though it still falls well within two standard errors of zero. -14- Roberds, Runkle, and Whiteman Periodicity and Information As noted above, since 1984 the overnight funds rate has displayed a marked periodic pattern over the two-week reserve maintenance period, while these same periodic patterns are absent for term Fed funds rates. To investigate the effect of this periodicity on estimates of equation (4), we estimated versions of equation (4) over subsamples that consist of observations on particular days of the reserve maintenance period* Representative results from this exercise are displayed in Tables 9 and 10. Table 9 shows that the choice of subsample makes a tremendous difference in the fit of equation (4), when the short rate is the overnight funds rate and the time period considered is the post-1984 borrowed-reserves/contemporaneous-reserve accounting regime. In these cases, i.e., in the first column of Table 9, the results for settlement Wednesdays display markedly higher values for both the estimate of the slope coefficient and for R2. This "settlement-day" effect is more muted or virtually nonexistent for short rates having a maturity of 30 days or more. It is also much harder to detect for the 1979-82 period, in that the results for Wednesdays (which were all settlement days) are extremely close to the results for the period as a whole (cf. Tables 6 and 9). Table 10 replicates the results of Table 9 for the repo market. The post-1984 results follow essentially the same pattern as for Fed funds, i.e., of better fits for the overnight rates on settlement Wednesdays, worse fits on non-settlement days, and few differences otherwise. Note that all of the slope coefficients when comparing overnight repos to longer-maturity repos are within two standard errors of unity. The 1979-82 results for Wednesdays differ little from the results over the entire 1979-82 sample (cf. Table 7). INTERPRETATIONS AND FINDINGS The results reported in Tables 1-10 are entirely consistent with the idea advanced by Mankiw and Miron (1986) that the information content of the term structure is strongly linked to the volatility of short-term interest rates. This effect shows up in two ways in our results. First, the estimates of the slope coefficient for equation (4) in Tables 6-10 are generally larger for the volatile 1979-82 period than is the case for the other subsamples. Second, both higher slope coefficient estimates and better fits are obtained for the relatively volatile subsample of settlement Wednesdays during the post-1984 period. The first of these two observations should be uncontroversial, as it has -15- Roberds, Runkle, and Whiteman already been reported by earlier studies, notably Hardouvelis (1988) and Simon (1990). To our knowledge, the second observation is unique to the present paper and thus merits additional discussion. From the standpoint of policy and evaluation of the current Fed operating procedure, the key question is "does the current operating procedure allow some information about short-term interest rates to be reflected in the term structure?" The answer to this question provided by Tables 5-10 is an unambiguous "yes, but ... •" The post-1984 results for equation (4) are certainly more favorable than the available 1975-79 results to the idea that rate spreads contain information about the future course of short-term rates. The down side of this generalization is that much of this information is of a limited, short-term nature. Tables 9 and 10 indicate that after 1984, equation (4) fits best for a particularly volatile subsample of our dataset, i.e., on settlement Wednesdays when the short rate is an overnight rate.14 This pattern of results is consistent with the widely held notion that while the Fed may loosen its grip on the overnight funds rate on settlement Wednesdays, the day following settlement will generally see the return of the overnight funds rate to previous target value. Our settlement-day results might be considered encouraging in the sense that it shows that the markets are not "spooked" by settlement-day pressures in the overnight Fed funds market. On the other hand, the value of this type of information is likely to be nil at other than very short horizons. Inspection of various entries of Tables 6-10 shows that such is in fact the case. For example, the second column of Tables 6,7,9, and 10 shows that in the post-1984 period, the 30-day/60-day and 30-day/ 90-day spreads do contain some information for future movements in 30-day rates. However, the 30-day/180-day spreads on Fed funds and repos do not have any forecasting power for future movements in 30-day rates. Similarly, the 60-day/180-day spreads on Fed funds and repos never provide information on the future course of 60-day rates. The same is true for the 90-day/180-day spreads on Fed funds and T-bills. In the case of repos, the 90-day/180-day spread does provide some information on future 90-day repo rates, but the sign of the slope 14. Since 1984, the overnight funds rate has also tended to be quite volatile around year-end, due to holiday cash demand and "window-dressing" pressures. Point estimates very similar to those obtained for settlement Wednesdays were obtained for a post-1984 subsample consisting of the two-week periods beginning on a Thursday and spanning the Christmas and New Year's holidays. -16- Roberds, Runkle, and Whiteman coefficient is the opposite of that predicted by the expectations hypothesis and the amount of variation explained is quite small. These last results suggest that despite the nominal distinctions between the post-1984 operating procedure and the funds-rate targeting regime of the late 1970s, relatively little information is being captured at the short end of the term structure. We find that what information available in the short end of the term structure vanishes at a horizon somewhere between 90 and 180 days, a finding consistent with the results of Hardouvelis (1988), whose data set extended only to 1985. That is, the net position of the markets, as reflected by the term structure, cannot be interpreted as having any predictive power beyond a horizon of roughly 90 days. To obtain a better idea of how this finding impacts on the market's expectations of monetary policy, we make use of an idea suggested by Simon (1991). Using 1983-88 data on 30- and 60-day term Fed funds rates, Simon (1991, p.334) finds that a version of equation (4) fits particularly well during the days immediately preceding and following FOMC meetings. Simon interprets this finding as supporting the notion that the policy intentions of the FOMC are quickly transmitted to financial markets. To implement this idea for our data set, we fit equation (4) to both Fed funds and repo data after 1984, restricting ourselves to the days immediately following FOMC meetings (or the second day of the meeting for 2-day meetings). These results are displayed in Tables 11 and 12, along with the analogous results for the 1979-82 period. In general the post-1984 results in Tables 11 and 12 do not differ radically from those reported in Tables 6 and 7. This is particularly true for the Fed funds market (cf. Tables 6 and 11). For repos, there is a somewhat better fit immediately post-FOMC for versions of equation (4) where the short rate is the overnight rate, or where the long rate has a horizon of 30, 60, or 90 days (cf. Tables 7 and 12). At a horizon of 180 days, there is no improvement in fit for the equations with a short rate having terms of 30, 60, or 90 days. These results suggest that interest rate spreads directly attributable to policy actions are not likely to be more informative than is usually the case, especially at horizons beyond 90 days. To summarize, our results indicate that in the current (post-1984) policy environment the information implied by the short end of the term structure vanishes at horizons beyond 90 days. This result is consistent with the Mankiw-Miron hypothesis in the sense that the available evidence from the 1979-82 period (which is necessarily limited -17- Roberds, Runkle, and Whiteman because of the short duration of the nonborrowed-reserves operating procedure) suggests that this was likely not the case when the Fed was less aggressively smoothing the funds rate. The fact that some information is contained in the post-1984 term structure for the very short term is consistent with the Mankiw-Miron hypothesis, in contrast to conjecture of Hardouvelis (1988, p.355). As is documented above, the volatility of the overnight funds rate on reserve settlement days is accompanied by an increase in the informativeness of the term structure. Since longer-term Fed funds and repo rates are generally not subject to the settlement day volatility, the Mankiw-Miron hypothesis would predict that the fit of equation (4) would fall with the maturity of the short rate. This is exactly what happens in the post-1984 sample. Our results are also complementary to those obtained by Campbell and Shiller (1991). Recall that Campbell and Shiller are unable to reject the expectations hypothesis restriction on equation (4) (i.e., that the slope coefficient equals one) when the long rate has a maturity greater than three years. For the post-1984 Fed funds and repo markets, our results imply that interest rate spreads are quite informative at very short horizons, although we can still formally reject the expectations hypothesis, excepting the repo market on settlement Wednesdays. Campbell and Shiller (1991, p.507) also note that for the Treasury market, the forecasting ability of equation (4) falls with the length of the forecasting horizon (the long rate maturity) at horizons of less than one year. We document a similar effect in the post-1984 Fed funds and repo markets. The information content of the yield curve in these markets begins to decline at a horizon of no more than two months, and vanishes at six months. CONCLUDING REMARKS The results discussed above, together with the term structure results obtained by Campbell and Shiller (1991), Fama (1984), and related papers, point to a remarkable empirical regularity associated with the recent U.S. term structure. In terms of the ability of the term structure to predict subsequent movements in short rates via equation (4), there is an "ozone hole" in the term structure beginning at a horizon of roughly six months and extending out to a horizon of two or three years. That is, the ability of the implicit forward rates to anticipate the future course of interest rates is severely curtailed at horizons between 3-6 months and 2-3 years. Our conjecture is that the cause of the "ozone hole" is the Fed's historically accommodative stance towards seasonal fluctuations in the demand for credit. At a horizon of -18- Roberds, Runkle, and Whiteman roughly six months, a policy incorporating seasonal accommodation has to come into conflict with the market-determination of short-term rates. Put another way, no one has yet invented a seasonally adjusted credit market• Without a well-specified model, it is not possible to analyze welfare implications of the results presented above. However, the operating procedure in place since 1984 has been only partially successful in terms of providing information to credit market participants concerning the future course of short-term interest rates. Further, the greatest amount of yield-curve information has been available during episodes associated with higher volatility of the overnight Fed funds rate. Finally, we conjecture that the Fed's historical policy of seasonal accommodation poses an inherent limitation, for better or for worse, on the ability of implicit forward rates to forecast future interest rates at horizons close to one year. -19- Roberds, Runkle, and Whiteman REFERENCES Batten, Dallas S., Blackwell, Michael P., Kim, In-Su, Nocera, Simon E., and Ozeki, Yuzuru. "The Conduct of Monetary Policy in the Major Industrial Countries: Instruments and Operating Procedures." International Monetary Fund Occasional Paper No. 70 (July, 1990). Campbell, John Y. and Shiller, Robert J. "Cointegration and Tests of Present Value Models," Journal of Political Economy, vol. 95 (1987), pp. 1062-88. Campbell, John Y. and Shiller, Robert J. "Yield Spreads and Interest Rate Movements: A Bird's Eye View," Review of Economic Studies, vol. 58 (1991), pp. 495-514. Cook, Timothy and Hahn, Thomas. "Interest Rate Expectations and the Slope of the Money Market Yield Curve," Federal Reserve Bank of Richmond Economic Review, vol. (1990), pp. 3-26. Cook, Timothy. "Determinants of the Federal Funds Rate 1979-82." Federal Reserve Bank of Richmond Economic Review, vol. 75 (January/February 1989), pp. 3-19. Fama, Eugene F. "The Information in the Term Structure," Journal Financial Economics, vol. 13 (1984), pp. 509-28. Friedman, Benjamin M. Journal of Economic of "Lessons on Monetary Policy from the 1980s," Perspectives, vol. 2 (1988), pp. 51-72. Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary Policy," Carnegie-Rochester Series on Public Policy 34 (Spring, 1991), pp. 7-30. Goodfriend, Marvin. "The Promises and Pitfalls of Contemporaneous Reserve Requirements for the Implementation of Monetary Policy." Federal Reserve Bank of Richmond Economic Review, vol. 70 (May 1984), pp. Goodfriend, Marvin and Whelpley, William. in Instruments of the Money Market, -20- "Federal Funds." Chapter 2 6th edition, edited by Timothy Roberds, Runkle, and Whiteman Q. Cook and Dimothy D. Rowe. of Richmond (1986). Richmond, Va.: Federal Reserve Bank Hansen, Lars Peter. "Large Sample Properties of Generalized Method of Moments Estimators," Econometric*, vol. 50 (1982), pp. 1029-54. Hansen, Lars Peter and Hodrick, Robert J. "Forward Exchange Rates as Optimal Predictors of Future Spot Rates: An Econometric Analysis," Journal of Political Economy, vol. 88 (1980), pp. 82953. Hansen, Lars Peter and Sargent, Thomas J. Rational Expectations Econometrics, With contributions by John Heaton, Albert Marcet, and William Roberds, Underground Classics in Economics, Boulder and Oxford: Westview Press, vol. (1991), pp. Hardouvelis, Gikas A. "The Predictive Power of the Term Structure during Recent Monetary Regimes," Journal of Finance, vol. 43 (1988), pp. 339-356. Heller, H. Robert. "Implementing Monetary Policy." Bulletin, vol. 74 (July, 1988), pp. 419-29. Federal Reserve Hodrick, Robert J. "Dividend Yields and Expected Stock Returns: Alternative Procedures for Inference and Measurement." National Bureau of Economic Research Technical Working Paper No. 108 (July, 1991). Kohn, Donald. "Policy Targets and Operating Procedures in the 1990s." Federal Reserve Bulletin, vol. 76 (January, 1990), pp. 1-7. Lumpkin, Stephen A. "Repurchase and Reverse Repurchase Agreements," in Cook, Timothy Q., and Rowe, Timothy D., eds, Instruments of the Money Market, Richmond: Federal Reserve Bank of Richmond (1986). Mankiw, N. Gregory, Miron, Jeffrey A. and Weil, David N. "The Adjustment of Expectations to a Change in Regime: A Study of the Founding of the Federal Reserve," American Economic Review, vol. 77 (1987), pp. 358-374. -21- Roberds, Runkle, and Whiteman Mankiw, N. Gregory and Miron, Jeffrey A. "The Changing Behavior of the Term Structure of Interest Rates,n Quarterly Journal of Economics, vol. 101 (1986), pp. 211-28. McCulloch, J. Houston. "U.S. Government Term Structure Data," in Friedman, Benjamin, and Hahn, Frank, eds., The Handbook of Monetary Economics, Amsterdam: North Holland (1990). Meulendyke, Anne-Marie. U.S. Monetary Policy and Financial New York: Federal Reserve Bank of New York (1989). Markets. Meulendyke, Anne-Marie. "A Review of Federal Reserve Policy Targets and Operating Guides in Recent Decades•" In Intermediate Targets and Indicators for Monetary Policy: A Critical Survey. New York: Federal Reserve Bank of New York (1990). Mishkin, Frederic S. "The Information in the Term Structure: Further Results," Journal of Applied Econometrics, Some vol. 3 (1988), pp. 307-14. Poole, William. "Interest Rates and the Conduct of Monetary Policy: A Comment,** Carnegie-Rochester Conference Series on Public Policy, vol. 34 (1991), pp. 31-41. Sargent, Thomas J. and Wallace, Neil. "The Real-Bills Doctrine versus the Quantity Theory: A Reconsideration," Journal of Political Economy, vol. 90 (1982), pp. 1212-36. Shiller, Robert J., Campbell, John Y. and Schoenholtz, Kermit L. "Forward Rates and Future Policy: Interpreting the Term Structure of Interest Rates," Brookings Papers on Economic Activity, vol. (1983), pp. Simon, David P. "Expectations and the Treasury Bill-Federal Funds Rate Spread over Recent Monetary Policy Regimes," Journal of Finance, vol. 45 (1990), pp. 467-77. Simon, David P. "Secrecy, Signalling and the Accuracy of Expectations during the Borrowed Reserves Operating Regime," Journal of Banking and Finance, vol. 15 (1991), pp. 329-41. -22- Roberds, Runkle, and Whiteman Stigum, Marcia. "The Repo and Reverse Markets," Jones-Irwin (1989). -23- Homewood: Dow Roberds, Runkle, and Whiteman 1. Data Series Abbreviation Series Availability 75: 1: 2 - 91: 7:24 FF30Y FF60Y FF90Y FF180Y Overnight Effective Federal Funds Rate 30-Day Fed Funds Rate 60-Day Fed Funds Rate 90-Day Fed Funds Rate 180-Day Fed Funds Rate 79:11:13 79:11:13 79:11:13 79:11:13 - 91: 91: 91: 91: 7:24 7:24 7:24 7:24 RPY RP30Y RP60Y RP90Y RP180Y Overnight Repo Rate 30-Day Repo Rate 60-Day Repo Rate 90-Day Repo Rate 180-Day Repo Rate 75: 1: 2 79: 8:27 79: 8:27 79: 8:27 79:11:13 - 91: 91: 91: 91: 91: 7:24 7:24 7:24 7:24 7:24 TBI One-month (28 Day)* T-bill Rate Three-month (91 Day)* T-bill Rate Six-month (182 Day)* T-bill Rate 75: 1: 2 - 91: 7:24 FFEY TB3 TB6 75: 1: 2 - 91: 7:24 75: 1: 2 - 91: 7:24 •Maturities of T-bills will fluctuate between auctions. above are for Thursdays. -24- Maturities ro 00"T>-noo"0-»iTJ-ti"0-n"0-Ti oooo o o o o p p -< Mill I I I I I I I I I I • I «00>00>00>0>OvO^O>0>0 Srt ^J# o 7C C i/» O 0) p * • 1 rt c II 2. I A O on <i a. -•» o o v> o> n i T S~ ~ V) <• 7* II i i i • i • i i i i i • i • i i i i i • i i i oo oo oo oo oo oo oo oo oo oo oo oo i i -#. ^J **J "^ "^ tr p> 3 <Q (0 o o> V o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o 20l§i§§lill| l~§°io§l|§l§ 1111 o o o o o o o o o o o • • o O -» • 0 0 ~ » 0 - » - » - » I M W roLLpQ-ip^o'^ww O O O O O O O O O O O O o o o o o o o o o o o o CO ID H» (0 O ft ID oooo iisiiiiiiiii oooo 3SSS o o o o o o o o o o o o o . o o o o . o a o o o-* -» - . _ » _ . ^ r\) ^ J - * oooootO"_ooo40ooouir\3r\) ^8? i i i i • i S « S^OIS. -»-»-* -» ro-» K>-» ru UJ r\* S» N32. VI - • rt ii 2 , ^ a 3* oo-»w O O-^ -» O - » - * - * 1/1 ->| . . . o o W9B-»« • O O O O O ^ I r J O O ooorvi m^jg^'§8S Sgk^kkk^«l IS^lS^SiSlgS ISgss | & 2 55 i f 5 w 3 i S*»«£giS552§3 S ^ k K K ^ S g gSS^gSSSsSS as§§ *1 ON-»C 5*^ o - i n a n oMl i cr ID CD rt Dl ft ID i 5 3 ft ^&S3SS3S2SaS SttggfiiS&gJLg S g ^ S S ' s i ! ^ O OO • O • O « O O W N . . O• O• O• • • • ^ 4 x a © 8*33fcS3SI5l2giB x 9 n H- o o o o o o o-»o**o* . . . . . . o • • • • • -^-T* CM\J-» IVJ • U t O O U J O O O w P> ^ u i o t / t o "" * ~ -H >v r> -H I 00 i * w c i ? oo oo oo oo oo oo oo oo oo oo oo o c o vt i •osssss-^ **-• l<^-<^-<-< So rt - I rt ... O -» rt O O » O II 7%. O O «||SIilli|3i iaiiaiiiiaii SHSHSSSH? nmmm* a^i ? D» 3 mrnmrnrnmrnrnmm rnrnmmrnmmrrirfirn m rn m m i < -1 » < rti m ftl r+i m -<-<-<-<-< 0 IX to O O 00 > | ( M f l M * J N - » oooo-^o^en-r-uiro-* OOOOOOOOOO l*OOOOOo'o-»0 O O O O O O O O O O I M O O O O O O O O O oo*^>oro**uio-^oo 0J 3 O. tt U1*^ U N - * 3 Q » O O O O O fr vn -!>• w r o - » *< O OO OO • • • . • tO - » 0 0 ^ oi/iuJO*otnoo-*o^i ro en en o p ro oo ro ^ ^ ifl^UJOO^UlUIOO^fO S O O S N ro £ 0 1 / 1 0 0 0* o i -» « j -» o •r* kurt 3 ~ 10.281 17.485 11.463 9.725 15.696 rt o utro o r o en oofo o* -» kurt fK W J - - * 00 *"* - * 00 t> •%! 0*0*^J**00*^"OUIUl c 00 Qv N O 0* to 82:10: 6 gouj «-*-» ro 8.072 13.210 25.497 -0.096 2.336 X s o o .r* -»ro 91: 7:2 o^enuiujroooo^joo " 09OOWN a rt O o o <0*^OQ0<r*U«<O<OQ0<0 ->»en-grooo a a 1: 2 to79:10: 3 C o tr 75: O0B fO^J O f O - J - * 0 - » * * ^ w * » mro - * o * -» *» to OCh(hVfltt skew 7t " 3 (0 oo oo o * * - * - » o *» enroroo p 1.998 1.981 0.562 1.067 2.706 C CO *- o o o o - * to <OUIV/IQO •f^OOCKOUl 3 79:10:11 O 0 0 N W ^ - » - » W W O . . . . . . . . . . X 3 IP 0 skew . "^ -* o o o o I ? s V) K- • < W M ^ N - * & < Fr «» & dev ->J^ J*00 -»0»O MO O * -»roc»ro<o *-* uir* o o <* •<-<-<-<-< CT (0 < to -i (0 rt to rt CL 0.644 0.289 0.070 0.498 0.048 _»_»_*cnroo-*->iro-» ">iboV»opouiV*ro"oo» oen-r-ouopoo^o oo->i<o-«j-4-r»UJoo<oui I O O O O O m m m m rtl dev o o o o o o o o o o en-»-»ro-»oororoen ui**Qo-r*<ooooo**Uiro ***»^<ov/iroooo»->jen O O O O O O O O O O CL U l - J IO M _ » - » - * (\J |\> U l Q . 0) -i 3d c P a x Hft N N - » 0 - » 0 0 0 0 - » 3 rO**UJ-*^ ^OOOWOO O O O O - * O C O N O S N N W N ^ <0-r*U1UJUJ <O-V|-NIOOOOQOC*OOOO O en enc* O uioo ->jro ro Ul * * *> .T» Ul woooto i X 3. 2.352 0.365 0.730 0.588 1.592 3 3 -1.501 -0.740 -0.649 -0.720 -3.427 ONN0a0B00 0»00<OO -»NNNNOON(M-* F - « 0 9 O N N t t W O W U1 UJ - * O* O W 00 00 - • -sj 3.934 1.511 2.565 3.001 2.717 r o o o o - » o o o o o go-*>o-»ro-»roroQo<o orJwro^iroui^rooJ — -* o r o - * o b'f O N W -2.342 -1.419 -2.048 -2.028 -2.991 0 0 0 > O O N U I ^ O U 1 - 0 00-»-*ONMW1WOW en0000OO^O00-»^| -» o o -* o o o r o -»ro oooNob»"ouiou»^j Ulttl CD -»oo-»oooro-*ro I roro ro 3* %U i 0009 to P» 3 .£> Q) - I -« - * -H -H < UlUl WW W -I — W O w » < pi 8 P( t D H ft 8 K & ooooo^joui.r>wro-* o o oo->io* ui ^ w ro-* oooooooooo oooooooooo O O O O O O O O O O O O O O O O O O O O 3 B -»-^OOOM-»NOO 3 i n .r> w ro U1-T" W N -* a §o£ ooooo ooooo o o o o o a O O O O O W O W N ? o - * oo w B 5 Ul #"* WfO -» I O O O O O O O O O O 3 V -*ro -*p-» w-» oF-» 3 rt a <t < oooooooooo </» . o o o o o oooooooooo o. o* o o 08 o> -*j ON p* oo * s <r» ^ o # * w 5* o* o i w o r J < oipo ^* o *s w w wroro o o o o o ro ro.r* ro"r* iw r o ro ui ro -» w o •- * r o rt a a. < ooooo a. skew oooooooooo uio^vnsu*^(>^sv/i •r* o OB oo w o o ^ i o ro oooooooooo 2.047 1.023 1.416 1.173 0.119 oooooooooo 8 H i ft» 3 3 ft ft*Q 3*8 O 8 8 3 50 PP 8 3 8 8 8 8 ft f t ri 3* < 8 D 8 8 *0 •< * M — 8 I P-O 3 8 ft*< 8 3 H 8 I 3 0 O P8 P» P» s o o o o o WO*>OtOOO-»WWUl -0.600 -0.330 -0.290 -0.330 -0.290 o o o o o o o o o o wLnuifo'-»wro'-»wro rt O UlUl ^1 ro UlUl rt O * ^ 0*00 W oooooooooo o o o o o o o o o o -*rororo^»'-*ro-*fofo vnww<r*~groo-»wro o o o o o o o o o o oooooooooo o o o o o § roo#* owoirooo» -roo* _» o -» - » o o bo In o po o o o o o max l/1Q\-^lfl-»^O-»000t O O O O O O O O O O 0.230 0.550 0.580 0.650 0.350 O->J-<<JOOOOOOOOOOOO O N N O Q0O0 OKQ* OOO OT €h \jn en 0* ON en oo -viro ro w o o o o o 9> ^ r o w oo oo ro I ff 79:10: 3 roo-soow*»o-» o u i o o ro rt O kurt a O rt O 17.242 6.486 9.541 10.150 5.159 rr rt O —ft 82:10: 6 ro •r> *>* o o - » o - * - * oo r o r o o o o o o V»o u t o o -0.890 -1.400 -1.230 -0.740 -0.600 oooooooooo ro-»-»ro-»-»-*-ftro-» oo^jwwoioo-*-^*" o o o o o o o o o o rt c ro kurt -» W W ^ -T* -*J * * 0 #- O WU1-»-»NO^NO-* o* * * o - * o o o - » O K o - * _»ro-»Ot-*-»00U4^4Ot ro 1.358 2.613 1.174 1.111 2.682 rvi _* ro -» -» o - * o ro w ?c -» oo o -r* o * * o - * v i - * - i *- 84: 2: 1 o o -» O - * O O O O - » O O </> • •• 10:11 „ . £ *^ skew JC rt K 0.289 0.269 0.207 0.064 0.694 w .r* o* ro * * o ro .r*-» o* OKoi - o - * o ^ i r o * * o o * u i o i - g . r * - » . r > - > j - * ^ j 3s 8 *J M 8 P-ft 0 8 a- 8 h a 8 3 a 09 •< a 41 pft 8 3 Roberds, Runkle, and Whiteman 5. VAR Tests of Expectations Model of the Term Structure Data Sample Long Rate NO. Of Lags(L) Short Rate Wald Test of Expectations Model t-X^L)] (P-value) 75-79 TB3 FFEY 21 76.1 (0.001) 75-79 TB3 FFEY 42 158.0 (0.00) 79-82 TB3 FFEY 21 48.6 (.224) 84-91 TB3 FFEY 21 65.0 (.0129) -28- Roberds, Runkle, and Whiteman Slope Coefficients in Term Structure Regressions! Fed Funds Market Data Sample: 84: 2: 2 - 91: 7:24 irr Overnight n= 30-Day 30-Day 0.71084 (0.87272E-01) [0.43868715] 60-Day 0.69768 (0.68365E-01) [0.45445806] 0.59242 (0.98340E-01) [0.15240927] 90-Day 0.63523 (0.83047E-01) [0.35558630] 0.39346 (0.14370) [0.065956] 180-Day 0.53355 (0.13379) [0.19241131] 0.21209 (0.28217) [0.01516] 60-Day 0.10545 (0.30458) [0.00284] 90-Day -0.14113 (0.60790) [0.00159] Data Sample: 79:10:11 - 82:10: 6 m* Overnight n= 30-Day 30-Day 0.66258 (0.13481) [0.18600642] 60-Day 0.80027 (0.24702) [0.15789838] 0.69623 (0.53894) [0.02302] 90-Day 0.88427 (0.25799) [0.15622411] 0.77503 (0.59434) [0.03277] 180-Day 0.87620 (0.19714) [0.15993267] 0.98028 (0.24848) [0.10281] 60-Day 0.97080 (0.33589) [0.08755] 90-Day 1.3203 (0.26700) [0.08581] 15. Note: m=Short Rate Maturity, n=Long Rate Maturity. Standard errors of slope coefficients in parentheses. R2's in brackets. -29- Roberds, Runkle, and Whiteman Slope Coefficients in Term Structure Regressions, Repo Market Data Sample: 84: 2: 2 - 91: 7:24 m= n= Overnight 30-Day 30-Day 0.81293 (0.76210E-01) [0.47094] 60-Day 0.80306 (0.83843E-01) [0.51417] 0.63056 (0.10172) [0.15022] 90-Day 0.74918 (0.10663) [0.42462] 0.58669 (0.13858) [0.11086] 180-Day 0.53700 (0.22351) [0.16325] -0.31357E-01 (0.34781) [0.23165E-03] 60-Day -0.37417 (0.31626) [0.25596E-01] 90-Day -1.2195 (0.44449) [0.11419] Data Sample: 79:10:11 - 82:10: 6 m* n= Overnight 30-Day 30-Day 0.73162 (0.76086E-01) [0.20980] 60-Day 0.55037 (0.10406) [0.90085E-01] 0.34580 (0.37397) [0.79651E-02] 90-Day 0.41506 (0.19693) [0.44481E-01] 0.33395 (0.48853) [0.59293E-02] 180-Day 0.15139 (0.29643) [0.62615E-02] 0.34026 (0.24751) [0.13233E-01] -30- 60-Day 0.44142 (0.26959) [0.18373E-01] 90-Day 0.56562 (0.22936) [0.16460E--01] Roberds, Runkle, and Whiteman Slope Coefficients in Term Structure Regressions, T-Bill Market Data Sample: 84: 2: 2 - 91: 7:24 m= n= Overnight (SR=Fedfunds) 28-Day 0.37178 (0.11692) [0.25126] 91-Day 0.67718 (0.13641) [0.35458] 182-Day 0.47338E-01 (0.66608E-01) [0.01087173] 91-Day -0.83553E-02 (0.28326E-01) [0.00135683] Data Sample: 79:10:11 - 82:10: 6 m* n= Overnight (SR==Fedf unds) 28-Day 0.54902 (0.84179E-01) [0.29063] 91-Day 1.2083 (0.24637) [0.41399] 182-Day 0.57291 (0.17944) [0.56781] 91-Day 0.18989 (0.11866) [0.16096] -31- Roberds, Runkle, and Whiteman 8. Continued Data Sample: 75: 1: 2 - 7 9 : 1 0 : 3 m« n= Overnight (SR=Fedfunds) 28-Day .22336 (.65927E-01) [.97948E-01] 91-Day .39235E-01 (.21011) [.015899E-02] 182-Day -.74599E-01 (.38467) [.26626E-02] 91-Day .43495 (.37102) [.049477E-01] -32- Roberds, Runkle, and Whiteman Slope Coefficients in Term Structure Regressions, Fed Funds Market Data Sample: 84: 2: 2 - 91: 7:24, Settlement Wednesdays m= Overnight n= 30-Day 30-Day 0.85147 (0.36168E-01) [0.87987] 60-Day 0.84914 (0.51974E-01) [0.81514] 0.75958 (0.13590) [0.17065] 90-Day 0.81140 (0.64996E-01) [0.73914] 0.29532 (0.13990) [0.32129E-01] 180-Day 0.79142 (0.69310E-01) [0.57071] 0.15565 (0.18608) [0.82124E-02] 60-Day 0.13821 (0.20262) [0.54109E-02] 90-Day -0.29710 (0.36759) [0.64982E-02] Data Sample: 84: 2: 2 - 91: 7:24, Wednesdays before Settlement m= n= Overnight 30-Day 30-Day 0.43032 (0.86797E-•01) [0.16910] 60-Day 0.43615 (0.11433) [0.10329] 0.56261 (0.70605E-01) [0.16686] 90-Day 0.31526 (0.10734) [0.51279E-•01] 0.41008 (0.93854E-01) [0.79219E-01] 180-Day 0.17994 (0.12682) [0.15009E--01] 0.12697 (0.17736) [0.51560E-02] -33- 60-Day 0.10924 (0.21018) [0.31432E-02] 90-Day -0.16026 (0.45797) [0.20927E-02 Roberds, Runkle, and Whiteman Continued Data Sample: 79:10:11 - 82:10: 6, Wednesdays m« n= Overnight 30-Day 30-Day 0.73740 (0.76320E-01) [0.39902] 60-Day 0.79166 (0.12191) [0.25797] 1.1530 (0.40904) [0.70105E-01] 90-Day 0.86129 (0.13293) [0.23771] 0.85878 (0.49648) [0.45068E-01] 180-Day 0.93573 (0.20738) [0.23866] 0.98905 (0.41132) [0.10274] -34- 60-Day 1.0780 (0.41318) [0.10083] 90-Day 1.4088 (0.53714) [0.87880E-01] Roberds, Runkle, and Whiteman 10. Slope Coefficients in Term Structure Regressions, Repo Market Data Sample: 84: 2: 2 - 91: 7:24, Settlement Wednesdays m* Overnight n= 30-Day 30-Day 0.88000 (0.41332E-01) [0.91237] 60-Day 0.89058 (0.59263E-01) [0.86838] 0.81124 (0.14040) [0.17430] 90-Day 0.86816 (0.73338E-01) [0.80642] 0.63977 (0.12307) [0.11341] 180-Day 0.82090 (0.99945E-01) [0.59916] -0.39116E-03 (0.30401) [0.31330E-07] 60-Day -0.35666 (0.35431) [0.22184E-01] 90-Day -1.1811 (0.42519) [0.10827] Data Sample: 84: 2: 2 - 91: 7:24, Wednesdays before Settlement m= n= Overnight 30-Day 30-Day 0.57175 (0.90018E- 01) [0.22958] 60-Day 0.57329 (0.12672) [0.16959] 0.94002 (0.10909) [0.17167] 90-Day 0.52235 (0.11224) [0.13487] 0.59400 (0.10046) [0.14026] 180-Day 0.19879 (0.16344) [0.17115E-•01] 0.56861E-01 (0.24290) [0.85628E-03] -35- 60-Day -0.27002 (0.29575) [0.13148E-01] 90-Day -1.1915 (0.42815) [0.98355E-01 Roberds, Runkle, and Whiteman 10. Continued Data Sample: 79:10:11 - 82:10: 6, Wednesdays m» n= Overnight 30-Day 30-Day 0.77928 (0.88908E-01) [0.27686] 60-Day 0.54894 (0.17192) [0.99390E-01] 0.36621 (0.46336) [0.69809E-02] 90-Day 0.48640 (0.24359) [0.64508E-01] 0.53541 (0.57689) [0.15607E-01] 180-Day 0.28259 (0.29999) [0.24219E-01] 0.24637 (0.44599) [0.73326E-02] -36- 60-Day 0.32061 (0.45207) [0.10550E-01] 90-Day Roberds, Runkle, and Whiteman 11. Slope Coefficients in Term Structure Regressions, Fed Funds Market Data Sample: 84: 2 : 2 - 91: 7:24, Day after FOMC meetings m= n= Overnight 30-Day 30-Day 0.74509 (0.83258E-01) [0.55146] 60-Day Q.71413 (0.10129) [0.53943] 0.71192 (0.17200) [0.28684] 90-Day 0.72288 (0.10569) [0.46611] 0.41039 (0.16886) [0.76046E-01] 180-Day 0.70486 (0.14008) [0.30932] 0.86951E-01 (0.18779) [0.29287E-02] 60-Day -0.20294E-01 (0.27085) [0.10258E-03] 90-Day -0.31490 (0.45530) [0.77648E-02] Data Sample: 79:10:11 - 82:10: 6, Day after FOMC meetings m= n= Overnight 30-Day 30-Day 1.1375 (0.12522) [0.55483] 60-Day 1.1084 (0.28921) [0.32927] 0.42279 (0.74265) [0.73300E-02] 90-Day 1.0367 (0.31461) [0.27301] 0.60534 (0.89888) [0.16951E-01] 180-Day 0.94613 (0.26896) [0.28244] 1.0098 (0.41959) [0.12071] -37- 60-Day 1.0853 (0.29268) [0.18424] 90-Day 1.2601 (0.48919) [0.10950] Roberds, Runkle, and Whiteman 12. Slope Coefficients in Term Structure Regressions, Repo Market Data Sample: 84: 2: 2 - 91: 7:24, Day after FOMC meetings m= n= Overnight 30-Day 30-Day 0.58366 (0.10163) [0.32624] 60-Day 0.69122 (0.90822E-01) [0.47256] 0.94945 (0.15035) [0.33104] 90-Day 0.69851 (0.10085) [0.40754] 0.81131 (0.15859) [0.19485] 180-Day 0.60386 (0.14340) [0.19960] 0.46057E-01 (0.32509) [0.47395E-03] 60-Day -0.44065 (0.38018) [0.31240E-01] 90-Day -1.4686 (0.47491) [0.14990] Data Sample: 79:10:11 - 82:10: 6, Day after FOMC meetings m= n= Overnight 30-Day 30-Day 1.1657 (0.19487) [0.56774] 60-Day 0.84924 (0.32370) [0.24901] 0.31677 (0.72952) [0.49330E-02] 90-Day 0.75639 (0.38906) [0.17021] 0.38240 (0.91514) [0.61153E-02] 180-Day 0.39520 (0.41207) [0.58096E-01] 0.38272 (0.52590) [0.15331E-01] -38- 60-Day 0.35125 (0.62701) [0.96151E-02] 90-Day 0.66613 (0.36359) [0.36161E-01] Roberds, Runkle, and Whiteman 1. Distribution of daily changes in effective Federal Funds rate, by operating regime. -39- Roberds, Runkle, and Whiteman 2. Distribution of daily changes in effective Federal Funds rate during the Fed Funds targeting period, by day of the settlement period (settlement day = day 5.) -0.5 0 hundreds of basis points 0.5 day 5 3. Distribution of daily changes in effective Federal Funds rate on settlement day during the nonborrowed reserves targeting period. -40- Roberds, Runkle, and Whiteman 4. Distribution of daily changes in effective Federal Funds rate during the borrowed reserves—lagged accounting regime, by day of the settlement period (settlement day = day 5.) -0.5 0 hundreds of basis points day 5 day 10 5. Distribution of* daily changes in effective Federal Funds rate during the borrowed reserves—contemporaneous accounting regime, by day of the settlement period (settlement day = day 10.) -41- FOMC OPERATING PROCEDURES AND THE TERM STRUCTURE: SOME COMMENTS Glenn D. Rudebusch1 This discussion is divided into two parts. The first describes some recent empirical results regarding the amount of information in the yield curve for forecasting future changes in short rates. My goal is to highlight several of the results of Roberts, Runkle, and Whiteman (1992) (henceforth RRW) and to compare their research to work done by previous authors. The second section contains an interpretation of this entire body of evidence in light of several characteristics of the Federal Reserve's monetary policy operating procedure. FACTS RRW provide a careful study of the Rational Expectations Hypothesis (REH) of the term structure using data at the short end of the maturity spectrum. Their paper contains much fascinating empirical detail, but this note focuses on a single issue: the predictive power of the term structure for future movements in interest rates. The REH of the term structure implies that the current spread between a long rate and a short rate should predict future changes in that short rate. I consider two special cases of such term structure predictions. First, I examine pairs of securities in which the maturity of the long-term debt instrument is twice that of the short-term debt instrument. Second, I consider the ability of the spread between the overnight rate and the one-month rate to predict future changes in the overnight rate. Let r(l) be the yield on a one-period bond, and let r(2) be the yield on a two-period bond. Then, the expectations hypothesis implies that 1. Glenn D. Rudebusch is on the staff of the Board of Governors of the Federal Reserve System, Division of Monetary Affairs. 2. In addition, in a separate appendix written after the conference, I provide a formal model of the propagation of changes in the federal funds rate out the yield curve in order to address some of the comments and questions that were generated by the results in RRW and several of the other conference papers. Rudebusch r(2) t « l/2[r(l)t + E t r(l) t+1 3 + X; (1) that is, the current two-period yield equals the average of the actual and expected one-period yields in sequence plus a term premium T. Assuming rational expectations, r(l)t+1 = V ( 1 ) t + 1 (2) + u t+l« where u +- is a forecast error orthogonal to information available at time t. Substituting (2) into (1) and rearranging provides a simple testable equation of the REH of the term structure: l/2[r(l) t+1 - r(l)t] = a + p[r(2) t - r(l)t] + e t + 1 . (3) Under the REH null hypothesis, p = 1 and a = -T; that is, after taking expectations of both sides of (3), one-half the optimal forecast of the change in the short rate should equal the spread between the long rate and short rate (minus a term premium). In addition, the error term is orthogonal to the right-hand side regressors, so ordinary least squares provides consistent coefficient estimates. Studies that have tested the REH using equation 3 have A obtained a wide variety of estimates of p. These p's are often significantly less than one; of particular note, however, is the dependence of the estimates on the maturity of the debt instruments being examined. Figure 1 provides estimates of P from eight studies prior to RRW that use data on yields of U.S. Treasury debt. The point estimates are arrayed as a function of the short (one-period) bond maturity, which ranges from two weeks to five years. Each observation is numbered according to its source. For example, the "l"'s in figure 1 are taken from Campbell and Shiller (1991), who 3. This corresponds to equation 4 in RRW with n = 2m. A 4. For example, the P's shown at the three-month maturity are obtained from a regression of the change in the yield of the threemonth bill on the yield spread between the six-month and three-month bills. Although equation (3) is a convenient form for expressing results across a range of maturities, it does not describe the frequency of observation. Typically, empirical studies have used overlapping observations that are more frequent than those separated by the maturity of the short bond. -2- Rudebusch provide a careful, exhaustive study spanning a large range of maturities. Based on this collection of point estimates from previous researchers, the shaded band in figure 1 provides an A informal summary of the relationship between the P's and the maturity of the short bond. Apparently, the forecast power of the term structure for changes in short rates is quite high for forecast horizons (i.e., shorter bond maturities) no longer than one month. As the horizon increases, forecast power initially disappears, as estimates of p fall essentially to zero over the range from three months to one year; however, with horizons longer than one year, forecast power starts to improve. The result is, according to Campbell and Shiller (1991), a "U-shaped" pattern of coefficients. The evidence of RRW is consistent with this U-shaped pattern for the short maturities that they investigate. For example, in their table 8, based on the spread between 3-month and 6-month Treasury bills, the estimates of P are not significantly different from zero. Also, based on the spread between the 30-day and 60-day term federal funds rates, RRW (table 6) report estimates of p equal to about 0.6 at the one-month horizon. Again, the predictive content of the yield curve, while substantial at very short maturities, appears to vanish at a forecast horizon of about three months. RRW also analyze the forecasting ability of yield spreads that involve the overnight federal funds rate. As above, the basic insight of the REH is that if the yield curve is steeply sloped, future short rates should on average be above the current short rate. Let the length of a period be a day, so r(l) is the overnight federal funds rate and r(30) is the yield on a 30-day bill; then the expectations hypothesis implies that (4) r(30)t = l/30[r(l)t + Et 29 L r(l) t + i ] + T. Assuming rational expectations, the analog to equation 3 is 5. Or, in their words, consistent with this "hole in the ozone layer," a metaphor that carries the gratuitous connotation of welfare loss . -3- Rudebusch (5) 29 29 l/30[ Z r(l).,,] - r(l). - a + p[r(30). - r(l).] + I e..,. i-0 i-1 Under the REH null hypothesis, p = 1; that is, the deviation of today's federal funds rate from its expected average level over the next month should equal the spread between the current 30-day A and one-day rates (minus a term premium). RRW find P to be fairly close to one. In their table 6, for example, using term federal funds data, they estimate p to be around 0.7. This high predictive power at short horizons generalizes to spreads between the overnight rate and the 60-day and 180-day bill yields and is broadly consistent with earlier work by Simon (1990). In summary, I characterize the evidence on the forecasting ability of the yield spread with four propositions: 51. Spreads between the overnight federal funds rate and one-month, two-month, and three-month yields are very good predictors of the change from the current daily rate to the average daily rate that prevails over the relevant time period. 52. Spreads between short-term bills--for example, 30-day and 60-day Treasury bills--are good predictors of the change in the short bill yield. 53. Spreads involving longer bill rates, say, the 3-month, 6-month, and 12-month yields, have essentially no predictive content for future changes in these bill rates. 54. Spreads involving medium and long maturity bonds-specifically, for maturities longer than two years--do appear to have some predictive content for movements in future interest rates. INTERPRETATION Simply put, propositions SI through S4 indicate that the yield curve is useful for forecasting future interest rates, but only at certain maturities. Recently, several authors have linked this finding to the behavior of the Federal Reserve. Specifically, they have asserted that the procedure by which the Fed controls the federal 6. RRW provide evidence consistent with the first three of these propositions. 7. See Mankiw and Miron (1986), Cook and Hahn (1989, 1991), and Goodfriend (1991) . -4- Rudebusch funds rate is responsible for the varying predictive power of the yield curve. This section clarifies and extends this argument. First, it is useful to describe the attributes of the Fed's operating procedure that are crucial for understanding the above results regarding the yield curve. Although the Fed's operating procedure has changed greatly over the past two decades--from direct federal funds rate targeting in the 1970s to indirect targeting through reserves in the 1980s --following Cook (1989) and Goodfriend (1991), I take as given that over this period the Fed has always taken an active interest in controlling the federal funds rate. Although the exact mechanism has changed over time, the following three attributes have characterized the Fed's underlying procedure for controlling interest rates: Fl. Transitory deviations from target are allowed on a daily basis. That is, the federal funds rate is not pegged to a target on an hourly or even a daily basis. Indeed, the Fed generally enters the federal funds market only once each day, in the late morning; thus, the intraday spot federal funds rate can display wide fluctuations. Furthermore, historically there has often been a target band for the federal funds rate rather than a single target value; using a band also allows transitory deviations from a mean value (usually the target band midpoint). F2. Targets are adjusted not continuously but in limited amounts at discrete intervals. Rather than immediately adjusting the target in response to each new piece of information, the Fed behaves as if it has a threshold whereby only after sufficient information has accumulated will a target change be triggered. Also, targets are usually adjusted by only 25 to 50 basis points at a time; thus, when new information requires a larger change, the Fed implements a series of smaller target adjustments that are separated by several days to a couple of weeks. .F3. Except when a quick succession of target changes is needed to make a large adjustment (as noted in F2), the target is set at a level the Fed expects to be able to maintain. These three attributes are apparent in figure 2, which shows federal funds rate targets from an illustrative episode of direct interest rate targeting. The targets are shown as a solid line; the actual daily effective federal funds rate appears as a dashed -5- Rudebusch line. Attribute Fl is apparent in the large but temporary deviations of the daily rate from the target. F2 is reflected in the "step-function" appearance of the target. Finally, the infrequent occurrence of target changes, shown by long steps or treads of many months duration, provides support for F3. The three attributes of the Fed's "interest rate smoothing" operating procedure can explain the term structure results SI through S3. First, the transitory daily deviations described by Fl, coupled with the target persistence of F3, imply substantial predictable variation in the overnight rate. If today's rate is unusually high, perhaps because the day is a settlement Wednesday with strong reserve pressures, tomorrow's rate (and future daily rates) will likely return to the target level. This occurrence explains why the spread between the 30-day term federal funds rate (which should be close to the current target rate) and the 1-day rate (which is transitorily high) moves with the spread between the average 1-day rate that will prevail over the next month (which is also close to the target rate) and the current 1-day rate. RRW provide confirming evidence for this interpretation: Their table 3 indicates that large changes on Wednesdays tend to be reversed on Thursdays, and their tables 9 and 10 indicate that SI holds best for 9 these volatile settlement Wednesdays. The predictive information described by S2, which is available at the very short end of the term structure, can be explained by the discrete nature of policy changes, attribute F2. Let us suppose that a significant piece of new information arrives-say, in the form of news about some macroeconomic variable--that clearly requires a major policy change. If the Fed accomplishes this policy change by a series of moderate target adjustments conducted, say, over a three-week span, then the gap between the time that new information influencing policy was released and the time that the full policy action is finished generates predictable 8. The target series was obtained by linking the expected federal funds rates given in the Federal Reserve Bank of New York's weekly "Report of Open Market Operations and Money Market Conditions" during the period. 9. Cook and Hahn (1989, footnote 6) also describe similar evidence. -6- Rudebusch changes in interest rates at horizons of less than one month. The lack of predictive information in the 3-month to 1-year maturity range of the term structure, noted in S3, reflects attribute F3. As Mankiw and Miron (1986) argued, if market participants (rationally) expect the Fed to maintain the current federal funds rate target, then current spreads will have no predictive power for actual future changes in interest rates. Thus, attributes Fl through F3 of the Fed's operating procedure appear to be responsible for yield spread results SI through S3. Further support for the reasoning linking F2 with S2 and F3 with S3 is provided by the analysis of Cook and Hahn (1989). They found that the 3-month, 6-month, and 12-month bill rates all moved on average about 50 basis points in response to a change of 1 percentage point in the federal funds rate target during 1974-79. That bill rates move by only about half of the target change suggests that target changes are forecastable to some extent (as implied by F2). However, that all three bill rates move by about the same amount means that new information about the federal funds target has little effect on the slope of the yield curve from 3 months to 12 months. This finding confirms the notion that the new federal funds rate target is expected to be maintained over that period. Finally, for completeness, let me examine S4, the proposition that spreads between long rates contain forecasting power. Since Fama and Bliss (1987), this proposition has been reduced to the issue of whether interest rates display a slow 10. Cook and Hahn (1990) stress a different, but related, aspect of F2 as an explanation of S2. They note that the information threshold required for a target change implies that news that suggests an imminent policy action is often available to the market. For example, if weak employment data that suggest the need for a policy easing are released, the Fed may wait a week for another reading on inflation before acting. This gap could also generate predictable target changes over a very short period. 11. Cook and Hahn (1989) show that this same reasoning holds if the funds target is expected to change in the near future (consistent with A2) and then to persist at its new level. Strictly speaking, if one assumes a constant term premium along with the persistence of targets, A the spread would be constant and the standard error of |J would be infinite. However, as shown in Mankiw and Miron (1986), with a timeA varying term premium, the P is biased toward zero. -7- Rudebusch reversion to mean at long horizons. The subsequent debate is summarized in Shea (1992). My own view on such issues, expressed in Rudebusch (1992, 1993), is that conclusions about the stationarity or nonstationarity of yields are very tenuous given the size of the available samples. However, although such deep conclusions cannot be made with any degree of certainty, at a practical level, S4 probably reflects the fact that over the sample period under consideration, markets have correctly anticipated that the Fed would be able to restrain inflation to moderate levels at business-cycle 12 frequencies. Coupled with a stationary real rate, the Fed's containment of inflation has probably generated the predictive power contained in long-maturity nominal interest rate spreads. CONCLUSION The explanation that RRW propose for linking Fed operating procedure to the empirical results on the predictive power of the term structure hinges on the Fed's elimination of seasonal fluctuations in interest rates, which would be predictable and which presumably would be reflected in spreads (see Hardouvelis (1988)). This interpretation appears flawed. The Fed's elimination of weekly, monthly, quarterly, and half-yearly seasonals would imply that figure 1 was a step function, displaying no predictive power up to six months and significant predictive power from twelve months and beyond. This implication does not accord with the U-shaped curve from the empirical evidence. Instead, I have provided an interpretation of the term structure results that relies on characteristics of the Fed's interest rate targeting operating procedure. A consideration of the normative value of this procedure would also be of interest. Goodfriend (1991) conjectures that the smoothing characteristics of the operating procedure facilitate the communication of Fed intentions to financial markets. Further research on this issue would be illuminating. 12. Again, the evidence in Cook and Hahn (1989) is instructive. They find that the 20-year bond yield responds by only about 10 basis points to a 1 percentage point change in the federal funds target. This is consistent with slow mean reversion at long horizons. -8- Rudebusch REFERENCES Campbell, John Y. and Shiller, Robert J. "Yield Spreads and Interest Rate Movements: A Bird's Eye View," Review of Economic Studies, 58 (1991), 495-514. Cook, Timothy. "Determinants of the Federal Funds Rate: 1979-1982," Federal Reserve Bank of Richmond Economic Review, (1989), 3-19. Cook, Timothy and Hahn, Thomas. "The Effect of Changes in the Federal Funds Rate Target on Market Interest Rates in the 1970s," Journal of Monetary Economics, 24 (1989), 331-351. Cook, Timothy and Hahn, Thomas. "Interest Rate Expectations and the Slope of the Money Market Yield Cure," Federal Reserve Bank of Richmond Economic Review, (1990), 3-26. Fama, Eugene F. "The Information in the Term Structure," Journal of Financial Economics, 13 (1984), 509-528. Fama, Eugene F. and Bliss, R.R. "The Information in Long-Maturity Forward Rates," American Economic Review, 77 (1987), 680-692. Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary Policy," Carnegie-Rochester Series on Public Policy 34 (Spring, 1991), 7-30. Hardouvelis, Gikas A. "The Predictive Power of the Term Structure during Recent Monetary Regimes," Journal of Finance, 43 (1988), 339-356. Mankiw, N. Gregory and Miron, Jeffrey A. "The Changing Behavior of the Term Structure of Interest Rates," Quarterly Journal of Economics, 101 (1986), 211-28. Mishkin, Frederic S. "The Information in the Term Structure: Some Further Results," Journal of Applied Econometrics, 3 (1988), 307-14 -9- Figure 2 Federal Funds Rate 6.5 Target Actual 6.0 c 5.5 XT c a n tr 5.0 4.5 4.0 Sep 1976 Oct Nov Dec Jan Feb Mar Apr May Jun Jul 1977 Aug APPENDIX PROPAGATION OF FEDERAL FUNDS RATE CHANGES TO LONGER-TERM RATES UNDER ALTERNATIVE POLICY REGIMES Glenn D. Rudebusch1 My comments above on Roberds, Runkle, and Whiteman (1992, henceforth RRW) focused on the relationship between FOMC operating procedures and the amount of information in the yield curve for forecasting future changes in short rates. However, several papers at the conference, including RRW, touched on the general issue of the propagation of changes in the short rate out the yield curve: most notably, Gagnon and Tryon (1992), Goodfriend (1992), Hess, Small, Brayton (1992), and Kasman (1992). Discussions both at the conference and subsequently tried to discern from this array of evidence whether adopting an alternative operating procedure that changed the variability of the funds rate might also affect the variability of long rates. In this appendix, I develop a model that links movements in long rates to policy and non-policy federal funds rate shocks in order to interpret some of the findings in earlier papers. In particular, my analysis clarifies the information content of RRW's evidence on the nature of the linkage between short and long rates. I show that the degree to which changes in the funds rate are transmitted to changes in long rates is unrelated to the predictive power of the yield spread. The model also provides a clear interpretation of changes in measures of interest rate volatility and correlation. Such changes are evident in the narrative history of Goodfriend (1992) as well as in the earlier evidence of Johnson (1981) . A MODEL OF SHORT AND LONG RATES This section describes a simple theoretical structure that links Federal Reserve actions to movements in the funds rate and in longer rates. There are two crucial elements of the model: First, 1. Glenn D. Rudebusch is on the staff of the Board of Governors of the Federal Reserve System, Division of Monetary Affairs. Rudebusch from equation 1. Substituting the first result into the second and 4 taking expectations yields E r ., *= E v(r + £ u , .+ £ ,,J ) *t t+k t t i . 1 t+i t+k (4) - rt. That is, the expected funds rate at any point in the future is equal to today's target. This result can be used in equation 3 to solve for the current long rate t-1 R. - (l/x)r. + (l/x)£. + (l/x) I r t x x x i=El x (5) + 9. x - r t + (l/x)£t + 6 t . The current long rate is thus equal to the current target plus a small fraction l/x of the current transitory funds rate shock plus a term premium. To express the long rate in terms of the shocks affecting the short rate, equation 5 can be rewritten as R - r , + u + (l/x)e + 9 . It is the policy shocks (u ) affecting r R because the term (l/x)£ unrelated to r . that will be reflected in is negligible and the term premium 8 is Accordingly, the amount of short-rate variation that is transmitted to the long rate will depend crucially on the size of G . u 4. This expression assumes that financial market participants can discern the current target rate on any given day; that is, they know e t and u at time t. Thus, in this model, there is no avenue for misperceptions of monetary policy, arising, say, from the adoption of new operating procedures, to affect the relationship between the funds rate and longer rates. A useful extension to the model would be to re-examine the results assuming market participants had difficulty distinguishing transitory daily fluctuations in the funds rate from true policy shifts. -4- Rudebusch INTERPRETING RRW'S HISTORICAL EVIDENCE The results of RRW indicated a clear difference across the pre-1979 and the 1979-1982 periods in the ability of the bill-funds yield spread to predict future changes in the funds rate. Their historical evidence was obtained by regressing actual changes in future short rates on the current spread between the funds rate and longer rates. In terms of the model outlined above, RRW's key regression can be obtained by subtracting r from both sides of (3), which after rearrangement gives X-l (6) (i/x)az V t + i ' c ^ - i ) / * ) ^ " (R t " rt} ' V Under the assumption of rational expectations, (7) Vt+i = r t+i + v t+i- where v + . is a forecast error orthogonal to information at time t. Then (6) can be rewritten as the regression X-l (8) (1/X)BI r t + i - ((x-l)/x)rt - P(Rt - rt) + e t . The dependent variable on the left-hand side requires the X-l construction of (1/x) I r++- * which is simply the average of funds i-1 x x rates observed after time t, and the use of (7) implies that P equals one. X-l The regression error, e = - 8 - L v ,., is correlated X X i = 1 T-M with R through the common term premium but is not correlated with r . Because of the correlation between the regression error and the X A regressor, the OLS estimate of p, denoted p, will be biased downward A from one. Still, the size of P has been interpreted, as in RRW, as measuring the ability of long rates to forecast future movements in short rates. A Given the model for the funds rate, the value of P (in the population) can be easily determined. By definition, 5. This regression is essentially equation 4 in RRW. -5- Rudebusch t-1 P - Cov[(l/x) I r t + ± - ( ( t - l ) / T ) r t . \ - rt]/Var(Rt - rt). The numerator of this term can be rewritten as T-1 i C o v [ ( l / x ) E ( I u t + i + e t + i ) - ((t-l)/x)e t . 6 t - ((x-l)/x)e t ] = ((x-i)/x) 2 o^. while the denominator can be rewritten as Var[r t + (l/x)e t + 6 t - r t - e t ] = a2Q - ((x-1)/x) 2 o 2 . A Thus, the population value of p is given by plim p - ((T-l)/t) 2 o|/[oJ - ((T-1)/T)2G2£] . - 1 / [X(X/(T-1)) 2 (9) where X s OQ/&1, - 1], the ratio of the variance of the term premium to the variance of the transitory shock. As can be seen from (9)» the estimate of P varies inversely A 2 with X. Thus, increases in O are reflected in a higher p. Intuitively, increased transitory deviations of today's funds rate from the target provide more predictable future variation in the funds rate because today's deviation will be eliminated, on average, the next day through reversion to target. This predictable future reversion to target is incorporated in the long rate, which boosts A 2 the value of p. In contrast, increases in a^ simply increase the noise in the long rate and lower p. Most importantly, however, the A crucial feature of (9) to note is that P does not depend on the variance of the policy shocks (o ) . Intuitively, this results because such shocks do not affect either side of the regression (8): the shocks are reflected completely in both r and R and so do not change the yield spread, and they are permanent and so do not show up in the difference between future and current funds rates. Thus, A evidence on the size of P places no restriction on the size of policy shocks to the funds rate and hence has nothing to say about -6- Rudebusch the extent to which the variability of the funds rate is reflected in long rates. The actual estimates of p from RRW during the 1975-1979 and 1979-1982 periods (with standard errors in parentheses) are (Fact 1) P* - 0.04 (.21) and j}** - 1.21 , (.25) where a single asterisk denotes the period before October 1979 and a double asterisk denotes the 1979-1982 period. The finding that p A** is less than P (10) X implies that in terms of the model variances (that is. C2Q ItT > X > ClQ /<T ). As we shall see below this evidence places no restriction on the volatility of long rates or on the correlation of movements in the long rates with movements in the funds rate. INTERPRETING JOHNSON'S HISTORICAL EVIDENCE In this section, I use the model of the funds rate (equations 1 and 2) and of the long rate (equation 5) to analyze the historical evidence provided in Johnson (1981) regarding changes in the behavior of interest rates after 1979. In particular, the next two subsections focus, in turn, on changes in (1) the volatility of the funds rate and longer-term rates and (2) the correlations between movements in the funds rate and in longer rates. The Volatility of the Funds Rate and Longer-Term Rates First, let us compute the volatility of the funds rate and of the long rate in terms of the parameters of the model. The variance of the change in the funds rate, denoted Ar - r - rr_i» is given by G Ar S Var ^Art) = Var(rt - r ^ + et - e ^ ) - Var(ut + e t - e t . 1 ) 6. These estimates are based on the spread between the three-month Treasury bill and the funds rate and are taken from table 8 in RRW. -7- Rudebusch -< (ID + *>i- Similarly, the variance of the change in the long rate is given by °2AR s Var(ARt) = Var[rt - r ^ + (1/x) (et - e ^ ) + 9t - 6^] - G2u + 2(l/x)2G* + 2 G Q . (12) Thus, the volatility of the funds rate depends on both the variance of the permanent policy shock, (T , and the variance of the temporary shock, G . The volatility of the long rate depends primarily on the variance of permanent policy shocks and the variance of the term premium, G Q ; the variance of the transitory shock is negligible because its weight, 2 (1/x) , is so small. The historical standard deviations of changes in the funds rate and in a variety of longer rates during the 1975-1979 and 19791982 periods are shown in table 1. Clearly, the volatility of all rates increased in the later period. Label this result as "fact 2"; that is. (Fact 2) G ^ < 02A" and C ^ < G^* , where again a single asterisk denotes the period before October 1979 and two asterisks denote the 1979-1982 period. The historical evidence on volatility expressed by fact 2 could be reconciled with equations 11 and 12 in a number of ways. A plausible set of assumptions about the variances of the model's shocks that could explain fact 2 are: (13) c\ < c\ . (14) <*<<". (15) <TQ < CQ . 7. This table extends some of the results in tables 4 and 9 of Johnson (1981) to daily data and a larger post-1979 sample. -8- Rudebusch Inequality 13 implies that the Federal Reserve allowed much larger temporary deviations from its implicit target rate in the 1979-1982 period than before. Inequality 14 implies that the Federal Reserve was much more aggressive in moving the target rate in the later period than before. The larger policy shocks reflected, in part the new emphasis on movements in Ml underlying Federal Reserve policy g actions. Finally, inequality 15 implies that the variance of the term premium also increased after 1979, which is consistent with 9 the increase in interest rate risk implicit in fact 1. Thus, the greater volatility of the funds rate in the 19791982 period reflects both-the looser control by the Federal Reserve over day-to-day movements in the funds rate (inequality 13) as well as larger permanent policy shifts (inequality 14). The greater volatility of longer rates reflects the greater policy shocks as well as the increase in the volatility of the term premium. The Correlation of Movements in the Funds Rate and Longer Rates How does the correlation between changes in the funds rate and the long rate depend on the model parameters? This correlation 10 u is given by p a Corr(Ar , AR ) - Cov(Art, ARt)/VVar(ARt)Var(Art) - (K + 2 / T ) / V ( K + 2)(K + 2X + 2/x 2} 9 (16) 2 2 here K s a /a , the ratio of the variances of the policy and transitory shocks, and where, as before, X s C Q / C T . The correlation 8. Both (13) and (14) are consistent with evidence presented in Balduzzi, Bertola, and Foresi (1992). 9. Inequality 15 is consistent with the evidence in Cook and Hahn (1990); the various proxies for the (ex ante) term premium that they display show a clear increase in volatility during the 1979-1982 period. 10. Note that the covariance of the two rate changes is given by Cov(Art, AR t ) - Et[(Art)(ARt)] - E t [(u t +e t -e t . 1 )(u t +(l/t)(e t -£ t . 1 )+G t -e t . 1 )] - c2 + U (2/T)O* £ -9- Rudebusch 2 p depends positively on K: increases in G increase the covariance of Ar and AR proportionately more than their variances and hence 2 boost p. while increases in G increase the variances more than the covariance and hence diminish p. Intuitively, the long and short rates will be more closely correlated when the shocks that drive them both (the permanent u policy shocks to r ) are more important than the shocks that essentially drive only the short rate (the £ shocks of r about r ) . The correlation also depends negatively on X because increases in the variability of the term premium (relative to G ) simply add noise to the long rate and leave the short rate unaffected. The actual correlations of changes in the funds rate with a variety of longer rates are shown in table 2. The correlations during the 1975-1979 period, p , are shown in the first column; the correlations during the 1979-1982 period, p , are shown in the second column. The correlations are all higher in the later period; that is, (Fact 3) p < p . In terms of the model variances, fact 3 implies that K < K (that is, a2 /cr < G2 /<r e u e ( t h a t i s , GQ ICT > GQ u ), or X > X o r , most l i k e l y , /tr ) b o t h 12 11. This extends some of the results in tables 10 through 13 in Johnson (1981) to daily data and a larger post-1979 sample. 12. One factor that has been ignored is duration (see Goodfriend, 1992) . The results in table 2 at maturities greater than one year are obtained from coupon securities, while the model considers only zerocoupon securities. The yield on a coupon security would move a bit more in response to a transitory shock than the yield on a zero-coupon security, and the correlation between the funds rate and the bond rate would be slightly higher. -10- Rudebusch CONCLUSION The change in Federal Reserve operating procedures in October 1979 ushered in, as had been expected, an era of increased funds rate volatility. At the time, many were surprised by how variable longer rates also became. In assessing whether a change in operating procedures will increase the variability of long rates, the key insight from the above model is that one must focus on the permanent policy shocks. Such a narrow focus is not surprising because only the long-lived shocks to the short rate will affect long rates under the rational expectations hypothesis of the term structure. Accordingly, whether future changes in the procedures governing the behavior of the funds rate will affect long rates depends in large part on whether the associated re-specified reaction function for policy has been linked to variables that are subject to more permanent shocks. More generally, in a world where the rational expectations hypothesis of the term structure may not hold, the variability of long rates depends on what market participants believe about changes in the size of the permanent shocks. -11- Rudebusch REFERENCES Balduzzi, Pierluigi, Bertola, Giuseppe, and Foresi, 'Silverio. "A Model of Target Changes and the Term Structure of Interest . Rates," (1992), manuscript. Cook, Timothy. "Determinants of the Federal Funds Rate: 1979-1982," Federal Reserve Bank of Richmond Economic Review, (1989), 3-19. Cook, Timothy and Hahn, Thomas. "The Effect of Changes in the Federal Funds Rate Target on Market Interest Rates in the 1970s," Journal of Monetary Economics, 24 (1989), 331-351. Cook, Timothy and Hahn, Thomas. "Interest Rate Expectations and the Slope of the Money Market Yield Cure," Federal Reserve Bank of Richmond Economic Review, (1990), 3-26. Gagnon, Joseph E. and Tryon, Ralph W., "Price and Output Stability Under Alternative Monetary Policy Rules," (1992), manuscript. Goodfriend, Marvin. "Interest Rates and the Conduct of Monetary Policy," Carnegie-Rochester Series on Public Policy 34 (Spring, 1991), 7-30. Goodfriend, Marvin, "Interest Rate Policy and the Inflation Scare Problem: 1979-1992," (1992), manuscript. Hess, Gregory D., Small, David H., and Brayton, Flint, "Nominal Income Targeting with the Monetary Base as Instrument: An Evaluation of McCallunTs Rule," (1992), manuscript. Johnson, Dana. "Interest Rate Volatility Under the New Operating Procedures and the Initial Response in Financial Markets," in New Monetary Control Procedures, vol. 1, February 1981, Board of Governors of the Federal Reserve System. Kasman, Bruce, "A Comparison of Monetary Policy Operating Procedures in Six Industrial Countries," (1992), manuscript. -12- Rudebusch Roberds. William, Runkle, David, and Whiteman, Charles H. "Another Hole in the Ozone Layer: Changes in FOMC Operating Procedure and the Term Structure," (1992), manuscript. -13- Rudebusch Table 1 Standard Deviations of Changes in Federal Funds Rates and in Rates on Various Treasury Securities (Daily data; percentage points) SflflipJ-e period Type of security Federal funds 3-month bill 6-month bill 1-year bill 3-year note 5-year note 10-year note 20-year bond 7W9 .32 .09 .07 .08 .06 .05 .04 .03 79-82 .96 .34 .31 .27 .21 .19 .17 .16 The sample period "75-79" includes data from January 6, 1975 through September 28, 1979; the sample period "79-82" includes data from October 15, 1979 through October 1, 1982. Table 2 Correlations of Changes in the Funds Rate with Changes in Rates on Various Treasury Securities (Daily data) Sample period Type of security 3-month bill 6-month bill 1-year bill 3-year note 5-year note 10-year note 20-year bond 75-79 .07 .05 .10 .06 .05 .04 .02 The sample periods are the same as in table 1. 79-S2 .21 .22 .20 .19 .16 .15 .15 Figure 1 Federal Funds Rate Target Actual 1976 1977 A POLICYMAKER'S GUIDE TO INDICATORS OP ECONOMIC ACTIVITY Charles Evans, Steven Strongin, and Francesca Eugeni The evaluation of economic indicators has often progressed with an odd independence for the way in which such indicators are actually used in practice in the economic policy process. The search is often for one "best" indicator, where "best" typically refers to winning in some narrowly defined contest of general purpose forecasting ability measured over some pre-selected time-span. The actuality of the policy process is far richer. Indicators are used in a kind of chaotic democracy, each indicator casting a vote based on its own forecast and the policymakers weighing each vote, based on their assessment of the current credibility of the indicator. This is quite different from the standard academic formulation of economic policy where a "true" model is developed and then policy run in a way optimizes the performance of the model. Understanding this difference in approach leads to very different ways of evaluating indicators. It is not just enough to produce a "best" model; rather, it is important to understand what type of information is contained in a given indicator so that its message can be properly evaluated and also how much weight to give that message given what else is also known. Indicators, like people, perform better or worse depending on the context in which they operate. Efficient usage requires matching indicators both with appropriate questions and with other complementary indicators. For instance, some indicators do far better at predicting short-run changes in activity, but do not do very well at pinning down the level of activity over longer time spans, while other indicators forecast short-run phenomena poorly, but do better at predicting average activity over longer time span. Also while some indicators have very close substitutes, such as the twenty or so short-term interest rates sometimes used in econometric studies, and thus provide little additional information beyond that already contained in other indicators, some indicators can provide substantial additional information, thus providing important confirming or contradicting information. The policymaker needs to know how to match questions with Evans, Strongin, and Eugeni indicators depending on the current policy context. A swiss army knife is a fine general purpose tool, but it is hardly a substitute for a we11-equipped workshop. This paper seeks to develop and implement a set of techniques for evaluating indicators of economic activity that more closely match the actual use of such indicators in the dayto-day policy process. We see that process as primarily involving the re-assessment of short- to medium-term economic activity based on indicator by indicator analysis with the primary decision matrix being whether it is necessary to ease or tighten policy in order to realize appropriate levels of economic activity. We do not address the longer run issues of assessing appropriate levels of economic activity or other issues involving inflation or the value of the dollar nor do we address the question of how best to implements those decisions. Evaluating indicators in this context has four primary parts; ranking candidate indicators, characterizing the nature of the information in those indicators, assessing their usefulness in practice and determining what relative weight should be given to each indicator. The idea is to develop the information that a policymaker needs in order to interpret information as it comes in and to choose which indicators to watch depending on the questions being asked. As policymakers typically use indicators one at a time, all of our analysis will be carried out on a bivariate basis. Multivariate regression models allow indicators to play off against one another so that if two indicators hold both common and independent information better statistical fits can usually be obtained by fitting one multivariate model rather than mixing 2 bivariate models. The advantage of using the mixing approach is that when one of the indicators begins to misbehave, which they do, you can, at least temporarily, just ignore that indicator. Second, by only using the primary information over-fitting is less of a worry. Third and most important, the mixing approach allows a much more precise assessment of exactly the type of information is contained in each indicator and thus allows policymakers to reoptimize their choice of indicator sets based on the type of -2- Evans, Strongin, and Eugeni question being asked. Beyond the focus on bivariate models, there are a number of other differences between our work and normal econometric practice that are worth noting. First, as will be shown in the paper different indicators are useful at different forecast horizons, so that we will not be suggesting one best model, but rather we will be suggesting ways of combining indicators depending on the precise policy question being asked. Second, along these same lines as we are more concerned with the interpretation of each of the individual indicators rather than the construction of a structural model of the economy, we will pay much more attention to characterizing the type of information in each individual indicator than is normally the case. Also, since the forecasts derived from the indicators typically get averaged together either informally in the policymaker's mind or formally in the mixing models shown in the last section of this paper, we analyze the degree to which one indicator can be said to have information which is independent from another. Policymakers are often faced with a variety of indicators pointing one way and another group pointing a different way, in such cases it is not only useful to know what weight would have produced the best forecast historically, but the degree to which the indicators are independent bits of information or the same information being repeated over and over again in a variety of guises. Policymakers quite rightly give greater weight to information which they see as independent confirmation. It is useful in this light to more fully analysis the independence of information in various indicators. It is also helpful to know if the indicator in question usually contains the type of information being sought. METHODOLOGY As noted above, the primary focus of this paper is the examination of various data series as indicators of changes in real economic activity, which we measure alternately as annualized log change in real GDP, employment and industrial production. In most cases results are supplied for all three measures of economic activity. -3- Evans, Strongin, and Eugeni The major focus will be on the forecasting real GDP, except in the sections of the paper which focus on issues of timing in which case employment will be used, since it is available at the monthly frequency allowing for more precise estimation of the pattern of impact over time. Throughout the paper the indicators are used to produce forecasts of economic activity. The specific functional form of the forecasting equation is always the same. One year of data for the indicator and one year of lagged economic activity is included in the regression. Thus, the exercise is strictly equivalent to a bivariate VAR with one year of lags, 4 lags for the real GDP models and 12 lags for the employment and industrial production models. The models are estimated in log differences and rates of change are annualized. Interest rates and interest rate spreads are used in their level form. In many of the tables an additional forecast is provided with the label "NONE". In this case, the forecast is based solely on the past history of economic activity, a pure auto-regressive model with one year of lagged data. This pure auto-regressive forecast is referred to as the no-indicator forecast. When the horizon of forecast is varied, we simply change the dependent variable in the regression rather than dynamically iterate the one period ahead forecast. This optimizes the parameterization for the forecast horizon in question, rather than multiplicatively combining estimation errors forward. Symbolically the forecasting equation can be written: y : . r y t =A(L)Ay M + B(D x.^ + co, where Yt it the log of economic activity at time t and It is the indicator at time t, k is the number of periods in the forecast horizon and A(L) and B(L) are polynomial in the lag operator L of order one year. The indicators are split into four groups, which we call families. Each family is meant to represent a natural division of indicators into groups which are likely to share similar characteristics. For example, the first family we examine is -4- Evans, Strongin, and Eugeni interest rates, the second is money-based measures, the third is interest rate spreads and the fourth is composite indicators, such as the Department of Commerce Leading Indicators and the S&P 500. The fourth group also contains those series which do not fit neatly into the overall classification scheme. The idea is to first examine the indicators within a family, finding out which indicators within each family produce the best forecasts and contain the most independent information and then taking these "best" indicators and examining what is to be gained by mixing the information from different families. number of purposes. This serves a First, by breaking the large list of potential indicators into smaller groups it makes each examination more manageable. Second, by using natural groupings it allows us to look at questions such as what is the best interest rate or the best money measure in a natural way. Third, one key issue for indicators is the degree to which they actually contain independent information. Focusing on groups which are already thought to have similar information provides a natural focus to learn if these preconceptions are accurate or if some of these groups contain more than one type of information. Lastly, by first selecting the best indicators at the family level and then mixing between families, we can produce a mixed forecast which, as noted above, closely approximates the way policy forecasting appears to be done in practice. Each family of indicators is subjected to the same analysis. First, each family of indicators is described and a table is presented which lists the indicators examined and their means, standard deviations and their correlations with the measures of economic activity. Then each of the indicators is subjected to four evaluations, 1.) Classical goodness of fit rankings, 2.) Characterization of fit, 3.) Indicators performance in practice and 4.) Encompassing tests. The classical goodness of fit rankings are based on simple full sample regressions estimated on data from the beginning of 1962 through the end of 1991. The results two of each family analysis section. -5- are presented in table Table two shows the rankings Evans, Strongin, and Eugeni for each indicators in the family based on the regression they produce. The idea is that the best indicators are the ones that produces the best fit where fit is measured by the R* of the regression or the standard deviation of the residual from the regression1. This closely approximates the oldest notions of evaluating the best indicators of economic activity for policy. It is also closely linked to the notion of Granger causality, which measures whether or not the indicators actually helps forecast economic activity. The p-value for this test is also included in the table. The second evaluation seeks to characterize the type of information in the indicator. Typically the question can be thought of as if the indicator goes up today how does that change my expectations about economic activity in the future. This is analyzed by calculating the dynamic response path of employment for each of the indicator forecasting equations, which shows how a one standard deviation2 increase in the indicator changes expectations about future growth rate of employment for each month for the next 3 6 months3. This allows us to characterize the information in the indicator based on how fast economic activity responds, how much it responds and how long the change in activity lasts. Figure 1 in each family section graphs the dynamic response path for each indicator in the family, as well as the 2 standard deviation bands on the estimates of the dynamic response paths to show the amount of uncertainty about the response path. 1. In the appendix tables which include sub-sample results are also presented. 2. The standard deviation measure used is the one from a bivariate VAR for the indicator and the measure of economic activity, this is used to approximate the average size of movement in the indicator series. 3 . This is basically the same as an impulse response function except that the identifying assumption is not derived from a specific decomposition of the error matrix, but on the assumed path of the actual series, i.e. the indicator changes given the level of current activity. This is arithmetically equivalent to an impulse response function using a Choleski decomposition with the indicator ordered last. -6- <*****, Evans, Strongin, and Eugeni Table 3 summarizes this information in terms of the maximum response for all three of the measures of economic activity, showing the timing, size and uncertainty of the maximum response of economic activity for each indicator in the family. The third evaluation switches the focus to how well the indicators are likely to work in practice. To this end, goodness of fit is reinterpreted in a way closer to the way forecasts are actually used. First, table 4 shows the goodness of fit ranking recalculated for a series of forecast horizons, so that we can get a better feel for what these indicators are good at. First, the single period horizon used in the first evaluation and then a onequarter horizon, a two-quarter horizon and a one-year horizon4. Table 5 in each section then repeats this analysis using forecasting equations which do not contain any prior information. Specifically, the forecasting equations are estimated sequentially using Kalman filtering techniques using only the sample information available prior to the period being forecast. This provides a more accurate assessment of how an indicator is likely to perform in practice. These forecasts are then ranked by the mean squared error (MSE) of the forecasts from 1972 onward. 2 R s are no longer well defined. The This analysis is followed up by Figure 2 in each section which graphs the cumulative residuals for Kalman forecasts from 1972 onward. This allows us to examine if these forecasts tend to perform badly during recessions or if there was some particular point in the past where they did especially well or poorly. It also tells us if the forecasts have tended to miss in some systematic fashion over time. The residuals are measures as the actual growth in economic activity minus the forecasted growth. Thus, a downward trend in the cumulative residuals would indicate a prolonged period of over- 4. It should be noted that these are not iterated VAR forecasts, rather the forecast parameters are chosen to maximize performance at the forecast horizon specified, this can either be thought of as a state space estimation minimize the t+k forecast variance or as simple OLS with the dependent variable the t+k growth rate. This avoids any problem that might result from a indicator that performs poorly at high frequencies having that failure interfere with longer frequency forecasting. -7- Evans, Strongin, and Eugeni predicting growth in activity. The fourth evaluation switches the focus to independence of information. As noted above one of the most important factors to understand about indicators is whether of not they contain independent information relative to some other indicator. This allows a policymakers to assess whether a new piece of information actually contains any additional information or whether it is simply the same information with a different label. The way to evaluate this is through a set of techniques called encompassing tests. encompass In the context of this paper, indicator A is said to indicator B, if given the forecast implicit based on A there is no additional information in indicator B. said to dominate encompass A. Indicator A is indicator B if A encompasses B and B does not The simplest way to test this is to run a regression with economic activity as the dependent variable and the forecast of activity based on indicator A and the forecast of activity based on indicator B as the independent variables. Symbolically this can be written AGDPr=<|> for(A)t+ (1-$) for(B)t +t Where for(A)t and for(B)t are the forecasts of GDPt based on indicators A and B respectively and <)> is relative weight OLS assigns to for(A)t and for(B)t. If <J> is significantly different from 0 then we can reject that for(A) is encompassed by for(B). Likewise if l-<(> is significantly different from 0 then we can reject that for(B) is encompassed by for(A). If neither is encompassed then both indicators contain independent information and a better forecast can be obtained by mixing both sets of information with the relative weight given by <(>. If only one is encompassed, then it is said to be dominated and only the other is necessary to produce an efficient forecast. If both are encompassed then either indicator alone can produce an efficient forecast, this occurs when there is a very high degree of collinearity and the standard error of the parameter estimate is large. In this case the indicator which has the best historical -8- Evans, Strongin, and Eugeni track record would likely be the superior choice. The generalization to longer horizons is straight forward, though the calculations of the standard errors is more complicated since the errors are no longer independent. Table 6 in each family section contains the encompassing test. The table is read as follows. The indicators are listed both along the top and along the side. The numbers in the table refer to the test that the indicator listed along the side is encompassed by the indicator along the top. The test statistics are the significance levels for the test the indicator along the top does in fact contain all the information in the indicator along the side. For the sake of readability values below .05 are left blank. The way to interpret these tables is that an indicator whose row is blank contains information that is independent of every other indicator in the family. An indictor whose column is full of high numbers is said to encompass those indicators. An indicator that did both would be said to dominate the family. In general, what we will search for is the set of indicators in each family which contain all the information in the family using as few indicators as possible. In general this will mean that the best variable from the previous tests will be included plus additional indicators which contain independent information i.e. the indicators that add the most. Formally this means that all indicators that are not encompassed by any other indicators in the family plus whatever additional indicators are necessary to fully encompasses or cover all the other indicators in the family. This is analogous to finding a set of minimum sufficient statistics. The indicators that make it through this process will then be tested in the mixing model section of the paper in betweenfamily encompassing tests, which examine whether or not there is independent information between families or not. Then a set of "best" indicators will be selected in order to develop a mixing models of indicators which contain independent information for each of the forecasting horizons. These models will contain estimates of the appropriate relative weights that should be -9- Evans, Strongm, and Eugeni applied to the individual indicator based forecasts. Completing the circle of policy forecasts, The mixing model will be timevarying to see if there is any gain from adjusting the weight applied to these individual forecasts based on recent performance. INTEREST RATE LEVELS Table 1.1 lists the nominal interest rates which were selected for investigation, along with some descriptive statistics. rates are expressed at annual rates: All of the the Federal Funds rate (FF), 3- and 6-month Treasury bill rates (TB03 and TB06), 1-, 3-, 5-, and 10-year constant maturity Treasury bond rates (CM01, CM03, CMOS, and CM10), the 3-month Eurodollar rate (EUR03), the 6-month Commercial Paper rate (CP6), and the BAA bond rate (BAAS). Each of these interest rates is negatively correlated with the economic activity variables. The interest rates with the largest absolute correlation with real GDP are the Federal Funds rate, the 3-month Eurodollar rate, and the 6-month Commercial Paper rate. Table 1.2 reports statistics for the one-period-ahead forecasting model. Notice that all of the interest rates provide significant predictive power for all three economic activities. The R2 fall within fairly narrow bands indicating that the relative rankings are not particularly important--all of these indicators are useful at the one-month forecast horizon. Figure 1.1 graphs the response of the employment growth forecast to a one-standard deviation change in information about last period's indicator. As with the F-tests in Table 1.2, the response paths are virtually identical across the interest rates considered: employment growth rises for three or four months and then falls, eventually asymptoting back to zero from below the axis. The confidence bounds on these responses are sufficiently wide that the initial response could be zero. For all of the interest rates, however, there is a point within the first year that employment growth is significantly negative: the largest such effects are for the 6-month Commercial Paper rate and the BAA bond rate. For all of the indicators across all of the activities, the maximum effect is negative and occurs within one -10- Evans, Strongin, and Eugeni year of the impulse. Tables 1.4 and 1.5 rank the indicator forecasts for m sample and out-of-sample forecasting behavior. Focusing on the out-of-sample results first, notice that for industrial production and employment at the one-month horizon, the no-indicator forecasts perform better than the interest rate forecasts. But for GDP all of the interest rate forecasts outperform the noindicator forecasts at all horizons. Focusing on GDP, the Federal Funds rate is ranked first at the four-quarter growth horizon; but the 3-month Eurodollar rate is best at the one- and twoquarter horizons. The success of the Eurodollar rate is also evident for industrial production and employment at all horizons beyond one-month. The 6-month Commercial Paper rate improves in forecasting accuracy as the horizon increases; this is true for GDP, industrial production, and employment (placing no worse than third at the one-year horizon). In general, the shorter maturity bills perform better than the longer maturity bonds (3-, 5-, and 10-year Treasuries). The in-sample results of Table 1.4 indicate that the Eurodollar rate increases in ranking due in part to its out-ofsample stability. In the out-of-sample rankings the Eurodollar rate is first for industrial production (3-, 6-, 12-months), employment (6- and 12-month), and GDP (one- and two-quarters). In 6 of these 7 instances, these represent an increase in ranking from the m-sample results. In contrast to this stability, the 6- month Commercial Paper rate does not fare as well. At the shorter forecast horizons, it goes from being ranked number 1 or 2 insample to either 6, 9, or 10 out-of-sample. For the industrial production and employment, the Federal Funds rate also experiences a substantial out-of-sample forecast deterioration at the shorter forecast horizons relative to the in-sample rankings. The cumulated residuals from the Kalman forecasts in Figure 1.2 show that, overall, the indicators in our interest rate family consistently underforecasted real GDP between 1974 and 1982. The upward trend in the cumulated residuals during this period can be explained in part by an unprecedented increase in inflation, which -11- Evans, Strongm f and Eugeni caused interest rates to rise without the normally anticipated decline in output. On the other hand, between 1983 and 1989, the Federal Funds rate, the 6-month Commercial Paper rate, the Eurodollar rate, and all of the Treasury bill rates performed well, as shown by the flattening of their cumulated residuals slopes during this period. Between 1990 and 1991, however, the indicators performance deteriorated again, as all of the interest rates missed the 1990-91 recession and consistently overforecasted real GDP. Table 1.6 reports the encompassing results for GDP. simplest case is for the 4-quarter horizon: The the Federal Funds rate dominates the other interest rates since it is unencompassed and it encompasses all other interest rates at this horizon. At the one- and two-quarter horizons, however, this domination does not hold; none of the interest rates are unencompassed at these horizons. Since all of the interest rates Granger-cause economic activity in Table 1.2, it is probably not surprising that each of the interest rates contains useful forecasting information. For example, at the one-quarter horizon the Federal Funds rate, the 3month Eurodollar rate and the 6-month Commercial Paper rate all can be said to encompass each other, i.e. if you know one interest rate based forecast knowing another is not much help. Since all of these interest rate forecasts are encompassed by at least one other interest rate forecast, the next criterion for selection is to determine if any one of the interest rate forecasts can cover all of the other interest rate forecasts. In fact, at the one-quarter horizon, the Federal Funds rate, the 3month Eurodollar rate, and the 6-month Commercial Paper rate all cover every other interest rate. The 3-month Eurodollar rate covers the Federal Funds rate and the 6-month Commercial Paper rate with higher levels of significance, and since, as noted above, the 3-month Eurodollar rate was the number one ranked indicator in the out-of-sample forecasts of GDP at the one-quarter horizon, the 3-month Eurodollar rate is selected as the best interest rate level indicator at the one-quarter horizon. -12- Similar Evans, Strongm, and Eugeni reasoning leads to the selection of the 3-month Eurodollar rate for the two-quarter horizon. -13- TABLE 1.1 - DESCRIPTIVE STATISTICS QUARTERLY (Jan 62 - Dec 91) MONTHLY (Jan 62 - Feb 92) Mean Std. Dev. Correlation with Real GDP -0.245 7.370 3.304 •0.353 -0.190 -0.222 6.620 2.686 -0.299 2.647 -0.186 -0.219 6.777 2.622 •0.295 7.265 2.872 -0.173 -0.215 7.282 2.849 -0.282 CM03 7.597 2.746 -0.162 -0.226 7.608 2.737 -0.257 CM05 7.736 2.708 -0.161 -0.231 7.744 2.705 -0.251 CM10 7.866 2.674 -0.154 -0.231 7.869 2.678 -0.237 EUR03 8.033 3.282 -0.224 -ff.254 8.055 3.232 -0.352 CP6 7.341 2.879 -0.223 -0.252 7.359 2.844 -0.342 BAA 9.588 3.108 -0.188 -0.286 9.590 3.120 -0.269 Correlation with Industrial Employment Production Indicator Mean Std.Dev. FF 7.352 3.345 -0.230 TB03 6.605 2.715 TB06 6.761 CM01 , f" TABLE 1.2 - CLASSICAL GOODNESS-OF-FIT STATISTICS MONTHLY (Jan 62 - Feb 92) MONTHLY (Jan 62 - Feb 92) INDUSTRIAL PRODUCTION EMPLOYMENT Change Indicator QUARTERLY (Jan 62 - Dec 91) GDP Change Change , R2 In R2 SEE P Value Rank 6 0.338 0.220 3.148 0.0000 3 0.0028 5 0.293 0.176 3.252 0.0001 6 2.291 0.0015 3 0.304 0.186 3.227 0.0000 5 0.054 2.294 0.0020 4 0.309 0.191 3.216 0.0000 4 0.047 2.307 0.0080 7 0.279 0.161 3.285 0.0002 7 0.268 0.150 3.310 0.0003 8 R2 lnR2 SEE P-Value Rank R2 lnR2 SEE P-Value Rank FF 0.259 0.063 8.965 0.0057 10 0.423 0.050 2.303 0.0052 TB03 0.263 0.067 8.940 0.0030 8 0.426 0.053 2.297 TB06 0.271 0.076 8.887 0.0007 5 0.429 0.055 CM01 0.273 0.078 8.875 0.0005 4 0.428 CM03 0.265 0.069 8.930 0 0022 7 0.421 i — •» ^ CM05 0.263 0.067 8.941 0.0030 9 0.419 0.046 2.310 0.0109 8 0.253 0.136 3343 0.0009 10 CM10 0.265 0.070 8.925 00020 6 0.417 0.043 2.315 0.0171 10 0.354 0.236 3.110 0.0000 1 EUR03 0.276 0.081 8.859 0.0003 3 0.431 0.057 2.287 0.0010 2 0.348 0.231 3.123 0.0000 2 CP6 0.286 0091 8.797 0.0001 1 0.438 0.065 2.273 0.0002 1 0.258 0.140 3.333 0.0007 9 BAA 0.283 0.087 8.818 0.0001 2 0.419 0.045 2.312 0.0124 9 TABLE 1.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS MONTHLY (Jan 62 - Feb 92) MONTHLY (Jan 62 - Feb 92) QUARTERLY (Jan 62 - Dec 91) GDP EMPLOYMENT INDUSTRIAL PRODUCTION Months to Max Max Impact Std. Dev. at Max Months to Max Max Impact Std. Dev. at Max Quarters to Max Max Impact Std. Dev. at Max FF 7 -1.349 0454 10 -0.406 0.129 3 -1.442 0 307 TB03 5 -1.577 0499 9 -0.335 0135 3 -1.250 0 280 TB06 5 -1.610 0.496 12 -0.365 0.130 3 -1.354 0 303 CM01 5 -1.655 0468 12 -0 411 0.123 3 -1.443 0 308 CM03 5 -1.550 0.482 12 -0.414 0.146 3 -1.383 0 326 CM05 5 -1.446 0480 10 -0.431 0141 3 -1.367 0 303 CM10 12 -1.270 0.488 12 -0.382 0.145 3 -1.332 0314 EUR03 5 -1.615 0.475 9 -0.494 0.124 3 -1605 0.290 CP6 5 -1.793 0.484 9 -0.464 0.123 3 -1.502 0.282 BAA 5 -1.973 0486 7 -0 395 0138 3 -1.280 0310 Indicator TABLE 1.4 • MULTIPERIOD FORECASTS (In-Sample) INDICATOR ( 1MON R2 RANK MONTHLY (Jan 62 - Feb 92) MONTHLY (Jan 62 - Feb 92) INDUSTRIAL PRODUCTION EMPLOYMENT 3MOS R2 RANK 6MOS R2 RANK 12MOS R2 RANK 1MON R2 RANK 3MOS R2 RANK QUARTERLY (Jan 6 2 - Dec 91) GDP 6MOS R2 RANK 12MOS R2 RANK lOTR R2 RANK 2QTRS R2 RANK 4QTRS *R2 RANK FF 0258 10 0J351 3 0400 3 0530 2 0423 6 0576 3 0571 3 0561 2 0338 3 0463 3 0 530 1 TB03 0263 8 0333 7 0337 6 0 477 5 0426 5 0564 7 0543 6 0529 4 0293 6 0402 5 0 496 3 TB06 0271 5 0346 5 0353 4 0483 4 0 429 3 0570 5 0 550 4 0528 5 0304 5 0406 4 0 487 5 CM01 0J?73 4 0341 6 0350 5 0 455 6 0 428 4 0567 6 0 547 5 0509 6 0309 4 0397 6 0443 6 CM03 0264 7 0325 8 0329 8 0410 7 0.421 7 0561 8 0540 8 0.486 7 0279 7 0350 7 0377 7 CMOS 0263 9 0318 9 0314 9 0 388 8 0419 8 0560 9 0536 9 0474 8 0268 8 0332 8 0346 8 CM10 0265 6 0307 10 0280 10 0346 10 0417 10 0550 10 0514 10 0442 10 0253 10 0296 10 0307 10 EUR03 0276 3 0371 2 0420 2 0509 3 0431 2 0581 2 0575 2 0546 3 0354 1 0 471 2 0 490 4 CP6 0286 1 0382 1 0438 1 0541 1 0 438 1 0592 1 0592 1 0564 1 0348 2 0 475 1 0516 2 BAA 0283 2 0348 4 0332 7 0361 9 0419 9 0570 4 0543 7 0468 9 0258 9 0329 9 0315 9 NONE 0196 11 0201 11 0115 11 0097 11 0373 11 0.489 11 0414 11 0269 11 0118 11 0123 11 0076 11 >-* •>v4 TABLE 1 5 - KAl MAN MULT1PEWO0 FORECASTS (Out of Sample) M O N T H . Y (Jul 73 Feb 92) INDUSTRIAL PRODUCTION COP EMPLOYMENT CMOS RMSE RANK 12MOS 1MON RMSE RANK RMSE RANK 11 7 141 6 4 778 3 2 761 11 2 162 11 1988 8 601 10 7353 11 4993 6 2707 9 2 129 10 8 8 301 7 7 076 7 4847 4 2644 7 2 074 10 664 7 8 198 5 6898 4 4882 5 2630 5 CMOS 10 577 4 8 154 3 6 897 3 5029 7 2604 CMOS 10 609 5 8229 6 6 973 5 5 131 8 CM10 10629 6 8 349 8 7172 10 5354 EURC3 10483 2 7899 1 6415 1 CPf 11 196 10 8426 9 6 801 BAA 10518 3 8 184 4 7 158 1MON 3MOS INDICATOR RMSE RANK FF 11232 11 8 845 TB03 11 168 9 TB06 10 735 CM01 1 00 1 RMSE RANK QUARTERLY (Jul 73I - D e c 9 l ) MONTHLY (Jul 73 • Feb 92) 12MOS 1QTR RMSE RANK RMSE RANK 2QTRS RMSE RANK 4QTRS RMSE RANK 10 1664 2 3793 2 2859 3 2160 1 2031 11 1728 6 3969 9 3075 6 22G0 5 8 1974 9 1702 4 3862 4 3000 5 2251 4 2 042 7 1932 5 1706 5 3826 3 2996 4 2356 6 3 2 011 5 1913 2 1732 7 3876 5 3094 7 2483 7 2599 2 2 010 4 1920 4 1757 8 3936 7 3144 8 2552 8 9 2625 4 2036 6 1969 8 1826 10 3949 8 3249 10 2683 9 4 531 1 2630 6 1993 2 1847 1 1610 1 3622 1 2754 1 2222 3 2 4 724 2 2752 10 2 124 9 1951 6 1686 3 3880 6 2827 2 2 216 2 9 5 495 11 2657 8 1998 3 1915 3 1784 9 4006 10 3197 9 2 725 10 3MOS CMOS RMSE RANK RMSE RANK o^oddsofioocvjuJs o o o o o o o o o o O ooo>^G5d>o>£»inc40) o o o o o o o o II I I I * - CO I I CO g J J I I (Q l§§§§S§gsf§ o o o o o o o o o o I I I I I I I - I I I I I I J I I I o o o <OCM^-^,»-CM<£>© o o o o o o o o r^ i^ co co co c o W O O B O O " * - co co co <o co o> to 8 8 § r^ 8 8 co !2 3 8 CM oo 0 0 0 0 0 0 0 I I I I I I I m 1 8 I ™ o co i o o o © o © o c o i i o o n u i n © O u to i 8 8 S 2 2 S £ 5$ © CM h- 288S88 «8£ 8 3 8 8 8 3 3 «S8 fc 0 0 0 0 0 0 0 f - w N t f c m CM II l I l 1o 2o2o8o8 ! ^ a C 0 0 0 0 0 0 c © © o o I I I I I I I I I I I I I I I I I I I I I I £ 1 I o I I I I Ireco 1 I c ' I I I I r»* I I I I I I I I I I Iret I c I I o 00 8 581 1 § 1 1 1 1 1 1 1 « & 1 II I C O I ll CM 1 0 0 ! 8$ I 00 o> o o CO 3 2 s I CM ^ ^ t«> O I O O O 1o * 8 I I 10 O CO CM ™ * - <— CM O O c O O O j « re33°i£ , o> J: ® 4 m ^ l CM O O 00 ! ? ^ S 5 ^ ! I 1 1 l 1 1 0 c 0 0 0 0 , ao tn o> in co *3Jn8&£ ^r co co 10 * - c CMinN.CMN.^O>?MCp ^O>(DtO(D^r«-CMC0 0 0 0 0 0 0 0 0 0 © © © © © l j I I I I 00 O 635 i 8 j I! c; ^» l§ • I I O I «£ re o O co N- I O O IO O CM coco*-<53co^^ com<o<o<0CMCMin 0 0 0 0 0 0 0 0 0 1 1 s I I s I I 1 0 8 re 1 j <= l I CO CO CM CM CO * 5 co 8 8 8 CM fc g © © © © © © e o 2 S $ 8 5 8 8 5 S 5 ^ a a m m i o m i A r ^ o o 0 0 0 0 0 0 0 0 0 co n m ^ n i A o O _ 0 0 0 0 0 ^ C ( D < o o 2 o o ^ a c j p < u . h h U U U U U J U f f l U.H-KOOOOUJOCD -19- ttt-i-ooooujom 1.1. Dynamic Response of Employment to Interest Rate Levels Fed Funds (FF) annualized percent growth rates 050 r 5 year Treasury bond (CM05) annualized percent growth rates 050 r -20- 1.2. Interest Rate Levels: Cumulated Kaiman Residuals in Forecasting Real GDP Fed funds (FF) cumulated Kaiman residuals 100 r 5 year Treasury bond (CM05) cumulated Kaiman residuals 75 r 10 year Treasury bond (CM 10) 75 r 6 month Treasury bill (TB06) 100 r 1 year Treasury bond (CM01) 100 r 6 month commercial paper (CP6) 100 3 year Treasury bond (CM03) 100 r 1973 -21- Evans, Strongm, and Eugeni THE MONETARY AGGREGATES Table 2.1 lists the monetary indicators which were selected for investigation, along with some descriptive statistics. For this family of indicators all but one of the variables are expressed as (log) growth rates: the monetary base (Board of Governors (MB) and St. Louis (MBSTL) versions), Ml, M2, M3, L, and long-term debt of nonfinancial institutions, as well as real Ml and real M2 (deflated by the consumer price index). The other monetary indicator is the ratio of nonborrowed reserves (this period) to total reserves (last period) (NBRX). Strongin (1991) has found that this normalized reserve aggregate contains much of the information about monetary policy actions which Sims (1991) attributes to innovations in the Federal Funds rate (orthogonalized relative to output and prices). Two observations about the descriptive statistics seem to be in order. First, these aggregates are plausible choices as monetary indicators of economic activity. Focusing on GDP, the aggregates tend to be correlated with GDP, and the highest correlations are with the real aggregates Ml and M2. In fact, it appears to be roughly the case that as the endogenous component of the monetary aggregate increases, the contemporaneous correlation with economic activity increases. This is loosely the causation/reverse causation debate--do the larger monetary aggregates influence activity more than the narrower aggregates, or are they influenced more? Second, for most of the aggregates the standard deviations are about one-half or less than the average growth rates; however, for real Ml and M2, the standard deviations are 2 and 6 times greater than the average growth rate. In turns out below, that these two aggregates, nominal M2, and the NBR/TR ratio are the most useful indicators. Table 2.2 reports statistics for the one-period-ahead forecasting model, an autoregression of the economic activity variable with lagged values of the indicator included. Focusing on GDP, notice that nominal M2, real Ml, real M2, and the NBR/TR ratio provide significant predictive power for GDP beyond the -22- Evans, Strongin, and Eugeni information contained in past values of GDP. These three indicators consistently provide predictive power for industrial Production and employment as well. For GDP the lowest ranking indicators tend to be nonfinancial debt, the monetary base, and the broad aggregate L. Figure 2.1 graphs the response of the employment growth forecast to a one-standard deviation change in information about last period's indicator. For all of the monetary indicators, a positive impulse eventually leads to a positive growth of employment. For most of these indicators, however, the imprecision of these forecasts is large enough so that the response is either not statistically significant for most of the response path (nominal Ml, M3, L and nonfinancial debt) or entirely insignificant (both monetary bases). have similar response patterns: response being a bit earlier. Real Ml and M2 all persistent and quick, with the Ml The responses of nominal M2 and the NBR/TR ratio are also persistent with a bit more raggedness than the responses to the real aggregates. the longest significant response. The NBR/TR ratio also has For all of the indicators and economic activity variables, the maximum one-period impact occurs within one year (reported in Table 2.3). Tables 2.4 and 2.5 rank the indicator forecasts for m sample and out-of-sample forecasts at various horizons. Turning to Table 2.5 first, notice that for the one-month forecast horizon for both industrial production and employment, the best forecast is one without any monetary indicators. For GDP there are four indicators which consistently provide additional information for forecasts: real M2 (which is always first), the NBR/TR ratio (always second), nominal M2 and real Ml. These indicators are also useful for industrial production and employment for six-month horizon and beyond. They are also the highest ranked indicators in Table 2.4 for the in-sample forecasts. The monetary aggregates which consistently provide no additional predictive power beyond the no-indicator model in the out-of-sample rankings are the two monetary base measures, nominal Ml, and L. They also do poorly in the in-sample rankings. -23- This Evans, Strongm, and Eugeni lack of information is stable across forecast horizons. The cumulated residuals from the Kalman forecasts shown in Figure 2.2 provide another perspective of the out-of-sample performance of our family of money based measures. In our case, the best indicator is again real M2 as its cumulated residuals path clearly stays near zero values, except for isolated periods of large forecast errors in 1978 and 1981, when real M2 underforecasted economic activity. Real M2's performance was again noticeably good between 1990 and 1991, when most of the other money based indicators clearly failed to predict the recession. The NBR/TR ratio was relatively stable from 1973 to 1981, but has shown a consistent pattern of overforecasting output growth since 1982. This deterioration may be due to increasing reluctance on the part of banks to borrow from the discount window. The performance of other monetary aggregates is less reliable and clearly more volatile than the behavior of real M2 and the NBR/TR ratio. For example, the two measures of the monetary base and Ml consistently underforecasted real GDP between 1974 and 1977, as shown by their upward sloping paths. Overall, the path of nominal aggregates plunged during the credit control program of 1980, overpredictmg output growth during the mild recession. From 1983 to 1988, these nominal aggregates performed fairly well, exhibiting uncharacteristic stability, except for Ml which did substantially worse between 1983 and 1984. Finally, between 1990 and 1991, there was a considerable deterioration in the performance of Ml, L, and the two measures of the montary base, as they consistently overpredicted economic growth. Table 2.6 reports the encompassing results for GDP. For each of the forecast horizons, we find that real M2 is not dominated by any of the other forecasts (reading across the real M2 row, the hypothesis is always rejected at low marginal significance levels). None of the other indicator forecasts can cover the information contained in real M2. Furthermore, the real M2 forecasts cover the information contained in all of the other indicator forecasts (reading down the real M2 column, the hypothesis that real M2 covers each forecast is not rejected). -24- Evans, Strongm, and Eugeni Therefore, real M2 is a dominant indicator within the class of monetary indicators selected here for GDP.5 -25- TABLE 2.1 - DESCRIPTIVE STATISTICS QUARTERLY (Jan 62 - Dec 91) MONTHLY (Jan 62 - Feb 92) Mean Std. Dev, Correlation with Real GDP -0.058 6 784 2.195 0.034 -0.021 -0.027 6.662 2.282 0.013 5.864 0 005 -0.033 6.055 3.730 0.157 7.750 4.082 0.119 0.013 7.769 3.292 0.236 M3 8.323 4.072 0.113 0.092 8.363 3.520 0.246 L 8.138 3.662 0.167 0.175 8.183 3.057 0.239 DBTNF 8.977 2.752 0.175 0.290 9.017 2.446 0.180 M1R 1.085 7.245 0.063 0:009 0.971 5.143 0.297 M2R 2.675 5837 0.156 0.053 2.685 4.868 0.353 NBRX 0.976 0.027 0.059 -0.026 0.983 0029 0.154 Correlation with Industrial Employment Production Indicator Mean Std.Dev. MBSTL 6.785 3.617 -0 014 MB 6.710 3.321 M1 6.160 M2 i i TABLE 2.2 - CLASSICAL GOODNESS-OF-FrT STATISTICS QUARTERLY (Jan 62 - Dec 91) MONTHLY (Jan 62 - Feb 92) MONTHLY (Jan 62 - Feb 92) GDP EMPLOYMENT INDUSTRIAL PRODUCTION ___ R2 Change lnR2 SEE P-Value Rank R2 Change lnR2 SEE P-Value Rank R2 Change lnR2 SEE P-Value Rank MBSTL 0.225 0029 9169 0 3997 8 0 393 0 019 2 363 0.5618 9 0.166 0 049 3 532 01744 7 MB 0222 0027 9182 0 4760 9 0.389 0 015 2 370 0.7445 10 0.145 0027 3 577 0 4734 9 M1 0 221 0026 9188 05172 10 0 401 0 028T 2.346 0.2192 8 0.172 0.055 3519 01284 5 M2 0.252 0057 9 003 0.0144 3 0442 0 069 2.264 0.0001 2 0.219 0101 3419 00084 4 M3 0229 0.033 9144 02755 7 0 412 0 039 2.324 0 0383 5 0.169 0 052 3525 01483 6 L 0.236 0041 9.100 01246 5 0 404 0.030 2.341 0.1486 6 0.164 0046 3538 01993 8 DBTNF 0.231 0036 9.128 0.2092 6 0.401 0.028 2.346 0.2156 7 0.124 0006 3 620 09352 10 M1R 0.244 0048 9 054 0.0477 4 0418 0044 2.314 0.0148 4 0.250 0.132 3351 0 0012 2 M2R 0284 0 089 8808 0.0001 1 0444 0.071 2.260 0.0001 1 0.346 0.228 3128 00000 1 NBRX 0 277 0.081 8854 0.0003 2 0 426 0.053 2.297 0 0028 3 0.249 0.131 3.352 0 0012 3 Indicator •27- TABLE 2.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS QUARTERLY (Jan 62 - Dec 91) MONTHLY (Jan 62 - Feb 92) MONTHLY (Jan 62 - Feb 92) GDP EMPLOYMENT INDUSTRIAL PRODUCTION Months to Max Max Impact Std. Dev. at Max Months to Max Max Impact Std. Dev. at Max Quarters to Max Max Impact Std. Dev. at Max MBSTL 4 1.054 0489 8 0 221 0148 2 0 787 0 323 MB 10 0.894 0508 5 0192 0.138 2 0 410 0300 Ml 7 1.145 0.515 3 0 370 0.126 2 0 671 0328 M2 7 1.755 0 476 9 0 705 0.137 2 0904 0309 M3 7 1.393 0538 9 0 500 0149 3 0 787 0344 L 7 1655 0507 9 0458 0.143 3 0 739 0333 DBTNF 2 1.214 0447 5 0 332 0.126 4 0113 0 301 MIR 7 1.371 0.513 5 0 485 0.124 2 1011 0 321 M2R 7 1567 0464 5 0 568 0.128 2 1069 0 289 NBRX 12 1.467 0496 8 0 449 0.148 3 1047 0308 Indicator TABLE Z4 - MULT1PERIOD FORECASTS (In-Sample) Indicator l 1MON R2 RANK MONTHLY (Jan 62 - Feb 92) MONTHLY (Jan 62 - Feb 92) INDUSTRIAL PRODUCTION EMPLOYMENT 3MOS R2 RANK 6MOS R2 RANK 12MOS R2 RANK 1MON R2 RANK 3MOS R2 RANK QUARTERLY (Jan 62 - Dec 91) GDP 6MOS R2 RANK 12MOS R2 RANK 10TR R2 RANK 2QTRS R2 RANK 4QTR$ R2 RANK MBSTL 0225 8 0230 8 0144 10 0113 10 0393 9 0506 8 0434 8 0274 10 0166 7 0154 8 0102 8 MB 0222 9 0226 9 0145 8 0135 8 0389 10 0499 9 0424 9 0276 9 0145 9 0144 9 0121 5 Ml 0221 10 0259 7 0199 6 0127 9 0401 8 0523 7 0454 7 0288 7 0172 5 0183 7 0096 10 M2 0252 3 0335 3 0333 3 0268 4 0442 2 0581 2 0540 2 0394 4 0219 4 0249 4 0186 4 M3 0229 7 0274 5 0211 5 0165 5 0412 5 0546 5 0487 5 0324 5 0169 6 0189 5 0107 7 L 0236 5 0269 6 0195 7 0146 7 0404 6 0527 6 0461 6 0291 6 0.164 8 0184 6 0097 9 DBTNF 0231 6 0225 10 0144 9 0151 6 0401 7 0496 10 0420 10 0285 8 0124 10 0133 10 0121 6 M1R 0244 4 0320 4 0304 4 0282 3 0418 4 0557 4 0517 4 0 415 3 0250 2 0288 3 0244 3 M2R 0284 1 0413 1 0478 1 0567 1 0 444 1 0609 1 0604 1 0573 1 0346 1 0447 1 0514 1 NBRX 0277 2 0365 2 0386 2 0417 2 0426 3 0562 3 0525 3 0446 2 0249 3 0327 2 0292 2 NONE 0196 11 0201 11 0115 11 0097 11 0373 11 0489 11 0414 11 0269 11 0118 11 0123 11 0076 11 i TABLE 2 5 KA11AAN MULTIPfniOO FORECASTS (OiH^>l San**>) MONTHLY (Jul 73 •F«b92) INDUSTRIAL PRODUCTION 1MON kxfcatar RMSE RANK MBSTL 10406 MB Ml 10275 10 452 B 5 10 6 M2 10 366 M3 10447 9 3MOS RMSE RANK 8317 8 142 8 218 7780 8 161 It 5 10 3 12MOS 1MON RMSE RANK RMSE RANK 7468 11 5 736 8 6 5 594 7273 6648 8 3 3MOS RMSE RANK CMOS RMSE RANK 12MOS 1QTR 2QTRS RMSE RANK RMSE RANK RMSE RANK .4QTRS RMSE RANK 2583 6 2040 to 2058 11 2 012 9 4 108 7 3 474 10 2904 8 7 2553 5 2011 6 2023 7 1986 7 4 114 8 3426 7 2840 7 5 760 9 2587 8 2022 7 2030 8 2023 10 4 149 10 3455 9 2992 11 5 425 4 2585 7 1931 3 1896 3 1894 3 3944 3 3252 3 2809 4 8 4 073 5 3394 6 2948 10 11 4 136 9 3432 8 2926 9 1955 6 4 242 11 3 495 11 2820 6 4 097 6 3285 4 2 775 3 3674 1 2844 1 2219 1 2 3003 2 2550 2 5 2006 6 1997 2028 8 ^050 10 2043 2034 9 2041 11 2064 11 2004 5 1905 4 3 1838 1 1731 1 1558 • 2 1903 2 1628 2 1711 8 7308 9 5 843 10 2610 9 7382 10 5843 11 2630 10 5 498 6 2535 4 10405 7 8 174 9 DBTNF 10112 4 8 157 6 7 270 7 MIR 10630 11 8 159 7 6952 4 5309 3 2671 M2R 10111 3 7368 2 5902 1 4229 1 2483 NBRX 9947 2 7362 1 6096 2 4594 2 2477 L GDP EMPLOYMENT CMOS RMSE RANK 7244 QUARTERLY (Jul 73 - &o c 9 l ) MONTH Y (Jul 73 r < * » 2 ) 1998 9 2 3799 I o I 3358 §3 X > ' - O O O O O O O O O ^8 i "» S trt 8 S$2> 8 8 3 © . I ^- c\» r ^ 0) ^ ^ s » S <r> o o o © oo 0 0 *•"* "m CO f*» ^~ o o o o o o o o o o I t ^ o n o t n w o o o o o o o o co CM oo oo r»» CM o> i © o o o o o o o o § c S S 5 > 0 ^ r ^ Q O | m c o ^ o ^ f ^ c n c o c o ™ c M N - O f l O ^ J O N CO o o o o o o o o co c SK8gRf2 O l C O C O ' V O O l O D O O n o o o o o o o o c o o o o o o o R t £ 8 o> o> oo (C co o> ^» o *"* o o o o o o o o o o P ^ o o o o 5 o > o o S o 3 J?^r c o o o o o o o o o © ( 0 CO CO CO 'Cf T CD Q N W N N 2> CD — CO CM K o o o o o o o fc ~- o o o o o o ! lie I ! ! i i c o o o C ' » » 1 1 rc i o o © c l i^ 1 l I I I I I 1 t (0 1 I I l £ l l N -. 8 -J *S <c co m c t-s ^ 83 i ! S«8 ! i ! © o o © fc © . I l/)CVI I I N W ffl I ' co <£> ™ ' O) . CM O , , , s 5! 3 28 ! ! ! o o o *- o « °1 CO tO if) i i * i o o o o oo © © ^ ^ **"• CO CD ~ •- w tc n o in o> i 00 <D <£> ) : tD OJ <£) ^•^ *•»» c *_. ^ s —^ • 1 « / W tf) 00 O) I i £ 2 I ; 88 o o o Ul i ? 8 • ! 2 2 $ !I o o I c S 2 c o oo §2 'SS! .S88 IOMO) *vi « s * v # « »/•» rvi o oo o oo 8 o 5 <^ CO ' 3 CM 85 £^ I co *- ^fc I -v8«! o i o o o o o !S o E 8- c i 8 i I I c i i I I u-> I £ I I I I o • • I I I l I l 8 I I "* I s 8§R i o o o i o a o »a r- o o o 8 S 2 5 i '-'-«> i o oo I I o l i l i re co 5 I vn CM 8 :: co <o i ** \n *- .2 5 51 z x P cc cc cc en CD CO ^ 2 2 2 eo CD * - C M CO w ? _ cc cc cc CD CD * - CM o CD ^ - _ 2 2 2 2 2 - J Q 2 2 Z 2 -J 0 2 2 Z -31- S2 ~ O CD i - N P P ccc c ccc c cic n CD <- CM GO 2 2 2 2 2-J Q 2 2 Z o z 2.1. Dynamic Response of Employment to Money Based Measures St. Louis monetary base (MBSTL) annualized percent growth rates 050 r 10 Nominal L (L) annualized percent growth rates 0 75 r 15 20 months 15 20 months -32- 25 2.2. Money Based Measures: Cumulated Kalman Residuals in Forecasting Real GDP St. Louts monetary base (MBSTL) Nominal L (L) cumulated Kalman residuals 75 i - cumulated Kalman residuals 75 50 h 50 25 25 •25 -25 » i •50 i i i i i i i i i i i i i i i « i FRB monetary base (MB) 75 Nominal nonfinancial debt (DBTNF) 75 r 50 J- 50 25 h 25 0 -25 h v v" t i i i i r i I 1 I I I I I I I I I I I I I I I 1 I •50 f v\/-~~v j k/^ X/ i •50 i t t i i i i i i i f Nominal M1 (Ml) 75 r L„l I I I I I I I I 1 I I 1 1 1 I 1 I RealMI (M1R) 75 r 50 f25 U -25 •50 Kr'Vv^v-v • i i i i i t i i i i i i i i i i i i Nominal M2 (M2) Real M2 (M2R) 50 100 r- V ^ / v •25 ' ' ' ' ' ' ' » ' * ' ^KT ' ' ' ' ' ' I I I "85 I I I "88 ' * ' NBR/TR ratio (NBRX) 25 1973 76 79 r 1973 -33- •J I I 76 I I 79 '82 '91 Evans, Strongin, and Eugeni INTEREST RATE SPREADS Recent research on financial market indicators of economic activity have brought renewed attention to interest rates spreads. Laurent (1988), Bernanke (1990), Friedman-Kuttner (1992), Estrella and Hardouvelis (1991), Kashyup-Stein-Wilcox (1991), and Stock- Watson (1989) have suggested and tested various interest rate spreads as predictors of economic activity with significant success. The idea behind most of these spreads is that the difference in yields between two different debt instruments provides information beyond that in the level of interest rates. The two primary types of interest rates spreads that have been used are risk-spreads which measure the difference in yield between a private debt instrument and the yield on a government bond of equivalent maturity and term-spreads which measure the difference in yield of government debt instruments of different maturities. Typically, the motivation for the risk spreads is that the risk in the private debt instrument is a measure of the market's assessment of the near term risk in the relevant business environment and that high risk implies a tough time for business ahead. Friedman-Kuttner have argued that this interpretation is probably flawed since the spreads are typically too large to be explained by any reasonable estimate of the risk inherent in the private debt instruments and suggest that liquidity considerations play a significant role in the pricing of public-private spreads. Following their lead, we also will refer to these spreads as public-private spreads. The term-spreads seek to measure the relative availability of credit through time. The convention is that the shorter maturity yield is subtracted from the longer. Thus, a positive spread would indicate that short term funding is available at a lower rate than longer term funding. The normal interpretation is that if short-term funds are especially cheap relative to longterm funds this will encourage borrowing and economic activity. An alternative explanation is that the higher long-term yields are -34- Evans, Strongm, and Eugeni signaling expectations of higher future credit demand resulting from increased economic activity. A third interpretation is that by taking the difference between a short- and long-term interest rate you are correcting the shorter term rate for changes in inflationary expectations and taxes, leaving a better measure of short-run credit conditions. In any case, all of these term- spread measures have the counter-intuitive implication that a rise in long-term interest rates is good for the near-term outlook of the economy. Estrella and Hardouvelis (1991) and Strongin (1990) attempt to reconcile the term-spread results with current theory with limited success. We test 3 public-private spreads and 5 term-spreads5. The specific measures we use are the TED or Eurodollar spread which is the 3-month Eurodollar rate minus the 3-month Treasury bill rate. The Commercial Paper spread which is the 6-month Commercial Paper spread minus the 6-month Treasury bill rate, and the Baa spread which the Baa yield minus the 10-year Treasury bond rate6. The five term-spreads contain three spreads based on the Federal Funds Rate, a short, medium and long spread -- the short spread is the 3-month bill rate minus the Federal Funds rate -- the medium spread is the 12-month bill rate minus the Federal Funds rate -the long spread is the 10-year bond rate minus the Federal Funds rate. There are two intermediate spreads as well the 12-month/3- month spread and the 10-year/l-year spread. Table 3.1 shows that as expected the public-private spreads all show a strong negative correlation with economic activity and the term-spreads all show a positive correlation with activity: the shorter the term-spread, the higher the correlation. Table 3.2 indicates that based on classical measures of fit 5. These are the only commonly used spreads available for the entire data sample. Other spreads are examined in the appendix for shorter samples, but the results are no different and the publicprivate spreads presented either dominate or at least equal any of those presented in the appendix. 6. The 10-year rate is used because the 7-year which might be preferred is not available for a sufficient time span. The appendix also includes the Baa/7-year spread and spread's using the AA bond yield. -35- Evans, Strongin, and Eugeni all of the spreads do fairly well in explaining all three measures of economic activity. The R2s for industrial production range from .236 to .339; the range for employment growth is .416 to .459; and the range for GDP is from .234 to .339. With the exception of the 12-month/3-month term-spread, every spread Granger causes activity at a high level of significance. The only exception is the 12-month/3-month spread which fails to Granger cause industrial production. The public-private spreads do a better job of predicting employment and industrial production with the Commercial Paper spread and the Baa spread ranking 1 and 2. For GDP the results are more mixed with the Commercial Paper spread and 12-month/Federal fund spread coming in 2nd. The dynamic response path graphs in Figure 3.1 shows substantial difference in the dynamic response of employment growth by type of spread. The response of employment to an increase in the Baa spread shows a quickly rise, peaking at only 3 months. The response then dies just as quickly. The response paths for the two shorter-term public-private spreads the Commercial Paper spread and the Eurodollar spread build rapidly then plateau for a number of months and then die quickly. The term-spread response paths, with exception of the 12-month/3-month spread, all build slowly, peak and then slowly die out. Only in the case of the Baa spread is there a well-defined peak in the response path, all of the other spreads show extended period of impact. This would suggest that the strength of the Baa spread will be in very short horizon forecasts, the strength of the Commercial Paper and Eurodollar spreads will be at short and middle horizons, while the strength of the term-spreads will be in longer term forecasts. Table 3.3 suggests similar conclusions with the Baa spread showing the quickest, largest and most tightly estimated peak for employment and industrial production. The longer horizon GDP results show the impact of the Baa spread falling off considerably though it is still very quick. Tables 3.4 and 3.5 strongly re-enforce these conclusion and provide some startling evidence on the effect of forecast horizons on indicator performance. First, in table 4 it is clear that the -36- Evans, Strongm, and Eugem performance of the Baa spread falls off dramatically as the forecast horizon is increased. Ranking 2nd for industrial production and employment at the one-month horizon, the rank drops to 6th and 7th for industrial production and employment respectively for the three-month horizon and is dead last by six months for all measures of activity. The Commercial Paper spread, on the other hand, does very well ranking 1st until the one-year horizon in both employment and industrial production, when it is superseded by a number of term-spreads. In forecasting GDP the Commercial Paper spread still does very well at the one-quarter horizon, but fades quickly falling to 4th at the six-month horizon and 6th at the one-year horizon. The Federal Funds rate based spreads do very well as the forecasting horizon lengthens. Starting out in the middle to back of the pack at the shortest horizons they rise to dominate the top of the ranking at the oneyear horizon with the 12-month/Federal Funds spread rising to 1st for all three measure of activity. The intermediate spreads rarely do well. Table 3.5, showing the out-of-sample results shows a very similar story in terms of rankings. The interesting additional fact is how well the spread models stand up to the no-indicator model. At every horizon except one-month the spread models strongly outperform the no-indicator model, though at the onemonth horizon the no-indicator model does outperforms all of the spread models except the Baa spread, which is only good at short horizons. Clearly the forecast horizon is extremely important to the evaluation of interest rate spread models. The cumulated residuals from the Kalman forecasts in Figure 3.2 show some striking similarities in the overall forecasting performance of our family of interest rate spreads. Except for the 3-, 6-, and 12-month/Federal Funds rate spreads, all of our spreads tend to overforecast real GDP, as shown by their consistently negative residuals. While the 3-, 6-, and 12- month/Federal Funds rate spreads performed fairly well from 1973 to 1980, they clearly failed during the last three recessions. In fact, they all underforecasted economic activity between 1980 and -37- Evans, Strongin, and Eugeni 1982, and then overpredicted real GDP between 1990 and 1991. Between 1982 and 1989, their path was conspicuously flat. This suggests that these spreads do well in forecasting normal periods of economic activity, but periodically fail in predicting recessions. Although the 5-year/ and 10-year/Federal Funds rate spreads follow a similar pattern between 1973 and 1981, after 1982 their cumulated residuals path never stabilized but plunged to persistently negative values. Our intermediate term spreads (12-month/3-month and 10-year/1-year spreads) failed during all of the recessions in our sample period (including the 1973-1975 recession), and developed a consistently negative bias after 1982, as they clearly overpredicted real GDP. All of the private/public spreads followed the same general pattern of mediocre performance from 1973 to 1981, and persistent overprediction of economic activity thereafter. In general, we conclude that, although a persistent bias in forecasting exists in all of the interest rate spreads we investigated, some of them did fairly well during most of our sample period, but failed during periods of large scale financial restructuring. The encompassing tests in Table 3.6 are exactly what would be expected given the previous results. To fully encompass all of the information in the interest rate spreads it is usually necessary to include both a public-private spread and a termspread. Also not surprisingly, the Commercial Paper spread and the 12-month/Federal Funds rate spreads dominate their respective groupings at the one- and two-quarter horizons. It is interesting to note that the Stock-Watson leading indicator index, which was designed to fit data at the six-month horizon, chose the Commercial Paper spread and the 10-year/l-year spread. For our sample period, the 12-month/Federal Funds spread narrowly dominates the 10-year/l-year spread. At the 4-quarter horizon the public-private spread no longer contains additional information beyond that contained in the 12-month/Federal Funds spread. The 12-month/Federal Funds spread, however, does not dominate since it fails to cover (only) the 10-year/l-year spread. We selected the 10-year/Federal Funds spread since it covers more spreads than the -38- Evans, Strongm, and Eugeni 10-year/l-year spread, covers the 10-year/l-year spread, and performs better out-of-sample. The selection of two term spreads is consistent with the previously noted results that the publicprivate spread do not contain as much long run information as the term-spreads. It is interesting to note that examination of the entire encompassing results indicate that the separation between the public-private spreads and the term-spreads is not very clear. At some horizons some term-spread encompass some public-private spreads while at other horizons the results reverse. This would seem to indicate that there are common multiple driving forces in the determination of these spreads and that those driver factors associated with longer horizon economic activity predominate in the term-spreads with the common factors that drive short-run performance dominate the public-private spreads. -39- TABLE 3.1 - DESCRIPTIVE STATISTICS QUARTERLY (Jan 62 - Dec 91) MONTHLY (Jan 62 - Feb 92) Mean Std. Dev. Correlation with Real GDP 0.252 -0.750 0.807 0449 0.292 0.255 -0 593 0.886 0.442 1.149 0 291 0.254 -0.580 1.082 0.425 0.384 1586 0.210 0.123 0.373 1511 0.321 CM10FF 0.514 1.791 0 200 0.114 0.499 1.713 0.309 TB12TB3 0.170 0.468 0.177 0.158 0.170 0.434 0225 CM10CM1 0.601 1.036 0083 0.001 0.588 0.997 0.170 EUR0TB3 1.428 0.931 -0.235 -0.248 1.434 0.885 -0378 CP6TB6 0579 0.489 -0.305 -0.297 0.582 0.461 -0.431 BAACM10 1.722 0 698 -0 248 -0.390 1.720 0.690 -0.297 Correlation with Employment Industrial Production Indicator Mean Sid. Dev. TB3FF -0747 0.864 0.291 TB6FF -0 591 0.948 TB12FF -0.577 CM05FF i .** o i TABLE 3.2 - CLASSICAL GOODNESS-OF-Ftf STATISTICS MONTHLY (Jan 62 - Feb 92) MONTHLY (Jan 62 - Feb 92) INDUSTRIAL PRODUCTION GDP EMPLOYMENT R2 Change lnR2 SEE P-Value Rank 3 0.327 0.209 3.174 0.0000 3 00012 5 0.321 0204 3.187 0.0000 4 2 291 0 0016 6 0.330 0.212 3.166 0.0000 2 0.042 2.318 0.0224 10 0.302 0.185 3.231 0.0000 6 0.416 0 043 2.316 0 0186 9 0.309 0191 3.216 0.0000 5 10 0 424 0.050 2.302 00047 7 0.238 0.120 3.377 0.0026 9 0.0207 9 0.417 0 044 2.315 0.0163 8 0.284 0.166 3.273 0.0001 8 8 870 0.0004 6 0.431 8.058 2.286 0.0009 4 0294 0.177 3.250 0.0001 7 0.144 8.462 0.0000 1 0 459 0.086 2.229 0,0000 1 0.339 0.221 3.145 0.0000 1 0.108 8.691 0.0000 2 0.437 0.064 2.274 0.0002 2 0234 0.116 3.386 0.0033 10 R2 Change lnR2 SEE 3 0 435 0.062 2.278 0 0004 0.0002 4 0.430 0.057 2.289 8.866 0.0004 5 0.429 0 055 0.061 8.981 0.0084 7 0 415 0.254 0.059 8.992 0.0111 8 TB12TB3 0.236 0.041 9.099 0.1224 CM10CM1 0250' 0.054 9.018 EUROTB3 0.274 0 079 CP6TB6 0.340 BAACM10 0.303 R2 Change lnR2 SEE TB3FF 0 291 0 095 8 769 0.0000 TB6FF 0.280 0.085 8.834 TB12FF 0.275 0.079 CM05FF 0.256 CM10FF Indicator QUARTERLY (Jan 62 - Dec 91) P-Value Rank P-Value Rank TABLE 3.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS MONTHLY (Jan 62 - Feb 92) QUARTERLY (Jan 62 - Dec 91) MONTHLY (Jan 62 - Feb 92) GDP EMPLOYMENT INDUSTRIAL PRODUCTION Months to Max Max Impact Std. Dev. at Max Months to Max Max Impact Std. Dev. at Max Quarters to Max Max Impact Std. Dev. at Max TB3FF 7 1591 0483 6 0.500 0.127 3 1.408 0.310 TB6FF 7 1.731 0468 6 0466 0.128 3 1.283 0 277 TB12FF 7 1.446 0466 6 0.412 0.125 3 1.259 0.285 CM05FF 7 1.091 0478 15 0.355 0.087 3 1.195 0.294 CM10FF 7 1.022 0.497 9 0.370 0.134 3 1.243 0.293 TB12TB3 14 1.375 0382 14 0534 0.111 5 0879 0 282 CM10CM1 5 1.471 0486 9- 0.342 0.134 3 1.243 0.296 EUROTB3 7 -2.083 0515 12 -0.553 0.146 3 -1.493 0.338 CPCTB6 9 -2.476 0480 8 -0.711 0.136 3 -1.449 0.312 BAACM10 3 -2.645 0473 3 -0.519 0.121 2 -0 833 0.312 Indicator i £ ' TABLE 3.4 - MULT1PERI0D FORECASTS (In-Sampla) Indicator 1MON R2 RANK MONTHLY (Jan 62 - Fab 92) MONTHLY (Jan 62 - Fab 92) INDUSTRIAL PRODUCTION EMPLOYMENT 3MOS R2 RANK 6MOS R2 RANK 12MOS R2 RANK 1MON R2 RANK 3MOS R2 RANK QUARTERLY (Jan 62 - Dae 91) GDP 6MOS R2 RANK 12MOS R2 RANK 10TR R2 RANK 2QTRS R2 RANK 4QTRS? R2 RANK TB3FF 0291 3 0402 2 0477 2 0524 3 0435 3 0591 3 0585 4 0549 6 0327 3 0446 3 0437 5 TB6FF 0280 4 0382 3 0456 3 0547 2 0430 5 0590 4 0595 3 0506 2 0321 4 0 459 2 0490 4 0275 5 0369 S 0438 4 0555 1 0429 6 0593 2 0604 2 0526 1 0330 2 0 470 1 0518 1 0334 7 0381 7 0488 4 0415 10 0569 7 0569 6 0585 4 0302 6 0 428 6 0498 2 6 0484 5 0.416 9 0572 5 0574 5 0567 3 0309 5 0 435 4 0491 3 TB12FF CM05FF 0256 7 CM10FF 0254 8 0337 6 0385 TB12TB3 0236 10 0286 10 0269 9 0358 8 0424 7 0567 8 0563 7 0572 5 0238 9 0333 9 0383 7 CM10CM1 0250 9 0305 9 0296 8 0361 7 0417 8 0552 10 0526 9 0 489 8 0284 8 0374 7 0396 6 EUROTB3 0274 6 0380 4 0413 5 0323 9 0431 4 0571 6 0542 8 0431 9 0294 7 0364 8 0230 9 CP6TB6 0340 1 0500 1 0559 1 0426 6 0459 1 0J634 1 0623 1 0506 7 0339 1 0 429 5 0289 8 BAACM10 0303 2 0324 8 0218 10 0168 10 0437 2 0553 9 0461 10 0293 10 0234 10 0.175 10 0138 10 NONE 0196 11 0201 11 0115 11 0097 11 0373 11 0.489 11 0414 11 0269 11 0118 11 0.123 11 0076 11 -P* I TABLE 35 - KALMAN MULTIPERIOO FORECASTS (Out of Sample) MONTHLY (Jul 73 -Feb 92) QUARTERLY (Jul 73 • Dec 91) MONTHLY (Jul 73. Feb 92) GDP INDUSTRIAL PRODUCTION Indicator 1MON RMSE RANK • MOS RMSE RANK 12MOS RMSE RANK 1MON RMSE RANK TB3FF 10161 5 7 421 3 5 763 2 4209 3 2581 5 TBSFF 10371 6 7768 4 6075 4 4081 1 2 624 7 TB12FF 10 582 8 8 164 8 6539 5 4 207 2 2 658 9 CMD5FF 10931 9 8386 9 6 742 7 4 517 4 2 708 10 CM10FF 10970 10 8 390 10 6656 6 4 542 5 2 740 11 TB12TB3 11054 11 8539 11 7169 11 4906 7 2653 8 CM10CM1 10 572 7 8 149 7 6891 8 5053 9 2622 6 EUROTB3 9898 3 7289 2 5937 3 5048 8 2 512 3 CPCTB* 9930 4 6703 BAACM10 9858 1 7 785 3MOS RMSE RANK 3MOS $MOS 12MOS 1QTR 2QTRS 4QTRS RMSE RANK RMSE RANK RMSE RANK RMSE RANK RMSE RANK RMSE RANK 1937 4 1785 3 1599 6 3609 1 2674 1 2253 5 I960 6 1775 2 1492 3 3691 3 2691 2 2081 2 1986 7 1791 4 1435 1 3 753 6 2 754 3 2015 1 2055 9 1866 8 1504 5 3 745 5 2811 6 2111 3 2063 11 1853 6 1495 4 3763 7 2785 5 2161 4 2030 8 1866 7 1479 2 4 197 11 3187 9 2 370 6 2055 10 1937 9 1674 7 3857 8 2970 8 2389 7 9 3698 4 2886 7 2 721 8 2 Z760 4 1926 1 4922 1 4806 6 2543 7025 9 5 575 11 2446 1834 5 1779 1632 1 1710 8 3656 2 744 9 1913 2 1972 11 1998 11 3963 9 3485 11 2846 11 1948 5 1953 10 1913 10 4 015 10 3 358 10 2819 10 1 790 5 3 4 1 1 I J> I 2463 2 TABUE 3.6 - MULUPERIOD ENCOMPASSING TESTS (Sample Period: Jan 62 - Dec 91) Probability Value for Null Hypothesis: X is Encompassed by Y GDP 1 qtr Y TB3FF TB6FF TB12FF CM05FF na 0462 0105 0 185 na 0 227 0 109 0062 0 333 0167 0 999 na 0389 0 231 0 797 0 106 — — — CM10FF TB12TB3 CM10CM1 EUROTB3 CP6TB6 BAACM10 Maximum P Value X __ TB3FF TB6FF TB12FF CM05FF CM10FF TB12TB3 CM10CM1 EUROTB3 CP6TB6 BAACM10 — — 0093 — 0125 0066 — — — — — — na 0215 0 430 0 454 — — — — — — — 0 540 na 0413 0 699 0098 0077 0115 na __ — — — — -- 0 186 na 0 104 0 185 0999 0 227 0 540 0 231 0 797 0699 0 186 0066 0 104 0 222 na 0936 0569 0798 0 155 0665 0 271 0 755 0 876 0 222 0 093 0 991 0560 na 0 575 0548 0197 0 027 0 176 0 720 0 576 0593 0989 0883 0 973 0 053 — — 0055 — na 0 072 0 156 0 109 — 0053 — GDP: 2qtrs i .£* en i TB3FF TB6FF TB12FF CM05FF CM10FF TB12TB3 CM10CM1 EUROTB3 CP6TB6 BAACM10 na 0070 — — — — — 0214 0093 0545 0 569 na 0155 0092 0055 0256 0 467 0 798 na 0337 0206 0 755 0126 na 0 271 0 370 0 876 0 665 na 0 353 0 710 — — na — 0071 na na 0 459 0 580 0908 0991 0436 0935 0 807 GDP 4qtrs TB3FF TB6FF TB12FF CM05FF CM10FF TB12TB3 CM10CM1 EUROTB3 CP6TB6 BAACM10 na — — — — — — 0752 0883 0500 0 322 na 0142 0989 0 870 0 392 0 548 0197 na 0 176 0144 0 576 0062 0979 0840 0442 NOTE* Values less than or equal to 0 05 are marked with a dash 0131 0056 0 092 na 0 720 0333 0 593 0899 0 774 0863 0 170 na 0261 0 230 0965 0 783 0930 — —__ — __ — na — 0094 — 0111 na 0428 0115 0973 0 569 3.1. Dynamic Response of Employment to Interest Rate Spreads 1 year T bill less 3 month T bill (TB12TB3) annualized percent growth rata* 0 75 r 3 month T bill less fed funds (TB3FF) annualized parcant growth rates 060 r 050 h 0 25 000 11 11 1 1 1 1 1 1 -0.20 1111111 • ' ' * ' ' ' ' ' ' ' * » • -0 25 10 year T bond less 1 year T bond ( C M 1 0 C M 1 ) 0 72 r 6 month T bill less fed funds (TB6FF) 060 r * * • 12 month T bill less fed funds (TB12FF) 080 3 month eurodollar less 3 month T bill ( E U R O T B 3 ) 0.25 r 5 year T bond less fed funds ( C M 0 5 F F ) 0 72 r 6 month commerciai paper less 6 month T bill ( C P 6 T B 6 ) 0 25 r 10 year T bond less fed funds ( C M 10FF) 060 r BAA corporate bond less 10 year T bond ( B A A C M 1 0 ) 044 r i 11 i i i i i i 11 i i 35 -46- 3.2. Interest Rate Spreads: Cumulated Kalman Residuals in Forecasting Real GDP 3 month T bill less fed funds (TB3FF) cumulated Kalman residuals 100 p 12 month T bill less 3 month T bill (TB12TB3) cumulated Kalman rtftiduais 75 r 75 k 50 ^ 50 U 25 L 25 h •25 -25 t -i i i i i t i -t i i i i i i i -50 6 month T bill less fed funds (TB6FF) 100 r I I I I I I t I t I I 1 I I I 1 » f I 10 year T bond less 1 year T bond (CM10CM1) 25 75 k 50 h 25 0 ^ V W •25 I I I I I I I I I I I I 111 12 month T bill less fed funds (TB12FF) .100 ' * ' ' ' ' 3 month eurodollar less 3 month T bill ( E U R O T B 3 ) 25 100 p 75 I50 U 25 h ^^\A/' -25 t ,1 I I l I I I 1 I 1 1 I I I I I 1 I 5 year T bond less fed funds (CM05FF) 25 r 6 mo. commercial paper less 6 mo. T bill ( C P 6 T B 6 ) 25 r 10 year T bond less fed funds ( C M 10FF) 25 r BAA corporate bond less 10 year T bond (B A A C M 10) 25 r 0 -25 -50 -75 .100 ' ' ' 1973 -47- ' * ' 76 ' ' 79 ' ' ' 82 ' ' ' *85 ' ' '8 •91 Evans, Strongin, and Eugeni COMPOSITE INDICATORS The composite indicator family consists of the NBER experimental leading indicator series (XLI) and the NBER experimental nonfinancial recession index (XRI2) (which measures the probability of a recession), the Department of Commerce leading indicators (lead), the National Association of Purchasing Managers Index (PMI), the change in the S&P 500 (S&P), changes in sensitive materials prices (SMPS), and the Kashyap-Stein-Wilcox "mix" (KSWMIX), which is the ratio of bank lending to the sum of bank lending and commercial paper lending (see Kashyap et al. (1991)). It should be noted that the NBER experimental index includes the 10-year/l-year interest rate spread and the Commercial Paper spread and that the Department of Commerce leading indicator index includes real M2, which have been used in previous sections. All three leading indicator composites are designed to predict economic activity at a six-month horizon, though the optimization for the Department of Commerce index is not as specific as either of the NBER indices. Table 4.1 shows that most of these series have the expected correlation with contemporaneous economic activity, except for the change in the S&P 500 which has small negative correlations with growth in industrial production and employment and only a small positive correlation with growth in real GNP. is positively correlated with real GDP: The KSWMIX variable one interpretation of this correlation is that increased (decreased) bank lending is associated with expansions (contractions). Table 4.2 shows that all of these series perform very well in classical regression analysis. They all produce high R2s. The 2 R s for industrial production range from .289 to .391; the range for employment growth is .434 to .527; and the range for GDP is from .205 to .455. Further, each of these indicators Granger causes activity at a high level of significance. In terms of ranking, the Department of Commerce leading indicators and NBER experimental index are 1st and 2nd for all of the measures of economic activity, with the Department of Commerce leading -48- Evans, Strongm, and Eugeni indicators coming in 1st for industrial production and employment and the NBER experimental index coming in 1st for real GNP. The change in S&P 500 comes in last in every category and the change in sensitive materials prices comes in next to last in every category. The dynamic response path graphs in Figure 4.1 show somewhat similar patterns.7 For all three leading indicators series — the NBER leading indicator, the NBER nonfinancial recession index and the Department of Commerce's leading indicators -- employment growth shows a rapid rise peaking at 5 months. From that peak all three graphs exhibit significantly different behavior. The NBER leading indicator graph plateaus for 4-5 months and then drops off before the end of the year. Employment's response to changes in the NBER nonfinancial recession index drops off steadily from the peak while the response to changes in the Department of Commerce's response path is in-between with a high initial peak followed by a stable period then a steady decline. The response of employment to the changes in the purchasing managers index and the change in sensitive material prices both show dramatic jumps in forecasted employment growth peaking at 3 and 2 months, respectively. Employment growth then falls steadily in the Purchasing Managers Index graph while it plateaus in the sensitive material prices graph. The S&P 500 graph is similar, showing a leap up followed by a steady decline, except it has a small initial drop in the first month. It is interesting to note that all of these dynamic response paths are barely significant at the one year mark, despite showing fairly precisely estimated effects earlier. As a group these series seem to hold a lot of information about short-run changes in economic activity, with most of that information centered at the 3-9 month horizon. Tables 4.4 and 4.5 which examine the forecasting ability of these indicators at different forecast horizons, in-sample and out-of-sample, show very stable rankings as the forecast horizon 7. The dynamic response of employment to the KSWMIX variable is not reported since it is only available on a quarterly basis. -49- Evans, Strongin, and Eugeni changes. The leading indicators series do best, posting very similar performances. The other series do not do as well, though the Purchasing Manager's Index does well at the one-month horizon for industrial production. Placing 3rd in-sample and 2na out-of- sample for the one-month horizon then falling off at longer horizons. Among the leading indicator series, the Department of Commerce series does best at horizons of less than six months, while the NBER index ranks first for horizons of 6 months and longer. For GDP, the NBER index always does better with the differential in performance increasing with horizon. The KSWMIX variable does reasonably well in-sample, but out-of-sample it performs worse than "NONE," the no-indicator forecast. The one anomaly in the tables is that the change in sensitive material prices does very well out-of-sample for GDP at the 4 quarter horizon, actually outperforming all of the other indicators except the NBER leading indicator series. The cumulated Kalman residuals in Figure 4.2 show some striking similarities and some differences in actual performance across these indicators. Except for KSWMIX, all of our composite indicators have overforecasted real GDP over time, as their cumulated residuals are consistently negative. This bias is clearly evident during recessions and becomes more dramatic after 1980. After 1982, while the negative bias is exacerbated in the NBER leading indicator and S&P 500, the path becomes somewhat more stable for most of our indicators. The NBER nonfinancial recession index is our best performer during this period, which is not surprising since the index was originally developed in response to the failure of the NBER leading indicator index to forecast the 1990-1991 recession. The encompassing results in Table 4.6 show that for horizons of two- and four-quarters the NBER index dominates this entire family of indicators, with the possible exception of the KSWMIX. At the one-quarter horizon both the Department of Commerce and NBER nonfinancial recession indices are not encompassed by any of the other forecasts. These results are not surprising in light of the ranking discussed earlier and the fact that the NBER -50- Evans, Strongin, and Eugeni leading indicator index was designed to provide a "best" forecast of economic activity at a six-month horizon, using virtually all of the macroeconomic data available. At the one- and two-quarter horizons, the KSWMIX is encompassed by the NBER index at the 5% significance level, but not the 10% level. We chose not to include the KSWMIX in the survivor list of indicators due to its poor out-of-sample performance in Table 4.5. -51- TABLE 4.1 - DESCRIPTIVE STATISTICS MONTHLY (Jan 63 - Feb 92) QUARTERLY (Jan 63 - Dec 91) Correlation with Industrial Employment Production Mean Std. Dev. Correlation with Real GDP 0.429 3.039 4.067 0.547 -0 556 -0.523 0.156 0.131 -0.649 11.148 0.454 0.249 2.993 8.832 0.600 53.380 7.668 0.524 0.681 53.400 7.473 0.632 S&P 6.463 42.410 -0.028 -0.050 6.451 24.588 0.185 SMPS 0.319 0.930 0.325 0.444 0.322 0.912 0.278 0.925 0.040 0.316 Indicator Mean Std. Dev. XLI 3.070 4.162 0.439 XRI2 0.157 0.139 LEAD 2.990 PMI i en i KSWMIX TABLE 4.2 - CLASSICAL GOODNESS-OF-FIT STATISTICS MONTHLY (Jan 63 - Feb 92) MONTHLY (Jan 63 - Feb 92) QUARTERLY (Jan 63 - Dec 91) EMPLOYMENT INDUSTRIAL PRODUCTION GDP R2 Change lnR2 SEE P-Value Rank R2 Change lnR2 SEE P-Value Rank R2 Change lnR2 SEE P-Value XLI 0.369 0.165 8.353 0.0000 2 0.484 0.104 2.198 0.0000 2 0.455 0.338 2893 0.0000 1 XRI2 0.338 0.134 8.555 0.0000 4 0.483 0.102 2.200 0.0000 3 0.385 0.268 3.073 0.0000 3 LEAD 0.391 0.187 8.204 0.0000 1 0.527 0.146 2.104 0.0000 1 0.405 0.288 3.022 0.0000 2 PMI 0.355 0.152 8.439 0.0000 3 0.463 0.083 2.241 0.0000 4 0.265 0.148 3.359 0.0005 4 S&P 0.289 0.085 8.864 0.0002 6 0.434 0.054 2.301 0.0029 6 0.205 0.089 3.493 0.0222 7 SMPS 0.300 0.096 8.795 0.0000 5 0.436 0.056 2.297 0.0019 5 0.232 0.115 3.433 0.0045 6 0.243 0.126 3.410 0.0023 5 Indicator Rank i en OJ i KSWMIX _ _ _ „ . TABLE 4.3 - MAXIMUM IMPACT OF DYNAMIC MULTIPLIERS MONTHLY (Jan 63 - Feb 92) MONTHLY (Jan 63 - Feb 92) QUARTERLY (Jan 63 - Dec 91) EMPLOYMENT INDUSTRIAL PRODUCTION GDP Months to Max Max Impact Std. Dev. at Max Months to Max Max Impact Std. Dev. at Max Quarters to Max Max Impact Std. Dev. at Max XLI 6 2.441 0.455 5 0.650 0.124 3 1.731 0.310 XRI2 3 -2.668 0.446 5 -0.811 0.137 2 -1.829 0.309 LEAD 5 2.475 0.453 5 0.766 0.132 2 1.851 0.274 PMI 2 2.655 0.454 3 0.617 0.124 2 1.223 0.314 S&P 5 2.310 0.487 7 0.596 0.145 2 0.935 0.322 SMPS 2 1.801 0.468 2 0.401 0.120 7 -0.679 0.246 KSWMIX . . _ _ _ 2 1.021 0.302 Indicator i en i „ TABLE 4.4 - MULTIPERIOD FORECASTS (In-Sample) Indicator 1MON R2 RANK MONTHLY (Jan 6 3 - F e b 92) MONTHLY (Jan 63 -Feb 92) INDUSTRIAL PRODUCTION EMPLOYMENT 3MOS R2 RANK 6MOS R2 RANK 12MOS R2 RANK 1MON R2 RANK 3MOS R2 RANK QUARTERLY (Jan 63 -Dec 91) GDP 6MOS R2 RANK 12MOS R2 RANK 1QTR R2 RANK 2QTRS R2 RANK 4 QTR^ R2 RANK XU 0369 2 0 555 2 0638 1 0 510 1 0 484 2 0675 2 0694 1 0608 1 0 455 1 0568 1 0401 1 XRI2 0 337 4 0419 3 0364 3 0 255 6 0 484 3 0646 3 0584 3 0417 3 0 382 3 0 316 3 0168 6 LEAD 0 391 1 0.569 1 0606 2 0460 2 0.527 1 0 705 1 0656 2 0500 2 0 405 2 0341 2 0 247 2 PMI 0 355 3 0389 4 0356 4 0 255 5 0463 4 0 580 4 0 483 5 0 325 6 0.265 4 0203 7 0173 5 0434 6 0 572 5 0 530 4 0.381 4 0.205 7 0.216 5 0152 7 0 436 5 0 547 6 0479 6 0 345 5 0.232 6 0206 6 0229 3 . 0 243 5 0 249 4 0193 4 S&P 0.289 6 0.364 5 0349 5 0 269 4 SMPS 0.300 5 0.346 6 0327 6 0334 3 . . KSWMIX • NONE I en en i . 0204 . 7 0204 7 0116 - 7 0088 . 7 - 0381 7 0493 - 0417 0282 0.117 0117 0072 TABLE 4.5 - KALMAN MULTIPERIOD FORECASTS (Out-of-Sample) MONTHLY (Jul 73 - Feb 92) MONTHLY (Jul 73 - Feb 92) INDUSTRIAL PRODUCTION EMPLOYMENT 6MOS RMSE RANK 1 2 MO S RMSE RANK 4QTtis 3MOS RMSE RANK XU 9441 3 6 293 2 4586 1 4226 1 2 405 3 1662 2 1481 1 1473 1 3 246 1 2 376 1 2 392 1 XRI2 9589 4 7 353 4 6604 6 5444 6 2 439 4 1801 3 1839 3 1897 4 3 427 3 3 026 3 2 758 5 LEAD 9057 1 6081 1 4899 2 4 449 2 2290 1 1614 1 1632 2 1717 2 3 307 2 3 024 2 2669 3 PMI 9172 2 7.163 3 6156 3 5101 4 2 402 2 1864 4 1935 6 1978 7 3838 4 3 319 6 2 736 4 S&P 9921 7 7.442 6 6 287 4 5124 5 2 522 7 1921 5 1884 4 1882 3 3964 6 3 253 4 2758 6 SMPS 9685 5 7 391 5 6 291 5 4 779 3 2 495 6 1944 6 1926 5 1915 5 3914 5 3 306 5 2612 2 . . . 4 078 8 3 377 8 2846 8 6 4 052 9894 6 7945 1MON RMSE RANK 3MOS RMSE RANK 1MON RMSE RANK NONE 12MOS RMSE RANK GDP Indicator KSWMIX 6MOS RMSE RANK QUARTERLY (Jul 73 - Dec 91) . 7 7125 - 7 5 575 7 2 467 1953 . - 1943 1QTR RMSE RANK 2QTRS RMSE RANK 3369 RMSE 9ANK 2 799 h o 8088R88 E -S •R > ~ CO *•» § r<*8 8co8h» d d d d d o d o o o o o o o d o o o d d d 2* l I 3|R l t ! ! ! o8 I I CO 28 £j CT CM I I I | i 8 & i £ g CM < i $2 5 i I coqwn J ^ o o o o o I I « I I I C | * - i :z i i in . o i 8 « 2 i i CM « j 1 1 1 ^ II I d i i i i CM CO CO ! SS i i ;: co 8SR8 I I d d d d . I I . cd *- <*> cQ <£ I J; CO CO N O I c O O O O 8 , H S S I © ' c d © o" d •o 2 CO CO * - tf> c o o o ' 3 ! 8 <M 8 ! l fc I d o d o o o 73 813 X 1 | o* o o o o "- o> N "«r r*- _, o> p co co N. <o S5S8838 £ 8 3 8 5 ? ; !£ o o o o o o CO d o d d d d s > X XU XRI2 LEAD PMI S&P SMPS KSWMIX TABLE 4.6-MULT1P Probability Value for CD !-*£! £2 < — Q. 0. ; D oc UJ S •» Z < x x 3 CL co co : -57- -o.2f Z • * Z CO a . co co * 4.1. Dynamic Response of Employment to Composite Indicators National Purchasing Managers' Index (PMI) annualized percent growth rates 1.05 r NBER Experimental Leading Index (XLI) annualized percent growth rates 1.00 r 0.70 h 0.35 000 •0.35 SAP 500 Stock Index (S&P) 1.00 r NBER Nonfinancia] Recession index (XRI2) 0.25 -1.25 II II I M l II II I III II III III I III M M I DOA Composite Index of Leading Indicators (LEAD) 1.20 -0.50 Change in sensitive materials prices (SMPS) 0.75 r 0.80 040 0.00 -0 40 -58- 4 . 2 . Composite Indicators: Cumulated Kaiman Residuals in Forecasting R e a l G D P NBER Experimental Leading Index (XLI) S A P 5 0 0 Stock Index (S&P) cumulated Kaiman residuals 25 r cumulated Kaiman residuals 25 r \ / & •25 •50 -75 t..< •100 •100 i • i i i i •• i i i i i i i NBER Nonfinancial Recession Index (XRI2) Change in sensitive materials prices (SMPS) 50 r 50 25 h i t r -25 * ' ' •4—L. I I I I I I. I DOC Composite Index of Leading Indicators (LEAD) Bank lending/(bank lending + CP) ratio (KSWMIX) 50 75 r 50 h 25 -25 National Purchasing Managers' Index (PMI) 50 r N^ .50 | I 1 1973 -59- I I 76 | j t 79 | | | '82 I I I 85 I I I '88 1 1 I '91 Evans, Strongin, and Eugeni MIXING MODELS FOR REAL GDP This section analyzes those indicators drawn from the previous sections that contain independent information and did well in the out-of-sample Kalman rankings. The indicators are subjected to another round of encompassing tests and rankings. Finally the usefulness of these final indicators are assessed in the context of a time-varying forecast-mixing model. Table 5.1 presents the Kalman forecast RMSE for the one-, two-, and four-quarter horizon forecasts of real GDP. For the one-quarter horizon the best indicators are the NBER composite indicators (XLI and XRI2), and the Department of Commerce Leading Indicators Index (LEAD). The spreads and real M2 do the worst at this short horizon, but all of the remaining indicators do contribute information beyond the own past history of GDP (NONE). At the two-quarter horizon, the best indicator is the NBER leading indicator index with the 12-month/Federal Funds rate spread coming in a distant second: the NBER leading indicator index is 14% more accurate than the 12-month/Federal Funds rate spread. This is not surprising since the NBER leading indicator index was constructed by Stock and Watson to produce the "best" forecast of the growth in economic activity over the six-month horizon considered here. Turning to the four-quarter horizon, it seems surprising that the NBER leading indicator index comes in last after the 12month/Federal Funds rate spread, the Federal Funds rate, the 10year/Federal Funds spread, and real M2. This demonstrates again that the choice of economic indicators depends critically upon the horizon being forecast-- at the four-quarter growth horizon, a different collection of interest rate spreads than the ones selected by Stock and Watson are useful. New encompassing results are displayed in Table 5.2. At this point, the purpose of these tests is to narrow the list of indicators in a structured manner. However, a rigid adherence to a statistical significance level is not maintained if an indicator is relatively useful and of independent interest. At the one- quarter horizon, the composite indicators the NBER leading -60- Evans, Strongm, and Eugeni indicator index, the NBER nonfinancial recession index, and the Department of Commerce leading indicators are each undominated and together sufficient. The two-quarter horizon is more interesting. Three indicators are clearly necessary. The NBER leading indicator index is undominated, and the 12-month/Federal Funds rate spread is undominated at the 10% level. The 3-month Eurodollar rate is not covered by these two indicators, and it is not dominated at the 11% significance level. included in this final cut for two reasons: Real M2 is also it is only covered by the NBER leading indicator index at the 14% significance level and it is of inherent interest as the best monetary aggregate considered here. Finally, notice that the 6-month Commercial paper spread (CP6TB6) did not make the final list at the twoquarter forecast horizon, but it is a component of the NBER leading indicator index. At the four-quarter horizon, three indicators are undominated: the Federal Funds rate, real M2, and the 12- month/Federal Funds rate spread. The NBER leading indicator index does not contain independent information beyond these indicators. The 10-year/Federal Funds rate spread is included in the final list for three reasons: it is undominated at the 15% significance level, it covers the NBER leading indicators index better than the shorter end of the term structure (12-month/Federal Funds rate spread), and it is interesting to include a long spread at this horizon since Stock and Watson found a long spread useful at the two-quarter horizon. The next step is to combine these forecasts into a forecasting model (for each horizon) which allows the weights on the indicators to vary over time depending upon their recent performance. Essentially we would like the model to take the following form: Ft = <(>u for(A)t^2c £or(B)t++u for(C) t where for(A) represents a forecast based upon indicator A and Ft is the combined forecast. and sum to one: The weights <|>lt should be non-negative in this case, the indicator's weight is a direct -61- Evans, Strongin, and Eugeni measure of its importance for the forecast. When the weights vary over time according to their forecast accuracy, the time path of the weights provide a direct measure of the indicators' reliability over time. way. Let eat2 We implement this model in the following be the sum of (recent) squared forecast errors based upon indicator i's model. In this paper, we take "recent" to be one year of known forecast errors (4 quarters). the average of the £lt2s at time t and m <3vgt(£lt2) over time. Let avgt(elt2) be is the average of £lt2 - Then <(>lt is defined to be: <(>iC = a, - p, (e?t - avg t (e 2 t ) - u,) , a, , p, > 0 where the parameters a and p can be estimated by a linear regression model if the non-negativity constraints are ignored, or nonlinear methods if the constraints are imposed.8 avgt(elt2) - m Since £lt2 - is mean zero by construction, the time-variation due to the P's nets out to zero over time. Consequently, the a estimates represent the average weight associated with each indicator forecast. However, over short periods of time when an indicator's forecast misbehaves, its errors elt2 will be larger than the average errors; this will lead to the indicator's forecast receiving a temporarily smaller weight. Table 5.3 displays the estimated a weights for these models. The one-quarter results indicate that the NBER leading indicator index is the most reliable, having an average weight of .533 in the combined forecast. The other indices (NBER Experimental Recession Index and the BEA Leading Indicators Index) received about equal shares of the remaining weight. are estimated to be zero; The P's in this case that is, there is no significant contribution to the forecast accuracy by allowing the weights to vary over time. The two-quarter results are more interesting. As was 8. The results in Table 5.3 were obtained by imposing nonnegativity constraints. Initially, each of the P's constrained to be positive. If the initial estimate was on boundary (zero), its corresponding time-varying component deleted from the estimation. The a's were constrained to positive and sum to one. -62- the was the was be Evans, Strongin, and Eugeni expected from the encompassing results, the NBER leading indicator index receives the bulk of the weight in the final forecast (61%). This agrees with the analysis of Stock and Watson who constructed the NBER leading indicator index explicitly for its ability to forecast at this two-quarter ahead horizon. We. do find that real M2 receives a substantial weight (19%), while the 12-month/Federal Funds rate spread is at 10% and the 3-month Eurodollar rate is 9%. Figure 5.1 graphs the time path of the <|> weights for these four indicators, as well as the two-quarter GDP forecast and actual. Notice first that the NBER leading indicator index forecasts have been quite reliable, only once dropping below a 50% weight in the combined forecast. Real M2, however, has varied dramatically in its usefulness, going negative on two occasions: immediately following the 1981-82 recession. in 1976 and During that recession, real M2 did not forecast negative growth at any time (although it did in the 1980 recession), whereas the 3-month Eurodollar rate, the 12-month/Federal Funds rate spread, and the NBER leading indicators index did forecast negative growth during some portion of this recession.9 This poor performance is captured in the time-varying model by decreasing the weight on the real M2 forecast temporarily until it begins to improve. On the other hand, during the most recent recession real M2 has gone above a 50% weight (keep in mind that the average weight for real M2 is .19). During this time, real M2 has grown only slowly and this lead to a forecast of slow growth during 1991 (see Figure 5.1). At this same time, the 3-month Eurodollar rate, the 12- month/Federal Funds rate spread, and the NBER leading indicators index signalled substantially higher growth than was realized. Each of these indicators is currently receiving less than its average weight. Consequently, the time-varying mixing model finds that real M2 has been an unusually useful indicator during the 9. It is useful to remember that the primary components of the NBER leading indicators index are the 6-month Commercial paper spread and the 10-year/l-year spread. So it should not be surprising that the NBER leading indicator index misbehaved during this period when the 3-month Eurodollar rate and the 12month/Federal Funds rate spreads also misbehaved. -63- Evans, Strongin, and Eugeni recent recession, despite its generally erratic performance at this horizon versus its relative failure at the twelve month horizon. By contrast the four-quarter horizon results appear to be a picture of stability. Real M2 and the 12-month/Federal Funds rate spread receive the largest unconditional weights, 41% and 37% respectively. The Federal Funds rate and the 10-year/Federal Funds rate spread receive considerably less (around 10% each). The graphs of the time-varying weights indicate that, at this horizon, real M2 and the 12-month/Federal Funds rate spread have been reasonably reliable indicators, always staying near their unconditional weight. On the other hand, the 10-year/Federal Funds spread has been extremely unreliable, going to zero or negative in 1987-88 and during the recent recession. The contrast between the dominance of the NBER leading indicator index at the six-month forecast horizon versus its lack of independent information at the twelve-month horizon demonstrates strongly the need for a different set of indicators for each forecast horizon. The usefulness of the 12-month/Federal Funds rate spread and real M2 for forecasting real GDP at the twelve-month horizon indicates that a different index would be constructed if this forecast horizon was the relevant objective. A note on standard errors is in order. Examination of Table 5.3 indicates that the standard errors associated with the parameters of these mixing models are fairly large. This is not surprising in light of the high degree of collinearity that would be expected of a set of reasonably successful forecasts. In fact, it is typically the case that only the strongest indicator at a given horizon is statistically significant. All this is saying is that the relative weights among successful indicators is subject to substantial uncertainty and that the marginal information after the first one or two indicators is quickly dropping toward 0. Nevertheless the point estimates and time paths of these relative weights provide a useful bench-mark, even though the precision they are estimated with would not change strongly held prior beliefs. -64- TABLE 5.1 - KALMAN RESIDUALS FOR SURVIVING INDICATORS Quarterly (Jul 73 • Dec 91) Real GDP 1Qtr RMSE Rank EUR03 3.622 4 2.754 3 n.a. n.a. n.a. n.a. n.a. n.a. 2.160 2 M2R 3.674 6 2.844 5 2219 4 CP6TB6 3.656 5 2.760 4 n.a. n.a. TB12FF 3.753 7 2.751 2 2.002 1 CM10FF n.a. n.a. n.a. n.a. 2.161 3 XLI 3246 1 2.376 1 2.392 5 XRI2 3.427 3 n.a. n.a. n.a. n.a. LEAD 3.307 2 n.a. n.a. n.a. n.a. NONE 4.052 8 3.369 FF 2Qtrs RMSE Rank 4Gtrs RMSE Rank Indicator n.a.: The indicator is not an initial survivor at this forecast horizon. -65- 2.799 -99- m r x x o j O S j m S B * ^ ^ a a EUR03 FF M2R CP6TB6 TB12FF CM10FF XU XRI2 LEAD 3 ° if l l l 1 0.270 0791 0.102 s l 1 5 ° ^33 3*i 8 9«g o o 1 o . »i i ° 2 • i 32 |S r x x 0 j 0 2 3 f n g P P i o . n > i sso1; i o o s o o i I f» ro s o i I I s 8 3 ~ ps I I I I II o o I <D <D I 00 U I I 3 o P Q O 3 0.105 0959 0960 8. s- I o o I :s 88 8 I I £ O O l I I I I I I | IS P l oo I I o> I 1 —- I s 3! O O i i » I I I I I I I I I I I I i i i CT> I I I i i I I I I I I I I O O O O O O O O O ( o o r o - o o D O O O 8S85S = S 3 S 8 I 3 01 I I 1 I I I 37 I I o 1 1 1 oS p 1 1 1 1o ^ 1i , i1 8 11 I I 3 p 1 I I I 1 1 1 1 1 1 1 1 1 1 I I I I I I I I I I o o o o o o o o o s u o b i b u ^ o } - * 23! I I I CO I I I I I I I I I » • 8 8 8 8fc3 3 I I o g i l l I I _ o o o o o o I | -; oo .*. ro -^ -* -* 1 23pSS28' i i p I I o o _ o o o o s b> ^ w o w - I 88^ O O I I I I I Q> I » I I 3 I I o o I 5B l I I p o p p3 o £ 8 en £ P 8 o © s s o I I 1p t I i | l | I 1 l II 1 1 11 o i i i 8 i i I II en I I o o o o o o o o o o o o bo •** ro '-+ <o —• •of < 21 M • 3 TABLE 5.3 - RELATIVE WEIGHTS IN MIXING REGRESSIONS Real GDP Indicator EUR03 FF 1Qtr * • CP6TB6 * TB12FF * CM10FF n.a. XRI2 LEAD 0.093 (0260) n.a. n.a. M2R XLI 2Qtrs 0.187 (0227) * .0.103 (0238) n.a. 0.617 (0.197) n.a. 0533 (0.174) 0214 (0.155) 0253 (0.206) n.a. 4Qtrs n.a. 0.105 (0209) 0.414 (0.178) n.a. 0.368 (0259) 0.114 (0212) • n.a. n.a. NOTES: - Numbers in parenthesis are standard errors. - n.a.: The indicator is not an initial survivor at this forecast horizon. - f ) : The indicator is encompassed by other indicators at this horizon. -67- 5.1. Mixing Results 2 Quarter Ahead Forecast vs. Actual Real M2 (M2R) annualized growth rates 9.0 3 month eurodollar (EUR03) annualized growth rates 90 r -6.0 12 month T bill less fed funds (TB12FF) 90 NBER Experimental Leading Index (XLI) 90 r Forecast Reliability Weight Real M2 (M2R) weight 0.9 r 3 month eurodollar (EUR03) weight 06 00 f I I I I I f I I t I I I I I I I f » -0 3 12 month T bill less fed funds (TB12FF) 09 r NBER Experimental Leading Index (XLI) 09 r 0.3 h 0.3 0.0 0.0 » 1973 i i i 76 i i f 79 i r i '82 i i i '85 i i i i * i '91 -0 3 -68- 1", I I I I I f I ,1 I I I I I I I I f I 1973 76 79 '82 '85 '91 -69- 88, S8, 28. 6L 91. Z161 16. T T roI 1 I I I I I 1 I I 1 I I 1 I I 16. 88. 98. 29, I I 1 I I I 1 I 1 I \ 6L 9L tl6i toI J I I » I 1 1 co- 20- co 00 J 90 ( d d 2 l 8 l ) spuni p«| ssa| ||iq j . MIUOUJ zv -J 90 I f 1 I 1 1 I I I 1 t 1 ] I I J 90 mala** (dd) spuni pej -i 90 (ddOlWO) spun* P«J ssoj puoq i JeeA o i WfrdAA Ai!|iqBi|eu lSBoaioj 16. mr 88. 98. 28. 6L 9L CZ61 09I I I I I 1 I I 1 I I I I » I oeoo oe J 06 (ddZlRL) spunj P9* » » l MM 1 MJUOUJ g i 06 (H2VM) 2IN l ^ d idomo J 0*6 sew MIMOJO p»zi|tnuue |en)3v SA jspoejo-j peeqy Jejjeno fr Joe wt*j IOMOJO p«zi|*nuue (dd) spunj p*d siinsey 6u;xi^ 'zs Evans, Strongin, and Eugeni CONCLUSION Four things become clear as the preceding analysis developed. First, the forecast horizon is an essential aspect of choosing and evaluating indicators. Second, substantial information resides in the term and public-private spreads and that both of these seemingly very different types of spreads seem to include significant common as well as distinct information sets. Third, while composite indicators may be extremely useful they are only as good as their design allows. The Stock-Watson NBER leading indicator series does very well at precisely what it was designed for, forecasting economic activity at a six-month horizon. Its usefulness beyond this is far more limited than prior analysis would have suggested. The analysis is also suggestive that the type of general purpose target variable that the old monetary targeting literature sought, probably does not exist at least in terms of real economic activity. Policymakers will continue to need to mix information according to their current focus. Mixing models of the sort used in this paper are meant to be preliminary work in this regard. The early results are intriguing. -70- Evans, Strongin, and Eugeni References Bernanke, Ben S., "On the predictive power of interest rates and interest and interest rate spreads," New England Economic Review, November-December, 1990, pp. 51-68. Chong, Y. and D. Hendry, "Econometric evaluation of linear macroeconomic models," Review of Economic Studies, 53, 1986, pp. 671-690. Estrella, A. and G. Hardouvelis, "The term structure as a predictor of real economic activity," Journal of Finance, 46, 1991, pp. 555-576. Friedman, B. and K. Kuttner, "Why does the paper-bill spread predict real economic activity?" forthcoming in James H. Stock and Mark W. Watson eds., New Research in Business Cycles, Indicators and Forecasting, University of Chicago Press and the NBER, 1992. Kashyap, A., J. Stein, and D. Wilcox, "Monetary policy and credit conditions: evidence from the composition of external finance," Federal Reserve Board, Working Paper No. 154, 1991. Laurent, Robert D., "An interest rate-based indicator of monetary policy," Economic Perspectives, Federal Reserve Bank of Chicago, January/February, 1988, pp. 3-14. National Bureau of Economic Research, Press Release, January 30, 1991. Sims, Christopher A., "Interpreting the macroeconomic time series facts: the effects of monetary policy," manuscript, 1991. Stock, J. and M. Watson, "Interpreting the evidence on moneyincome causality," Journal of Econometrics, Vol. 40, 1989a, pp. 161-182. Stock, J. and M. Watson, "New indexes of coincident and leading economic indicators," in NBER Macroeconomics Annual, edited by O. Blanchard and S. Fischer, the MIT Press, 1989b, pp. 351-409. Strongin, Steven, "Macroeconomic models and the term structure of interest rates," Federal Reserve Bank of Chicago, Working Paper No. 90-14, 1990. Strongin, Steven, "The identification of monetary policy disturbances: explaining the liquidity puzzle," Federal Reserve Bank of Chicago, Working Paper No. 91-24, 1991. -71- DISCUSSION OF A POLICYMAKER'S GUIDE TO INDICATORS OF ECONOMIC ACTIVITY Richard W. Kopcke1 This paper examines various indicators, seeking those that are correlated most highly with the future course of economic activity. First, the indicators are arranged into "natural groups.H Second, the paper selects the most promising indicators from each group. Third, the forecasts of the selected indicators are combined to produce mixed forecasts. The paper includes many, but not all, of the popular indicators. Given that it confines itself to indicators of real economic activity, perhaps the paper should drop nominal Ml and nominal M2 (which apparently perform poorly) from its list to make room for model forecasts, real interest rates, stock prices, consumer confidence, and other indicators mentioned so much in the press. Although the paper's strategy avoids using explicit economic models, in my opinion, it does not escape the consequences of measurement without theory. On the most elementary level, the paper's horse races should include the forecasts produced by economic models. The mean squared errors of the indicators appears to be high compared to those of the forecasting services surveyed by Stephen McNees. On a deeper level, the paper's strategy seems to require or presume implicit models which remain, undiscussed, behind the findings. What determines the natural groups of indicators? Interest rates are gathered into one such group, presumably because they are all called interest rates. But they do not seem to be a natural group. The federal funds rate, for example, principally reflects monetary policy. The Baa rate reflects general economic conditions. Perhaps the federal Richard W. Kopcke is Vice President and Economist of the Federal Reserve Bank of Boston. Richard W. Kopcke funds rate, reserves, etc. constitute a more natural group, while the Baa rate, real M2, etc. constitute another. Although the battery of tests performed on the indicators in each group are reasonably thorough, they are not entirely convincing without the benefit of the analysis that accompanies models. More importantly, it is not clear how the results of these tests are useful to policymakers. Historical correlations reflect some mix of fiscal and monetary policies here and abroad as well as some mix of changing aggregate supply and demand. As these mixes vary in the future, these correlations will likely change. Specifically, if the average historical mix should not prevail in this recovery, the indicators may yield poor forecasts of the course of economic activity in the next few years. Historical correlations between indicators and economic activity may not be a good guide in the future if we know, for example, that: (i) fiscal policy will be unusually restrictive for a recovery in coming years, (ii) the growth rate of the labor force will be less than one-half that prevailing since World War II, (iii) changing demographics will reduce the potential magnitude of a housing boom, (iv) the GDP gap differs from that at the inception of the average recovery, or (v) our economy is now more open to foreign trade than it had been in previous recoveries. Indeed, in the four-quarter forecasts (Chart 5.2, upper graphs), the indicators, too often, are negatively correlated with the course of economic activity during the current business cycle. According to these indicators, the average business cycle is a poor guide to this cycle. In the 1940s Haavelmo and Duesenberry explained that the correlations among state variables (which include both indicators and economic activity) could not be interpreted outside a model. Because these correlations are unstable when economic conditions change, the remedy requires the modelling of economic behavior, which entails descriptions of how these correlations are likely to change. Whatever the weakness of these models,"however competently they describe the way businesses, consumers, and governments make decisions, these models provide a structure needed for private or public policy analysis. Correlation coefficients are functions of partial correlation -2- Richard W. Kopcke coefficients that might be more stable; nonlinearities are allowed. If the Federal Reserve should change its operating procedures (perhaps following some of these indicators), we cannot anticipate how the correlations among the federal funds rate, real M2, and economic activity will change without a model. To illustrate further the difficulties that interpreting the correlations between indicators and economic activity pose for policymakers, consider the federal funds rate (Chart 5.2, upper leftmost graph) . The correlation between the federal funds rate and activity may be relatively low for three reasons: (i) monetary policy has worked well as a shock absorber, offsetting potential disruptions, smoothing the ride; (ii) monetary policy has not reacted to short-run economic conditions; or (iii) monetary policy has been "out of phase" with the business cycle. The correlations of the indicators with activity, by themselves, do not tell us whether operating procedures should change, or how they should change. Setting aside the problems of structural changes, without a model the correlations among state variables remain dubious guides. The paper's bivariate horse races, for example, do not necessarily select the best indicators. Bivariate correlations do not predict the order in which variables are added to or removed from step-wise regressions, and the results of Granger tests depend on the variables included in the regression. Therefore, an indicator which is deemed the best single candidate in its group may be inferior to another member of its group when more than one indicator (drawing from any group) is to be considered at a time. These problems might diminish if a model were used to form natural groups from the start, but if we ultimately are to consider multivariate forecasts, we ought to begin with multivariate techniques. In forming multivariate forecasts, the information in each indicator should not be represented simply by its forecast from its bivariate regression with economic activity. These first-stage regressions restrict the information provided by each indicator, so the multivariate regression cannot make full use of the correlations among indicators to describe economic activity. Constraining the weights of the forecasts to be positive or to sum to one in the multivariate -3- Richard W. Kopcke regression also prevents the full consideration of all the information in the indicators. No explicit model dictates these restrictions. Indicators may be valuable to bond traders and others who want instant forecasts, who want inexpensive forecasts, who have little interest in describing the workings of the economy behind the forecasts, or who do not require the most accurate forecast, either because they only need a rough projection or because they make new forecasts very frequently. For the purposes of making policy, however, indicators are not so attractive. Suppose real M2 and the slope of the yield curve foretell unacceptable growth of GDP. give policymakers? What guidance do these indicators Should policy change? If so, how much? policy influence M2, the slope of the yield curve, and GDP? How does I am reminded of the comment that we must control GDP in order to control M2. A dilemma also would confront policymakers when, as is often the case, the indicator that forecasts one horizon best seems too far out of line with the indicator for a slightly shorter horizon. Because the paper concludes that there is no indicator for all seasons, policymakers need a model or metaindicator to interpret the signals. For want of a model, indicators also seem to be poor guides for policymakers, because they provide no framework for setting either the objectives or the instruments of policy. For example, indicators do not show what paths for GDP are feasible or which paths are consistent with goals for inflation. Indicators, without a model, do not suggest how policymakers should react to economic conditions either to achieve a dynamically stable course for policy or to avoid increasing the volatility of GDP and prices. In order to integrate consistently forecasts with policy, we build models; yet, indicators retain some allure. in indicators remains for one of two reasons. Perhaps the interest First, although our models are not producing forecasts that are clearly inferior, we may not take proper advantage of these models for analyzing the consequences of policy. Second, the goals of policy may not be specified sufficiently clearly (perhaps for want of agreement) in order to use the models as a guide. In this second case, indicators appear to be useful surrogates; they fail to stir passions while bridging the potentially disparate beliefs of policymakers. -4- Richard W. Kopcke REFERENCES Haavelmo, T. "The Probability Approach in Economics," Econometrica, vol. 12, Special Supplement, July 1944, esp. pp. 12-39. Duesenberry, J. S. "Income-Consumption Relations and Their Implications," Income, Employment, and Public Policy, Essays in Honor of Alvin Hansen, (W. W. Norton, 1948), pp. 54-81, reprinted in M. G. Mueller, ed., Readings in Macroeconomics and Winston, 1966 and 1971), pp. 61-76. -5- (Holt, Rinehart, DISCOUNT WINDOW BORROWING AND LIQUIDITY W. J. Coleman. C. Gilles, and P. Labadie1 Three features seem centra] to understanding the relationship between U.S. monetary policy and the comovements of open market operations, monetary aggregates, and interest rates. First, shocks to bank reserves affect interest rates in ways that axe not tightly linked to the Fisherian fundamentals (expected inflation, marginal rate of substitution, and marginal productivity of capital). Second, banks often respond to reserve shocks by adjusting their borrowing at the Federal Reserve's discount window. Third, the Federal Reserve often conducts open market operations to smooth interest rates that would otherwise react to private-sector demand shocks. In this paper, we study a stochastic general equilibrium model that incorporates these features in an effort to understand important empirical regularities involving monetary aggregates and interest rates. The empirical regularities we have in mind are those documented in the vast literature aimed at uncovering a negative correlation between short-term interest rates and exogenous policy shocks to nominal monetary aggregates, a relationship often referred to as the liquidity effect. Cagan (1972) and Cagan and Gandolfi (1969), among many others, have reported finding negative correlations between Ml itself and various short-term interest rates. Subsequent studies have reported similar correlations with innovations in Ml backed out using a Choleski decomposition of the residuals in a vector autoregression (for a variety of orderings). More recently, however, Leeper and Gordon (forthcoming) have made a strong case that these innovations probably do not represent exogenous monetary policy shocks, as the money supply may be endogenously 1 Board of Governors, Federal Reserve System. We gratefully acknowledge helpful discussions with Jim Clouse and Josh Feinman. Coleman, Gilles, and Labadie determined in ways that are not captured by any Choleski decomposition. To support their claim, they noted that the statistical properties of these innovations are sensitive to the other endogenous variables included in the VAR, the sample period, and the measure of money selected for analysis. Some researchers, for example Bernanke and Blinder (1990) and Sims (forthcoming), have responded to such criticism by assuming that innovations to interest rates reflect policy shocks, to which the supply of money responds endogenously. For our purpose, however, this strategy does not resolve the central question: if there exists a liquidity effect, then why are these interest rate innovations not robustly negatively correlated with monetary aggregates (an observation also made by Leeper and Gordon)? Christiano and Eichenbaum (1991) and Strongin (1991) have tried to obtain robust negative correlations by using nonborrowed reserves as the measure of money. This approach contrasts with that of Leeper and Gordon, who experimented with monetary aggregates that are at least as broad as the monetary base. Christiano and Eichenbaum's rationale for using nonborrowed reserves is based on the widely held perception that the Fed controls this aggregate. For this reason they associated policy shocks with innovations to nonborrowed reserves, which they then showed to be negatively correlated with the federal funds rate. In fact, using nonborrowed reserves as the measure of money, they found evidence of a negative correlation regardless of whether money innovations or interest rates innovations were identified as the policy shocks, and they showed that these correlations are remarkably robust to the sample time period. To explain why the innovations to broader monetary aggregates do not exhibit a similar correlation, they noted that these aggregates are largely endogenously determined by the banking system. For example, they argued that total reserves may be inelastic in the short run, and therefore not correlated with interest rates at all. In this example, policy shocks to nonborrowed reserves do not affect total reserves immediately. Strongin refined this argument; he argued that innovations to nonborrowed reserves that are not reflected in shocks to total reserves should be identified as the policy shocks. He asserted, in essence, that shocks to required reserves lead to an adjustment in both - 2 - Coleman, Gilles. and Labadie nonborrowed and total reserves, whereas open market operations lead to an adjustment in only nonborrowed reserves. We develop a model that is rich enough to address the empirical issues presented above. To do this, we introduce a banking system, reserve requirements, and a discount window into a model of liquidity based on the works of Grossman and Weiss (1983), Rotemberg (1984), Lucas (1990) and Fuerst (1992). In these models, and here, the term liquidity effect refers not merely to a negative correlation between monetary policy shocks and interest rates but more generally to any non-Fisherian effect on interest rates. Interest rates deviate from their Fisherian fundamentals because of shocks to the demand for bank deposits from businesses to finance new investment projects and perhaps also because of monetary policy shocks. In our model, the interest rate is also the cost (both pecuniary and nonpecuniary) of borrowing reserves from the discount window, so that over time there is a well defined relationship between borrowed reserves and the interest rate. Monetary policy designed to smooth interest rates then leads to rather complicated mutual dependencies among open market operations, both broad and narrow monetary aggregates, and interest rates; in particular, monetary policy can lead to positive correlations between broad monetary aggregates and interest rates in spite of the liquidity effect. When policy shocks are correctly identified, however, the model suggests that broad monetary aggregates are negatively correlated with interest rates, showing evidence of the liquidity effect. Furthermore, the model always generates a negative correlation between nonborrowed reserves and short-term interest rates, regardless of what the policy shocks are and how they are identified. Such a result is due to the way the discount window is operated. In light of this model, one interpretation of Christiano-Eichenbaum and Strongin's results is that they identified the discount window policy. Since this policy implies a negative correlation between nonborrowed reserves and interest rates whether or not the model incorporates a liquidity effect, their results shed little light on the presence of such an effect. - 3 - Coleman. Gilles, and Labadie THE MODEL DESCRIPTION. To get an overview of the model, consider the following accounting of the assets and liabilities of banks. Their liabilities comprise demand deposits of firms and households as well as savings deposits of households. Their assets are made up of reserves and a portfolio of government securities and loans to firms. Banks are required to hold as reserves a fraction of their demand deposits;'to avoid a deficiency, they can borrow reserves at the discount window. Borrowed reserves incur pecuniary and nonpecuniary costs. To start building a model around this balance sheet, think of households as dividing their deposits between demand deposits, which can be used to buy goods, and savings deposits, which cannot. Assume that this division is made before the value of the open maxket operation is known, resulting in a liquidity effect as described by Lucas (1990) and Fuerst (1992). Also assume, as Fuerst (1992) did, that firms must finance their purchases of investment goods with demand deposits, so that these deposits represent intermediated capital, as in Freeman and Huffman (1991). To view the model in more detail, consider a representative household that ranks stochastic consumption and leisure streams {ct,lt} according to the utility function Lt=0 \t=0 / where /3{ is the date-i realization of the random discount factor; /3*+i is unknown at the beginning of period t but is revealed later during that period. The household begins period t with money balances Mt in an interest-bearing savings account. It immediately transfers amount Zt to a checking account which bears no interest but can be used during the period to finance consumption ct; only one transfer during the period is allowed. The household must choose Zt before it knows the realization of any of the current shocks, or prices for that matter. Its purchases of goods are subject to the finance constraint Ptct < Zt. -4- Coleman, Gilles. and Labadie At the end of the period, Mt — Zt remains in the household's savings account and Zt — PfCt in its checking account. The household derives income from several sources. It provides labor to the firm, working a fraction of time equal to 1 — it at wage rate Wt] it earns interest at rate r\ on the amount Mt — Zt in its savings account; it collects a transfer Xt from the government; finally, as owner of both the firm and the bank, it collects Il( and II*, the period's proceeds from the sale of output net of all costs and bank profit respectively. The household receives its income, including income from labor performed during the period, at the beginning of the next period, when it is directly deposited into the savings account. With unspent checking account balances being transferred back into the savings account, the law of motion for Mt is Mt+i = Zt - Ptct + (Mt - Zt)(l + r{) + Wi(l - It) + Xt+ Ii{ + II*. The firm, the second agent in the economy, combines, capital and labor inputs to produce a homogeneous product sold to buyers of consumption and capital goods. The production function is Vt = F(kt,nt,0t), where yt is the output, kt and nt are the inputs of capital and labor, and 9% is a technological shock. The firm owns the capital stock kt and hires labor at rate Wt] it makes wage payments at the beginning of the next period using the receipts from the sale of output. The firm must also acquire investment goods it; it purchases these goods from other firms in the goods market but cannot use its sales receipts for this purpose. Instead, it finances investment by borrowing Bt from a bank, which charges interest at rate r*. The bank provides this financing by crediting the amount to the firm's checking account, increasing the balance from its starting level of zero. The firm's finance constraint is Bt > Ptit. At the end of the period, the firm has spent Ptit on investment goods and deposits its current sales receipts, PtVt, leaving Bt -f Pt(yt — U) in its checking -5- Coleman, Gilles, and Labadie account. At the beginning of the next period, the firm repays its bank loan and transfers wages into the worker's savings account. The amount left in the firm's account, 11^, is paid to the firm's owner as dividend: n / = Ptyt - Wtm - Ptit - rtBt. The stock of capital depreciates at the constant rate 6. so that its law of motion obeys fct+i = (1 -6)kt + it. The firm makes all its decisions (namely, J3t, it, and rtt) with full knowledge of the current shocks and prices. The bank, the third agent in the economy, starts period t with liabilities equal to Mt (the household's savings account) and holds an equal amount of vault cash as an offsetting asset (we write "vault cash" for definiteness; Mt could also be thought of as an account at the central bank). The household immediately transfers Zt from its savings to its checking account, without affecting the bank's total liabilities or assets. The bank pays interest r\ on Mt — Zt, the amount left in the savings account, but pays no interest on checking deposits. By lending Bt to the firm, an amount that is credited to the firm's checking account, the bank increases both its liabilities and its assets from Mt to Mt -r Bt. To buy government bonds and to honor checks written to finance purchases of consumption and investment goods, the bank depletes its holding of vault cash, Mt] but it replenishes this cash position by the amount of the checks that firms receive for selling their output, checks that they deposit in their account. The amount of vault cash that the bank holds at the end of the period counts as reserves. Note that for an individual competitive bank, the loan of Bt to a firm drains reserves (when the firm spends the proceeds) just as much as if the bank had spent an equal amount to purchase government securities; therefore, at the same rate of interest, the bank is indifferent between the two types of lending. For the banking system as a whole, however, loans to firms involve no net loss of reserves, but merely a transfer from the borrower's bank to the bank of the producer of investment goods. -6- Coleman. Gilles, and Labadie Reserves. VJ, pay no interest and are subject to a reserve requirement, a fixed fraction p of the amount of checking deposits on the books of the bank at the end of the period: (1) Vt > p x [(Zt - Pta) + (Bt - Ptit -t- Pm)]. If the bank cannot satisfy the reserve requirement with the amount of vault cash it has at the end of the period (after checks have cleared), it can borrow the shortfall from the government at the discount window. Therefore, the following accounting identity must hold (2) Mt r D t = qtGt + Pt(it T*- yt) -f Vu where G% is the number of one-period pure discount government bonds the bank acquires, at a unit cost of qt = 1/(1 + rt), and D% is the amount it borrows at the discount window. Government bonds, private loans, and discount window borrowing carry the same rate of interest rt. The bank's objective is to maximize its period profit, which is given by (3) n j = Tt(Bt + qtGt - Dt) - r\(Mt - Zt\ The government, the fourth agent in the economy, sells one-period bonds in the securities market and redeems them at the beginning of the following period, operates the discount window, and makes transfers to the household's bank account. During period i, the government announces the open market operation Gt and the amount of transfers Xt after the household chooses Zt but before any other decision by any agent has to be made. All money flowing between the government and the private sector, as well as within the banking industry, takes the form of fiat money. The bank starts period t with an amount of fiat money (which it calls vault cash) equal to Mt. Nonborrowed reserves Vt — Dt is the amount left in vault cash after the purchase of government bonds and check clearing but before borrowing at the discount window; in equilibrium, Vt — Dt = Mt — qtGt as can be seen from eq. (2). Let Ht denote the outstanding supply of fiat money at the beginning of period t (Mt is best thought of as the demand for fiat money, so that in - 7 - Coleman. Gilles, and Labadie equilibrium Ht = Mt). The law of motion for Ht, which can also be thought of as the government budget constraint, is as follows: i?t+1 = Ht T Tt{qtGt — Dt) — Xt. Think of government policy as a rule that generates the values of Gt and Xt and that also sets the rate of interest at the discount window. Assume that the government lends reserves at the discount window according to an upward-sloping function if> : [0, oo) —• [0, oo) that relates the rate of interest it charges to the fraction of total reserves that it lends. Banks cannot lend at the discount window, so that when the equilibrium rate of interest is lower than the minimum rate at which the government is willing to lend, V>(0), there is no discount window activity: rt = i)(Dt/Vt) r t < V^O) whenever Dt > 0; whenever Dt = 0. The argument of if) ought to be the amount supplied at the window, which in equilibrium turns out to be equal to Dt, the amount demanded. Incorporating this equilibrium relationship directly simplifies the notation, but keep in mind that banks take as given all interest rates, including the rate they face at the discount window (which is equal to the rate on government securities). When the Federal Reserve lends at the discount window, the borrowing bank pays the discount rate plus a nonpecuniary cost; at the margin, this sum must equal the cost of borrowing from other banks, which is the federal funds rate. The marginal nonpecuniary cost is thus captured by the difference between the federal funds rate and the discount rate, called the spread. Historically, the policy of the Federal Reserve seems to have been to supply funds at the discount window at an increasing nonpecuniary cost (spread), which is precisely what the function tp assumes. This type of discount-window policy has been documented in the empirical literature, and is commonly modeled in the theoretical literature. 2 Chart 1, which graphs the monthly time series for 2 See for example Polakoff(1960), Goldfeld and Kane (1966), and more recently Goodfriend (1983), Dutkowsky (1984), and Waller (1990). In particular, Fig. 1, p. 346 in Goodfriend depicts an assumed ip function that is strikingly similar to the function that would best fit the scatter plot of our Chart 2. -8- Coleman, Gilles. and Labadie the federal funds rate and the nonborrowed reserve ratio (the mirror image of the borrowed reserve ratio), reveals the basis for the findings of the empirical studies. On closer inspection, a picture of the function ib emerges in a scatter plot of the borrowed reserve ratio against the spread, shown in Chart 2. Since this picture suggests that the Federal Reserve is ready to lend its first dollar at a zero spread, the value of t^(0) corresponds to the discount rate. With this interpretation of ^(0), the model simply assumes a constant discount rate. A word about terminology is in order. Vt is total reserves in the banking system; Dt is borrowed reserves; the difference Vt — Dt is nonborrowed reserves; and required reserves is p x [Zt + Bt + Pt(yt — it — ct)]. Besides total reserves, it is possible to identify the analogues of several monetary aggregates. M% (or Ht) corresponds to the monetary base, MO; the analogue of Ml is the sum of all reservable accounts, Zt + B%\ the total libilities of the banking sector at the end of the period, Mt + B^ correspond to M2 (strictly speaking, Ml and M2 both should include Pt{yt — ct — U) as well, but this is equal to zero in equilibrium); finally, the difference between M2 and MO, which is Bt, is inside money. It is now useful to summarize the timing of information and decisions. During period i, the realizations of four random variables shock the economy—the technological shock 0t> the preference shock /3t+i, the open market operation Gt, and the government transfer Xt. At the beginning of the period, the household must decide how much to put into its checking account, not knowing the current realization of 0t, /3t+i, Gt, or Xt, and therefore not knowing what interest rates, prices, output, or consumption will be. After it makes this decision, all four shocks are revealed and prices are set. On the basis of these shocks and these prices, the household decides how much to consume and how much to work; the firm decides how much to borrow, how much to invest, and how much labor to hire; and the bank decides how much to lend to the firm and to the government. Then trading takes place and checks clear. The bank monitors its reserve position and borrows at the discount window to cover any reserve deficiency (the bank can be thought of as borrowing at the same time it invests in government bonds or lends to firms, because it - 9 - Coleman, Gilles. and Labadie has the same information when it engages in any of these activities). At the start of next period, the firm pays its wage bill, repays its bank loan, and pays out its earnings to its shareholder; the government makes transfers to the household's savings account and redeems the bonds that the bank holds; the bank pays interest on its savings account, settles its discount window debt, and pays out its earnings. These activities determine the new initial balance in the household's savings account. Then a new cycle starts. The activities of the four agents that have been described above must, of course, satisfy the following standard market-clearing conditions. yt = a + it goods market; nt = 1 — it labor market; Ht = Mt money market. The economy is competitive, and agents have rational expectations. An equilibrium is a set of state-contingent prices and interest rates such that markets clear when all agents solve their optimization problems, treating prices as given. In the next subsection, we are more explicit about what this means. THE MODEL AS A RECURSIVE SYSTEM The household solves a dynamic program, which is recursive under standard assumptions about preferences, technology, and the stochastic environment. ASSUMPTION 1. tiate, The period utility function U is twice continuously strictly increasing in both arguments, and strictly ASSUMPTION 2. concave. The production function F has the form F(k, n, 0) = 9f{k, n), where f is twice continuously guments, differen- differentiate, concave, and homogeneous strictly increasing in both ar- of degree one. (Stochastic constant returns to scale.) ASSUMPTION 3. The preference shocks {/3f} and the technological shocks {6t} are generated by independent first-order Markov processes. The support of &t is contained in (0,1) and that of &t is contained in (0, oo). Monetary policy consists of a rule that dictates the value of open market operations, the size of government transfers, and the level of the discount rate; -10- Coleman. Gilles. and Labadie these instruments are not completely independent of each other. The operation of the discount window is modeled through a fixed function w that relates the discount rate to borrowed reserves. Think of the government as announcing this function and keeping it fixed in all periods, leaving the discount rate itself endogenousiy determined by the demand for borrowed reserves. Given the function V>, the values of Gt and Xt in period t are implied by the choices of the ratios gt = Gt/Ht and 7* = iift+i/^t- To induce stationarity and recursivity, choose (gujt) ASSUMPTION 4. order Markov as the policy variables and make the following assumption. The monetary policy shocks { # , 7*} are generated by a firstprocess. Starting with the optimization problem faced by the bank simplifies both the notation and the analysis. The bank maximizes its period profit, given in (3), by choosing an optimal portfolio (Sf,Gt,i?t, Vi), subject to the legal reserve constraint (1), and the accounting identity (2). Clearly, optimization requires that V% = p[Zt + Bt + Pt(yt — it — ct)] (no excess reserves) if r* > 0. A zero-profit condition, the result of perfect competition and constant returns to scale in the banking industry, implies that r\ = [{Mt + Bt — Vt)/(Mt — Zt)] x rt; this condition in turn yields r\ = r t [l + (l — p)(Zt + Bt)/(Mt — Zt% which holds whether or not r* > 0. To obtain the last expression, recall the market-clearing condition yt = ct-r itSince the firm and the bank belong to the household, it is possible to integrate the problems faced by the firm, the bank, and the household. Because money growth induces a trend in nominal variables, stationarity of the equilibrium requires that nominal variables—denoted by uppercase letters— be divided by the supply of fiat money. The new variables are denoted by the corresponding lowercase letters; thus, nit = Mt/Ht, zt = Zt/Ht, and so forth. Under assumptions 3 and 4, the evolution of the shocks is determined at the beginning of period t by the vector (/3t,0t-i> 5t-ij7t-i)> which consists of the latest known realizations of the shocks. The state of the economy at that time can then be expressed as st = («t?/3t, 0t-i?5t-i)7i-i)> where Kt is the aggregate per capita stock of capital (as opposed to fct, which is the individual firm's holding). In equilibrium, of course, individual decisions determine - 11- Coleman. Gilles, and Labadie aggregate outcomes, so that K% = kf. A solution is a set of functions p, w. and r such that pt = p(st,st+i), wt = tu(st,.st+i), and rt = r(s t ,«st+i) yield the equilibrium values of the normalized price level, the normalized wage rate, and the rate of interest on date t (again, pt = Pt/Ht and wt = Wt/Ht). Since qt = 1/(1 +Tt)} the equilibrium function r determines a function q satisfying 9t = g ( j t i * t + i ) . Given such pricing functions, let J(m, &, s) denote the value of the optimal discounted stream of utility for a household starting a given period with money balances m, while the firm owns capital stock k and the economy is in state s = («,/?, 0,(/, 7). The household first chooses z, which is the transfer from its savings to its checking account, expressed as a fraction of the outstanding supply of fiat money. Then (/9^0^if^7 , ) are revealed (a prime denotes the realization of a variable that was unknown at the beginning of the period), and these shocks determine the current price, wage rate, and rate of interest, as well as the next-period state s'. To determine s1, the household must know how the evolution of the aggregate capital stock depends on the state of the economy. In equilibrium, of course, this law of motion follows from the individual optimal decisions. On the basis of an assumed law of motion for K and of p(s,s ; ), w(s1s')} and r(s,s'), the*household makes its consumption and leisure decisions and the firm makes its labor and investment decisions. What these optimal decisions are can be studied by considering the Bellman equation characterizing J, the value function. J(m,k,s) = max-Ej max {C/(c,£)' + / 3 J ( m U V ) } subject to (4) 2>P<:; n* = pO1 f(k1 n) — (1 -J- r)pi — wn; Jfe' = ( l - * ) J b + i; w(l - I) + (m - z)(l + rb) -f x1 + 7Tf -»- (z - pc) , m = -12- Coleman. Gilles, and Labadie the last constraint on the problem is the law of motion for K. Here p, w. and r are short for p($, s'), w(s, s1), and r(s, s'), and E9 is the expectation conditional on 5. Using the results of the bank's optimization problem, the market-ciearing condition 6' f{k,n) = c + i, and the firm's optimization condition b = pi. we have r > 0, v > p(z + 6), and v = p(z + b) if r > 0. OPTIMIZATION AND EQUILIBRIUM CONDITIONS. The Bellman equation for J includes two maximization operators; the first refers to the choice of z, which is conditional only on s, and the second refers to the choice of (c, £, n, i) which is conditional on both s and sf. Corresponding to the latter choice, we have the following four first-order conditions: (<0 u>(j,5') u>(a,s') = p ( * , a ' y / 2 ( * , n ) ; (») (0 y J 2 ( m \ Jb',,') = p(*,,')[! + r(s, M')]Jl{m''?'J); where A is the Kuhn-Tucker multiplier associated with the finance constraint (4), so that A(z — pc) = 0. Indexes to the functions U and J denote partial derivatives; therefore, U\, for example, is the partial derivative of U with respect to its first argument, consumption. The first-order condition associated with the choice of z is (*) E. Ui(c,£) 3 S IP( > ')\ = E. 0[l + rh(s,*')] Ji(m',k',s') r To solve the dynamic programming problem, we need the following envelope conditions, which give the marginal values of money and capital: (m) / l ( m , 4 , ( ) . & [ ^ ] . (*) J 8 (m, * , . ) = E. [(U,(c,t) - pX) ( « 7 i ( * , n ) + (1 + r)(l - * ) ) ] ; where p is short for p(s, s1), and similarly for w and r. -13- Coleman, Gilles, and Labadie Finally, an equilibrium in this economy is a set of functions w(s,s!), p(s,$'), and r(s, s1) [or equivalently 9(3, s1)] and a law of motion for the aggregate capital stock K such that the associated solution of the dynamic programming problem—that is, values for (z, A, c, /, n, 2, i/, d) that solve the first-order and envelope conditions—satisfies the following equilibrium conditions: c + i = tC/(i, n); l - * = n; qg + v - (f = m; rn = 1; fc' = « ' ; r6 = m —2 xr; d x r = (fx ip(d/v). The last equation states that, when the monetary authorities lend at the discount window (d > 0), they do so in accordance with their supply behavior, so that r = ^{d/v). In the third equilibrium condition, qg1 + v — d = m, v is equal to p(z + pi) unless r = 0, in which case v can exceed required reserves. SOLVING THE MODEL Consider initially a slightly simplified version of the model in which labor is inelastically supplied (I = 0) and money supply'is constant (7 = 1). To solve this simplified model, first reduce the system of equations that determines the equilibrium to only three equations in the three unknown functions c, z, and (a transformation of) J\. To simplify the notation, define £(/9,.s') = /3Ji(l,/c',y). 3 Then the firstorder condition (c) becomes Ui(c) = (\ + t)PRecall that K is one of the arguments of 5, so that the function £ is well defined. -14- Coleman, Gilles. and Labadie Here and below £ stands for f(/3,s'); accordingly £' below stands for £(/?', s"). Using this equation and the constraint z > pc, which holds with equality whenever A > 0, isolate p as (5) p = nun ^ { } Substitute this equation in £ = (3E5i [U^c^/p1], which follows from the definition of £ and the envelope condition (m), to obtain ( = 0E, max (6) {***.<} this equation is the first of the set of three to be solved (£ now replaces J\). The second equation follows from substituting the expression (5) for p into the first-order condition (z), obtaining (7) £.|ma*{^,*}]=2<;.l(l + r )t}. h The last equation in the system follows from the first-order equation (i) and the envelope condition (Jk): mm {**«} l£l (8) z'e, Ux{c') \ {9"h{k') + (1 + r')(l - 6)) = 0qEs nun < — <O}CJ To write (6) - (8) solely in terms of c, z, and £, express r and r in terms of these functions as follows: T = i>(dlv); and (9) r> = where -15- Coleman, Gilles, and Labadie d = qg ~ v - 1; v = p(z + 6); 6 = m | _ _ _ | m ; and finally, i = 0'/(Jfe)-c. These equations hold provided d > 0 and r > 0; if d = 0, then r < ^(0), while if r = 0, then v > p{z + 6). Rather than solving this model explicitly, which can be done numerically using the methodology presented by Coleman (1992), we devise an example which admits a closed-form solution. This example highlights all the features of the model that are useful in interpreting the empirical regularities mentioned earlier. AN EXAMPLE To develop an intuitive understanding of the model, it is instructive to consider a parametrization that allows a closed-form solution. Suppose that (a) utility is logarithmic; (b) production satisfies f(k) = fca, for 0 < a < 1; (c) capital depreciates completely over each period; and (d) the technological shocks 5, the policy shocks g, and the preference shocks 0 are all iid (although not necessarily independent of each other). Now, conjecture that no excess cash is ever held in the goods market and that z is constant at z. ; circumstances, 6 = zi/c, i = fc , and equations (6)-(8) simplify to z (10) (11) 1 = E, 'P(l+rb) " = 0'qEsl ) '6"a(k')a-1' J c where the interest rate r satisfies <*> -*r»'' —1 6 - Under these Coleman. Gilles, and Labadie and rb is given by (9). Further conjecture that the consumption function can be written as Q = TTzrt—r$ k , 1 + Wq) for some function h. Note that because k!/c = h(/3'g), the function h can be thought of as the investment to consumption ratio. Since h depends only on flq and since q = 1/(1 + r), (12) determines r a s a function of 5, /3', and g1. Write this function, which implies that r and q are iid and independent of s, as r = RJ^z^ff^g1) and correspondingly q = Q{z^P\g9)\ now substitute these equations into (9), and the resulting equation into (10), to obtain ! = £ , tt[i + [i^->n*««*<i'-™t)m!,M) This equation has the important implication that z does not depend on 5, because s enters only through the conditional expectation, and /?' and g1 are iid. This observation verifies the conjecture z(a) = z. Tofindfc,substitute the conjecture about the consumption function into (11) and simplify to obtain h(l3,q) = a(3'q(l + Esl[h(/3"q')}). Using the fact that 0' and q are iid (because q — Q(z,0',g'), and (@,g) is iid), this equation implies H{l3q) -l-Ela0<qy where E[. ] is the unconditional expectation, taken over the constant distribution of (y9',g). It is then straightforward to verify that the finance constraint in the goods market is always binding; therefore, all the initial conjectures were correct. This example leads to a sharp characterization of the response of monetary aggregates and the interest rate to supply and demand shocks. equilibrium value of k'/c =fc,rewrite (12) as (13) U^^f^Sl-^d-ElaP'q)) q ^\ pz(l+a0'q-E[a(3'q}) -17- Using the Coleman, Gilles, and Labadie Consider first the effect of technological shocks, &. .Such shocks do not affect r, as (13) makes clear, and thus they do not affect any of the monetary aggregates. They have real effects, of course, since they affect output, consumption, and investment. But they fail to move nominal interest rates (although real rates certainly do) because the demand for consumption and investment goods shift proportionately. This feature is due to the choice of utility and production functions, and is not a general feature of the model. It indicates, however, that in the general case productivity shocks can affect interest rates and monetary aggregates in either direction. Before turning to the effect of other shocks, it is helpful to list the relevant equations. The first is (13), which determines the correlation between each shock and the nominal rate of interest. The others are: (14) total reserves: v = pz[l + h(/3'q)]] (15) nonborrowed reserves: (16) borrowed reserves: (17) Ml: z + b=z[l (18) M2: 1 + 6 = 1 + 2&(0'g). t; — d = 1 — qg1] d = v x ^-1(r); + h(0'q)]] To isolate the effect of policy shocks, assume first that there are no other shocks (a similar procedure will uncover the effect of preference shocks). Note that the left side of (13) is decreasing in g, while the right side is increasing both in q and in g1 (recall that T/J is increasing); therefore g' and q vary inversely. For the same reason, but considering the right side as a function of q and qg\ q and qgf vary inversely also. Hence, gf, r, and qg1 all move in the same direction. In view of (15), then, policy shocks induce a negative correlation between the nominal rate of interest r and nonborrowed reserves v — d. They also induce a negative correlation between r and v, total reserves, as (14) reveals since h increases in q. The correlation between r and v can be entirely attributed to the variance of inside money, z/i(/3'g); this variance also induces a negative correlation between r and the broader monetary aggregates Ml and M2, as shown by (17) and (18). From (16), it is clear that the ratio of borrowed to total reserves is positively correlated with the interest rate, a relation which -18- Coleman, Gilles, and Labadie has nothing to do with the source of the shock but is due exclusively to the form of ^, that is, to the operation of the discount window. If total reserves did not respond to the policy shock (an assumption which is sometimes made in empirical work), the form of ifr alone would induce a positive correlation between the interest rate and borrowed reserves. Suppose now that shocks to /3 are the only shocks in the system. The left side of (13) is decreasing in g, while the right side is increasing in q and decreasing in /3'g; therefore, q and /3'q (and therefore q and /3' also) move in opposite directions, while {31 and /3'q move in the same direction. Equations (14)—(18) then show that preference shocks induce a positive correlation between the interest rate and any of the reserve or monetary aggregates (total, nonborrowed, and borrowed reserves; inside money, Ml, and M2). It is now possible to use the example to study more complicated policies. Suppose that in response to positive preference shocks that would otherwise increase interest rates, the government chooses its open market operation to keep the rate constant, which corresponds to a small realization of g1 (in this case, /3 and g are still iid, but not independent of each other). With the interest rate constant, /?' high and gf low, all the reserve and monetary aggregates are high (but the borrowed reserve ratio is constant). If the policy response only partially offsets the preference shock, all reserve and monetary aggregates may still rise, while the rate of interest rises also. In that case, despite the presence of a liquidity effect in the model, open market operations could be seen as "inducing" a positive correlation between interest rates and various monetary aggregates (and nonborrowed reserves as well). CONCLUSION: INTERPRETING THE EMPIRICAL LITERATURE As mentioned in the introduction, the empirical literature directed to measuring the effect of monetary policy shocks on interest rates is replete with seemingly conflicting results. The model provides a framework for thinking about these results and for interpreting the literature; the example brings out the important features of the model. First, the model highlights the role of inside money creation as an avenue for total reserves to respond to open market -19- Coleman. Gilles, and Labadie operations. In this sense, the model fails to support Strongin's identifying restrictions that total reserves do not respond to open market operations within a month or a quarter. Second, the model suggests that the operation of the discount window, summarized by a fixed and positively sloped supply function, can alone generate a negative correlation between nonborrowed reserves and the federal funds rate. Such a correlation has been documented by Christiano and Eichenbaum (1991). While they identified policy shocks as innovations to nonborrowed reserves, the model suggests an alternative explanation that has nothing to do with policy shocks. Third, although the model is designed to have a liquidity effect, a policy of interest-rate smoothing hinders efforts to detect its presence. This could explain the difficulties econometricians have had in measuring this effect. To identify policy shocks, it is not sufficient to identify a variable (such as nonborrowed reserves) that is under the control of the Fed, since the Fed may use its instrument to achieve particular objectives. In this sense, the model points to the familiar need, and provides a framework for, identifying demand and supply shocks to estimate a liquidity effect. -20- Coleman, Gilles, and Labadie REFERENCES Bernanke. B., and A. Blinder. "The Federal Funds Rate and the Channels of Monetary Transmission," Working Paper No. 3487. New York: National Bureau of Economic Research, October 1990. Cagan, P. The Channels of Monetary Effects on Interest Rates. New York: National Bureau of Economic Research, 1972. Cagan, P., and A. Gandolfi. "The Lag in Monetary Policy as Implied by the Time Pattern of Monetary effects on Interest Rates," American Economic Review, vol. 59 (Papers and Proceedings, 1969), 277-84. Christiano, L. J., and M. Eichenbaum. "Identification and the Liquidity effect of a Monetary Policy Shock." Unpublished manuscript, Federal Reserve Bank of Minneapolis, November 1991. Coleman, W. J. "Solving Nonlinear Dynamic Models on Parallel Computers," Institute for Empirical Economics working paper. Federal Reserve Bank of Minneapolis, 1992. Dutkowsky, D. "The Demand for Borrowed Reserves: A Switching Regression Model," Journal of Finance, vol. 39 (1984), 407-24. Freeman, S., and G. W. Huffman. "Inside Money, Output, and Causality," International Economic Review, vol. 32 (1991), 645-67. Fuerst, T. S. "Liquidity, Loanable Funds, and Real Activity," Journal of Monetary Economics, vol. 29 (1992), 3-24. Goldfeld, S. M., and E. J. Kane. "The Determinants of Member-Bank Borrowing: An Econometric Study," vol. 21 (1966), 499-514. Goodfriend, M. "Discount Window Borrowing, Monetary Policy, and the PostOctober 6, 1979 Federal Reserve Operating Procedure," Journal of Monetary Economics, vol. 12 (1983), 343-56. Grossman, S. J., and L. Weiss. "A Transaction-based Model of the Monetary Transmission Mechanism, " American Economic Review, vol. 73 (1983), 871-80. King, R., and C. Plosser. "Money, Credit, and Prices in a Real Business Cycle," American Economic Review, vol. 74 (1984), 363-80. -21- Coleman. Gilles, and Labadie Leeper, E. M., and D. B. Gordon. "In Search of the Liquidity Effect," Journal of Monetary Economics (forthcoming). Lucas, R. E., Jr. "Liquidity and Interest Rates," Journal of Economic Theory, vol. 50 (1990), 237-64. Polakoff, M. E. "Reluctance Elasticity, Least Cost, and Member-Bank Borrowing: A Suggested Integration," Journal of Finance, vol. 15 (1960), 1-18. Rotemberg, J. J. "A Monetary Equilibrium Model with Transaction Costs," Journal of Political Economy, vol. 92 (1984), 40-58. Sims, C. A. "Interpreting the Macroeconomic Time Series Facts: The Effects of Monetary Policy," European Economic Review (forthcoming). Strongin, S. "The Identification of Monetary Disturbances: Explaining the Liquidity Puzzle." Unpublished manuscript, Federal Reserve Bank of Chicago, December 1991. Waller, C. J. "Administering the Window: A Game-Theoretic Model of DiscountWindow Borrowing," Journal of Monetary Economics, vol. 25 (1990), 273-87. -22- Chart 1. Federal Funds Rate and Nonborrowed Reserves Ratio Monthly, January 1961 - July 1992 percent 1960 1965 1970 1975 1980 1985 1990 i Chart 2. The Psi Function; 1961 (1)-1992(7). Ratio of Borrowed To Total Reserves U.1U • 0.08 — • 0.06 • • • • • • • • 0.04 *• • • • : • • • • • • • § • • • • •• • • • • • • ••• • •• • • • • / •• • • • •• • • •- • • • • 1 • • • 1 • • • • 0.02 • % 0.00 hi • . : *• .wA * * i * '„•.•••• 1 •• •• •• • . l_ -2 finrftarl fFed Funds - Discount Ratol L_ l_J Comments on "Discount Window Borrowing and Liquidity" by Coleman, Gilles, and Labadie Michael Dotsey I have been asked to discuss "Discount Window Borrowing and Liquidity" which I view as very interesting but preliminary work toward examining "liquidity effects" in a framework that incorporates a fairly (primitive) reserves market. I use the term primitive with regard to the reserves market since no interesting dynamic behavior is present in this market. Viewing work on BRd, especially that of Goodfriend (1983) this is a shortcoming that I hope will be addressed by later generations of the model. The paper, however, is very rigorous and state of the art on other dimensions and the authors deserve a lot of credit for moving the liquidity effects literature in this direction. The empirical motivation for the paper can be traced to work by Christiano and Eichenbaum (1992) and especially to that of Strongin (1991). Strongin's work is fairly persuasive and indicates that in order for any model to replicate data on liquidity type effects reserve market behavior is likely to be a crucial ingredient. This is because the liquidity effect only shows up in NBR's or to be more accurate, in the part of NBR that represents independent monetary policy. This paper's novel inclusion of reserve market behavior represents a commendable extension of this basic line of research.1 In reading this paper, I found that it raised at least as many questions as it answered. Much of my confusion is not the 1. One thing I would like to see done in these estimations is removing settlement day data. This data could potentially contaminate the results. Suppose for instance the Fed misforecasts float or treasury balances believing there will be more of these funds available than are actually there. NBR will be low on the settlement day and the funds rate will be high, perhaps by a substantial amount. Two such occurrences in a month (at least 25% probability) could make monthly average NBR a little low and monthly average rF a little high. While I doubt this is the reason for Strongin's results it would be nice to purge the data of what is merely an interbank friction. Michael Dotsey result nor the fault of this paper in particular, but rather comes from a lack of understanding and perhaps misgivings of this literature in general. In my comments I will discuss some of these misgivings and, hopefully, my comments will lead to some discussion from the rest of the audience. The paper extends a branch of research that is attempting to understand the effect of monetary policy on interest rates and real activity. In particular these papers7 search for a mechanism that will explain (1) how contractionary monetary policy raises short-term interest rates and (2) how it causes declines in economic activity. This literature received its impetus from Lucas's (1990) influential paper. A common feature of most of this literature involves cash-in-advance constraints that constrain the amount of money available for use in a loan or securities market, however, no two papers seem to use the same exact specification. Lucas's original setup and CGL (1991) envision bond traders as only having limited funds and, therefore, open market operations affect the price of bonds .and thus interest rates. The appeal of Lucas's setup is that it eliminates the differential wealth effect of open market operations that were present in earlier literature (eg Grossman and Weiss and Rotemberg). Fuerst (1991) extends Lucas's setup to a production economy that places a CIA constraint on both investment and labor expenditures. Unlike households' portfolio decisions, production decisions are made after the stochastic state of the economy is known. Since individuals must choose the portion of their portfolio to lend to firms via intermediaries prior to observing the monetary transfer or the market clearing interest rate, the monetary transfer can affect the tightness or looseness of the loan market. Hence liquidity effects that have real consequences result from monetary policy. Christiano (1991) subjects the Fuerst model and an alternative version of that model in which investment decisions - 2 - Michael Dotsey are also made prior to the realization of shocks to a statistical comparison with a RBC model that contains a standard CIA constraint. For reasonable parameter specifications the Fuerst • model can not produce a liquidity effect that dominates anticipated inflation effects on the nominal interest rate while the sluggish capital model can produce a dominant liquidity effect. Both these models produce too much variability in consumption and the counterfactual result that consumption and prices move in opposite directions. They also produce very low interest elasticities of money demand and monetary policy has very little effect on variations in output. Furthermore, anticipated inflation has much too large an effect on labor, consumption, and output. To remedy this last result, Christiano and Eichenbaun (1992) relax the CIA constraint on investment. They also split the period into two parts allowing firms to adjust their hiring decision after observing open market operations while initial hiring and investment decisions are made prior to observing open market operations. They do this with the hope of magnifying the response of employment and output to liquidity effects. In CGL's current paper firms face a CIA constraint on investment but can pay workers out of end of period revenues. Also, monetary transfers are made directly to consumers after their portfolio decision has been made. Thus these transfers do not affect the funds available in the credit market and, therefore, do not give rise to a "liquidity effect." Because there is a CIA constraint on capital, monetary policy can have inflation tax effects as well. As their work progresses separating liquidity effects from inflation tax effects will be important. Not all of these scenarios can be correct. constraints placed where they are? These assumptions of infinite transactions costs are not innocuous. in these models. Why are CIA They are the driving force It seems that rather than trying to incorporate a realistic financial structure into a dynamic macro model and - 3 - Michael Dotsey then testing the model, investigators are trying to find a mathematical structure that produces the correlations they desire. Apart from Christiano (1991) very little effort is made to see if these models are an improvement on basic RBC models or even if they produce counterfactural predictions along other dimensions. Since other classes of models can produce negative correlations between NBR and the funds rate, examining how CIA models fit the data along other dimensions will be important if the CIA approach is to gain widespread acceptance. For example a model like that in Goodfriend's (1987) paper can potentially produce correlations of the type this literature is seeking. In that model, which has no rigidities, purposeful behavior by the Fed can set up negative correlations between the funds rate and NBR. If the Fed wishes to reduce inflation, it can do so by reducing the future money supply and in particular future NBR. Due to anticipated inflation effects, the nominal interest rate would fall increasing the demand for money and total reserves. If the Fed wishes to reduce price level surprises it can supply the necessary NBR to prevent price level movements. Thus this policy sets up the requisite negative correlation. If that was all that was going on one would expect this negative correlation to carry over to broader aggregates. However, M2-M1 components of M2 which involve a large savings motive should be positively correlated with the real rate of interest and movements in BR, which are highly variable and positively correlated with the funds rate, could cause TR to be positively correlated on net as well. Alternatively say the Fed is following an exogenous upward movement in the real rate of interest in an attempt to target inflation. If the own rate on money balances is sticky then money (Ml) and hence total reserves will decline along their demand curve. (Also, M2 could be rising with the real rates.) This would set up a negative relationship between NBR and the funds - 4 - Michael Dotsey rate. As rm adjusted, total reserve demand would increase as would NBR as the Fed defended the new higher funds rate. If the Fed did not react instantaneously or vigorously enough to the increased reserve demand the funds rate could rise further and then fall as nonborrowed reserves were pumped into the system reinforcing the initial negative correlation. Also, sticky price models may be able to generate some of the correlations displayed in the data as well. Also, the question of what constitutes a period is somewhat fuzzy in this literature. Is it a day or perhaps a week? Most people make some form of cash management decision weekly and I can not think of any time where a shortage of cash has affected my real consumption for more than a day or two. Perhaps I'm taking the CIA constraint too literally, but if the period is rather short, as I believe it is, then the propagation mechanisms needed to match the data would seem incredible by RBC model standards. I have strayed a little far afield so let me return to this paper more specifically. My primary confusion is linking the author's major contribution which shows how different measures of money can have different correlations with interest rates with the motivation for their paper which appears to be the results found in Strongin. In this paper money (1) M t ^ - Mt + rt(Gt-Dt) + xt. The xt portion of measured money provides no liquidity effects. The 6t portion, that is open market operations has the standard liquidity effects since it influences the portion of firm borrowing that must be financed by discount window loans. The equilibrium condition that is being used is (2) NBRt = Vt - Dt = Mt - Gt where Vt * 0(M t +B t ). An increase in 6t (an open market sale) requires more discount window borrowing and an increase in interest rates since r =0(D/TR) is increasing. - 5 - Using Mt+1 can Michael Dotsey contaminate regression results since it rises by r t (G t -D t ), which will in general be positive in this model and no liquidity effect will be present. Furthermore, growth in money via transfers wiVI further bias econometric results. For econometric purposes I see no useful way of isolating any aggregate to uncover liquidity effects- Xt type disturbances, in reality, involve transfers from the Treasury. These involve a reduction in Treasury accounts at the Fed and an increase in NBR. What the model here indicates is that one wants to examine only changes in reserves that involve changes in the public's asset positions and that exclude any interest or lump sum payments. While these decompositional problems are important for this model and may in fact be important more generally, they seem to have little to do with Strongin's empirical strategy nor do they affect interpretations in other models. Strongin tries to separate "pure" supply movements in NBR from those engendered by policy responses to changes in TR. Whether his identification procedure is a good one or not could be debated, but he is not concerned with measurement or decompositional problems in various reserve measures. The decompositional problem arises in CGL because of their modeling of xt as having no liquidity effects. In Fuerst or Christiano and Eichenbaun, there is only xt and it enters the model in a way that produces liquidity effects. That is NBR supply disturbances that are not responses to TR shocks produce liquidity effects. It seems that Strongin's methodology is more closely aligned with these models. Whether decompositional problems are important or not, I don't know. They arise in this model by a specification that at this point seems somewhat arbitrary. It is no more arbitrary than any other specification in the literature, but that does not make it convincing. I believe the author's need to make a convincing argument as to why some forms of morjey creation are more likely to - 6- Michael Dotsey involve liquidity effects than others if their message is to carry weight. After all, in this model one could easily reverse the roles of Xt and Gt or make them complimentary. The discussion on page 11 regarding the estimation of rp is also a little confusing. With (3) n-1 -± -±£ v V they claim that 0 can be estimated no matter what the shock. But is that relevant? We would like to know how j> is influenced contingent on different shocks. Here a positive V shock induced by a shift in the demand for loans causes 0 to rise and n to fall, while a decline in NBR due to an open market sale (G up) also causes n to fall and i> to rise. It is only the latter effect that one has in mind when discussing liquidity effects, so perhaps the ratio is not the correct variable to focus on. Rather, in this model it should be the relationship between the level of NBR and the funds rate. Also in estimating 0, one would expect shifts in the function over time since administration of the discount window has changed over time. For example, I believe window administration was more lax when the Fed faced a membership problem. I would also downplay somewhat figure one. The interest rate of consequence is the spread between the funds rate and the discount rate. When one looks at this graph the correlations seem at least as pronounced. But has anything but a borrowed reserve demand function been uncovered? Finally, the discussion concerning adjustably pegging the interest rate based solely on technological disturbances raises questions concerning the nominal determinacy of the model (see McCallum (1981, 1986)). Overall, I thought this paper was interesting and represents a nice attempt to start thinking about how behavior in - 7 - Michael Dotsey the market for reserves influences the correlations we observe between various monetary measures and the funds rate. Given my qualms concerning this methodology's ability to explain anything at business cycle frequencies, I would suggest directing the model in an alternative direction. Perhaps this framework could be used to help explain short-term term structure movements in interest rates and examine the so-called "ozone hole." This line of inquiry would be interesting since it could integrate reserve market behavior and a tight specification of policy in a fully developed general equilibrium model. - 8 - Michael Dotsey REFERENCES Christiano, Lawrence J., "Modeling the Liquidity Effect of a Money Shock," Federal Reserve Bank of Minneapolis Quarterly Review, Winter 1991. 15(1). 3-34. Christiano, L.J., and M. Eichenbaum, (1992) "Liquidity Effects, Monetary Policy and the Business Cycle," unpublished ms.. Federal Reserve Bank of Minneapolis, Northwestern University, NBER and Federal Reserve Bank of Chicago, July 1992. Fuerst, T.S. "Liquidity, Loanable Funds, and Real Activity," Journal of Monetary Economics, vol. 29 (1992), 3-24. Goodfriend, Marvin. "Discount Window Borrowing, Monetary Policy, and the Post-October 6, 1979 Federal Reserve Operating Procedure," Journal of Monetary Economics, vol. 12 (1983), 343-56. , "Interest Rate Smoothing and Price Level TrendStationarity, " Journal of Monetary Economics, March 1987, 19, 335-48. Grossman, S. J., and L. Weiss. "A Transact ion-based Model of the Monetary Transmission Mechanism," American Economic Review, vol. 73 (1983), 871-80. Lucas, R.E., Jr. "Liquidity and Interest Rates," Journal Economic Theory, vol. 50 (1990), 237-64. of McCallum, Bennett, T. (1981) "Price Level Determinancy with an Interest Rate Policy Rule and Rational Expectations," Journal of Monetary Economics, vol. 8 (November). , (1986) " Some Issues Concerning Interest Rate Pegging, Price Level Determinacy, and the Real Bills Doctrine," Journal of Monetary Economics vol. 17 (January). Rotemberg, J.J. "A Monetary Equilibrium Model with Transaction Costs," Journal of Political Economy, vol. 92 (1984), 40-58. Strongin, S. "The Identification of Monetary Disturbances: Explaining the Liquidity Puzzle." Unpublished manuscript, Federal Reserve Bank of Chicago, December 1991. Credit Conditions and External Finance: Interpreting the Behavior of Financial Flows and Interest Rate Spreads Kenneth N.Kuttner1 Aflurryof recent macroeconomic research has drawn attention to the relationship between monetary policy, credit conditions, and the markets for short-term debt Two recent papers have focused onfirms'substitution between bank and non-bank externalfinancein particular, proposing macroeconomic indicators based onfinancialmarket activity. Kashyap, Stein, and Wilcox (1992) employ quantity data directly, arguing that the share of bank loans out of firms' total short-term finance is an informative index of Federal Reserve policy and loan availability more generally. In a complementary line of research, Friedman and Kuttner (1992) identify monetary policy and bank lending as potential sources offluctuationsin the spread between yields on commercial paper and Treasury bills. While both papers have demonstrated solid empirical links between these financial indicators and real economic activity, neither hasrigorouslyassessed the extent to which fluctuations in these indicators actually represent exogenous changes in credit conditions, rather than endogenous responses to changing economic conditions. This paper's goal is to provide such an assessment. The paper begins with a sketch of the mechanism through which credit conditions affect firms' short-termfinancing,drawing a distinction between the effects of the Federal Reserve's open market operations and other factors influencing banks' willingness to lend. The second section summarizes the reduced-form relationships between real output, the interest rate, and three alternative indices 1. Senior Economist, Federal Reserve Bank of Chicago. I am grateful to Benjamin Friedman and David Wilcox for their comments and suggestions. -1- Kuttner of credit conditions: the composition of external finance, the spread between the loan rate and the commercial paper rate, and the analogous spread between commercial paper and Treasury bills. The third section turns to a closer examination of the impact of monetary policy and loan availability on bank and non-bank finance using structural VAR techniques. Identifying monetary policy with innovations to non-borrowed reserves and controlling for firms'financingrequirements, the first of the three models estimates the dynamic effects of monetary and lending shocks on the composition of external finance, the interest rate, and real output. The second structural VAR system assesses the effects of reserves and lending shocks on the paper-bill spread. The third model identifies lending shocks with innovations in the loan-paper spread. Estimates of these models confirm that all three variables respond appropriately to reserves shocks. In addition, lending shocks, whether identified through financial flows or via fluctuations in the loan spread, induce a substitution between bank and non-bank finance. Less clear is the extent to which any of these measures exclusively reflects the effects of changing loan availability. The fact that positive lending shocks are associated with increases in the interest rate and the paper-bill spread suggests that changes in the composition of external finance have more to do with firms' financing requirements than with exogenous changes in banks' willingness to lend. Another slightly puzzling observation is that the largest source of changes to the composition of external finance seems to be wholly unrelated to both reserves and bank lending. Together, these two results suggest that while credit conditions are one important determinant of firms' choice of financing, short-term debt flows may be informative for reasons other than those involving the substitution between bank/non-bank substitution. Although its implications for real activity are rather weak, the loan spread appears to be a plausible alternative measure of credit conditions. A model offinancialflowsand interest rate spreads How do the markets for short-term bank and non-bank finance respond to monetary impulses? And how do non-monetary shocks affect these markets? And how might one construct an index of the availability of intermediated funds? -2- Kuttner As a first step towards answering these questions, this section analyzes a simple model of the markets for commercial paper, bank loans, and Treasury bills in the style of Brainard (1964) or Bosworth and Duesenberry (1973). While not as detailed as either of those models, it is adapted to highlight firms' tradeoff between bank and non-bank finance. It also draws an important distinction between purely monetary influences acting through open market operations, and credit conditions defined more broadly, which may include other factors affecting banks' willingness to lend. One of the model's more obvious properties is that an injection of reserves causes the interest rate to fall — the familiar "liquidity effect." Reserves injections also cause the spread between the interest rates on bank lending and commercial paper to fall, and leads to increased reliance on bank finance. Lending shocks, which are assumed to affect only banks' preferences over alternative assets, turn out to have similar effects on the loan-paper spread and the composition of firms' finance. Lending shocks, by contrast, have no effect on the level of interest rates — only the spreads. The model also identifies two other factors with implications for the money market. First, firms' demand for external finance may induce changes in the relevant interest rate spreads and consequently the composition of finance; controlling for this demand-side influence turns out to be a major challenge to the construction of an empirical measure of credit availability. Similarly, the stock of outstanding Treasury bills may have tangible effects on the spreads and the composition of finance. The three players in the money market are households, banks, and firms, who participate in the markets for reserves, commercial paper, Treasury bills, and loans. Specifically, households' portfolios include demand deposits (DD), commercial paper (P), and Treasury bills (B) according to DD* = <Krp) W, 4>' < 0 df df P* s flrp, rB)W, — > 0 and — < 0 brp drg B^ s (1 - <J) - / f o rB)) W, -3- Deposit demand Paper demand Bill demand Kuttner where W is the sum of deposits, paper, and bills held by households. Households' demand for non-interest-bearing bank deposits is a decreasing function of the prevailing paper rate, rP. A key assumption is that households view commercial paper and Treasury bills as imperfect substitutes, so that changes in their relative supplies affect their respective yields.2 Households require a higher paper rate (or a lower bill rate) to hold a larger share of their portfolio as commercial paper. Demand deposits are banks' sole liability. Their assets are divided among Treasury bills, loans (L), and deposits at the Federal Reserve (R) according to: K* s p(rp)DDt p ' < 0 Ld = Sin, rP, \)DDf — > 0 and — < 0 drL drP B*b s (1 - p(rP) - g(rLf rP, K))DD. Reserve demand Loan demand Bill demand Banks' demand for non-interest-bearing reserves falls with the prevailing paper rate, while loan demand is increasing in the loan rate and decreasing in the paper rate.3 The stock of reserves is set at R' by the Federal Reserve; discount window borrowing is ignored. Banks' demand for loans is also allowed to depend on the variable X, representing any other factors affecting banks' willingness to lend. These "lending" shocks lead banks to shift the composition of their portfolios between bills and loans; negative shifts in X may be interpreted as "credit crunch" episodes. These may occur in reaction to a perceived deterioration in borrowers' creditworthiness, or to more stringent capital requirements as suggested by Bernanke and Lown (1991). They may also be the result of the "moral suasion" instrument of monetary policy; Owens and Schreft (1992) identify a number of episodes in which banks contracted their lending in response to Federal Reserve pressure. Whatever the source, the key feature of these "lending" shocks is that they need not be accompanied by overt monetary policy in the form of open market operations.4 2. Friedman and Kuttner (1992) discuss some possible reasons for this imperfect substitutability. Lawler (1978) also finds evidence for imperfect substitutability at seasonal frequencies. 3. Note that throughout the paper, assets are "demanded" while liabilities are "supplied." Hence, banks "demand" loans and bills, while firms "supply" loans and paper. 4. This point is stressed by Friedman (1991). -4- Kuttner Finally,firmschoose between bank lending and paper issuance as sources of short-term finance according to fth hh dri orp V a h(rL, rP)F, —- < 0 and —- > 0 Loan supply P* » (1 - h(rL, rP))E Paper supply For simplicity, the amount to befinanced,F, is assumed to be exogenous with respect to the various interest rates. Becausefirmsview loans and paper as imperfect substitutes, they willfinancesome portion of F through bank lending even though rL generally exceeds />; as discussed by Kashyap, Stein and Wilcox (hereafter KSW), this presumably reflects some intangible benefit accruing to the firm from maintaining a relationship with a bank. Firms * share of bankfinance(the KSW "mix") responds predictably to the loan and paper rates: an increase in the loan rate (or a decrease in the paper rate), leadsfirmsto substitute away from bankfinancetowards non-bank external finance.5 In equilibrium, the demand for the four assets equals their supply, p(rpMrP)W = l? frurpiKftW-hirurpyF^O Kr»rg)W-{l-h(n,r,)yFmO (1 - g(rL, rPt \))$W+ (1 - / f a rB) - +)W = B*9 determining yields and quantities as functions of the exogenous /?', X, F, and B*. Walras' law allows the bill market equation to be dropped. Further simplification is possible by assuming the asset demand and supply functions to be homogeneous of degree zero with respect to the assets' 5. This model embodies the assumption that bank and commercial paperfinanceare viable alternatives for an economically relevant group offirms.However, there is increasing evidence that this set offirmsis rather small, and that much of the observed variation in the aggregate composition of finance is due to the relative availability offinanceto small and largefirms;see Gertler and Gilchrist (1992) and Oliner and Rudebusch (1992). -5- Kuttner yields, so that (for example) g(n +c,rp + c, X) = gin, />, X) for any constant c. In this case, the/, g and h functions can be specified in terms of interest rate spreads, and the system reduces to: gizLP,mrpW-h(zu>)F = 0 (I) KzpBW-(l-h(zu>))F = 0 where zLP and zPB denote the loan-paper and paper-bill spreads. Analyzing*the comparative statics of (1) is simplified by its (somewhat artificial) recursive structure. The interest rate level is entirely determined by supply and demand in the market for reserves; the fall in reserves resulting from a contractionary open market operation requires a higher rate to equilibrate the reserves market, as illustrated in Figure l. 6 This higher interest rate leads in turn to a shrinkage of demand deposits and the banking system as a whole. Banks respond by raising the loan-paper spread, prompting some of its borrowers to switch to alternative forms of finance— short-term paper in this model. The increased supply of paper (relative to bills) leads to a widening spread between the paper and bill rates. The effects of an adverse lending shock resemble those of a reserves contraction in that both produce a rising loan spread and a substitution towards non-bank finance. Although both shocks produce similar effects on banks' portfolios, they differ in one important respect: reserves shocks affect the level of the short-term interest rate, while lending shocks leave the paper rate unchanged. A fall in X leads banks to shift the composition of their portfolios away from loans and into Treasury bills, leaving their reserve demand and the paper rate (and consequently deposits and the banking system's size) unchanged. Banks increase their spreads relative to the paper rate in order to reduce their stock of loans. As before,firms'increased reliance on commercial paper drives up the paperbill spread. 6. Total wealth is held constant in an open market operation, as the withdrawal of reserves is offset by a sale of Treasury securities. -6- Kuttner The observation that both reserves and lending shocks may contribute to real economic fluctuations is one explanation of the widespread interest in constructing a broader measure of credit conditions than reserves or the interest rate in isolation, which reflect largely those shocks originating from the reserves market The attractive feature of the credit conditions indicators discussed here is their ability to detect the effects of changes in loan availability and reservesfluctuations:in this model, the "mix," the loan-paper spread, and the paper-bill all reflect the impact of both types of shocks. In fact, in the absence of any other shocks, all three of these measures should respond to monetary and credit factors in qualitatively similar ways. One problem common to all three of these measures (and the interest rate itself) is their susceptibility to contamination from changes infirms'overall demand forfinancing,which may alter yield spreads and the composition of externalfinancefor reasons having nothing to do with to exogenous changes in credit conditions.7 This can be illustrated by examining the comparative statics of (1) in response to an increase in F, the dollar amount of fundsfirmswish to raise from the short-term credit markets. A greater demand for loanable funds unambiguously increases the prevailing interest rate, />. Its effects on the loan-paper spread (and therefore the composition of external finance) is ambiguous, as it depends onfirms9share of bankfinance(/t) relative to households* wealth fraction in bank deposits (<|>), and the share of banks' portfolios held as loans (g). When h(zLP) > tyrp)g (as is presumably the case), increases in F cause loan demand growth in excess of deposit growth, driving up the relative cost of bankfinanceand the share of paper infirms'external finance.8 The same inequality is also relevant for the paper-bill spread; a second sufficient condition for a rising spread is that (1 - h(zLP)) > J{zpB\ so that the increasing paper demand would require households to hold a larger share of paper in their portfolios. 7. Under most of the Federal Reserves' post-Accord operating procedures, non-borrowed reserves may also be contaminated in this way; see Strongin (1991). 8. A special feature of the KSW model is that changingfinancingrequirements affect loans and paper proportionally, leaving the "mix" unchanged. -7- Kuttner One additional complication for interpreting the paper-bill spread as a measure of credit conditions is that it may be affected by changes in the outstanding stock of Treasury bills. In addition, the wealth effects associated with changes in the volume of Treasuryfinancemay alter the level of interest rates and loan spread, and consequently the composition of external finance.9 In this model, an increase in the supply of bills reduces the paper-bill spread, as investors require higher returns to entice them to hold the additional stock of bills. This increase in banks' demand for loans leads to a fall in the loan rate relative to the paper rate, and increased reliance on bank finance. To summarize, the model's main implications are: • Both reserves and lending shocks alter the relative price of bank and non-bank finance, inducing a substitution between alternative forms of external finance. • By affecting the supply of commercial paper, this substitution also affects the relative yields on Treasury bills and commercial paper. • Changes in reserves affect the level of interest rates, while lending shocks leave the level unchanged. • Firms' overallfinancingrequirements may affect interest rate spreads and their composition of short-term finance. The goal of the paper's subsequent empirical work is to explore these implications. Specifically, it attempts to identify lending shocks through their impact on the composition of externalfinanceand interest rate spreads, while controlling for reserves and the overall demand for loanable funds. Short-term credit markets and real economic activity One desirable feature of any index of credit conditions is a systematic link between it and subsequent fluctuations in real economic activity.10 The results below summarize the predictive properties of the KSW "mix," the prime-paper spread, and the paper-bill spread. The results show that the "mix" 9. Of course, this assumes that households view government bonds as net wealth; see Barro (1974). 10. Economists and market observers have long recognized the cyclical properties of commercial paper, bank lending, and their relative yields; see, for example, Foulke (1931), Selden (1963), and Stigum (1990). -8- Kuttner and the paper-bill spread are good predictors of future changes in real GDP (although this alone does not justify their interpretation as measures of credit availability). "Causality" tests Table 1 examines the incremental information content of the three measures for future changes in real GDP in the presence of traditional measures of monetary policy: non-borrowed reserves and the commercial paper rate. Regressions 1-3 are four-variate reduced-form equations of the form 4 4 4 4 Ax, a Ho + Hi* + ] T OjAx^ + ^T pi[A ln(J?),w + ] T Y,Arj>^ + ] T 6 , A ^ + e, where x is the logarithm of real GDP, R is non-borrowed reserves adjusted for extended credit and deflated by the GDP deflator, i> is the commercial paper rate, and q denotes, in turn, the "mix", the loan-paper spread, and the paper-bill spread. As in KSW, the "mix" is computed as the observed ratio of bank lending to the sum of lending to commercial paper, or L/(L + P).n The results use the six-month commercial paper and Treasury bill yields, and the prime rate (from the Federal Reserve H.1S release) is used as the lending rate. The table reports F-tests for the exclusion of the four 6, terms for the entire 1960:2-1991:4 sample, as well as two shorter samples. One truncated sample begins in 703, when Regulation Q was eliminated forroostlarge CDs.12 Another begins in 1975:1. Although this date is somewhat arbitrary, it corresponds roughly to the beginning of a rapid expansion of the commercial paper market, during which it became a more popular vehicle for non-financialfirms'short-term finance.13 11. The augmented Dickey-Fuller u statistic (computed with eight lags) for the stationarity of the "mix" is -4.10, rejecting the null hypothesis of nonstationarity at the 1% level. Consequently, it is included here in levels along with a linear trend term. 12. Regulation Q interest rate ceilings on 30-89 day CDs in denominations of $100,000 were eliminated on June 24, 1970. Ceilings on CDs with maturities in excess of 90 days remained in place until March 16,1973. -9- Kuttner The 1975-91 sample also excludes the Penn Central and Franklin National disruptions of 1970 and 1974, and covers the period in which ratings were assigned to commercial paper issues.14 The results of thefirstregression corroborate the strong link between the "mix" and real output found by KSW, supporting theirfindingthat the composition of finance has significant predictive power for future real economic activity, even in the presence of reserves and interest rates. The poor performance of the loan-paper spread in the second regression (again in the presence of reserves and the commercial paper rate) is consistent with the notion that banks' lending rates are relatively uninformative.15 The third regression demonstrates the incremental information content of the paper-bill spread — at least in the earlier samples. Impulse responses While the F-statistics for "causality" give some indication of the strength of the predictive power of thesefinancialindicators, they give no indication of the size or direction of their impact. The impulse response functions plotted in Figure 2 provide a richer description of the effects of innovations to the financial indicators. Each of the three rows of graphs is from the VAR corresponding to regressions 1-3 in Table 1. In each case, the system has been orthogonalized (according to the triangular Cholesky decomposition) with the credit conditions index in last place. Three responses are plotted for each regression: thefinancialindicator's effects on output and the interest rate, and the effect of reserves innovations on thefinancialindicator. The dotted lines depict the approximate 95% confidence bounds. Panels (a) and (b) from the first specification show that "mix" innovations indeed act like reasonable measures of credit conditions; reserves injections increase the share of bank loans, and 13. At the end of 1974, non-financial commercial paper accounted for only 13.5 billion dollars. By 1982, thisfigurehad grown 325.2 percent to 57.4 billion. See Hurley (1977,1982), and Stigum (1990). 14. Moody's and Standard and Poor's began rating commercial paper in 1974. 15. Similar results are obtained with the average of large banks' lending rates obtained from the Federal Reserve Survey of Terms of Bank Lending reported in release E.2. -10- Kuttner output rises in response to positive "mix" shocks, which might be interpreted as the pure lending component of credit conditions. The panel (c) plot, however, is something of a puzzle. It shows that "mix" innovations are associated with a rising commercial paper rate—not what one would expect from an increased willingness to lend on the part of banks, and inconsistent with the implications of the model presented earlier.16 However, this pattern is consistent with banks passively supplying more loans in response to rising demand for credit. The second row of plots confirm the generally weak relationship between the prime-paper spread and real output. One interesting feature of the loan spread is that it initially rises in response to a reserves innovation — clearly inconsistent with the loosening of credit conditions implied by the reserves injection. The loan spread ultimately falls, however, suggesting that this response is due to a certain sluggishness in the way banks adjust their lending rates. The impulse response functions from the paper-bill spread regression are all consistent with what one would expect from an indicator of credit conditions: positive shocks to the spread generate declining real output, while reserves injections reduce the spread. Furthermore, unlike the "mix", innovations in the spread itself have essentially no impact on the level of interest rates. Comparing the "mix" and the paper-bill spread Because regressions 1-3 included each of the credit conditions measures in isolation, the results raise an important question: to what extent are the three indicators measuring the same phenomenon? An obvious way to address this question is to include more than one indicator in the same regression to see if the presence of one vitiates the predictive power of the other. The results from two additional regressions (numbered 4 and 5) are reported in Table 1. The results from specification 4, which includes both the "mix" and the loan spread, are not surprising given the weak performance of the loan spread in isolation — the F-statistics for the "mix" remain virtually unchanged. Somewhat more surprising are the results from specification 5, in which both the "mix" and the paper-bill spread appear. Here, the relationship between the two variables and real 16. The "mix" terms are significant in the interest rate equation at the 10% level. -11- Kuttner output is uniformly stronger (judged by the F-statistics) than when they are included individually. Qearly, one (or both) of the indicators is doing something other than simply summarizing the state of credit market conditions. The roles of commercial paper and bank loans The model sketched earlier suggests that flows of commercial paper and bank lending are informative to the extent that they reflect the substitution between the two forms of finance in response to a monetary or a lending shock. KSW exploit this insight by looking at the ratio of bank loans to the sum of loans and paper, shocks that affect both forms of debt proportionally are presumed to stem from sources other than loan availability. A useful check on this specification is to verify that paper and lending flows enter an unrestricted regression in such a way that the "mix" is the variable that matters. This is easily accomplished by differentiating the "mix" (designated h) with respect to time, P — L dt = (I+P)* L P (I+P) 2 = h{\ - h%/L - /i(l - h)P/P9 decomposing its movements into distinct lending and paper contributions. In discrete time, the analogous decomposition, AA, - AM(1 - AKI)AL/1M - A M (1 - A,-i)AP/P M - <&L - Afip expresses A/i as a weighted sum of commercial paper and bank loan growth rates, denoted tJxL and tJip. If AA were in fact the appropriate measure of the impact of credit conditions on the real economy, the two components would enter real output regressions with equal and opposite signs; the regression itself would "choose" the KSW specification. Table 2 displays the results of this experiment. Panel (a) reports the outcome of a regression of first-differenced log real GDP on four lags of output, tJxL and tJiP over the 1960:2-91:4 sample. Judged by the F-statistics, the commercial paper terms are much more informative than the lending -12- Kuttner terms; tJip is significant at the 0.01 level, while the tJiL terms are not significant at even the 0.10 level.17 The sum of the estimated coefficients on lending is negative, but statistically insignificant The regression in panel (b) refines the test by specifying the regression in terms of tJi and tJip — simply a transformation of the regression in panel (a). Excluding the four lags of &hP is equivalent to restricting the coefficients on iskL and tJip to have equal and opposite signs. Here, the tJi terms are statistically insignificant, while the AhP terms are significant at the 0.05 level. Moreover; the negative estimated sum of the "mix" coefficients is inconsistent with the substitution hypothesis, although this sum is again statistically insignificant. To guard against the possibility that the results in the first two panels are an artifact of the differenced specification, panel (c) reports the results of a regression that includes a linear trend and h in levels. While not tJ\P terms are not as strong in the levels specification, the coefficients on the h terms remain statistically insignificant. These experiments show that the "mix" owes its predictive power in large part to something other than the substitution between bank and paper finance. In unrestricted equations, h terms are generally insignificant, while the hypothesis that commercial paper in isolation does not matter for predicting real output can be rejected. This observation suggests a closer examination of lending and commercial paper flows individually, and their relation to monetary policy and credit conditions. A structural approach to identifying lending shocks The atheoretical results in the preceding section provided some evidence in favor of interpreting the financing "mix" and the paper-bill spread as measures of credit conditions, although innovations in the composition offinancewere, contrary to the simple model, are associated with a rising interest rate. One reason for this pattern may be the result of inadequately controlling for the overall demand for short-term finance. As demonstrated earlier, an increase in the amount to befinancedneed not raise bank and non-bankfinanceproportionally. In this case, if increases infirms'demand for funds 17. This is consistent with the results of King (1986). -13- Kuttner are accommodated primarily through bank lending, the "mix" may rise for reasons unrelated to credit conditions. Figure 3 plots thefinancinggap (defined as the difference betweenfirms'capital expenditures less inventory IVA and after-tax internal funds) along with commercial paper and bank loan flows, demonstrating the close relationship between thefinancinggap and the volume of bank lending (although commercial paper appears to have become more sensitive to thefinancinggap in the later part of the sample). To control for credit demand, the results in this section include the financing gap as an additional determinant offirms'debt issuance. A more interesting alternative hypothesis is that is that the substitution mechanism inadequately explains the joint behavior of commercial paper and bank lending, and that factors other than monetary policy are what drive the observedfluctuationsin the composition of short-term external finance. The apparent asymmetry between the effects of loan and paper flows uncovered in Table 2 provides some circumstantial evidence for this view. The results presented in this section attempt to address these issues by separately analyzing flows of lending and commercial paper in a structural VAR setting that controls for the overall demand for loanable funds. Moving to a more structural approach also addresses the possibility that the interest rate's odd response to "mix" shocks is as an artifact of the artificial triangular structure of the Cholesky decomposition employed earlier. Thefirstmodel focuses on the response of lending and paper flows to reservesfluctuations,and examines the properties of the innovations identified as lending shocks. The second describes the response of the paper-bill spread to thefinancialflows generated by reserves and lending shocks. The third usesfluctuationsin the loan-paper spread as an alternative means of identifying lending shocks. A review of structural VARs Beginning with an unrestricted i-variate dynamic simultaneous equation system, T -14- Kuttner the standard VAR achieves identification by restricting the contemporaneous relationships between the elements of y, i.e., by setting BQ = 0 and A = /, while placing no restrictions on the covariance matrix of v, ie., £(w') = Q. The structural VAR introduced by Blanchard and Watson (1986) and Bernanke (1986) achieves identification by allowing some nonzero elements in thcB0 matrix, while restricting the covariance matrix of v, the structural disturbances, to be diagonal. Off-diagonal elements in A can be introduced to allow distinct elements of y to depend on common structural shocks. Thus, structural VARs differ from traditional structural models by replacing the assumption of an exogenous instrument set with the assumption of orthogonal structural shocks. At the same time, the dynamics of the system are left unrestricted, as in the conventional VAR. Another interpretation of the structural VAR is as a decomposition of the covariance matrix of VAR residuals. If the structural disturbances are uncorrected with one another, Lc, £(w') = D, Q, the covariance matrix of the VAR errors becomes a nonlinear function of the structural parameters: Q=£(J#Av,v,'A'J#) mBfADA'Bf. If the system is just-identified, the above equality is exact; B^AD™ is a matrix square root of Q, and A'1 B0 diagonalizes Q.18 Reserves, lending, and short-term debt flows Thefirstmodel is a just-identified six-variable system involvingfinancinggap (F)> bank lending, non-financial commercial paper (P) the commercial paperrate(r>), real GDP (x), and non-borrowed reserves adjusted for extended credit (R). The interest rate is differenced, while reserves and GDP enter as log differences. The lending and paper data are again taken from the Flow of Funds accounts for the non-farm, non-financial corporate and noncorporate sectors. With F, P and L expressed 18. With a total of 2A2 elements in A and B0 and only &(&+1)/2 unique elements in Q, it is clear that the stnicmral parameters are not identified without additional restrictions on A and£0. The Cholesky decomposition, which is equivalent to setting B0 = / and making A lower triangular, is but one possibility. In overidentified systems, the problem becomes one of choosing the stnicmral parameters in Bo and A to generate the bestfitbetween thefittedand the observed covariance matrices. -15- Kuttner as shares of the total dollar volume of outstanding paper and loans, changes in the "mix" can be constructed as the weighted average of the two flows: v 'L+P L+P The substance of the model is contained in the six equations describing the contemporaneous relationships between the variables, R as bU6x + vx Reserves (2a) F as bxxR + bz& + V2 Financing gap (2b) rP ss bx\R + bxiF + V3 Interest rate (2c) L as b<xR + b^iF + d<30» + v4 P ss b^R + fci2P+*s^> + *MV4 + v5 x as d 0 r P + 644I + fc^P + v6 Lending (2J) Paper (2e) Output (20. No restrictions are placed on the dynamics of the system; consequently, terms dated t-\ and before are omitted, but implicit. Equation 2a allows the Federal Reserve to vary reserves contemporaneously with real GDP in a primitive feedback relationship. Thefinancinggap (equation 2b) also depends on the level of real economic activity. Consistent with the model presented earlier, the commercial paper rate in 2c is a function of reserves and thefinancinggap. The model's key equations are 2d and 2e, describing the behavior of bank lending and commercial paper flows as a function of thefinancinggap, reserves, and the interest rate. The coefficients on F measure the proportion of the currentfinancinggap satisfiedfinancedthrough loans and paper. The two equations' coefificients on R determine the immediate response, ceteris paribus, of the two forms of short-termfinanceto changes the banking system's reserve position. The v4 term in the lending equation represents lending shocks that are orthogonal to reserve andfinancinggap innovations, which would include factors such as credit crunches. For this interpretation of v4 to be -16- Kuttner legitimate, one of two conditions has to hold: either the observed financing gap must appropriately control for firms' demand for funds, or the amount of funds banks have available is fixed in the current quarter. The v4 innovation also appears in the commercial paper equation with the coefficient a^, allowing commercial paper to respond directly to lending shocks. This parameter determines the extent to which lending shocks are "recycled" into the commercial paper market within the current quarter. The v5 term in the commercial paper equation accounts for shocks to paper issuance uncorrected with the other structural disturbances. The final equation for real GDP is a reducedform equation describing the economy's response to the reserves and credit shocks in the preceding equations. The parameter estimates in Table 3 summarize the model's contemporaneous behavior, while the impulse responses functions plotted in Figure 4 describe its dynamics of the system whose orthogonalization is implicit in equations 2a-2f. Like the earlier reduced-form regressions, these results provide some evidence to support the use of lending flows as an indicator of credit conditions, while confirming the doubts raised in the atheoretical VARs. First, The negative estimate of the coefficient on R in the lending equation (2d) contradicts the hypothesis that the primary effect of monetary policy is a substitution between bank and non-bank finance; the contemporaneous response of an injection of non-borrowed reserves, ceteris paribus, is a fall in bank lending. However, because of the contemporaneous relationship from reserves to the financing gap and short-term finance via the interest rate and output, the coefficients on R in equations 2d and 2e do not by themselves determine the overall response of the "mix" to a reserves shock. The actual responses can be read from the impulse response function, plotted in the top panel of Figure 4.19 This shows that the net effect of a reserves injection is initially rather small, with the loan share gradually rising after two to three quarters. 19. The sample average values of h are used to compute the approximate response of the "mix" from the impulse response functions of the underlying variables. -17- Kuttner Figure 4 also shows that lending shocks seem to have a considerably larger impact on the composition of externalfinancethan reserves for thefirstfour quarters. Lending shocks9 effect is strengthened somewhat by the statistically significant negative estimate offl$4>which is consistent with roughly 10% of the lending shock being "recycled" into the paper market in the current quarter. The coefficients on F in the paper and lending equations show that neither responds immediately to fluctuations in thefinancinggap. A strong liquidity effect is associated with injections of non-borrowed reserves; the paper rate falls contemporaneously (the negative coefficient on R in equation 2c) and over a longer horizon (the center panel of Figure 4). These results also confirm the curious positive relation between the "mix" and the level of interest rates highlighted earlier in the paper. The center panel shows that positive lending innovations imply a rising interest rate, contradicting the theoretical model's implications for the effects of lending shocks. Both monetary and lending shocks are important sources of output fluctuations. Increased bank lending is contemporaneously associated with more rapid real GDP growth in the short run, as shown by both the positive (but not quite significant) coefficient on L and the impulse response function. What about shocks to commercial paper, v5? The top panel of Figure 4 shows that these shocks — which are, by construction, orthogonal to the system's other structural disturbances — have the largest and most persistent impact on the composition of external finance. Interestingly, the center panel shows that these innovations have essentially no implications for the interest rate, although they do seem to have a small, negative impact on real output. Financial flows and interest-rate spreads Recent papers by Bernanke (1990) and Friedman and Kuttner (1992) suggest that the substitution between bank and non-bank debt is an important source of fluctuations in the paper-bill spread. As discussed earlier, monetary contractions reduce lending by shrinking the stock of deposits, leading -18- Kuttner firms to raise the loan rate relative to the paper rate, discouraging intermediated borrowing. Similarly, adverse lending shocks cause banks to shift from loans to Treasury bills. Asfirmsturn to the paper market to satisfy theirfinancingneeds, the paper supply rises and bill supply to households falls, raising the paper-bill spread. If this is the way in which credit conditions affect the spread, one would expect tofindthe mechanism operating through the volume of outstanding non-financial commercial paper. The second structural VAR is designed to detect the operation of this mechanism. It augments thefirstmodel (equations 2a-2f) with the addition of a seventh equation for the paper-bill spread, x a b&rP + b^JL + b^sP + b^(rP - rB) + v6 Output (3/) rP-rB = bitiR + 6 7f2 F+b^rp + 67,4^ + bn,$P + v?, Paper-Bill spread (3g) and also includes the spread in the output equation. The remaining five equations are identical to those in the earlier model (2a-e). The parameter estimates reported in Table 4 provide weak evidence for bank/non-bank substitution as a source of paper-bill spread. The positive and marginally significant on the paper term shows that flows of non-financial paper do exert an influence on the spread.20 However, the very large, significant coefficient on reserves shows indicates that a great deal of the impact of monetary policy is transmitted to the spread via other routes. The impulse responses in the top panel of Figure 5 confirm the spread's strong reaction to non-borrowed reserves innovations. Positive shocks to thefinancinggap also drive up the spread, as predicted, while paper shocks have little or no impact. Lending shocks again pose a problem, however. If the lending innovations identified by the VAR correspond to changes in the availability of loans, the model suggests that positive shocks should be associated with a falling paper-bill spread. The opposite is true: lending shocks imply a rising spread. Again, this pattern is consistent 20. By contrast, the results in Table 10 of Friedman and Kuttner (1992) using the total volume of commercial paper outstanding are consistent with a stronger link between paper issuance and the spread. -19- Kuttner with bank lending responding passively to changes in the demand for funds inadequately captured by the financing gap. The bottom panel of Figure 5 suggests something other than bank/non-bank substitution is driving the paper-bill spread. Despite the inclusion of a variety of financial variables purporting to capture the impact of monetary policy on credit markets, the graph shows that the spread continues have strong implications for future output — comparable in magnitude to those of non-borrowed reserves. Even accounting for reserves, lending, and paper shocks, orthogonal spread innovations still result in falling real economic activity. Identifying lending shocks with loan spread innovations In light of the conclusion that lending flows (and the "mix") may in part represent endogenous response to firms' financing demands, the third structural VAR uses an alternative assumption to identify lending shocks, attributing (orthogonalized) innovations in the loan-paperspread to changes in banks9 willingness to lend. In the context of the simple model presented earlier, the loan spread should embody exactly the same information as the "mix." In practice, as KSW note, the loan rate is likely to be a poor measure of the true cost of bank finance, an observation that motivates their use of the quantity variables. Indeed, the sluggish response of the loan rate to changes in the paper rate corroborates this view. The weak response of output to the loan-paper spread makes this approach seem even less promising. With these reservations in mind, thefirststructural VAR can be adapted to incorporate the loanpaper spread. An equation for the loan spread is added to the system, and lending and paper flows are allowed to depend on this spread, as well as on reserves and the financing gap. The covariation between the flows that is a function of credit conditions is a result of their common dependence on the loan spread. This identification scheme will work if the financing gap is an imperfect proxy for the overall demand for funds so long as banks passively accommodate firms' funding requirements within the quarter at the going spread (that is, if their demand for loans is elastic). -20- Kuttner The modified system is: R s bi& + Vi Reserves'(4fl) F m bZ\R + bZ6x + V2 Financing gap (4b) r> « buR + b^JF + v3 Interest rate (4c) rL-rP = b<tR + b^F+b^rP Loan spread (4rf) I =fcMrt+ 6 ^ + b & r p + 65,4(^1 - rP) • v5 P « 6Mtf + *42F+fc^r/. -•- A^fo. - TP) + **sV5 + v6 * s ^73r/» + ^ f a , - rP) + 67,5^ • th,*? + ^7 Lending (4e) Paper (4/) Output (4g). Under the assumptions outlined above, the innovations to the loan spread equation are now associated with changes in credit conditions, while the v5 lending innovations represent shocks to firms' loan supply (that is, their demand for funds). The parameter estimates in Table 5 accord surprisingly well with the implications of the model. Although its sluggish response makes the loan spread is subject to large, transitory effects from the paper rate and reserves, the negative estimated 65,4 and the positive b$4 show that loan and paper volume respond as they should to the spread. Furthermore, reserves have no discernible independent impact on financial flows. A rising loan spread is contractionary, although again, the effect is statistically weak. The corresponding impulse response functions appear in Figure 6. The top panel again illustrates the consequences of sluggish loan rate adjustment, with reserves injections causing the loan spread to rise sharply in the current quarter. Over time, reserves innovations produce a falling spread. The center panel shows the familiar liquidity effect, and the positive impact of lending innovations on the commercial paper rate. In this model, however, with innovations to loan volume interpreted as shocks to firms' funding requirements, the result is perfectly natural. By contrast, innovations in the loan spread have quite mild effects on the paper rate. -21- Kuttner Conclusions This paper has examined the relationship between monetary policy, loan availability, and alternative indicators of credit market activity. One of is mainfindingsis that the substitution between bank and non-bank finance is indeed an identifiable effect of monetary policy as measured by innovations to non-borrowed reserves. This substitution is, however, not the only factor affecting financial flows. One of the major contributors to the aggregate composition offirms'short-term obligations is flows of commercial paper unrelated to lending shocks. Furthermore, the portion of bank lending not attributable to monetary policy is associated with increases in the commercial paper rate and the paper-bill spread, suggesting that the behavior of the KSW "mix" is in part due to changes in firms' demand for loanable funds. Despite its apparent slow adjustment to changes in market interest rates, the loan-paper spread is a plausible alternative indicator of credit conditions. The paper-bill spread responds appropriately to monetary shocks, rising in response to a reserves contraction. However, the strength of its response cannot entirely be accounted for by flows of non-financial paper, suggesting that its informativeness as a predictor of real economic activity may be due to other sources, such as changes in banks' issuance of negotiable CDs. This is consistent with the observation that non-financial commercial paper comprises a tiny share of the relevant market—only 25% of total commercial paper, and less than 9% of the sum of paper, CDs and Treasury bills.21 Understanding how Federal Reserve policy and credit conditions affect the paper-bill spread will require expanding the model to take into account the behavior of other relevant assets, such as CDs andfinancialpaper. 21. These figures are for 1991:4. The share of non-financial commercial paper is even smaller earlier in the sample. -22- Kuttner References Barro, Robert (1974), "Are Government Bonds Net Wealth?** Journal of Political Economy .82, pp. 1095-1118. Bernanke, Ben S. (1986), "Alternative Explanations of the Money-Income Correlation,** Carnegie Rochester Conference Series on Public Policy 25, pp. 49-100. Bernanke, Ben S. (1990), "On the Predictive Power of Interest Rates and Interest Rate Spreads,** New England Economic Review November-December, pp. 51-68. Bernanke, Ben S. and Cara S. Lown (1991), "The Credit Crunch,** Brookings Papers on Economic Activity 2, pp. 205-39. Blanchard, Olivier, and Mark Watson (1986), "Are Business Cycles All Alike?** in Robert A. Gordon, ed., The American Business Cycle: Continuity and Change. Chicago: The University of Chicago Press and the NBER. Bosworth, Barry and James S. Duesenberry (1973), "A Row of Funds Model and its Implications*' in Issues in Federal Debt Management, Federal Reserve Bank of Boston Conference Series 10, pp. 39-149. Brainard, William C. (1964), "Financial Intermediaries and a Theory of Monetary Control,** Yale Economic Essays 4, pp. 431-82. Foulkc, Roy A. (1931), The Commercial Paper Market New York The Bankers Publishing Company. Friedman, Benjamin M. (1991), "Comments on Bernanke and Lown,** Brookings Papers on Economic Activity 2, pp. 240-44. Friedman, Benjamin M. and Kenneth N. Kuttner (1992), "Why Does the Paper-Bill Spread Predict Real Economic Activity?** forthcoming in James H. Stock and Mark W. Watson eds., New Research in Business Cycles, Indicators and Forecasting, Chicago: University of Chicago Press and the NBER. Hurley, Evelyn (1977), "The Commercial Paper Market,** Federal Reserve Bulletin 63, June, pp. 525-536. Hurley, Evelyn (1982), "The Commercial Paper Market since the Mid-Seventies** Federal Reserve Bulletin 68, June, pp. 327-333. Gertler, Mark and Simon Gilchrist (1992), "The Role of Credit Market Imperfections in the Monetary Transmission Mechanism: Arguments and Evidence,*' Manuscript. Kashyap, Anil, Jeremy C. Stein, and David Wilcox (1992), "Monetary Policy and Credit Conditions: Evidence from the Composition of External Finance,*' NBER Working Paper #4015, Cambridge: National Bureau of Economic Research. King, Stephen R. (1986), "Monetary Transmission: Through Bank Loans or Bank Liabilities?" Journal of Money, Credit and Banking 18, August, pp. 290-303. Lawler, Thomas A. (1978), "Seasonal Movements in Short-term Yield Spreads,*' Federal Reserve Bank of Richmond Economic Review, July/August,. -23- Kuttner Oliner, Stephen D. and Glenn D. Rudebusch (1992), "The Transmission of Monetary Policy to Small and Large Firms," Manuscript. Owens, Raymond E. and Stacey L. Schreft (1992), "Identifying Credit Crunches," Federal Reserve Bank of Richmond Working Paper #92-1. Selden, Richard T. (1963), Trends and Cycles in the commercial Paper Market," National Bureau of Economic Research Occasional Paper #85. Stigum, Marcia (1990), The Money Market Homewood: Dow Jones Irwin. Strongin, Steven H. (1991), "The Identification of Monetary Policy Disturbances: Explaining the Liquidity Puzzle," Federal Reserve Bank of Chicago Working Paper #91-24. -24- Kuttner 1. F-Statistics for Alternative Measures of Credit Conditions in Quarterly Real Output Equations 60:2-91:4 70:3-91:4 75:1-91:4 (1) "Mix" alone 3.36" 2.09* 286" (2) Loan spread alone an 0.30 0.46 (3) Paper-bill spread alone 3.81"* 2.71" 1.81 (4) "Mix" + loan spread "mix" terms loan spread terms 3.37" 0.23 1.81 0.14 3.07" 0.79 (5) "Mix" + paper-ttll spread "mix" terms paper-bill spread terms 4.17"* 4.62'" 2.46 3.07" 4.39"* 3.30" Specification * ** *** Notes: Significant at the 10% level Significant at the 5% level Significant at the 1% level The regressions are based on qtiarterly data for the sample indicated. In addition to the variables indicated, each regression includes four lags of real GDP growth, real non-borrowed reserves growth, the differenced commercial paper rate, plus constant and trend terms. -25- Kuttner 2. Decomposing Changes in the Composition of External Finance (a) Regression with separate commercial paper and bank lending terms Exclusion F-stat (p-value) Commercial paper (AA/>) Bank lending (AAL) Sum of coefficients (p-value) 4.00 (0.005) -0.51 (004) 1.39 (Q24) -0.90 (0.22) (b) Regression with the differenced "mix" and commercial paper Exclusion F-stat (p-value) "Mix" (AA) Commercial paper (iJip) 1.45 (0.22) 264 (0.04) Sum of coefficients (p-value) -0.91 (0.23) -1.38 (0.04) (c) Regression with the "mix" in levels, commercial paper, and linear trend Exclusion F-stat (p-value) "Mix" (h) Commercial paper (A/i/>) Notes: 1.48 (0-21) 227 (0.07) Sum of coefficients (p-value) 0.11 (0.16) -1.77 (0.01) The regressions are based on quarterly data for 1960:2 through 1991:4. The specifications include four lags of each included variable and a constant term. -26- Kuttner 3. Structural VAR Estimates, Credit Conditions Identified via Lending Flows (equations 2a-2f) 2a. 2b. 2c. 2d. 2e. 2f. Notes: R= -0.625 x+ v, (1.94) F= 0.159 r+ 1.974 (1.05) (4.11) />=-0.208 R+ 0.037 (6.96) (210) L = -0.396 * - 0.022 (1.51) (ttl6) P = ttl35 rt + 0.027 (1.29) (0.51) x = -0.023 r+ 0.019 (0.23) (1.50) x+v* F+vj F+ 2125 r, + v4 (3.23) F + 0.444 r>- 0.094 v4 + v5 (1.68) (271) 1 + 0.004 P + v6 (ttl4) Estimates are based on quarterly data for 1960:2 through 1991:4. Regressions include three lags of each variable, constant and trend terms. Numbers in parentheses are /-statistics. -27- Kuttner 4. Structural VAR Estimates of the Effects of Lending Shocks on the Paper-Bill Spread (equations 3a-3g) 3a. 3b. 3c. 3d. 3e. 3f. 3g. Notes: /?=-0.558 x+V! (1.51) F = 0.048 r+ 1.882 (033) (3.70) r, * -0.216 R + 0.027 (7.63) (1.57) L = -0.306 R- 0.022 (1.18) (017) P = 0.110/?+ 0.038 (1.05) (071) x= 0.123 r+ 0.016 (1.20) (1.35) r/,-r,= x+vj F+ v, F+ 2051 r, + v4 (3.05) F+ 0.470 r , - 0.091 v4+ v5 (1.73) (261) 1+ 0.023 P - 0.897 (rP-rB) + v6 (076) (3.20) 0.016/?+ 0.181 r+ 0.000 F+ 0.002 1 + (1.40) (6.20) (007) (046) 0.016 P + v, (1.71) Estimates are based on quarteriy data for 1960:2 through 1991:4. Regressions include three lags of each variable, constant and trend terms. Numbers in parentheses are /-statistics. -28- Kuttner 5. Structural VAR Estimates, Credit Conditions Identified via the Loan Spread (equations 4a-2g) 4a. 4b. 4c. 4d. 4e. 4f. 4gNotes: R= -0.347 x+ v, (0.85) F= 0.122 r+ 2132 x + (a89) (4.46) i> = -0.202 rt + 0.016 F+ (aiO) (0.96) r t - r P = 0.070/?+ 0.014 F (4.99) (1.77) 1 = -0.039 rt + a021 F + (0.14) (ai5) P= 0.030 J? + a003 F + V2 »* a261 (6.56) a783 (0.96) 0.722 (218) (a27) (ao6) x= -0.047 r - 0.362 ( n - rP)-f 0.016 (0.39) (1.54) (1.28) /> + v4 r, - 4.727 (rL - »>) + vs (299) /> + 1.176 (rL-rP)- 0.085 v5 + v6 (1.83) (242) 1+ 0.015 P+ v7 (0.46) Estimates are based on quarterly data for 1960:2 through 1991:4. Regressions include three lags of each variable, constant and trend terms. Numbers in parentheses are /-statistics. -29- Kuttner Figure 1 reserves market Rs Cf«0 l-fp *"f "*"* loan market paper market \yS -30- (I w (HHi) Figure 2 Impulse Response Functions of Credit Conditions Indicators (a) mix -> output (b) reserves -> mix 0.0048 (c) mix -> interest rate 0.0032 0.00361 0.00161 0.0024 0.00121 0.0000 T^-r 0.0000 -.0012 3 6 1 9 1—r- 0 (d) loan spread -> output I I 3 I I I (e) reserves -> loan spread 0.0020 -.0016 6 0.0036 (0 loan spread -> interest rate i CO -.00201 i—i • i i (g) paper-bill spread -> output 0.0016 (h) reserves -> paper-bill spread 0.0007 n 0.0000 -.0016 -.0032 -.0046 .0040 • L -i—i—i—i—i—i—r- r- « i—i 3 i r - t • 9 r • 0) paper-bill spread -> interest rate 0.00501 Figure 3: Financing Gap and Financial Flows bank lending and paper issuance, four-quarter moving average 1 v> • 2 w c •o2 18 -36 .• I ' I ' I ' I ' I ' I ' I ' I • I ' I ' I • I • I • I • I • I • I • I ' I ' I ' I ' I ' I ' I ' I ' I ' I ' I 60 63 66 69 72 75 78 81 84 87 90 Kuttner Figure 4: credit conditions = lending shocks response of the Mix 0.16 0.08 H o.oo -.08 -j reserves lending •.16 -| paper -fin gap— .24 T 1 1 P I I 1 8 9 1 10 11 response of the interest rate 0.035 0.000 .035 H .070 i 1 7 8 11 reserves Ien3ing paper — 0.050 - ^^"^ — — .-.-—.. .-v^—-• 0.025 - 0.000 —' 1 i 10 response of real output n f\7K —i Kj.UfD -.025 - 1 9 ****** . , , **"* , 2 . 3 , 4 , 5 6 - 33 - , 7 , 8 , 9 , 1 10 11 Kixttner Figure 5: credit conditions = lending shocks response of the paper-bill spread 0.008 0.000 -.008 H -.016 l 9 r 10 11 response of real output 0.10 0.05 H 0.00 J~^ \ •.05 1 0 1 1 1 1 2 n 3 4 1 5 1 6 - 3H - 1 7 1 8 1 9 1 1 10 11 Kuttner Figure 6: credit conditions = loan spread shocks response of the loan spread reserves lending paper response of the interest rate s s / " •••••i^>% response of real output - 35 - COMMENTS ON CREDIT CONDITIONS AND EXTERNAL FINANCE: INTERPRETING THE BEHAVIOR OF FINANCIAL FLOWS AND INTEREST RATE SPREADS David Wilcox Two opposing views have animated much recent research on the transmission channels of monetary policy. One view (stated in its extreme form) is that the impulses of monetary policy are transmitted to the real economy exclusively via the market for reserves. By manipulating the quantity of available reserves, the Federal Reserve is able to change the relative supply of money and bonds. Given this change in relative supply, the interest rate must change in order to clear the markets for money and bonds. In turn, the change in the interest rate alters the user cost of capital, and so influences the investment decisions of businesses and the spending decisions of households. An essential assumption implicit in this so-called "money" view of the transmission mechanism is that bank loans, market-intermediated privately-issued debt such as commercial paper and corporate bonds, and privately-held government debt can be treated as perfect substitutes. Indeed, this assumption is embedded in the conventional IS-LM model, where the aggregate non-money financial asset is simply labelled "bonds" for convenience. According to the money view, the reduction in bank loans that accompanies a reduction in reserves is of no particular significance in itself because firms can satisfy any unmet demand for external finance by issuing market-intermediated debt which is indistinguishable from bank debt. For this reason, the money view often is summarized by the proposition that bank loans are not "special." The opposing view of the transmission mechanism assigns a central role to bank loans. According to this view, bank loans, market-intermediated privately-issued debt, and government debt are not perfect substitutes. The reduction in the volume of bank loans that accompanies a move toward a more restrictive monetary policy is 1. David Wilcox is on the staff of the Board of Governors of the Federal Reserve System. Wilcox contractionary in itself, even controlling for any associated change in interest rates. In effect, bank loans behave as if they were a factor of production. A reduction in their availability increases their relative price (the spread between the loan rate and the openmarket rate increases). In response, firms seek cheaper alternatives for their external finance. However, given the imperfect substitutability of other forms of debt for bank loans, the reduction in loan availability implies a contraction in real activity. The important distinction between the money view and the loans view is that the latter implies that the impulses of monetary policy are transmitted not only through the overall level of interest rates, but also through the relative prices and relative quantities of bank loans and other forms of external finance. If the loans view is right, fluctuations in the quantities and prices of bank loans, commercial paper, other private debt, and government debt will be worth keeping track of separately because they will be informative for either the current or future state of the economy, or both. Moreover, the loans view suggests, as Kuttner (this volume) and Friedman (1991) emphasize, that there is no reason for being uniquely interested in changes in the stance of monetary policy; other factors (including but not restricted to the stringency of regulatory oversight) will also be worthy of study to the extent that they bear on loan availability. THE IDENTIFICATION PROBLEM One approach to investigating the empirical significance of the loans channel has been to regress some measure of real activity (such as industrial production or GNP) on current and lagged measures of bank loans. A positive correlation between bank loans and real activity has sometimes been interpreted as contradicting the money view and supporting the existence of a separate loans channel. The flaw in this argument is not hard to spot: A positive correlation between bank loans and real activity could simply reflect an endogenous response of the demand for bank loans to changes in real activity rather than an exogenous cause of changes in real activity. Even a finding of a positive correlation between bank loans and subsequent changes in activity (as opposed to contemporaneous ones) would not be convincing evidence of a separate loans channel; such a phenomenon could reflect, for example, a need to secure financing some months or -2- Wilcox even quarters before the bulk of the associated activity is to take place. An important challange taken up in the more recent literature has been to solve this identification problem in a convincing manner.2 SUMMARY OF KUTTNER'S PAPER Ken Kuttner's paper makes two important contributions to the literature on the monetary policy transmission mechanism: one theoretical, the other empirical. On the theoretical front, he presents a very nice compact model of the flow of funds in a simple economy. He distinguishes five financial instruments in his model (in contrast to the usual two): deposits ("money"), bank loans, commercial paper, reserves, and government debt. He posits the existence of a representive firm, a representative bank, and a representative household, and endows each of them with standard portfolio behavior (households* demand for money is declining in the opportunity cost of holding money, and so forth). Then he derives the implications of changes in the stance of monetary policy, changes in banks' willingness to lend, and changes in firms' demand for external finance for three quantities: the mix of external finance, the spread between the loan rate and the commercial paper rate, and the spread between the paper' rate and the Treasury bill rate. The beauty of Kuttner's model is that it delivers sensible results very directly. For example, a reduction in banks' willingness 4 to lend causes the loan-paper spread to rise. In response, firms 2. The approach proposed in Kashyap, Stein, and Wilcox (1992) is to focus on changes in the composition of external finance rather than fluctuations in any one component alone. Intuitively, one would not expect changes in the volume of bank loans relative to the volume of other debt to be informative for current or future changes in real activity if bank debt is a perfect substitute for non-bank debt. 3. Implicitly, other corporate liabilities such as medium- and long-term bonds are treated as perfect substitutes for commercial paper. 4. Kuttner interprets "negative shifts in X as 'credit crunch' episodes." He notes, however, that a negative shift in X could reflect a "perceived deterioration in borrowers' creditworthiness." In my opinion, it would be more useful to reserve the term "credit (Footnote continues on next page) -3- Wilcox shift the mix of external finance away from bank loans and toward market-mediated debt. The increased issuance of commercial paper drives up the spread between commercial paper rates and bill rates. With respect to these three key variables, the effects of a reduction in banks' willingness to lend are identical to the effects of a move by the Federal Reserve toward a more restictive monetary policy, suggesting that any one of the three might be useful as an index of loan availability. In fact, it turns out that these three variables also respond in qualitatively the same manner to the other two exogenous factors in Kuttner's model (monetary policy and the demand for external finance). That is, no matter what the conceptual experiment being run in Kuttner's model, the loan-paper spread will always move in the same direction as the paper-bill spread, and the two spreads will always move in the opposite direction of the mix. In light of these predictions from his theoretical model, Kuttner's finding that the loan-paper spread significantly underperforms the mix and the paper-bill spread as indicators for future real GNP is interesting and a bit puzzling. Kashyap, Stein, (Footnote continued from previous page) crunch" for periods in which some potential borrowers are turned away even though, with Identical characteristics in every respect (including "credit worthiness"), they would.have been granted credit in "normal" times. 5. In Kuttner's model, the commercial paper rate is taken as the benchmark rate over which the Federal Reserve has direct control in the reserves market. As a result, a reduction in banks' willingness to lend has no effect on the JeveJ of the commercial paper rate. As was noted in the text, however, it does increase the loans-paper spread. As a result, the volume of commercial paper outstanding rises and the paper-bill spread increases. Given the fixity of the paper rate in the face of this experiment, it must be that the bill rate has declined. If the bill rate (rather than the paper rate) were assumed to clear the market for reserves, all the essential results still would hold (the mix would shift away from loans, the loans-paper spread and the paper-bills spread both would rise), but the bill rate would be fixed and the paper rate would rise. 6. Kuttner notes that the effects of a shift in monetary policy are not identical in every respect to the effects of a shift in banks' willingness to lend: The former affects the level of the interest rate in the market for reserves, whereas the latter does not. -4- Wilcox and Wilcox (1992) argued that the mix might be preferable to the loanpaper spread as an indicator of loan availability (because the stated loan rate would not adequately reflect changes in non-price terms of loan contracts such as collateral requirements), but then proceeded to find in their sample that the predictive power of the two variables was roughly comparable. It would be worth attempting to reconcile Kuttner's results with those of KSW, and (assuming Kuttner's results hold up) attempting to verify the KSW hypothesis about why the loanpaper spread might be an inferior performer. On the empirical side, Kuttner's paper introduces a new approach to solving the identification problem. He posits several simple "structual vector autoregression" models of the markets for reserves, bank loans, and commercial paper. Kuttner is bold enough to supply sufficient prior restrictions on the specification of the various equations, and finds that, for the most part the estimates that follow are well in line with the predictions that were outlined in his theoretical section. The major exception--and one that deserves further investigation--is that increases in banks' willingness to lend (counterintuitively) appear to cause Increases in interest rates. AN ASYMMETRIC-INFORMATION-BASED ACCOUNT OF SUBSTITUTION BETWEEN LOANS AND PAPER In line with most of its recent predecessors, Kuttner's paper adopts an aggregate perspective: The model is inhabited by representative banks, households, and non-bank firms, and the empirical work is conducted using aggregate data. As in the earlier papers, this perspective--through no fault of the author--sets up certain tensions of both an expositional sort and a substantive sort. On the expositional side, the most natural way to tell the story of the loans channel involves an appeal to heterogeneity among firms: Some are capable of issuing commercial paper while others are not. Obviously, a story such as this is difficult to link up directly to a model with a single representative non-bank firm. On the substantive side, the representative-agent approach to modelling the problem fuels the intuition that some firms should be observed to be on the margin between bank loans and commercial paper. The purpose of the rest of these comments is to sketch verbally a model that allows for -5- Wilcox heterogeneity among firms, and then to point out two important implications of such an approach. The loans view is predicated on the assertion that non-bank debt is not perfectly substitutable for bank debt. That imperfect substitutability can be motivated as reflecting market imperfections that arise when borrowers have more information about their economic prospects than do prospective lenders. Banks specialize in "information-intensive" lending--that is, in lending to customers (such as small businesses) for whom the asymmetric-information problem is more acute, and hence more difficult for arms-length capital markets to solve. A contractionary shift in the stance of monetary policy will cause banks to reduce the size of their loan portfolios. Banks will tend to cut off their most risky customers and continue to service their most creditworthy ones. Firms that are denied credit by banks may be unable to borrow from any other lender. Certainly, they will not be able to issue debt in arms-length capital markets: nor will they be able to attract financing from other non-bank sources simply by announcing their willingness to pay a higher rate of interest on the debt, because potential lenders will recognize that only the riskiest firms would be willing to offer a higher rate of return. In the end, these firms are likely to be particularly vulnerable to the monetary contraction. After a monetary contraction, a larger fraction of total external finance will be provided via arms-length capital markets and a smaller fraction through bank loans. This change in composition may reflect either (or both) of two factors: First, it may reflect increased issuance of trade credit by large, financially secure firms to their smaller, less creditworthy suppliers. An increase in commercial paper borrowing would be used, in effect, to finance the rise in trade credit. Large firms may be willing to act, in effect. as financial intermediaries because they will have accumulated substantial inside information about the financial stability of their suppliers in the course of having interacted with them before the 7. A lower level of reserves will only support a lower level of deposits. The lower level of deposits (which comprise banks* liabilities) implies that assets will have to decline as well. Given that banks view loans and securities as imperfect substitutes, some of that decline in assets will be absorbed in loans. -6- Wilcox credit crunch. Alternatively, the increase in the share of commercial paper in total external finance may reflect that large firms tend to expand when their smaller rivals are weakened by financial stringency; the large firms take the opportunity to seize some portion of the product market, financing the larger scale of their operations with the increase in commercial paper issuance. These two mechanisms show that bank loans and commercial paper can be substitutes at the aggregate level even though not so for any individual firm. Failure to observe firms operating on the margin between bank loans and other market-mediated debt does not constitute evidence against the heterogeneous-firms version of the loans channel. IMPLICATIONS OF THE ASYMMETRIC-INFORMATION-BASED APPROACH The informal discussion in the previous section points to two important implications for future research. First, the very motivation of banks specializing in information-intensive lending suggests that further progress probably would flow from the analysis of models that allow for heterogeneous non-bank firms. In particular, it seems likely that most such models will imply that, when the Federal Reserve adopts a more restrictive monetary policy, banks will shrink their loan portfolios by refusing credit to their riskiest (least financially stable) customers. Commercial paper issuance will rise because firms already issuing paper will issue more--either to finance their own expanded operations, or to finance the passthrough of trade credit to their suppliers. By contrast, a representative-firm model suggests that all firms should be on the margin between bank debt and commercial paper, and that when the Federal Reserve tightens we should observe a rebalancing of liabilities taking place at the individual firm level. The implausibility of this account is obvious, given that fewer than 1300 firms in the United States have commercial paper programs rated by Moody's. 8. Firms that are growing in size will, at some point, find it possible to issue commercial paper for the first time. If the profitability of commercial paper issuance is an inverse function of bank-loan availability, establishment of commercial paper programs will tend to be bunched into periods immediately following tightenings of monetary policy. Historically, of course, the commercial paper market was not always as well-developed as it is now; as the market deepened and became more efficient, even firms that had been large and creditworthy for a long time established new programs. -7- Wilcox The second implication of the disaggregated approach is that future empirical work should focus on micro-level datasets. Such investigations will be essential for: (1) establishing the identity of bank customers who are denied credit in the wake of a tightening by the Federal Reserve; and (2) establishing the source of the accompanying increase in commercial paper issuance. DISCOUNT WINDOW BORROWING AND LIQUIDITY W. J. Coleman. C. Gilles, and P. Labadie1 Three features seem centra] to understanding the relationship between U.S. monetary policy and the comovements of open market operations, monetary aggregates, and interest rates. First, shocks to bank reserves affect interest rates in ways that axe not tightly linked to the Fisherian fundamentals (expected inflation, marginal rate of substitution, and marginal productivity of capital). Second, banks often respond to reserve shocks by adjusting their borrowing at the Federal Reserve's discount window. Third, the Federal Reserve often conducts open market operations to smooth interest rates that would otherwise react to private-sector demand shocks. In this paper, we study a stochastic general equilibrium model that incorporates these features in an effort to understand important empirical regularities involving monetary aggregates and interest rates. The empirical regularities we have in mind are those documented in the vast literature aimed at uncovering a negative correlation between short-term interest rates and exogenous policy shocks to nominal monetary aggregates, a relationship often referred to as the liquidity effect. Cagan (1972) and Cagan and Gandolfi (1969), among many others, have reported finding negative correlations between Ml itself and various short-term interest rates. Subsequent studies have reported similar correlations with innovations in Ml backed out using a Choleski decomposition of the residuals in a vector autoregression (for a variety of orderings). More recently, however, Leeper and Gordon (forthcoming) have made a strong case that these innovations probably do not represent exogenous monetary policy shocks, as the money supply may be endogenously 1 Board of Governors, Federal Reserve System. We gratefully acknowledge helpful discussions with Jim Clouse and Josh Feinman. Coleman, Gilles, and Labadie determined in ways that are not captured by any Choleski decomposition. To support their claim, they noted that the statistical properties of these innovations are sensitive to the other endogenous variables included in the VAR, the sample period, and the measure of money selected for analysis. Some researchers, for example Bernanke and Blinder (1990) and Sims (forthcoming), have responded to such criticism by assuming that innovations to interest rates reflect policy shocks, to which the supply of money responds endogenously. For our purpose, however, this strategy does not resolve the central question: if there exists a liquidity effect, then why are these interest rate innovations not robustly negatively correlated with monetary aggregates (an observation also made by Leeper and Gordon)? Christiano and Eichenbaum (1991) and Strongin (1991) have tried to obtain robust negative correlations by using nonborrowed reserves as the measure of money. This approach contrasts with that of Leeper and Gordon, who experimented with monetary aggregates that are at least as broad as the monetary base. Christiano and Eichenbaum's rationale for using nonborrowed reserves is based on the widely held perception that the Fed controls this aggregate. For this reason they associated policy shocks with innovations to nonborrowed reserves, which they then showed to be negatively correlated with the federal funds rate. In fact, using nonborrowed reserves as the measure of money, they found evidence of a negative correlation regardless of whether money innovations or interest rates innovations were identified as the policy shocks, and they showed that these correlations are remarkably robust to the sample time period. To explain why the innovations to broader monetary aggregates do not exhibit a similar correlation, they noted that these aggregates are largely endogenously determined by the banking system. For example, they argued that total reserves may be inelastic in the short run, and therefore not correlated with interest rates at all. In this example, policy shocks to nonborrowed reserves do not affect total reserves immediately. Strongin refined this argument; he argued that innovations to nonborrowed reserves that are not reflected in shocks to total reserves should be identified as the policy shocks. He asserted, in essence, that shocks to required reserves lead to an adjustment in both - 2 - Coleman, Gilles. and Labadie nonborrowed and total reserves, whereas open market operations lead to an adjustment in only nonborrowed reserves. We develop a model that is rich enough to address the empirical issues presented above. To do this, we introduce a banking system, reserve requirements, and a discount window into a model of liquidity based on the works of Grossman and Weiss (1983), Rotemberg (1984), Lucas (1990) and Fuerst (1992). In these models, and here, the term liquidity effect refers not merely to a negative correlation between monetary policy shocks and interest rates but more generally to any non-Fisherian effect on interest rates. Interest rates deviate from their Fisherian fundamentals because of shocks to the demand for bank deposits from businesses to finance new investment projects and perhaps also because of monetary policy shocks. In our model, the interest rate is also the cost (both pecuniary and nonpecuniary) of borrowing reserves from the discount window, so that over time there is a well defined relationship between borrowed reserves and the interest rate. Monetary policy designed to smooth interest rates then leads to rather complicated mutual dependencies among open market operations, both broad and narrow monetary aggregates, and interest rates; in particular, monetary policy can lead to positive correlations between broad monetary aggregates and interest rates in spite of the liquidity effect. When policy shocks are correctly identified, however, the model suggests that broad monetary aggregates are negatively correlated with interest rates, showing evidence of the liquidity effect. Furthermore, the model always generates a negative correlation between nonborrowed reserves and short-term interest rates, regardless of what the policy shocks are and how they are identified. Such a result is due to the way the discount window is operated. In light of this model, one interpretation of Christiano-Eichenbaum and Strongin's results is that they identified the discount window policy. Since this policy implies a negative correlation between nonborrowed reserves and interest rates whether or not the model incorporates a liquidity effect, their results shed little light on the presence of such an effect. - 3 - Coleman. Gilles, and Labadie THE MODEL DESCRIPTION. To get an overview of the model, consider the following accounting of the assets and liabilities of banks. Their liabilities comprise demand deposits of firms and households as well as savings deposits of households. Their assets are made up of reserves and a portfolio of government securities and loans to firms. Banks are required to hold as reserves a fraction of their demand deposits;'to avoid a deficiency, they can borrow reserves at the discount window. Borrowed reserves incur pecuniary and nonpecuniary costs. To start building a model around this balance sheet, think of households as dividing their deposits between demand deposits, which can be used to buy goods, and savings deposits, which cannot. Assume that this division is made before the value of the open maxket operation is known, resulting in a liquidity effect as described by Lucas (1990) and Fuerst (1992). Also assume, as Fuerst (1992) did, that firms must finance their purchases of investment goods with demand deposits, so that these deposits represent intermediated capital, as in Freeman and Huffman (1991). To view the model in more detail, consider a representative household that ranks stochastic consumption and leisure streams {ct,lt} according to the utility function Lt=0 \t=0 / where /3{ is the date-i realization of the random discount factor; /3*+i is unknown at the beginning of period t but is revealed later during that period. The household begins period t with money balances Mt in an interest-bearing savings account. It immediately transfers amount Zt to a checking account which bears no interest but can be used during the period to finance consumption ct; only one transfer during the period is allowed. The household must choose Zt before it knows the realization of any of the current shocks, or prices for that matter. Its purchases of goods are subject to the finance constraint Ptct < Zt. -4- Coleman, Gilles. and Labadie At the end of the period, Mt — Zt remains in the household's savings account and Zt — PfCt in its checking account. The household derives income from several sources. It provides labor to the firm, working a fraction of time equal to 1 — it at wage rate Wt] it earns interest at rate r\ on the amount Mt — Zt in its savings account; it collects a transfer Xt from the government; finally, as owner of both the firm and the bank, it collects Il( and II*, the period's proceeds from the sale of output net of all costs and bank profit respectively. The household receives its income, including income from labor performed during the period, at the beginning of the next period, when it is directly deposited into the savings account. With unspent checking account balances being transferred back into the savings account, the law of motion for Mt is Mt+i = Zt - Ptct + (Mt - Zt)(l + r{) + Wi(l - It) + Xt+ Ii{ + II*. The firm, the second agent in the economy, combines, capital and labor inputs to produce a homogeneous product sold to buyers of consumption and capital goods. The production function is Vt = F(kt,nt,0t), where yt is the output, kt and nt are the inputs of capital and labor, and 9% is a technological shock. The firm owns the capital stock kt and hires labor at rate Wt] it makes wage payments at the beginning of the next period using the receipts from the sale of output. The firm must also acquire investment goods it; it purchases these goods from other firms in the goods market but cannot use its sales receipts for this purpose. Instead, it finances investment by borrowing Bt from a bank, which charges interest at rate r*. The bank provides this financing by crediting the amount to the firm's checking account, increasing the balance from its starting level of zero. The firm's finance constraint is Bt > Ptit. At the end of the period, the firm has spent Ptit on investment goods and deposits its current sales receipts, PtVt, leaving Bt -f Pt(yt — U) in its checking -5- Coleman, Gilles, and Labadie account. At the beginning of the next period, the firm repays its bank loan and transfers wages into the worker's savings account. The amount left in the firm's account, Il£, is paid to the firm's owner as dividend: n / = Ptyt - Wtm - Ptit - rtBt. The stock of capital depreciates at the constant rate 6. so that its law of motion obeys fct+i = (1 -6)kt + it. The firm makes all its decisions (namely, J3t, it, and rtt) with full knowledge of the current shocks and prices. The bank, the third agent in the economy, starts period t with liabilities equal to Mt (the household's savings account) and holds an equal amount of vault cash as an offsetting asset (we write "vault cash" for definiteness; Mt could also be thought of as an account at the central bank). The household immediately transfers Zt from its savings to its checking account, without affecting the bank's total liabilities or assets. The bank pays interest r\ on Mt — Zt, the amount left in the savings account, but pays no interest on checking deposits. By lending Bt to the firm, an amount that is credited to the firm's checking account, the bank increases both its liabilities and its assets from Mt to Mt -r Bt. To buy government bonds and to honor checks written to finance purchases of consumption and investment goods, the bank depletes its holding of vault cash, Mt] but it replenishes this cash position by the amount of the checks that firms receive for selling their output, checks that they deposit in their account. The amount of vault cash that the bank holds at the end of the period counts as reserves. Note that for an individual competitive bank, the loan of Bt to a firm drains reserves (when the firm spends the proceeds) just as much as if the bank had spent an equal amount to purchase government securities; therefore, at the same rate of interest, the bank is indifferent between the two types of lending. For the banking system as a whole, however, loans to firms involve no net loss of reserves, but merely a transfer from the borrower's bank to the bank of the producer of investment goods. -6- Coleman. Gilles, and Labadie Reserves. VJ, pay no interest and are subject to a reserve requirement, a fixed fraction p of the amount of checking deposits on the books of the bank at the end of the period: (1) Vt > p x [(Zt - Pta) + (Bt - Ptit -t- Pm)]. If the bank cannot satisfy the reserve requirement with the amount of vault cash it has at the end of the period (after checks have cleared), it can borrow the shortfall from the government at the discount window. Therefore, the following accounting identity must hold (2) Mt r D t = qtGt + Pt(it T*- yt) -f Vu where G% is the number of one-period pure discount government bonds the bank acquires, at a unit cost of qt = 1/(1 + rt), and D% is the amount it borrows at the discount window. Government bonds, private loans, and discount window borrowing carry the same rate of interest rt. The bank's objective is to maximize its period profit, which is given by (3) n j = Tt(Bt + qtGt - Dt) - r\(Mt - Zt\ The government, the fourth agent in the economy, sells one-period bonds in the securities market and redeems them at the beginning of the following period, operates the discount window, and makes transfers to the household's bank account. During period i, the government announces the open market operation Gt and the amount of transfers Xt after the household chooses Zt but before any other decision by any agent has to be made. All money flowing between the government and the private sector, as well as within the banking industry, takes the form of fiat money. The bank starts period t with an amount of fiat money (which it calls vault cash) equal to Mt. Nonborrowed reserves Vt — Dt is the amount left in vault cash after the purchase of government bonds and check clearing but before borrowing at the discount window; in equilibrium, Vt — Dt = Mt — qtGt as can be seen from eq. (2). Let Ht denote the outstanding supply of fiat money at the beginning of period t (Mt is best thought of as the demand for fiat money, so that in - 7 - Coleman. Gilles, and Labadie equilibrium Ht = Mt). The law of motion for Ht, which can also be thought of as the government budget constraint, is as follows: i?t+1 = Ht T Tt{qtGt — Dt) — Xt. Think of government policy as a rule that generates the values of Gt and Xt and that also sets the rate of interest at the discount window. Assume that the government lends reserves at the discount window according to an upward-sloping function if> : [0, oo) —• [0, oo) that relates the rate of interest it charges to the fraction of total reserves that it lends. Banks cannot lend at the discount window, so that when the equilibrium rate of interest is lower than the minimum rate at which the government is willing to lend, V>(0), there is no discount window activity: rt = i)(Dt/Vt) r t < V^O) whenever Dt > 0; whenever Dt = 0. The argument of if) ought to be the amount supplied at the window, which in equilibrium turns out to be equal to Dt, the amount demanded. Incorporating this equilibrium relationship directly simplifies the notation, but keep in mind that banks take as given all interest rates, including the rate they face at the discount window (which is equal to the rate on government securities). When the Federal Reserve lends at the discount window, the borrowing bank pays the discount rate plus a nonpecuniary cost; at the margin, this sum must equal the cost of borrowing from other banks, which is the federal funds rate. The marginal nonpecuniary cost is thus captured by the difference between the federal funds rate and the discount rate, called the spread. Historically, the policy of the Federal Reserve seems to have been to supply funds at the discount window at an increasing nonpecuniary cost (spread), which is precisely what the function tp assumes. This type of discount-window policy has been documented in the empirical literature, and is commonly modeled in the theoretical literature. 2 Chart 1, which graphs the monthly time series for 2 See for example Polakoff(1960), Goldfeld and Kane (1966), and more recently Goodfriend (1983), Dutkowsky (1984), and Waller (1990). In particular, Fig. 1, p. 346 in Goodfriend depicts an assumed ip function that is strikingly similar to the function that would best fit the scatter plot of our Chart 2. -8- Coleman, Gilles. and Labadie the federal funds rate and the nonborrowed reserve ratio (the mirror image of the borrowed reserve ratio), reveals the basis for the findings of the empirical studies. On closer inspection, a picture of the function ib emerges in a scatter plot of the borrowed reserve ratio against the spread, shown in Chart 2. Since this picture suggests that the Federal Reserve is ready to lend its first dollar at a zero spread, the value of t^(0) corresponds to the discount rate. With this interpretation of ^(0), the model simply assumes a constant discount rate. A word about terminology is in order. Vt is total reserves in the banking system; Dt is borrowed reserves; the difference Vt — Dt is nonborrowed reserves; and required reserves is p x [Zt + Bt + Pt(yt — it — ct)]. Besides total reserves, it is possible to identify the analogues of several monetary aggregates. M% (or Ht) corresponds to the monetary base, MO; the analogue of Ml is the sum of all reservable accounts, Zt + B%\ the total libilities of the banking sector at the end of the period, Mt + B^ correspond to M2 (strictly speaking, Ml and M2 both should include Pt{yt — ct — U) as well, but this is equal to zero in equilibrium); finally, the difference between M2 and MO, which is Bt, is inside money. It is now useful to summarize the timing of information and decisions. During period i, the realizations of four random variables shock the economy—the technological shock 0t> the preference shock /3t+i, the open market operation Gt, and the government transfer Xt. At the beginning of the period, the household must decide how much to put into its checking account, not knowing the current realization of 0t, /3t+i, Gt, or Xt, and therefore not knowing what interest rates, prices, output, or consumption will be. After it makes this decision, all four shocks are revealed and prices are set. On the basis of these shocks and these prices, the household decides how much to consume and how much to work; the firm decides how much to borrow, how much to invest, and how much labor to hire; and the bank decides how much to lend to the firm and to the government. Then trading takes place and checks clear. The bank monitors its reserve position and borrows at the discount window to cover any reserve deficiency (the bank can be thought of as borrowing at the same time it invests in government bonds or lends to firms, because it - 9 - Coleman, Gilles. and Labadie has the same information when it engages in any of these activities). At the start of next period, the firm pays its wage bill, repays its bank loan, and pays out its earnings to its shareholder; the government makes transfers to the household's savings account and redeems the bonds that the bank holds; the bank pays interest on its savings account, settles its discount window debt, and pays out its earnings. These activities determine the new initial balance in the household's savings account. Then a new cycle starts. The activities of the four agents that have been described above must, of course, satisfy the following standard market-clearing conditions. yt = a + it goods market; nt = 1 — it labor market; Ht = Mt money market. The economy is competitive, and agents have rational expectations. An equilibrium is a set of state-contingent prices and interest rates such that markets clear when all agents solve their optimization problems, treating prices as given. In the next subsection, we are more explicit about what this means. THE MODEL AS A RECURSIVE SYSTEM The household solves a dynamic program, which is recursive under standard assumptions about preferences, technology, and the stochastic environment. ASSUMPTION 1. tiate, The period utility function U is twice continuously strictly increasing in both arguments, and strictly ASSUMPTION 2. concave. The production function F has the form F(k, n, 0) = 9f{k, n), where f is twice continuously guments, differen- differentiate, concave, and homogeneous strictly increasing in both ar- of degree one. (Stochastic constant returns to scale.) ASSUMPTION 3. The preference shocks {/3f} and the technological shocks {6t} are generated by independent first-order Markov processes. The support of &t is contained in (0,1) and that of &t is contained in (0, oo). Monetary policy consists of a rule that dictates the value of open market operations, the size of government transfers, and the level of the discount rate; -10- Coleman. Gilles. and Labadie these instruments are not completely independent of each other. The operation of the discount window is modeled through a fixed function w that relates the discount rate to borrowed reserves. Think of the government as announcing this function and keeping it fixed in all periods, leaving the discount rate itself endogenousiy determined by the demand for borrowed reserves. Given the function V>, the values of Gt and Xt in period t are implied by the choices of the ratios gt = Gt/Ht and 7* = iift+i/^t- To induce stationarity and recursivity, choose (gujt) ASSUMPTION 4. order Markov as the policy variables and make the following assumption. The monetary policy shocks { # , 7*} are generated by a firstprocess. Starting with the optimization problem faced by the bank simplifies both the notation and the analysis. The bank maximizes its period profit, given in (3), by choosing an optimal portfolio (Sf,Gt,i?t, Vi), subject to the legal reserve constraint (1), and the accounting identity (2). Clearly, optimization requires that V% = p[Zt + Bt + Pt(yt — it — ct)] (no excess reserves) if r* > 0. A zero-profit condition, the result of perfect competition and constant returns to scale in the banking industry, implies that r\ = [{Mt + Bt — Vt)/(Mt — Zt)] x rt; this condition in turn yields r\ = r t [l + (l — p)(Zt + Bt)/(Mt — Zt% which holds whether or not r* > 0. To obtain the last expression, recall the market-clearing condition yt = ct-r itSince the firm and the bank belong to the household, it is possible to integrate the problems faced by the firm, the bank, and the household. Because money growth induces a trend in nominal variables, stationarity of the equilibrium requires that nominal variables—denoted by uppercase letters— be divided by the supply of fiat money. The new variables are denoted by the corresponding lowercase letters; thus, nit = Mt/Ht, zt = Zt/Ht, and so forth. Under assumptions 3 and 4, the evolution of the shocks is determined at the beginning of period t by the vector (/3t,0t-i> 5t-ij7t-i)> which consists of the latest known realizations of the shocks. The state of the economy at that time can then be expressed as st = («t?/3t, 0t-i?5t-i)7i-i)> where Kt is the aggregate per capita stock of capital (as opposed to fct, which is the individual firm's holding). In equilibrium, of course, individual decisions determine - 11- Coleman. Gilles, and Labadie aggregate outcomes, so that K% = kf. A solution is a set of functions p, w. and r such that pt = p(st,st+i), wt = tu(st,.st+i), and rt = r(s t ,«st+i) yield the equilibrium values of the normalized price level, the normalized wage rate, and the rate of interest on date t (again, pt = Pt/Ht and wt = Wt/Ht). Since qt = 1/(1 +Tt)} the equilibrium function r determines a function q satisfying 9t = g ( j t i * t + i ) . Given such pricing functions, let J(m, &, s) denote the value of the optimal discounted stream of utility for a household starting a given period with money balances m, while the firm owns capital stock k and the economy is in state s = («,/?, 0,(/, 7). The household first chooses z, which is the transfer from its savings to its checking account, expressed as a fraction of the outstanding supply of fiat money. Then (/9^0^if^7 , ) are revealed (a prime denotes the realization of a variable that was unknown at the beginning of the period), and these shocks determine the current price, wage rate, and rate of interest, as well as the next-period state s'. To determine s1, the household must know how the evolution of the aggregate capital stock depends on the state of the economy. In equilibrium, of course, this law of motion follows from the individual optimal decisions. On the basis of an assumed law of motion for K and of p(s,s ; ), w(s1s')} and r(s,s'), the*household makes its consumption and leisure decisions and the firm makes its labor and investment decisions. What these optimal decisions are can be studied by considering the Bellman equation characterizing J, the value function. J(m,k,s) = max-Ej max {C/(c,£)' + / 3 J ( m U V ) } subject to (4) 2>P<:; n* = pO1 f(k1 n) — (1 -J- r)pi — wn; Jfe' = ( l - * ) J b + i; w(l - I) + (m - z)(l + rb) -f x1 + 7Tf -»- (z - pc) , m = -12- Coleman. Gilles, and Labadie the last constraint on the problem is the law of motion for K. Here p, w. and r are short for p($, s'), w(s, s1), and r(s, s'), and E9 is the expectation conditional on 5. Using the results of the bank's optimization problem, the market-ciearing condition 6' f{k,n) = c + i, and the firm's optimization condition b = pi. we have r > 0, v > p(z + 6), and v = p(z + b) if r > 0. OPTIMIZATION AND EQUILIBRIUM CONDITIONS. The Bellman equation for J includes two maximization operators; the first refers to the choice of z, which is conditional only on s, and the second refers to the choice of (c, £, n, i) which is conditional on both s and sf. Corresponding to the latter choice, we have the following four first-order conditions: (<0 u>(j,5') u>(a,s') = p ( * , a ' y / 2 ( * , n ) ; (») (0 y J 2 ( m \ Jb',,') = p(*,,')[! + r(s, M')]Jl{m''?'J); where A is the Kuhn-Tucker multiplier associated with the finance constraint (4), so that A(z — pc) = 0. Indexes to the functions U and J denote partial derivatives; therefore, U\, for example, is the partial derivative of U with respect to its first argument, consumption. The first-order condition associated with the choice of z is (*) E. Ui(c,£) 3 S IP( > ')\ = E. 0[l + rh(s,*')] Ji(m',k',s') r To solve the dynamic programming problem, we need the following envelope conditions, which give the marginal values of money and capital: (m) / l ( m , 4 , ( ) . & [ ^ ] . (*) J 8 (m, * , . ) = E. [(U,(c,t) - pX) ( « 7 i ( * , n ) + (1 + r)(l - * ) ) ] ; where p is short for p(s, s1), and similarly for w and r. -13- Coleman, Gilles, and Labadie Finally, an equilibrium in this economy is a set of functions w(s,s!), p(s,$'), and r(s, s1) [or equivalently 9(3, s1)] and a law of motion for the aggregate capital stock K such that the associated solution of the dynamic programming problem—that is, values for (z, A, c, /, n, 2, i/, d) that solve the first-order and envelope conditions—satisfies the following equilibrium conditions: c + i = tC/(i, n); l - * = n; qg + v - (f = m; rn = 1; fc' = « ' ; r6 = m —2 xr; d x r = (fx ip(d/v). The last equation states that, when the monetary authorities lend at the discount window (d > 0), they do so in accordance with their supply behavior, so that r = ^{d/v). In the third equilibrium condition, qg1 + v — d = m, v is equal to p(z + pi) unless r = 0, in which case v can exceed required reserves. SOLVING THE MODEL Consider initially a slightly simplified version of the model in which labor is inelastically supplied (I = 0) and money supply'is constant (7 = 1). To solve this simplified model, first reduce the system of equations that determines the equilibrium to only three equations in the three unknown functions c, z, and (a transformation of) J\. To simplify the notation, define £(/9,.s') = /3Ji(l,/c',y). 3 Then the firstorder condition (c) becomes Ui(c) = (\ + t)PRecall that K is one of the arguments of 5, so that the function £ is well defined. -14- Coleman, Gilles. and Labadie Here and below £ stands for f(/3,s'); accordingly £' below stands for £(/?', s"). Using this equation and the constraint z > pc, which holds with equality whenever A > 0, isolate p as (5) p = nun ^ { } Substitute this equation in £ = (3E5i [U^c^/p1], which follows from the definition of £ and the envelope condition (m), to obtain ( = 0E, max (6) {***.<} this equation is the first of the set of three to be solved (£ now replaces J\). The second equation follows from substituting the expression (5) for p into the first-order condition (z), obtaining (7) £.|ma*{^,*}]=2<;.l(l + r )t}. h The last equation in the system follows from the first-order equation (i) and the envelope condition (Jk): mm {**«} l£l (8) z'e, Ux{c') \ {9"h{k') + (1 + r')(l - 6)) = 0qEs nun < — <O}CJ To write (6) - (8) solely in terms of c, z, and £, express r and r in terms of these functions as follows: T = i>(dlv); and (9) r> = where -15- Coleman, Gilles, and Labadie d = qg ~ v - 1; v = p(z + 6); 6 = m | _ _ _ | m ; and finally, i = 0'/(Jfe)-c. These equations hold provided d > 0 and r > 0; if d = 0, then r < ^(0), while if r = 0, then v > p{z + 6). Rather than solving this model explicitly, which can be done numerically using the methodology presented by Coleman (1992), we devise an example which admits a closed-form solution. This example highlights all the features of the model that are useful in interpreting the empirical regularities mentioned earlier. AN EXAMPLE To develop an intuitive understanding of the model, it is instructive to consider a parametrization that allows a closed-form solution. Suppose that (a) utility is logarithmic; (b) production satisfies f(k) = fca, for 0 < a < 1; (c) capital depreciates completely over each period; and (d) the technological shocks 5, the policy shocks g, and the preference shocks 0 are all iid (although not necessarily independent of each other). Now, conjecture that no excess cash is ever held in the goods market and that z is constant at z. ; circumstances, 6 = zi/c, i = fc , and equations (6)-(8) simplify to z (10) (11) 1 = E, 'P(l+rb) " = 0'qEsl ) '6"a(k')a-1' J c where the interest rate r satisfies <*> -*r»'' —1 6 - Under these Coleman. Gilles, and Labadie and rb is given by (9). Further conjecture that the consumption function can be written as Q = TTzrt—r$ k , 1 + Wq) for some function h. Note that because k!/c = h(/3'g), the function h can be thought of as the investment to consumption ratio. Since h depends only on flq and since q = 1/(1 + r), (12) determines r a s a function of 5, /3', and g1. Write this function, which implies that r and q are iid and independent of s, as r = RJ^z^ff^g1) and correspondingly q = Q{z^P\g9)\ now substitute these equations into (9), and the resulting equation into (10), to obtain ! = £ , tt[i + [i^->n*««*<i'-™t)m!,M) This equation has the important implication that z does not depend on 5, because s enters only through the conditional expectation, and /?' and g1 are iid. This observation verifies the conjecture z(a) = z. Tofindfc,substitute the conjecture about the consumption function into (11) and simplify to obtain h(l3,q) = a(3'q(l + Esl[h(/3"q')}). Using the fact that 0' and q are iid (because q — Q(z,0',g'), and (@,g) is iid), this equation implies H{l3q) -l-Ela0<qy where E[. ] is the unconditional expectation, taken over the constant distribution of (y9',g). It is then straightforward to verify that the finance constraint in the goods market is always binding; therefore, all the initial conjectures were correct. This example leads to a sharp characterization of the response of monetary aggregates and the interest rate to supply and demand shocks. equilibrium value of k'/c =fc,rewrite (12) as (13) U^^f^Sl-^d-ElaP'q)) q ^\ pz(l+a0'q-E[a(3'q}) -17- Using the Coleman, Gilles, and Labadie Consider first the effect of technological shocks, &. .Such shocks do not affect r, as (13) makes clear, and thus they do not affect any of the monetary aggregates. They have real effects, of course, since they affect output, consumption, and investment. But they fail to move nominal interest rates (although real rates certainly do) because the demand for consumption and investment goods shift proportionately. This feature is due to the choice of utility and production functions, and is not a general feature of the model. It indicates, however, that in the general case productivity shocks can affect interest rates and monetary aggregates in either direction. Before turning to the effect of other shocks, it is helpful to list the relevant equations. The first is (13), which determines the correlation between each shock and the nominal rate of interest. The others are: (14) total reserves: v = pz[l + h(/3'q)]] (15) nonborrowed reserves: (16) borrowed reserves: (17) Ml: z + b=z[l (18) M2: 1 + 6 = 1 + 2&(0'g). t; — d = 1 — qg1] d = v x ^-1(r); + h(0'q)]] To isolate the effect of policy shocks, assume first that there are no other shocks (a similar procedure will uncover the effect of preference shocks). Note that the left side of (13) is decreasing in g, while the right side is increasing both in q and in g1 (recall that T/J is increasing); therefore g' and q vary inversely. For the same reason, but considering the right side as a function of q and qg\ q and qgf vary inversely also. Hence, gf, r, and qg1 all move in the same direction. In view of (15), then, policy shocks induce a negative correlation between the nominal rate of interest r and nonborrowed reserves v — d. They also induce a negative correlation between r and v, total reserves, as (14) reveals since h increases in q. The correlation between r and v can be entirely attributed to the variance of inside money, z/i(/3'g); this variance also induces a negative correlation between r and the broader monetary aggregates Ml and M2, as shown by (17) and (18). From (16), it is clear that the ratio of borrowed to total reserves is positively correlated with the interest rate, a relation which -18- Coleman, Gilles, and Labadie has nothing to do with the source of the shock but is due exclusively to the form of ^, that is, to the operation of the discount window. If total reserves did not respond to the policy shock (an assumption which is sometimes made in empirical work), the form of ifr alone would induce a positive correlation between the interest rate and borrowed reserves. Suppose now that shocks to /3 are the only shocks in the system. The left side of (13) is decreasing in g, while the right side is increasing in q and decreasing in /3'g; therefore, q and /3'q (and therefore q and /3' also) move in opposite directions, while {31 and /3'q move in the same direction. Equations (14)—(18) then show that preference shocks induce a positive correlation between the interest rate and any of the reserve or monetary aggregates (total, nonborrowed, and borrowed reserves; inside money, Ml, and M2). It is now possible to use the example to study more complicated policies. Suppose that in response to positive preference shocks that would otherwise increase interest rates, the government chooses its open market operation to keep the rate constant, which corresponds to a small realization of g1 (in this case, /3 and g are still iid, but not independent of each other). With the interest rate constant, /?' high and gf low, all the reserve and monetary aggregates are high (but the borrowed reserve ratio is constant). If the policy response only partially offsets the preference shock, all reserve and monetary aggregates may still rise, while the rate of interest rises also. In that case, despite the presence of a liquidity effect in the model, open market operations could be seen as "inducing" a positive correlation between interest rates and various monetary aggregates (and nonborrowed reserves as well). CONCLUSION: INTERPRETING THE EMPIRICAL LITERATURE As mentioned in the introduction, the empirical literature directed to measuring the effect of monetary policy shocks on interest rates is replete with seemingly conflicting results. The model provides a framework for thinking about these results and for interpreting the literature; the example brings out the important features of the model. First, the model highlights the role of inside money creation as an avenue for total reserves to respond to open market -19- Coleman. Gilles, and Labadie operations. In this sense, the model fails to support Strongin's identifying restrictions that total reserves do not respond to open market operations within a month or a quarter. Second, the model suggests that the operation of the discount window, summarized by a fixed and positively sloped supply function, can alone generate a negative correlation between nonborrowed reserves and the federal funds rate. Such a correlation has been documented by Christiano and Eichenbaum (1991). While they identified policy shocks as innovations to nonborrowed reserves, the model suggests an alternative explanation that has nothing to do with policy shocks. Third, although the model is designed to have a liquidity effect, a policy of interest-rate smoothing hinders efforts to detect its presence. This could explain the difficulties econometricians have had in measuring this effect. To identify policy shocks, it is not sufficient to identify a variable (such as nonborrowed reserves) that is under the control of the Fed, since the Fed may use its instrument to achieve particular objectives. In this sense, the model points to the familiar need, and provides a framework for, identifying demand and supply shocks to estimate a liquidity effect. -20- Coleman, Gilles, and Labadie REFERENCES Bernanke. B., and A. Blinder. "The Federal Funds Rate and the Channels of Monetary Transmission," Working Paper No. 3487. New York: National Bureau of Economic Research, October 1990. Cagan, P. The Channels of Monetary Effects on Interest Rates. New York: National Bureau of Economic Research, 1972. Cagan, P., and A. Gandolfi. "The Lag in Monetary Policy as Implied by the Time Pattern of Monetary effects on Interest Rates," American Economic Review, vol. 59 (Papers and Proceedings, 1969), 277-84. Christiano, L. J., and M. Eichenbaum. "Identification and the Liquidity effect of a Monetary Policy Shock." Unpublished manuscript, Federal Reserve Bank of Minneapolis, November 1991. Coleman, W. J. "Solving Nonlinear Dynamic Models on Parallel Computers," Institute for Empirical Economics working paper. Federal Reserve Bank of Minneapolis, 1992. Dutkowsky, D. "The Demand for Borrowed Reserves: A Switching Regression Model," Journal of Finance, vol. 39 (1984), 407-24. Freeman, S., and G. W. Huffman. "Inside Money, Output, and Causality," International Economic Review, vol. 32 (1991), 645-67. Fuerst, T. S. "Liquidity, Loanable Funds, and Real Activity," Journal of Monetary Economics, vol. 29 (1992), 3-24. Goldfeld, S. M., and E. J. Kane. "The Determinants of Member-Bank Borrowing: An Econometric Study," vol. 21 (1966), 499-514. Goodfriend, M. "Discount Window Borrowing, Monetary Policy, and the PostOctober 6, 1979 Federal Reserve Operating Procedure," Journal of Monetary Economics, vol. 12 (1983), 343-56. Grossman, S. J., and L. Weiss. "A Transaction-based Model of the Monetary Transmission Mechanism, " American Economic Review, vol. 73 (1983), 871-80. King, R., and C. Plosser. "Money, Credit, and Prices in a Real Business Cycle," American Economic Review, vol. 74 (1984), 363-80. -21- Coleman. Gilles, and Labadie Leeper, E. M., and D. B. Gordon. "In Search of the Liquidity Effect," Journal of Monetary Economics (forthcoming). Lucas, R. E., Jr. "Liquidity and Interest Rates," Journal of Economic Theory, vol. 50 (1990), 237-64. Polakoff, M. E. "Reluctance Elasticity, Least Cost, and Member-Bank Borrowing: A Suggested Integration," Journal of Finance, vol. 15 (1960), 1-18. Rotemberg, J. J. "A Monetary Equilibrium Model with Transaction Costs," Journal of Political Economy, vol. 92 (1984), 40-58. Sims, C. A. "Interpreting the Macroeconomic Time Series Facts: The Effects of Monetary Policy," European Economic Review (forthcoming). Strongin, S. "The Identification of Monetary Disturbances: Explaining the Liquidity Puzzle." Unpublished manuscript, Federal Reserve Bank of Chicago, December 1991. Waller, C. J. "Administering the Window: A Game-Theoretic Model of DiscountWindow Borrowing," Journal of Monetary Economics, vol. 25 (1990), 273-87. -22- Chart 1. Federal Funds Rate and Nonborrowed Reserves Ratio Monthly, January 1961 - July 1992 percent 1960 1965 1970 1975 1980 1985 1990 i Chart 2. The Psi Function; 1961 (1)-1992(7). Ratio of Borrowed To Total Reserves U.1U • 0.08 — • 0.06 • • • • • • • • 0.04 *• • • • : • • • • • • • § • • • • •• • • • • • • ••• • •• • • • • / •• • • • •• • • •- • • • • 1 • • • 1 • • • • 0.02 • % 0.00 hi • . : *• .wA * * i * '„•.•••• 1 •• •• •• • . l_ -2 finrftarl fFed Funds - Discount Ratol L_ l_J Comments on "Discount Window Borrowing and Liquidity" by Coleman, Gilles, and Labadie Michael Dotsey I have been asked to discuss "Discount Window Borrowing and Liquidity" which I view as very interesting but preliminary work toward examining "liquidity effects" in a framework that incorporates a fairly (primitive) reserves market. I use the term primitive with regard to the reserves market since no interesting dynamic behavior is present in this market. Viewing work on BRd, especially that of Goodfriend (1983) this is a shortcoming that I hope will be addressed by later generations of the model. The paper, however, is very rigorous and state of the art on other dimensions and the authors deserve a lot of credit for moving the liquidity effects literature in this direction. The empirical motivation for the paper can be traced to work by Christiano and Eichenbaum (1992) and especially to that of Strongin (1991). Strongin's work is fairly persuasive and indicates that in order for any model to replicate data on liquidity type effects reserve market behavior is likely to be a crucial ingredient. This is because the liquidity effect only shows up in NBR's or to be more accurate, in the part of NBR that represents independent monetary policy. This paper's novel inclusion of reserve market behavior represents a commendable extension of this basic line of research.1 In reading this paper, I found that it raised at least as many questions as it answered. Much of my confusion is not the 1. One thing I would like to see done in these estimations is removing settlement day data. This data could potentially contaminate the results. Suppose for instance the Fed misforecasts float or treasury balances believing there will be more of these funds available than are actually there. NBR will be low on the settlement day and the funds rate will be high, perhaps by a substantial amount. Two such occurrences in a month (at least 25% probability) could make monthly average NBR a little low and monthly average rF a little high. While I doubt this is the reason for Strongin's results it would be nice to purge the data of what is merely an interbank friction. Michael Dotsey result nor the fault of this paper in particular, but rather comes from a lack of understanding and perhaps misgivings of this literature in general. In my comments I will discuss some of these misgivings and, hopefully, my comments will lead to some discussion from the rest of the audience. The paper extends a branch of research that is attempting to understand the effect of monetary policy on interest rates and real activity. In particular these papers7 search for a mechanism that will explain (1) how contractionary monetary policy raises short-term interest rates and (2) how it causes declines in economic activity. This literature received its impetus from Lucas's (1990) influential paper. A common feature of most of this literature involves cash-in-advance constraints that constrain the amount of money available for use in a loan or securities market, however, no two papers seem to use the same exact specification. Lucas's original setup and CGL (1991) envision bond traders as only having limited funds and, therefore, open market operations affect the price of bonds .and thus interest rates. The appeal of Lucas's setup is that it eliminates the differential wealth effect of open market operations that were present in earlier literature (eg Grossman and Weiss and Rotemberg). Fuerst (1991) extends Lucas's setup to a production economy that places a CIA constraint on both investment and labor expenditures. Unlike households' portfolio decisions, production decisions are made after the stochastic state of the economy is known. Since individuals must choose the portion of their portfolio to lend to firms via intermediaries prior to observing the monetary transfer or the market clearing interest rate, the monetary transfer can affect the tightness or looseness of the loan market. Hence liquidity effects that have real consequences result from monetary policy. Christiano (1991) subjects the Fuerst model and an alternative version of that model in which investment decisions - 2 - Michael Dotsey are also made prior to the realization of shocks to a statistical comparison with a RBC model that contains a standard CIA constraint. For reasonable parameter specifications the Fuerst • model can not produce a liquidity effect that dominates anticipated inflation effects on the nominal interest rate while the sluggish capital model can produce a dominant liquidity effect. Both these models produce too much variability in consumption and the counterfactual result that consumption and prices move in opposite directions. They also produce very low interest elasticities of money demand and monetary policy has very little effect on variations in output. Furthermore, anticipated inflation has much too large an effect on labor, consumption, and output. To remedy this last result, Christiano and Eichenbaun (1992) relax the CIA constraint on investment. They also split the period into two parts allowing firms to adjust their hiring decision after observing open market operations while initial hiring and investment decisions are made prior to observing open market operations. They do this with the hope of magnifying the response of employment and output to liquidity effects. In CGL's current paper firms face a CIA constraint on investment but can pay workers out of end of period revenues. Also, monetary transfers are made directly to consumers after their portfolio decision has been made. Thus these transfers do not affect the funds available in the credit market and, therefore, do not give rise to a "liquidity effect." Because there is a CIA constraint on capital, monetary policy can have inflation tax effects as well. As their work progresses separating liquidity effects from inflation tax effects will be important. Not all of these scenarios can be correct. constraints placed where they are? These assumptions of infinite transactions costs are not innocuous. in these models. Why are CIA They are the driving force It seems that rather than trying to incorporate a realistic financial structure into a dynamic macro model and - 3 - Michael Dotsey then testing the model, investigators are trying to find a mathematical structure that produces the correlations they desire. Apart from Christiano (1991) very little effort is made to see if these models are an improvement on basic RBC models or even if they produce counterfactural predictions along other dimensions. Since other classes of models can produce negative correlations between NBR and the funds rate, examining how CIA models fit the data along other dimensions will be important if the CIA approach is to gain widespread acceptance. For example a model like that in Goodfriend's (1987) paper can potentially produce correlations of the type this literature is seeking. In that model, which has no rigidities, purposeful behavior by the Fed can set up negative correlations between the funds rate and NBR. If the Fed wishes to reduce inflation, it can do so by reducing the future money supply and in particular future NBR. Due to anticipated inflation effects, the nominal interest rate would fall increasing the demand for money and total reserves. If the Fed wishes to reduce price level surprises it can supply the necessary NBR to prevent price level movements. Thus this policy sets up the requisite negative correlation. If that was all that was going on one would expect this negative correlation to carry over to broader aggregates. However, M2-M1 components of M2 which involve a large savings motive should be positively correlated with the real rate of interest and movements in BR, which are highly variable and positively correlated with the funds rate, could cause TR to be positively correlated on net as well. Alternatively say the Fed is following an exogenous upward movement in the real rate of interest in an attempt to target inflation. If the own rate on money balances is sticky then money (Ml) and hence total reserves will decline along their demand curve. (Also, M2 could be rising with the real rates.) This would set up a negative relationship between NBR and the funds - 4 - Michael Dotsey rate. As rm adjusted, total reserve demand would increase as would NBR as the Fed defended the new higher funds rate. If the Fed did not react instantaneously or vigorously enough to the increased reserve demand the funds rate could rise further and then fall as nonborrowed reserves were pumped into the system reinforcing the initial negative correlation. Also, sticky price models may be able to generate some of the correlations displayed in the data as well. Also, the question of what constitutes a period is somewhat fuzzy in this literature. Is it a day or perhaps a week? Most people make some form of cash management decision weekly and I can not think of any time where a shortage of cash has affected my real consumption for more than a day or two. Perhaps I'm taking the CIA constraint too literally, but if the period is rather short, as I believe it is, then the propagation mechanisms needed to match the data would seem incredible by RBC model standards. I have strayed a little far afield so let me return to this paper more specifically. My primary confusion is linking the author's major contribution which shows how different measures of money can have different correlations with interest rates with the motivation for their paper which appears to be the results found in Strongin. In this paper money (1) M t ^ - Mt + rt(Gt-Dt) + xt. The xt portion of measured money provides no liquidity effects. The 6t portion, that is open market operations has the standard liquidity effects since it influences the portion of firm borrowing that must be financed by discount window loans. The equilibrium condition that is being used is (2) NBRt = Vt - Dt = Mt - Gt where Vt * 0(M t +B t ). An increase in 6t (an open market sale) requires more discount window borrowing and an increase in interest rates since r =0(D/TR) is increasing. - 5 - Using Mt+1 can Michael Dotsey contaminate regression results since it rises by r t (G t -D t ), which will in general be positive in this model and no liquidity effect will be present. Furthermore, growth in money via transfers wiVI further bias econometric results. For econometric purposes I see no useful way of isolating any aggregate to uncover liquidity effects- Xt type disturbances, in reality, involve transfers from the Treasury. These involve a reduction in Treasury accounts at the Fed and an increase in NBR. What the model here indicates is that one wants to examine only changes in reserves that involve changes in the public's asset positions and that exclude any interest or lump sum payments. While these decompositional problems are important for this model and may in fact be important more generally, they seem to have little to do with Strongin's empirical strategy nor do they affect interpretations in other models. Strongin tries to separate "pure" supply movements in NBR from those engendered by policy responses to changes in TR. Whether his identification procedure is a good one or not could be debated, but he is not concerned with measurement or decompositional problems in various reserve measures. The decompositional problem arises in CGL because of their modeling of xt as having no liquidity effects. In Fuerst or Christiano and Eichenbaun, there is only xt and it enters the model in a way that produces liquidity effects. That is NBR supply disturbances that are not responses to TR shocks produce liquidity effects. It seems that Strongin's methodology is more closely aligned with these models. Whether decompositional problems are important or not, I don't know. They arise in this model by a specification that at this point seems somewhat arbitrary. It is no more arbitrary than any other specification in the literature, but that does not make it convincing. I believe the author's need to make a convincing argument as to why some forms of morjey creation are more likely to - 6- Michael Dotsey involve liquidity effects than others if their message is to carry weight. After all, in this model one could easily reverse the roles of Xt and Gt or make them complimentary. The discussion on page 11 regarding the estimation of rp is also a little confusing. With (3) n-1 -± -±£ v V they claim that 0 can be estimated no matter what the shock. But is that relevant? We would like to know how j> is influenced contingent on different shocks. Here a positive V shock induced by a shift in the demand for loans causes 0 to rise and n to fall, while a decline in NBR due to an open market sale (G up) also causes n to fall and i> to rise. It is only the latter effect that one has in mind when discussing liquidity effects, so perhaps the ratio is not the correct variable to focus on. Rather, in this model it should be the relationship between the level of NBR and the funds rate. Also in estimating 0, one would expect shifts in the function over time since administration of the discount window has changed over time. For example, I believe window administration was more lax when the Fed faced a membership problem. I would also downplay somewhat figure one. The interest rate of consequence is the spread between the funds rate and the discount rate. When one looks at this graph the correlations seem at least as pronounced. But has anything but a borrowed reserve demand function been uncovered? Finally, the discussion concerning adjustably pegging the interest rate based solely on technological disturbances raises questions concerning the nominal determinacy of the model (see McCallum (1981, 1986)). Overall, I thought this paper was interesting and represents a nice attempt to start thinking about how behavior in - 7 - Michael Dotsey the market for reserves influences the correlations we observe between various monetary measures and the funds rate. Given my qualms concerning this methodology's ability to explain anything at business cycle frequencies, I would suggest directing the model in an alternative direction. Perhaps this framework could be used to help explain short-term term structure movements in interest rates and examine the so-called "ozone hole." This line of inquiry would be interesting since it could integrate reserve market behavior and a tight specification of policy in a fully developed general equilibrium model. - 8 - Michael Dotsey REFERENCES Christiano, Lawrence J., "Modeling the Liquidity Effect of a Money Shock," Federal Reserve Bank of Minneapolis Quarterly Review, Winter 1991. 15(1). 3-34. Christiano, L.J., and M. Eichenbaum, (1992) "Liquidity Effects, Monetary Policy and the Business Cycle," unpublished ms.. Federal Reserve Bank of Minneapolis, Northwestern University, NBER and Federal Reserve Bank of Chicago, July 1992. Fuerst, T.S. "Liquidity, Loanable Funds, and Real Activity," Journal of Monetary Economics, vol. 29 (1992), 3-24. Goodfriend, Marvin. "Discount Window Borrowing, Monetary Policy, and the Post-October 6, 1979 Federal Reserve Operating Procedure," Journal of Monetary Economics, vol. 12 (1983), 343-56. , "Interest Rate Smoothing and Price Level TrendStationarity, " Journal of Monetary Economics, March 1987, 19, 335-48. Grossman, S. J., and L. Weiss. "A Transact ion-based Model of the Monetary Transmission Mechanism," American Economic Review, vol. 73 (1983), 871-80. Lucas, R.E., Jr. "Liquidity and Interest Rates," Journal Economic Theory, vol. 50 (1990), 237-64. of McCallum, Bennett, T. (1981) "Price Level Determinancy with an Interest Rate Policy Rule and Rational Expectations," Journal of Monetary Economics, vol. 8 (November). , (1986) " Some Issues Concerning Interest Rate Pegging, Price Level Determinacy, and the Real Bills Doctrine," Journal of Monetary Economics vol. 17 (January). Rotemberg, J.J. "A Monetary Equilibrium Model with Transaction Costs," Journal of Political Economy, vol. 92 (1984), 40-58. Strongin, S. "The Identification of Monetary Disturbances: Explaining the Liquidity Puzzle." Unpublished manuscript, Federal Reserve Bank of Chicago, December 1991. Credit Conditions and External Finance: Interpreting the Behavior of Financial Flows and Interest Rate Spreads Kenneth N.Kuttner1 Aflurryof recent macroeconomic research has drawn attention to the relationship between monetary policy, credit conditions, and the markets for short-term debt Two recent papers have focused onfirms'substitution between bank and non-bank externalfinancein particular, proposing macroeconomic indicators based onfinancialmarket activity. Kashyap, Stein, and Wilcox (1992) employ quantity data directly, arguing that the share of bank loans out of firms' total short-term finance is an informative index of Federal Reserve policy and loan availability more generally. In a complementary line of research, Friedman and Kuttner (1992) identify monetary policy and bank lending as potential sources offluctuationsin the spread between yields on commercial paper and Treasury bills. While both papers have demonstrated solid empirical links between these financial indicators and real economic activity, neither hasrigorouslyassessed the extent to which fluctuations in these indicators actually represent exogenous changes in credit conditions, rather than endogenous responses to changing economic conditions. This paper's goal is to provide such an assessment. The paper begins with a sketch of the mechanism through which credit conditions affect firms' short-termfinancing,drawing a distinction between the effects of the Federal Reserve's open market operations and other factors influencing banks' willingness to lend. The second section summarizes the reduced-form relationships between real output, the interest rate, and three alternative indices 1. Senior Economist, Federal Reserve Bank of Chicago. I am grateful to Benjamin Friedman and David Wilcox for their comments and suggestions. -1- Kuttner of credit conditions: the composition of external finance, the spread between the loan rate and the commercial paper rate, and the analogous spread between commercial paper and Treasury bills. The third section turns to a closer examination of the impact of monetary policy and loan availability on bank and non-bank finance using structural VAR techniques. Identifying monetary policy with innovations to non-borrowed reserves and controlling for firms'financingrequirements, the first of the three models estimates the dynamic effects of monetary and lending shocks on the composition of external finance, the interest rate, and real output. The second structural VAR system assesses the effects of reserves and lending shocks on the paper-bill spread. The third model identifies lending shocks with innovations in the loan-paper spread. Estimates of these models confirm that all three variables respond appropriately to reserves shocks. In addition, lending shocks, whether identified through financial flows or via fluctuations in the loan spread, induce a substitution between bank and non-bank finance. Less clear is the extent to which any of these measures exclusively reflects the effects of changing loan availability. The fact that positive lending shocks are associated with increases in the interest rate and the paper-bill spread suggests that changes in the composition of external finance have more to do with firms' financing requirements than with exogenous changes in banks' willingness to lend. Another slightly puzzling observation is that the largest source of changes to the composition of external finance seems to be wholly unrelated to both reserves and bank lending. Together, these two results suggest that while credit conditions are one important determinant of firms' choice of financing, short-term debt flows may be informative for reasons other than those involving the substitution between bank/non-bank substitution. Although its implications for real activity are rather weak, the loan spread appears to be a plausible alternative measure of credit conditions. A model offinancialflowsand interest rate spreads How do the markets for short-term bank and non-bank finance respond to monetary impulses? And how do non-monetary shocks affect these markets? And how might one construct an index of the availability of intermediated funds? -2- Kuttner As a first step towards answering these questions, this section analyzes a simple model of the markets for commercial paper, bank loans, and Treasury bills in the style of Brainard (1964) or Bosworth and Duesenberry (1973). While not as detailed as either of those models, it is adapted to highlight firms' tradeoff between bank and non-bank finance. It also draws an important distinction between purely monetary influences acting through open market operations, and credit conditions defined more broadly, which may include other factors affecting banks' willingness to lend. One of the model's more obvious properties is that an injection of reserves causes the interest rate to fall — the familiar "liquidity effect." Reserves injections also cause the spread between the interest rates on bank lending and commercial paper to fall, and leads to increased reliance on bank finance. Lending shocks, which are assumed to affect only banks' preferences over alternative assets, turn out to have similar effects on the loan-paper spread and the composition of firms' finance. Lending shocks, by contrast, have no effect on the level of interest rates — only the spreads. The model also identifies two other factors with implications for the money market. First, firms' demand for external finance may induce changes in the relevant interest rate spreads and consequently the composition of finance; controlling for this demand-side influence turns out to be a major challenge to the construction of an empirical measure of credit availability. Similarly, the stock of outstanding Treasury bills may have tangible effects on the spreads and the composition of finance. The three players in the money market are households, banks, and firms, who participate in the markets for reserves, commercial paper, Treasury bills, and loans. Specifically, households' portfolios include demand deposits (DD), commercial paper (P), and Treasury bills (B) according to DD* = <Krp) W, 4>' < 0 df df P* s flrp, rB)W, — > 0 and — < 0 brp drg B^ s (1 - <J) - / f o rB)) W, -3- Deposit demand Paper demand Bill demand Kuttner where W is the sum of deposits, paper, and bills held by households. Households' demand for non-interest-bearing bank deposits is a decreasing function of the prevailing paper rate, rP. A key assumption is that households view commercial paper and Treasury bills as imperfect substitutes, so that changes in their relative supplies affect their respective yields.2 Households require a higher paper rate (or a lower bill rate) to hold a larger share of their portfolio as commercial paper. Demand deposits are banks' sole liability. Their assets are divided among Treasury bills, loans (L), and deposits at the Federal Reserve (R) according to: K* s p(rp)DDt p ' < 0 Ld = Sin, rP, \)DDf — > 0 and — < 0 drL drP B*b s (1 - p(rP) - g(rLf rP, K))DD. Reserve demand Loan demand Bill demand Banks' demand for non-interest-bearing reserves falls with the prevailing paper rate, while loan demand is increasing in the loan rate and decreasing in the paper rate.3 The stock of reserves is set at R' by the Federal Reserve; discount window borrowing is ignored. Banks' demand for loans is also allowed to depend on the variable X, representing any other factors affecting banks' willingness to lend. These "lending" shocks lead banks to shift the composition of their portfolios between bills and loans; negative shifts in X may be interpreted as "credit crunch" episodes. These may occur in reaction to a perceived deterioration in borrowers' creditworthiness, or to more stringent capital requirements as suggested by Bernanke and Lown (1991). They may also be the result of the "moral suasion" instrument of monetary policy; Owens and Schreft (1992) identify a number of episodes in which banks contracted their lending in response to Federal Reserve pressure. Whatever the source, the key feature of these "lending" shocks is that they need not be accompanied by overt monetary policy in the form of open market operations.4 2. Friedman and Kuttner (1992) discuss some possible reasons for this imperfect substitutability. Lawler (1978) also finds evidence for imperfect substitutability at seasonal frequencies. 3. Note that throughout the paper, assets are "demanded" while liabilities are "supplied." Hence, banks "demand" loans and bills, while firms "supply" loans and paper. 4. This point is stressed by Friedman (1991). -4- Kuttner Finally,firmschoose between bank lending and paper issuance as sources of short-term finance according to fth hh dri orp V a h(rL, rP)F, —- < 0 and —- > 0 Loan supply P* » (1 - h(rL, rP))E Paper supply For simplicity, the amount to befinanced,F, is assumed to be exogenous with respect to the various interest rates. Becausefirmsview loans and paper as imperfect substitutes, they willfinancesome portion of F through bank lending even though rL generally exceeds />; as discussed by Kashyap, Stein and Wilcox (hereafter KSW), this presumably reflects some intangible benefit accruing to the firm from maintaining a relationship with a bank. Firms * share of bankfinance(the KSW "mix") responds predictably to the loan and paper rates: an increase in the loan rate (or a decrease in the paper rate), leadsfirmsto substitute away from bankfinancetowards non-bank external finance.5 In equilibrium, the demand for the four assets equals their supply, p(rpMrP)W = l? frurpiKftW-hirurpyF^O Kr»rg)W-{l-h(n,r,)yFmO (1 - g(rL, rPt \))$W+ (1 - / f a rB) - +)W = B*9 determining yields and quantities as functions of the exogenous /?', X, F, and B*. Walras' law allows the bill market equation to be dropped. Further simplification is possible by assuming the asset demand and supply functions to be homogeneous of degree zero with respect to the assets' 5. This model embodies the assumption that bank and commercial paperfinanceare viable alternatives for an economically relevant group offirms.However, there is increasing evidence that this set offirmsis rather small, and that much of the observed variation in the aggregate composition of finance is due to the relative availability offinanceto small and largefirms;see Gertler and Gilchrist (1992) and Oliner and Rudebusch (1992). -5- Kuttner yields, so that (for example) g(n +c,rp + c, X) = gin, />, X) for any constant c. In this case, the/, g and h functions can be specified in terms of interest rate spreads, and the system reduces to: gizLP,mrpW-h(zu>)F = 0 (I) KzpBW-(l-h(zu>))F = 0 where zLP and zPB denote the loan-paper and paper-bill spreads. Analyzing*the comparative statics of (1) is simplified by its (somewhat artificial) recursive structure. The interest rate level is entirely determined by supply and demand in the market for reserves; the fall in reserves resulting from a contractionary open market operation requires a higher rate to equilibrate the reserves market, as illustrated in Figure l. 6 This higher interest rate leads in turn to a shrinkage of demand deposits and the banking system as a whole. Banks respond by raising the loan-paper spread, prompting some of its borrowers to switch to alternative forms of finance— short-term paper in this model. The increased supply of paper (relative to bills) leads to a widening spread between the paper and bill rates. The effects of an adverse lending shock resemble those of a reserves contraction in that both produce a rising loan spread and a substitution towards non-bank finance. Although both shocks produce similar effects on banks' portfolios, they differ in one important respect: reserves shocks affect the level of the short-term interest rate, while lending shocks leave the paper rate unchanged. A fall in X leads banks to shift the composition of their portfolios away from loans and into Treasury bills, leaving their reserve demand and the paper rate (and consequently deposits and the banking system's size) unchanged. Banks increase their spreads relative to the paper rate in order to reduce their stock of loans. As before,firms'increased reliance on commercial paper drives up the paperbill spread. 6. Total wealth is held constant in an open market operation, as the withdrawal of reserves is offset by a sale of Treasury securities. -6- Kuttner The observation that both reserves and lending shocks may contribute to real economic fluctuations is one explanation of the widespread interest in constructing a broader measure of credit conditions than reserves or the interest rate in isolation, which reflect largely those shocks originating from the reserves market The attractive feature of the credit conditions indicators discussed here is their ability to detect the effects of changes in loan availability and reservesfluctuations:in this model, the "mix," the loan-paper spread, and the paper-bill all reflect the impact of both types of shocks. In fact, in the absence of any other shocks, all three of these measures should respond to monetary and credit factors in qualitatively similar ways. One problem common to all three of these measures (and the interest rate itself) is their susceptibility to contamination from changes infirms'overall demand forfinancing,which may alter yield spreads and the composition of externalfinancefor reasons having nothing to do with to exogenous changes in credit conditions.7 This can be illustrated by examining the comparative statics of (1) in response to an increase in F, the dollar amount of fundsfirmswish to raise from the short-term credit markets. A greater demand for loanable funds unambiguously increases the prevailing interest rate, />. Its effects on the loan-paper spread (and therefore the composition of external finance) is ambiguous, as it depends onfirms9share of bankfinance(/t) relative to households* wealth fraction in bank deposits (<|>), and the share of banks' portfolios held as loans (g). When h(zLP) > tyrp)g (as is presumably the case), increases in F cause loan demand growth in excess of deposit growth, driving up the relative cost of bankfinanceand the share of paper infirms'external finance.8 The same inequality is also relevant for the paper-bill spread; a second sufficient condition for a rising spread is that (1 - h(zLP)) > J{zpB\ so that the increasing paper demand would require households to hold a larger share of paper in their portfolios. 7. Under most of the Federal Reserves' post-Accord operating procedures, non-borrowed reserves may also be contaminated in this way; see Strongin (1991). 8. A special feature of the KSW model is that changingfinancingrequirements affect loans and paper proportionally, leaving the "mix" unchanged. -7- Kuttner One additional complication for interpreting the paper-bill spread as a measure of credit conditions is that it may be affected by changes in the outstanding stock of Treasury bills. In addition, the wealth effects associated with changes in the volume of Treasuryfinancemay alter the level of interest rates and loan spread, and consequently the composition of external finance.9 In this model, an increase in the supply of bills reduces the paper-bill spread, as investors require higher returns to entice them to hold the additional stock of bills. This increase in banks' demand for loans leads to a fall in the loan rate relative to the paper rate, and increased reliance on bank finance. To summarize, the model's main implications are: • Both reserves and lending shocks alter the relative price of bank and non-bank finance, inducing a substitution between alternative forms of external finance. • By affecting the supply of commercial paper, this substitution also affects the relative yields on Treasury bills and commercial paper. • Changes in reserves affect the level of interest rates, while lending shocks leave the level unchanged. • Firms' overallfinancingrequirements may affect interest rate spreads and their composition of short-term finance. The goal of the paper's subsequent empirical work is to explore these implications. Specifically, it attempts to identify lending shocks through their impact on the composition of externalfinanceand interest rate spreads, while controlling for reserves and the overall demand for loanable funds. Short-term credit markets and real economic activity One desirable feature of any index of credit conditions is a systematic link between it and subsequent fluctuations in real economic activity.10 The results below summarize the predictive properties of the KSW "mix," the prime-paper spread, and the paper-bill spread. The results show that the "mix" 9. Of course, this assumes that households view government bonds as net wealth; see Barro (1974). 10. Economists and market observers have long recognized the cyclical properties of commercial paper, bank lending, and their relative yields; see, for example, Foulke (1931), Selden (1963), and Stigum (1990). -8- Kuttner and the paper-bill spread are good predictors of future changes in real GDP (although this alone does not justify their interpretation as measures of credit availability). "Causality" tests Table 1 examines the incremental information content of the three measures for future changes in real GDP in the presence of traditional measures of monetary policy: non-borrowed reserves and the commercial paper rate. Regressions 1-3 are four-variate reduced-form equations of the form 4 4 4 4 Ax, a Ho + Hi* + ] T OjAx^ + ^T pi[A ln(J?),w + ] T Y,Arj>^ + ] T 6 , A ^ + e, where x is the logarithm of real GDP, R is non-borrowed reserves adjusted for extended credit and deflated by the GDP deflator, i> is the commercial paper rate, and q denotes, in turn, the "mix", the loan-paper spread, and the paper-bill spread. As in KSW, the "mix" is computed as the observed ratio of bank lending to the sum of lending to commercial paper, or L/(L + P).n The results use the six-month commercial paper and Treasury bill yields, and the prime rate (from the Federal Reserve H.1S release) is used as the lending rate. The table reports F-tests for the exclusion of the four 6, terms for the entire 1960:2-1991:4 sample, as well as two shorter samples. One truncated sample begins in 703, when Regulation Q was eliminated forroostlarge CDs.12 Another begins in 1975:1. Although this date is somewhat arbitrary, it corresponds roughly to the beginning of a rapid expansion of the commercial paper market, during which it became a more popular vehicle for non-financialfirms'short-term finance.13 11. The augmented Dickey-Fuller u statistic (computed with eight lags) for the stationarity of the "mix" is -4.10, rejecting the null hypothesis of nonstationarity at the 1% level. Consequently, it is included here in levels along with a linear trend term. 12. Regulation Q interest rate ceilings on 30-89 day CDs in denominations of $100,000 were eliminated on June 24, 1970. Ceilings on CDs with maturities in excess of 90 days remained in place until March 16,1973. -9- Kuttner The 1975-91 sample also excludes the Penn Central and Franklin National disruptions of 1970 and 1974, and covers the period in which ratings were assigned to commercial paper issues.14 The results of thefirstregression corroborate the strong link between the "mix" and real output found by KSW, supporting theirfindingthat the composition of finance has significant predictive power for future real economic activity, even in the presence of reserves and interest rates. The poor performance of the loan-paper spread in the second regression (again in the presence of reserves and the commercial paper rate) is consistent with the notion that banks' lending rates are relatively uninformative.15 The third regression demonstrates the incremental information content of the paper-bill spread — at least in the earlier samples. Impulse responses While the F-statistics for "causality" give some indication of the strength of the predictive power of thesefinancialindicators, they give no indication of the size or direction of their impact. The impulse response functions plotted in Figure 2 provide a richer description of the effects of innovations to the financial indicators. Each of the three rows of graphs is from the VAR corresponding to regressions 1-3 in Table 1. In each case, the system has been orthogonalized (according to the triangular Cholesky decomposition) with the credit conditions index in last place. Three responses are plotted for each regression: thefinancialindicator's effects on output and the interest rate, and the effect of reserves innovations on thefinancialindicator. The dotted lines depict the approximate 95% confidence bounds. Panels (a) and (b) from the first specification show that "mix" innovations indeed act like reasonable measures of credit conditions; reserves injections increase the share of bank loans, and 13. At the end of 1974, non-financial commercial paper accounted for only 13.5 billion dollars. By 1982, thisfigurehad grown 325.2 percent to 57.4 billion. See Hurley (1977,1982), and Stigum (1990). 14. Moody's and Standard and Poor's began rating commercial paper in 1974. 15. Similar results are obtained with the average of large banks' lending rates obtained from the Federal Reserve Survey of Terms of Bank Lending reported in release E.2. -10- Kuttner output rises in response to positive "mix" shocks, which might be interpreted as the pure lending component of credit conditions. The panel (c) plot, however, is something of a puzzle. It shows that "mix" innovations are associated with a rising commercial paper rate—not what one would expect from an increased willingness to lend on the part of banks, and inconsistent with the implications of the model presented earlier.16 However, this pattern is consistent with banks passively supplying more loans in response to rising demand for credit. The second row of plots confirm the generally weak relationship between the prime-paper spread and real output. One interesting feature of the loan spread is that it initially rises in response to a reserves innovation — clearly inconsistent with the loosening of credit conditions implied by the reserves injection. The loan spread ultimately falls, however, suggesting that this response is due to a certain sluggishness in the way banks adjust their lending rates. The impulse response functions from the paper-bill spread regression are all consistent with what one would expect from an indicator of credit conditions: positive shocks to the spread generate declining real output, while reserves injections reduce the spread. Furthermore, unlike the "mix", innovations in the spread itself have essentially no impact on the level of interest rates. Comparing the "mix" and the paper-bill spread Because regressions 1-3 included each of the credit conditions measures in isolation, the results raise an important question: to what extent are the three indicators measuring the same phenomenon? An obvious way to address this question is to include more than one indicator in the same regression to see if the presence of one vitiates the predictive power of the other. The results from two additional regressions (numbered 4 and 5) are reported in Table 1. The results from specification 4, which includes both the "mix" and the loan spread, are not surprising given the weak performance of the loan spread in isolation — the F-statistics for the "mix" remain virtually unchanged. Somewhat more surprising are the results from specification 5, in which both the "mix" and the paper-bill spread appear. Here, the relationship between the two variables and real 16. The "mix" terms are significant in the interest rate equation at the 10% level. -11- Kuttner output is uniformly stronger (judged by the F-statistics) than when they are included individually. Qearly, one (or both) of the indicators is doing something other than simply summarizing the state of credit market conditions. The roles of commercial paper and bank loans The model sketched earlier suggests that flows of commercial paper and bank lending are informative to the extent that they reflect the substitution between the two forms of finance in response to a monetary or a lending shock. KSW exploit this insight by looking at the ratio of bank loans to the sum of loans and paper, shocks that affect both forms of debt proportionally are presumed to stem from sources other than loan availability. A useful check on this specification is to verify that paper and lending flows enter an unrestricted regression in such a way that the "mix" is the variable that matters. This is easily accomplished by differentiating the "mix" (designated h) with respect to time, P — L dt = (I+P)* L P (I+P) 2 = h{\ - h%/L - /i(l - h)P/P9 decomposing its movements into distinct lending and paper contributions. In discrete time, the analogous decomposition, AA, - AM(1 - AKI)AL/1M - A M (1 - A,-i)AP/P M - <&L - Afip expresses A/i as a weighted sum of commercial paper and bank loan growth rates, denoted tJxL and tJip. If AA were in fact the appropriate measure of the impact of credit conditions on the real economy, the two components would enter real output regressions with equal and opposite signs; the regression itself would "choose" the KSW specification. Table 2 displays the results of this experiment. Panel (a) reports the outcome of a regression of first-differenced log real GDP on four lags of output, tJxL and tJiP over the 1960:2-91:4 sample. Judged by the F-statistics, the commercial paper terms are much more informative than the lending -12- Kuttner terms; tJip is significant at the 0.01 level, while the tJiL terms are not significant at even the 0.10 level.17 The sum of the estimated coefficients on lending is negative, but statistically insignificant The regression in panel (b) refines the test by specifying the regression in terms of tJi and tJip — simply a transformation of the regression in panel (a). Excluding the four lags of &hP is equivalent to restricting the coefficients on iskL and tJip to have equal and opposite signs. Here, the tJi terms are statistically insignificant, while the AhP terms are significant at the 0.05 level. Moreover; the negative estimated sum of the "mix" coefficients is inconsistent with the substitution hypothesis, although this sum is again statistically insignificant. To guard against the possibility that the results in the first two panels are an artifact of the differenced specification, panel (c) reports the results of a regression that includes a linear trend and h in levels. While not tJ\P terms are not as strong in the levels specification, the coefficients on the h terms remain statistically insignificant. These experiments show that the "mix" owes its predictive power in large part to something other than the substitution between bank and paper finance. In unrestricted equations, h terms are generally insignificant, while the hypothesis that commercial paper in isolation does not matter for predicting real output can be rejected. This observation suggests a closer examination of lending and commercial paper flows individually, and their relation to monetary policy and credit conditions. A structural approach to identifying lending shocks The atheoretical results in the preceding section provided some evidence in favor of interpreting the financing "mix" and the paper-bill spread as measures of credit conditions, although innovations in the composition offinancewere, contrary to the simple model, are associated with a rising interest rate. One reason for this pattern may be the result of inadequately controlling for the overall demand for short-term finance. As demonstrated earlier, an increase in the amount to befinancedneed not raise bank and non-bankfinanceproportionally. In this case, if increases infirms'demand for funds 17. This is consistent with the results of King (1986). -13- Kuttner are accommodated primarily through bank lending, the "mix" may rise for reasons unrelated to credit conditions. Figure 3 plots thefinancinggap (defined as the difference betweenfirms'capital expenditures less inventory IVA and after-tax internal funds) along with commercial paper and bank loan flows, demonstrating the close relationship between thefinancinggap and the volume of bank lending (although commercial paper appears to have become more sensitive to thefinancinggap in the later part of the sample). To control for credit demand, the results in this section include the financing gap as an additional determinant offirms'debt issuance. A more interesting alternative hypothesis is that is that the substitution mechanism inadequately explains the joint behavior of commercial paper and bank lending, and that factors other than monetary policy are what drive the observedfluctuationsin the composition of short-term external finance. The apparent asymmetry between the effects of loan and paper flows uncovered in Table 2 provides some circumstantial evidence for this view. The results presented in this section attempt to address these issues by separately analyzing flows of lending and commercial paper in a structural VAR setting that controls for the overall demand for loanable funds. Moving to a more structural approach also addresses the possibility that the interest rate's odd response to "mix" shocks is as an artifact of the artificial triangular structure of the Cholesky decomposition employed earlier. Thefirstmodel focuses on the response of lending and paper flows to reservesfluctuations,and examines the properties of the innovations identified as lending shocks. The second describes the response of the paper-bill spread to thefinancialflows generated by reserves and lending shocks. The third usesfluctuationsin the loan-paper spread as an alternative means of identifying lending shocks. A review of structural VARs Beginning with an unrestricted i-variate dynamic simultaneous equation system, T -14- Kuttner the standard VAR achieves identification by restricting the contemporaneous relationships between the elements of y, i.e., by setting BQ = 0 and A = /, while placing no restrictions on the covariance matrix of v, ie., £(w') = Q. The structural VAR introduced by Blanchard and Watson (1986) and Bernanke (1986) achieves identification by allowing some nonzero elements in thcB0 matrix, while restricting the covariance matrix of v, the structural disturbances, to be diagonal. Off-diagonal elements in A can be introduced to allow distinct elements of y to depend on common structural shocks. Thus, structural VARs differ from traditional structural models by replacing the assumption of an exogenous instrument set with the assumption of orthogonal structural shocks. At the same time, the dynamics of the system are left unrestricted, as in the conventional VAR. Another interpretation of the structural VAR is as a decomposition of the covariance matrix of VAR residuals. If the structural disturbances are uncorrected with one another, Lc, £(w') = D, Q, the covariance matrix of the VAR errors becomes a nonlinear function of the structural parameters: Q=£(J#Av,v,'A'J#) mBfADA'Bf. If the system is just-identified, the above equality is exact; B^AD™ is a matrix square root of Q, and A'1 B0 diagonalizes Q.18 Reserves, lending, and short-term debt flows Thefirstmodel is a just-identified six-variable system involvingfinancinggap (F)> bank lending, non-financial commercial paper (P) the commercial paperrate(r>), real GDP (x), and non-borrowed reserves adjusted for extended credit (R). The interest rate is differenced, while reserves and GDP enter as log differences. The lending and paper data are again taken from the Flow of Funds accounts for the non-farm, non-financial corporate and noncorporate sectors. With F, P and L expressed 18. With a total of 2A2 elements in A and B0 and only &(&+1)/2 unique elements in Q, it is clear that the stnicmral parameters are not identified without additional restrictions on A and£0. The Cholesky decomposition, which is equivalent to setting B0 = / and making A lower triangular, is but one possibility. In overidentified systems, the problem becomes one of choosing the stnicmral parameters in Bo and A to generate the bestfitbetween thefittedand the observed covariance matrices. -15- Kuttner as shares of the total dollar volume of outstanding paper and loans, changes in the "mix" can be constructed as the weighted average of the two flows: v 'L+P L+P The substance of the model is contained in the six equations describing the contemporaneous relationships between the variables, R as bU6x + vx Reserves (2a) F as bxxR + bz& + V2 Financing gap (2b) rP ss bx\R + bxiF + V3 Interest rate (2c) L as b<xR + b^iF + d<30» + v4 P ss b^R + fci2P+*s^> + *MV4 + v5 x as d 0 r P + 644I + fc^P + v6 Lending (2J) Paper (2e) Output (20. No restrictions are placed on the dynamics of the system; consequently, terms dated t-\ and before are omitted, but implicit. Equation 2a allows the Federal Reserve to vary reserves contemporaneously with real GDP in a primitive feedback relationship. Thefinancinggap (equation 2b) also depends on the level of real economic activity. Consistent with the model presented earlier, the commercial paper rate in 2c is a function of reserves and thefinancinggap. The model's key equations are 2d and 2e, describing the behavior of bank lending and commercial paper flows as a function of thefinancinggap, reserves, and the interest rate. The coefficients on F measure the proportion of the currentfinancinggap satisfiedfinancedthrough loans and paper. The two equations' coefificients on R determine the immediate response, ceteris paribus, of the two forms of short-termfinanceto changes the banking system's reserve position. The v4 term in the lending equation represents lending shocks that are orthogonal to reserve andfinancinggap innovations, which would include factors such as credit crunches. For this interpretation of v4 to be -16- Kuttner legitimate, one of two conditions has to hold: either the observed financing gap must appropriately control for firms' demand for funds, or the amount of funds banks have available is fixed in the current quarter. The v4 innovation also appears in the commercial paper equation with the coefficient a^, allowing commercial paper to respond directly to lending shocks. This parameter determines the extent to which lending shocks are "recycled" into the commercial paper market within the current quarter. The v5 term in the commercial paper equation accounts for shocks to paper issuance uncorrected with the other structural disturbances. The final equation for real GDP is a reducedform equation describing the economy's response to the reserves and credit shocks in the preceding equations. The parameter estimates in Table 3 summarize the model's contemporaneous behavior, while the impulse responses functions plotted in Figure 4 describe its dynamics of the system whose orthogonalization is implicit in equations 2a-2f. Like the earlier reduced-form regressions, these results provide some evidence to support the use of lending flows as an indicator of credit conditions, while confirming the doubts raised in the atheoretical VARs. First, The negative estimate of the coefficient on R in the lending equation (2d) contradicts the hypothesis that the primary effect of monetary policy is a substitution between bank and non-bank finance; the contemporaneous response of an injection of non-borrowed reserves, ceteris paribus, is a fall in bank lending. However, because of the contemporaneous relationship from reserves to the financing gap and short-term finance via the interest rate and output, the coefficients on R in equations 2d and 2e do not by themselves determine the overall response of the "mix" to a reserves shock. The actual responses can be read from the impulse response function, plotted in the top panel of Figure 4.19 This shows that the net effect of a reserves injection is initially rather small, with the loan share gradually rising after two to three quarters. 19. The sample average values of h are used to compute the approximate response of the "mix" from the impulse response functions of the underlying variables. -17- Kuttner Figure 4 also shows that lending shocks seem to have a considerably larger impact on the composition of externalfinancethan reserves for thefirstfour quarters. Lending shocks9 effect is strengthened somewhat by the statistically significant negative estimate offl$4>which is consistent with roughly 10% of the lending shock being "recycled" into the paper market in the current quarter. The coefficients on F in the paper and lending equations show that neither responds immediately to fluctuations in thefinancinggap. A strong liquidity effect is associated with injections of non-borrowed reserves; the paper rate falls contemporaneously (the negative coefficient on R in equation 2c) and over a longer horizon (the center panel of Figure 4). These results also confirm the curious positive relation between the "mix" and the level of interest rates highlighted earlier in the paper. The center panel shows that positive lending innovations imply a rising interest rate, contradicting the theoretical model's implications for the effects of lending shocks. Both monetary and lending shocks are important sources of output fluctuations. Increased bank lending is contemporaneously associated with more rapid real GDP growth in the short run, as shown by both the positive (but not quite significant) coefficient on L and the impulse response function. What about shocks to commercial paper, v5? The top panel of Figure 4 shows that these shocks — which are, by construction, orthogonal to the system's other structural disturbances — have the largest and most persistent impact on the composition of external finance. Interestingly, the center panel shows that these innovations have essentially no implications for the interest rate, although they do seem to have a small, negative impact on real output. Financial flows and interest-rate spreads Recent papers by Bernanke (1990) and Friedman and Kuttner (1992) suggest that the substitution between bank and non-bank debt is an important source of fluctuations in the paper-bill spread. As discussed earlier, monetary contractions reduce lending by shrinking the stock of deposits, leading -18- Kuttner firms to raise the loan rate relative to the paper rate, discouraging intermediated borrowing. Similarly, adverse lending shocks cause banks to shift from loans to Treasury bills. Asfirmsturn to the paper market to satisfy theirfinancingneeds, the paper supply rises and bill supply to households falls, raising the paper-bill spread. If this is the way in which credit conditions affect the spread, one would expect tofindthe mechanism operating through the volume of outstanding non-financial commercial paper. The second structural VAR is designed to detect the operation of this mechanism. It augments thefirstmodel (equations 2a-2f) with the addition of a seventh equation for the paper-bill spread, x a b&rP + b^JL + b^sP + b^(rP - rB) + v6 Output (3/) rP-rB = bitiR + 6 7f2 F+b^rp + 67,4^ + bn,$P + v?, Paper-Bill spread (3g) and also includes the spread in the output equation. The remaining five equations are identical to those in the earlier model (2a-e). The parameter estimates reported in Table 4 provide weak evidence for bank/non-bank substitution as a source of paper-bill spread. The positive and marginally significant on the paper term shows that flows of non-financial paper do exert an influence on the spread.20 However, the very large, significant coefficient on reserves shows indicates that a great deal of the impact of monetary policy is transmitted to the spread via other routes. The impulse responses in the top panel of Figure 5 confirm the spread's strong reaction to non-borrowed reserves innovations. Positive shocks to thefinancinggap also drive up the spread, as predicted, while paper shocks have little or no impact. Lending shocks again pose a problem, however. If the lending innovations identified by the VAR correspond to changes in the availability of loans, the model suggests that positive shocks should be associated with a falling paper-bill spread. The opposite is true: lending shocks imply a rising spread. Again, this pattern is consistent 20. By contrast, the results in Table 10 of Friedman and Kuttner (1992) using the total volume of commercial paper outstanding are consistent with a stronger link between paper issuance and the spread. -19- Kuttner with bank lending responding passively to changes in the demand for funds inadequately captured by the financing gap. The bottom panel of Figure 5 suggests something other than bank/non-bank substitution is driving the paper-bill spread. Despite the inclusion of a variety of financial variables purporting to capture the impact of monetary policy on credit markets, the graph shows that the spread continues have strong implications for future output — comparable in magnitude to those of non-borrowed reserves. Even accounting for reserves, lending, and paper shocks, orthogonal spread innovations still result in falling real economic activity. Identifying lending shocks with loan spread innovations In light of the conclusion that lending flows (and the "mix") may in part represent endogenous response to firms' financing demands, the third structural VAR uses an alternative assumption to identify lending shocks, attributing (orthogonalized) innovations in the loan-paperspread to changes in banks9 willingness to lend. In the context of the simple model presented earlier, the loan spread should embody exactly the same information as the "mix." In practice, as KSW note, the loan rate is likely to be a poor measure of the true cost of bank finance, an observation that motivates their use of the quantity variables. Indeed, the sluggish response of the loan rate to changes in the paper rate corroborates this view. The weak response of output to the loan-paper spread makes this approach seem even less promising. With these reservations in mind, thefirststructural VAR can be adapted to incorporate the loanpaper spread. An equation for the loan spread is added to the system, and lending and paper flows are allowed to depend on this spread, as well as on reserves and the financing gap. The covariation between the flows that is a function of credit conditions is a result of their common dependence on the loan spread. This identification scheme will work if the financing gap is an imperfect proxy for the overall demand for funds so long as banks passively accommodate firms' funding requirements within the quarter at the going spread (that is, if their demand for loans is elastic). -20- Kuttner The modified system is: R s bi& + Vi Reserves'(4fl) F m bZ\R + bZ6x + V2 Financing gap (4b) r> « buR + b^JF + v3 Interest rate (4c) rL-rP = b<tR + b^F+b^rP Loan spread (4rf) I =fcMrt+ 6 ^ + b & r p + 65,4(^1 - rP) • v5 P « 6Mtf + *42F+fc^r/. -•- A^fo. - TP) + **sV5 + v6 * s ^73r/» + ^ f a , - rP) + 67,5^ • th,*? + ^7 Lending (4e) Paper (4/) Output (4g). Under the assumptions outlined above, the innovations to the loan spread equation are now associated with changes in credit conditions, while the v5 lending innovations represent shocks to firms' loan supply (that is, their demand for funds). The parameter estimates in Table 5 accord surprisingly well with the implications of the model. Although its sluggish response makes the loan spread is subject to large, transitory effects from the paper rate and reserves, the negative estimated 65,4 and the positive b$4 show that loan and paper volume respond as they should to the spread. Furthermore, reserves have no discernible independent impact on financial flows. A rising loan spread is contractionary, although again, the effect is statistically weak. The corresponding impulse response functions appear in Figure 6. The top panel again illustrates the consequences of sluggish loan rate adjustment, with reserves injections causing the loan spread to rise sharply in the current quarter. Over time, reserves innovations produce a falling spread. The center panel shows the familiar liquidity effect, and the positive impact of lending innovations on the commercial paper rate. In this model, however, with innovations to loan volume interpreted as shocks to firms' funding requirements, the result is perfectly natural. By contrast, innovations in the loan spread have quite mild effects on the paper rate. -21- Kuttner Conclusions This paper has examined the relationship between monetary policy, loan availability, and alternative indicators of credit market activity. One of is mainfindingsis that the substitution between bank and non-bank finance is indeed an identifiable effect of monetary policy as measured by innovations to non-borrowed reserves. This substitution is, however, not the only factor affecting financial flows. One of the major contributors to the aggregate composition offirms'short-term obligations is flows of commercial paper unrelated to lending shocks. Furthermore, the portion of bank lending not attributable to monetary policy is associated with increases in the commercial paper rate and the paper-bill spread, suggesting that the behavior of the KSW "mix" is in part due to changes in firms' demand for loanable funds. Despite its apparent slow adjustment to changes in market interest rates, the loan-paper spread is a plausible alternative indicator of credit conditions. The paper-bill spread responds appropriately to monetary shocks, rising in response to a reserves contraction. However, the strength of its response cannot entirely be accounted for by flows of non-financial paper, suggesting that its informativeness as a predictor of real economic activity may be due to other sources, such as changes in banks' issuance of negotiable CDs. This is consistent with the observation that non-financial commercial paper comprises a tiny share of the relevant market—only 25% of total commercial paper, and less than 9% of the sum of paper, CDs and Treasury bills.21 Understanding how Federal Reserve policy and credit conditions affect the paper-bill spread will require expanding the model to take into account the behavior of other relevant assets, such as CDs andfinancialpaper. 21. These figures are for 1991:4. The share of non-financial commercial paper is even smaller earlier in the sample. -22- Kuttner References Barro, Robert (1974), "Are Government Bonds Net Wealth?** Journal of Political Economy .82, pp. 1095-1118. Bernanke, Ben S. (1986), "Alternative Explanations of the Money-Income Correlation,** Carnegie Rochester Conference Series on Public Policy 25, pp. 49-100. Bernanke, Ben S. (1990), "On the Predictive Power of Interest Rates and Interest Rate Spreads,** New England Economic Review November-December, pp. 51-68. Bernanke, Ben S. and Cara S. Lown (1991), "The Credit Crunch,** Brookings Papers on Economic Activity 2, pp. 205-39. Blanchard, Olivier, and Mark Watson (1986), "Are Business Cycles All Alike?** in Robert A. Gordon, ed., The American Business Cycle: Continuity and Change. Chicago: The University of Chicago Press and the NBER. Bosworth, Barry and James S. Duesenberry (1973), "A Row of Funds Model and its Implications*' in Issues in Federal Debt Management, Federal Reserve Bank of Boston Conference Series 10, pp. 39-149. Brainard, William C. (1964), "Financial Intermediaries and a Theory of Monetary Control,** Yale Economic Essays 4, pp. 431-82. Foulkc, Roy A. (1931), The Commercial Paper Market New York The Bankers Publishing Company. Friedman, Benjamin M. (1991), "Comments on Bernanke and Lown,** Brookings Papers on Economic Activity 2, pp. 240-44. Friedman, Benjamin M. and Kenneth N. Kuttner (1992), "Why Does the Paper-Bill Spread Predict Real Economic Activity?** forthcoming in James H. Stock and Mark W. Watson eds., New Research in Business Cycles, Indicators and Forecasting, Chicago: University of Chicago Press and the NBER. Hurley, Evelyn (1977), "The Commercial Paper Market,** Federal Reserve Bulletin 63, June, pp. 525-536. Hurley, Evelyn (1982), "The Commercial Paper Market since the Mid-Seventies** Federal Reserve Bulletin 68, June, pp. 327-333. Gertler, Mark and Simon Gilchrist (1992), "The Role of Credit Market Imperfections in the Monetary Transmission Mechanism: Arguments and Evidence,*' Manuscript. Kashyap, Anil, Jeremy C. Stein, and David Wilcox (1992), "Monetary Policy and Credit Conditions: Evidence from the Composition of External Finance,*' NBER Working Paper #4015, Cambridge: National Bureau of Economic Research. King, Stephen R. (1986), "Monetary Transmission: Through Bank Loans or Bank Liabilities?" Journal of Money, Credit and Banking 18, August, pp. 290-303. Lawler, Thomas A. (1978), "Seasonal Movements in Short-term Yield Spreads,*' Federal Reserve Bank of Richmond Economic Review, July/August,. -23- Kuttner Oliner, Stephen D. and Glenn D. Rudebusch (1992), "The Transmission of Monetary Policy to Small and Large Firms," Manuscript. Owens, Raymond E. and Stacey L. Schreft (1992), "Identifying Credit Crunches," Federal Reserve Bank of Richmond Working Paper #92-1. Selden, Richard T. (1963), Trends and Cycles in the commercial Paper Market," National Bureau of Economic Research Occasional Paper #85. Stigum, Marcia (1990), The Money Market Homewood: Dow Jones Irwin. Strongin, Steven H. (1991), "The Identification of Monetary Policy Disturbances: Explaining the Liquidity Puzzle," Federal Reserve Bank of Chicago Working Paper #91-24. -24- Kuttner 1. F-Statistics for Alternative Measures of Credit Conditions in Quarterly Real Output Equations 60:2-91:4 70:3-91:4 75:1-91:4 (1) "Mix" alone 3.36" 2.09* 286" (2) Loan spread alone an 0.30 0.46 (3) Paper-bill spread alone 3.81"* 2.71" 1.81 (4) "Mix" + loan spread "mix" terms loan spread terms 3.37" 0.23 1.81 0.14 3.07" 0.79 (5) "Mix" + paper-ttll spread "mix" terms paper-bill spread terms 4.17"* 4.62'" 2.46 3.07" 4.39"* 3.30" Specification * ** *** Notes: Significant at the 10% level Significant at the 5% level Significant at the 1% level The regressions are based on qtiarterly data for the sample indicated. In addition to the variables indicated, each regression includes four lags of real GDP growth, real non-borrowed reserves growth, the differenced commercial paper rate, plus constant and trend terms. -25- Kuttner 2. Decomposing Changes in the Composition of External Finance (a) Regression with separate commercial paper and bank lending terms Exclusion F-stat (p-value) Commercial paper (AA/>) Bank lending (AAL) Sum of coefficients (p-value) 4.00 (0.005) -0.51 (004) 1.39 (Q24) -0.90 (0.22) (b) Regression with the differenced "mix" and commercial paper Exclusion F-stat (p-value) "Mix" (AA) Commercial paper (iJip) 1.45 (0.22) 264 (0.04) Sum of coefficients (p-value) -0.91 (0.23) -1.38 (0.04) (c) Regression with the "mix" in levels, commercial paper, and linear trend Exclusion F-stat (p-value) "Mix" (h) Commercial paper (A/i/>) Notes: 1.48 (0-21) 227 (0.07) Sum of coefficients (p-value) 0.11 (0.16) -1.77 (0.01) The regressions are based on quarterly data for 1960:2 through 1991:4. The specifications include four lags of each included variable and a constant term. -26- Kuttner 3. Structural VAR Estimates, Credit Conditions Identified via Lending Flows (equations 2a-2f) 2a. 2b. 2c. 2d. 2e. 2f. Notes: R= -0.625 x+ v, (1.94) F= 0.159 r+ 1.974 (1.05) (4.11) />=-0.208 R+ 0.037 (6.96) (210) L = -0.396 * - 0.022 (1.51) (ttl6) P = ttl35 rt + 0.027 (1.29) (0.51) x = -0.023 r+ 0.019 (0.23) (1.50) x+v* F+vj F+ 2125 r, + v4 (3.23) F + 0.444 r>- 0.094 v4 + v5 (1.68) (271) 1 + 0.004 P + v6 (ttl4) Estimates are based on quarterly data for 1960:2 through 1991:4. Regressions include three lags of each variable, constant and trend terms. Numbers in parentheses are /-statistics. -27- Kuttner 4. Structural VAR Estimates of the Effects of Lending Shocks on the Paper-Bill Spread (equations 3a-3g) 3a. 3b. 3c. 3d. 3e. 3f. 3g. Notes: /?=-0.558 x+V! (1.51) F = 0.048 r+ 1.882 (033) (3.70) r, * -0.216 R + 0.027 (7.63) (1.57) L = -0.306 R- 0.022 (1.18) (017) P = 0.110/?+ 0.038 (1.05) (071) x= 0.123 r+ 0.016 (1.20) (1.35) r/,-r,= x+vj F+ v, F+ 2051 r, + v4 (3.05) F+ 0.470 r , - 0.091 v4+ v5 (1.73) (261) 1+ 0.023 P - 0.897 (rP-rB) + v6 (076) (3.20) 0.016/?+ 0.181 r+ 0.000 F+ 0.002 1 + (1.40) (6.20) (007) (046) 0.016 P + v, (1.71) Estimates are based on quarteriy data for 1960:2 through 1991:4. Regressions include three lags of each variable, constant and trend terms. Numbers in parentheses are /-statistics. -28- Kuttner 5. Structural VAR Estimates, Credit Conditions Identified via the Loan Spread (equations 4a-2g) 4a. 4b. 4c. 4d. 4e. 4f. 4gNotes: R= -0.347 x+ v, (0.85) F= 0.122 r+ 2132 x + (a89) (4.46) i> = -0.202 rt + 0.016 F+ (aiO) (0.96) r t - r P = 0.070/?+ 0.014 F (4.99) (1.77) 1 = -0.039 rt + a021 F + (0.14) (ai5) P= 0.030 J? + a003 F + V2 »* a261 (6.56) a783 (0.96) 0.722 (218) (a27) (ao6) x= -0.047 r - 0.362 ( n - rP)-f 0.016 (0.39) (1.54) (1.28) /> + v4 r, - 4.727 (rL - »>) + vs (299) /> + 1.176 (rL-rP)- 0.085 v5 + v6 (1.83) (242) 1+ 0.015 P+ v7 (0.46) Estimates are based on quarterly data for 1960:2 through 1991:4. Regressions include three lags of each variable, constant and trend terms. Numbers in parentheses are /-statistics. -29- Kuttner Figure 1 reserves market Rs Cf«0 l-fp *"f "*"* loan market paper market \yS -30- (I w (HHi) Figure 2 Impulse Response Functions of Credit Conditions Indicators (a) mix -> output (b) reserves -> mix 0.0048 (c) mix -> interest rate 0.0032 0.00361 0.00161 0.0024 0.00121 0.0000 T^-r 0.0000 -.0012 3 6 1 9 1—r- 0 (d) loan spread -> output I I 3 I I I (e) reserves -> loan spread 0.0020 -.0016 6 0.0036 (0 loan spread -> interest rate i CO -.00201 i—i • i i (g) paper-bill spread -> output 0.0016 (h) reserves -> paper-bill spread 0.0007 n 0.0000 -.0016 -.0032 -.0046 .0040 • L -i—i—i—i—i—i—r- r- « i—i 3 i r - t • 9 r • 0) paper-bill spread -> interest rate 0.00501 Figure 3: Financing Gap and Financial Flows bank lending and paper issuance, four-quarter moving average 1 v> • 2 w c •o2 18 -36 .• I ' I ' I ' I ' I ' I ' I ' I • I ' I ' I • I • I • I • I • I • I • I ' I ' I ' I ' I ' I ' I ' I ' I ' I ' I 60 63 66 69 72 75 78 81 84 87 90 Kuttner Figure 4: credit conditions = lending shocks response of the Mix 0.16 0.08 H o.oo -.08 -j reserves lending •.16 -| paper -fin gap— .24 T 1 1 P I I 1 8 9 1 10 11 response of the interest rate 0.035 0.000 .035 H .070 i 1 7 8 11 reserves Ien3ing paper — 0.050 - ^^"^ — — .-.-—.. .-v^—-• 0.025 - 0.000 —' 1 i 10 response of real output n f\7K —i Kj.UfD -.025 - 1 9 ****** . , , **"* , 2 . 3 , 4 , 5 6 - 33 - , 7 , 8 , 9 , 1 10 11 Kixttner Figure 5: credit conditions = lending shocks response of the paper-bill spread 0.008 0.000 -.008 H -.016 l 9 r 10 11 response of real output 0.10 0.05 H 0.00 J~^ \ •.05 1 0 1 1 1 1 2 n 3 4 1 5 1 6 - 3H - 1 7 1 8 1 9 1 1 10 11 Kuttner Figure 6: credit conditions = loan spread shocks response of the loan spread reserves lending paper response of the interest rate s s / " •••••i^>% response of real output - 35 - COMMENTS ON CREDIT CONDITIONS AND EXTERNAL FINANCE: INTERPRETING THE BEHAVIOR OF FINANCIAL FLOWS AND INTEREST RATE SPREADS David Wilcox Two opposing views have animated much recent research on the transmission channels of monetary policy. One view (stated in its extreme form) is that the impulses of monetary policy are transmitted to the real economy exclusively via the market for reserves. By manipulating the quantity of available reserves, the Federal Reserve is able to change the relative supply of money and bonds. Given this change in relative supply, the interest rate must change in order to clear the markets for money and bonds. In turn, the change in the interest rate alters the user cost of capital, and so influences the investment decisions of businesses and the spending decisions of households. An essential assumption implicit in this so-called "money" view of the transmission mechanism is that bank loans, market-intermediated privately-issued debt such as commercial paper and corporate bonds, and privately-held government debt can be treated as perfect substitutes. Indeed, this assumption is embedded in the conventional IS-LM model, where the aggregate non-money financial asset is simply labelled "bonds" for convenience. According to the money view, the reduction in bank loans that accompanies a reduction in reserves is of no particular significance in itself because firms can satisfy any unmet demand for external finance by issuing market-intermediated debt which is indistinguishable from bank debt. For this reason, the money view often is summarized by the proposition that bank loans are not "special." The opposing view of the transmission mechanism assigns a central role to bank loans. According to this view, bank loans, market-intermediated privately-issued debt, and government debt are not perfect substitutes. The reduction in the volume of bank loans that accompanies a move toward a more restrictive monetary policy is 1. David Wilcox is on the staff of the Board of Governors of the Federal Reserve System. Wilcox contractionary in itself, even controlling for any associated change in interest rates. In effect, bank loans behave as if they were a factor of production. A reduction in their availability increases their relative price (the spread between the loan rate and the openmarket rate increases). In response, firms seek cheaper alternatives for their external finance. However, given the imperfect substitutability of other forms of debt for bank loans, the reduction in loan availability implies a contraction in real activity. The important distinction between the money view and the loans view is that the latter implies that the impulses of monetary policy are transmitted not only through the overall level of interest rates, but also through the relative prices and relative quantities of bank loans and other forms of external finance. If the loans view is right, fluctuations in the quantities and prices of bank loans, commercial paper, other private debt, and government debt will be worth keeping track of separately because they will be informative for either the current or future state of the economy, or both. Moreover, the loans view suggests, as Kuttner (this volume) and Friedman (1991) emphasize, that there is no reason for being uniquely interested in changes in the stance of monetary policy; other factors (including but not restricted to the stringency of regulatory oversight) will also be worthy of study to the extent that they bear on loan availability. THE IDENTIFICATION PROBLEM One approach to investigating the empirical significance of the loans channel has been to regress some measure of real activity (such as industrial production or GNP) on current and lagged measures of bank loans. A positive correlation between bank loans and real activity has sometimes been interpreted as contradicting the money view and supporting the existence of a separate loans channel. The flaw in this argument is not hard to spot: A positive correlation between bank loans and real activity could simply reflect an endogenous response of the demand for bank loans to changes in real activity rather than an exogenous cause of changes in real activity. Even a finding of a positive correlation between bank loans and subsequent changes in activity (as opposed to contemporaneous ones) would not be convincing evidence of a separate loans channel; such a phenomenon could reflect, for example, a need to secure financing some months or -2- Wilcox even quarters before the bulk of the associated activity is to take place. An important challange taken up in the more recent literature has been to solve this identification problem in a convincing manner.2 SUMMARY OF KUTTNER'S PAPER Ken Kuttner's paper makes two important contributions to the literature on the monetary policy transmission mechanism: one theoretical, the other empirical. On the theoretical front, he presents a very nice compact model of the flow of funds in a simple economy. He distinguishes five financial instruments in his model (in contrast to the usual two): deposits ("money"), bank loans, commercial paper, reserves, and government debt. He posits the existence of a representive firm, a representative bank, and a representative household, and endows each of them with standard portfolio behavior (households* demand for money is declining in the opportunity cost of holding money, and so forth). Then he derives the implications of changes in the stance of monetary policy, changes in banks' willingness to lend, and changes in firms' demand for external finance for three quantities: the mix of external finance, the spread between the loan rate and the commercial paper rate, and the spread between the paper' rate and the Treasury bill rate. The beauty of Kuttner's model is that it delivers sensible results very directly. For example, a reduction in banks' willingness 4 to lend causes the loan-paper spread to rise. In response, firms 2. The approach proposed in Kashyap, Stein, and Wilcox (1992) is to focus on changes in the composition of external finance rather than fluctuations in any one component alone. Intuitively, one would not expect changes in the volume of bank loans relative to the volume of other debt to be informative for current or future changes in real activity if bank debt is a perfect substitute for non-bank debt. 3. Implicitly, other corporate liabilities such as medium- and long-term bonds are treated as perfect substitutes for commercial paper. 4. Kuttner interprets "negative shifts in X as 'credit crunch' episodes." He notes, however, that a negative shift in X could reflect a "perceived deterioration in borrowers' creditworthiness." In my opinion, it would be more useful to reserve the term "credit (Footnote continues on next page) -3- Wilcox shift the mix of external finance away from bank loans and toward market-mediated debt. The increased issuance of commercial paper drives up the spread between commercial paper rates and bill rates. With respect to these three key variables, the effects of a reduction in banks' willingness to lend are identical to the effects of a move by the Federal Reserve toward a more restictive monetary policy, suggesting that any one of the three might be useful as an index of loan availability. In fact, it turns out that these three variables also respond in qualitatively the same manner to the other two exogenous factors in Kuttner's model (monetary policy and the demand for external finance). That is, no matter what the conceptual experiment being run in Kuttner's model, the loan-paper spread will always move in the same direction as the paper-bill spread, and the two spreads will always move in the opposite direction of the mix. In light of these predictions from his theoretical model, Kuttner's finding that the loan-paper spread significantly underperforms the mix and the paper-bill spread as indicators for future real GNP is interesting and a bit puzzling. Kashyap, Stein, (Footnote continued from previous page) crunch" for periods in which some potential borrowers are turned away even though, with Identical characteristics in every respect (including "credit worthiness"), they would.have been granted credit in "normal" times. 5. In Kuttner's model, the commercial paper rate is taken as the benchmark rate over which the Federal Reserve has direct control in the reserves market. As a result, a reduction in banks' willingness to lend has no effect on the JeveJ of the commercial paper rate. As was noted in the text, however, it does increase the loans-paper spread. As a result, the volume of commercial paper outstanding rises and the paper-bill spread increases. Given the fixity of the paper rate in the face of this experiment, it must be that the bill rate has declined. If the bill rate (rather than the paper rate) were assumed to clear the market for reserves, all the essential results still would hold (the mix would shift away from loans, the loans-paper spread and the paper-bills spread both would rise), but the bill rate would be fixed and the paper rate would rise. 6. Kuttner notes that the effects of a shift in monetary policy are not identical in every respect to the effects of a shift in banks' willingness to lend: The former affects the level of the interest rate in the market for reserves, whereas the latter does not. -4- Wilcox and Wilcox (1992) argued that the mix might be preferable to the loanpaper spread as an indicator of loan availability (because the stated loan rate would not adequately reflect changes in non-price terms of loan contracts such as collateral requirements), but then proceeded to find in their sample that the predictive power of the two variables was roughly comparable. It would be worth attempting to reconcile Kuttner's results with those of KSW, and (assuming Kuttner's results hold up) attempting to verify the KSW hypothesis about why the loanpaper spread might be an inferior performer. On the empirical side, Kuttner's paper introduces a new approach to solving the identification problem. He posits several simple "structual vector autoregression" models of the markets for reserves, bank loans, and commercial paper. Kuttner is bold enough to supply sufficient prior restrictions on the specification of the various equations, and finds that, for the most part the estimates that follow are well in line with the predictions that were outlined in his theoretical section. The major exception--and one that deserves further investigation--is that increases in banks' willingness to lend (counterintuitively) appear to cause Increases in interest rates. AN ASYMMETRIC-INFORMATION-BASED ACCOUNT OF SUBSTITUTION BETWEEN LOANS AND PAPER In line with most of its recent predecessors, Kuttner's paper adopts an aggregate perspective: The model is inhabited by representative banks, households, and non-bank firms, and the empirical work is conducted using aggregate data. As in the earlier papers, this perspective--through no fault of the author--sets up certain tensions of both an expositional sort and a substantive sort. On the expositional side, the most natural way to tell the story of the loans channel involves an appeal to heterogeneity among firms: Some are capable of issuing commercial paper while others are not. Obviously, a story such as this is difficult to link up directly to a model with a single representative non-bank firm. On the substantive side, the representative-agent approach to modelling the problem fuels the intuition that some firms should be observed to be on the margin between bank loans and commercial paper. The purpose of the rest of these comments is to sketch verbally a model that allows for -5- Wilcox heterogeneity among firms, and then to point out two important implications of such an approach. The loans view is predicated on the assertion that non-bank debt is not perfectly substitutable for bank debt. That imperfect substitutability can be motivated as reflecting market imperfections that arise when borrowers have more information about their economic prospects than do prospective lenders. Banks specialize in "information-intensive" lending--that is, in lending to customers (such as small businesses) for whom the asymmetric-information problem is more acute, and hence more difficult for arms-length capital markets to solve. A contractionary shift in the stance of monetary policy will cause banks to reduce the size of their loan portfolios. Banks will tend to cut off their most risky customers and continue to service their most creditworthy ones. Firms that are denied credit by banks may be unable to borrow from any other lender. Certainly, they will not be able to issue debt in arms-length capital markets: nor will they be able to attract financing from other non-bank sources simply by announcing their willingness to pay a higher rate of interest on the debt, because potential lenders will recognize that only the riskiest firms would be willing to offer a higher rate of return. In the end, these firms are likely to be particularly vulnerable to the monetary contraction. After a monetary contraction, a larger fraction of total external finance will be provided via arms-length capital markets and a smaller fraction through bank loans. This change in composition may reflect either (or both) of two factors: First, it may reflect increased issuance of trade credit by large, financially secure firms to their smaller, less creditworthy suppliers. An increase in commercial paper borrowing would be used, in effect, to finance the rise in trade credit. Large firms may be willing to act, in effect. as financial intermediaries because they will have accumulated substantial inside information about the financial stability of their suppliers in the course of having interacted with them before the 7. A lower level of reserves will only support a lower level of deposits. The lower level of deposits (which comprise banks* liabilities) implies that assets will have to decline as well. Given that banks view loans and securities as imperfect substitutes, some of that decline in assets will be absorbed in loans. -6- Wilcox credit crunch. Alternatively, the increase in the share of commercial paper in total external finance may reflect that large firms tend to expand when their smaller rivals are weakened by financial stringency; the large firms take the opportunity to seize some portion of the product market, financing the larger scale of their operations with the increase in commercial paper issuance. These two mechanisms show that bank loans and commercial paper can be substitutes at the aggregate level even though not so for any individual firm. Failure to observe firms operating on the margin between bank loans and other market-mediated debt does not constitute evidence against the heterogeneous-firms version of the loans channel. IMPLICATIONS OF THE ASYMMETRIC-INFORMATION-BASED APPROACH The informal discussion in the previous section points to two important implications for future research. First, the very motivation of banks specializing in information-intensive lending suggests that further progress probably would flow from the analysis of models that allow for heterogeneous non-bank firms. In particular, it seems likely that most such models will imply that, when the Federal Reserve adopts a more restrictive monetary policy, banks will shrink their loan portfolios by refusing credit to their riskiest (least financially stable) customers. Commercial paper issuance will rise because firms already issuing paper will issue more--either to finance their own expanded operations, or to finance the passthrough of trade credit to their suppliers. By contrast, a representative-firm model suggests that all firms should be on the margin between bank debt and commercial paper, and that when the Federal Reserve tightens we should observe a rebalancing of liabilities taking place at the individual firm level. The implausibility of this account is obvious, given that fewer than 1300 firms in the United States have commercial paper programs rated by Moody's. 8. Firms that are growing in size will, at some point, find it possible to issue commercial paper for the first time. If the profitability of commercial paper issuance is an inverse function of bank-loan availability, establishment of commercial paper programs will tend to be bunched into periods immediately following tightenings of monetary policy. Historically, of course, the commercial paper market was not always as well-developed as it is now; as the market deepened and became more efficient, even firms that had been large and creditworthy for a long time established new programs. -7- Wilcox The second implication of the disaggregated approach is that future empirical work should focus on micro-level datasets. Such investigations will be essential for: (1) establishing the identity of bank customers who are denied credit in the wake of a tightening by the Federal Reserve; and (2) establishing the source of the accompanying increase in commercial paper issuance. TABLE OF CONTENTS VOLUME 1 Session 1. Historical Overview Ann-Marie Meulendyke "Federal Reserve Tools in the Monetary Policy Process in Recent Decades" Comments: Robert L. Hetzel Marvin Goodfriend "Interest Rate Policy and the Inflation Scare Problem: 1979-1992" Comments: R. Alton Gilbert Session 2. International Comparisons John Morton and Paul Wood "Interest Rate Operating Procedures of Foreign Central Banks" Bruce Kasman "A Comparison of Monetary Policy Operating Procedures in Six Industrial Countries" Comments (on both John Morton and Paul Wood and on Bruce Kasman' Stephen A. Meyer Robert B. Kahn and Linda S. Kole "Monetary Transmission Channels in Major Foreign Industrial Countries" Comments: Craig S. Hakkio Session 3: Time Series Econometric Issues William Roberds, David Runkle, and Charles H. Whiteman "Another Hole in the Ozone Layer: .Changes in FOMC Operating Procedure and the Term Structure Comments: Glenn D. Rudebusch Charles Evans, Steven Strongin, and Francesca Eugeni "A Policymaker's Guide to Indicators of Economic Activity" Comments: Richard W. Kopcke -2- Session 4: Operating Issues Related to Banking Wilbur John Coleman II, Christian Gilles, and Pamela Labadie "Discount Window Borrowing and Liquidity" Comments: Michael Dotsey Kenneth N. Kuttner "Credit Conditions and External Finance: Interpreting the Behavior of Financial Flows and Interest Rate Spreads" Comments: David Wilcox VOLUME 2 Session 5: Reserve Targeting. Interest Rate Targeting, and Term Structure Volatility Joseph E. Gagnon and Ralph W. Tryon "Price and Output Stability Under Alternative Monetary Policy Rules" Comments: Satyajit Chatterjee Steven Russell "Monetary Policy Experiments in a Stochastic Overlapping Generations Model of the Term Structure" Comments: Eric M. Leeper Jeffrey Fuhrer and George Moore "Inflation Persistence" Comments: John B. Taylor Session 6: Adapting to Regulatory Change Allan D. Brunner and Cara S. Lown "Implementing Short-Run Monetary Policy with Lower Reserve Requirements" Comments: Edward J. Stevens John Wenninger and William Lee "Federal Reserve Operating Procedures and Institutional Change" Comments: Daniel L. Thornton -3- Session 7; Feedback Rules for Monetary Policy John P. Judd and Brian Motley "Controlling Inflation with an Interest Rate Instrument" Comments: Evan F. Koenig Gregory D. Hess, David H. Small, and Flint Brayton "Nominal Income Targeting with the Monetary Base as Instrument: An Evaluation of McCallum's Rule" (with an appendix by Richard D. Porter) Comments: Bennett T. McCallum Summary and Overview John B. Taylor "New Directions in Monetary Policy Research: Comments on the Federal Reserve System's Special Meeting on Operating Procedures" Bennett T. McCallum "Concluding Observations" PRICE AND OUTPUT STABILITY UNDER ALTERNATIVE MONETARY POLICY RULES Joseph E. Gagnon and Ralph W. Tryon This paper is an empirical study of alternative monetary policy regimes in the United States using stochastic simulation of the MX3 multicountry rational-expectations macro model developed by the staff of the Board of Governors. We focus on the implications of interest rate smoothing and incomplete information for the stability of prices, output, and long-term interest rates when the monetary authority targets nominal income. We also conduct a limited number of simulations with a modified version of our model that incorporates staggered real price contracts in the manner of Fuhrer and Moore (1992). The paper builds on the methods and results of an earlier paper (Gagnon and Tryon (1992)) that examined monetary policy rules using stochastic simulations of the MX3 model. There are several findings. First, we confirm our earlier result that the variabilities of prices and output are roughly equal whether the monetary authority targets the monetary base or nominal income, and we obtain confidence intervals for this result. Second, we find that interest rate smoothing provides a significant reduction in interest rate variability with almost no increase in the variability of price and output. Third, it appears that random errors in the observation of the target variable may not significantly increase the variability of price and output. Fourth, while staggered real price contracts tend to increase the size and persistence of price and output deviations, they do not lead to different conclusions about the relative effects of nominal income targetting with and without interest rate smoothing. Finally, we find that the variability of the long-term interest rate is much lower than that of the short-term interest rate in all the monetary regimes studied. 1. Division of International Finance, Board of Governors of the Federal Reserve System. We are grateful to Mark Unferth for very capable assistance in running the simulations and preparing the tables. Gagnon and Tryon STOCHASTIC SIMULATION FRAMEWORK The paper uses stochastic simulations of the MX3 multicountry model to evaluate different monetary policy rules. MX3 is a medium-sized rationalexpectations model of the United States, Japan, Germany, and the rest of 2 the world. We analyze the effectiveness of different monetary policy rules in stabilizing the economy. The policy rules are simple feedback relations between the short-term interest rate and deviations of the target variable from its target value. The target value in each case is the baseline path for the target variable; the baseline path is the deterministic solution for the model. The functional form and the parameters of the policy rules are chosen arbitrarily at plausible values, rather than as the solution to an optimization problem. We simulate the MX3 model for multiple replications of each rule, using random shocks drawn from a joint normal distribution using the estimated covariance matrix of the model residuals for the period 1976-88. The simulation range for each replication is over 20 quarters, from 1989 through 1993. 3 The baseline path is the simulation over the same period without any stochastic shocks, converging toward a steady state. For each replication, we calculate the deviation of each variable of interest, including the (log) levels and growth rates of income, prices, and interest rates, from the baseline values. The root-mean-squared deviation (RMSD) across replications is calculated for each rule; this measure of variability is compared across rules. Comparison of rules Using the RMSD to compare different rules implies that the monetary authority's objective is stated in terms of the second moments, rather than the first moments, of the data. The choice of this objective reflects our conviction that the average levels of real economic variables are invariant to any well-specified monetary policy rule in the long run. 2. For a description of the theory and estimation of the model, see Gagnon (1991). We do not believe that any of our results are strongly dependent on the use of a multicountry model rather than a purely domestic model. Nonetheless, it is a property of the model that foreign responses to U.S. shocks can have feedback effects on the United States through the exchange rate and the trade balance. 3. For a description of the historical residuals and their calculation, see Gagnon and Tryon (1992). - 2- Gagnon and Tryon Although nominal variables do depend on monetary policy, this study ignores the factors involved in choosing a long-run inflation rate and focuses solely on deviations from the long-run rate. The use of second moments as measures of economic performance may be rationalized on two grounds. First, fluctuations of variables around their expected values give rise to adjustment costs as agents adapt their behavior to the new conditions. Second, agents may be risk averse, so that their utility is increased when monetary policy succeeds in reducing the variance of an important variable. Of course it is possible that, by reducing the variance of one variable, policy may increase the variance of some other variable. In conducting the analysis it is necessary to con- sider all of the most important variables. Implicitly or explicitly, policymakers may have to weigh stabilization of one variable against the destabilization of another. The transition from one policy regime to another is likely to involve significant costs as agents learn gradually about the new regime. It would be of interest to consider the problem of making such a regime shift less costly, but we do not pursue that topic here. The assumption behind all the stochastic simulations in this paper is that the regime shift is understood perfectly by the private sector and is fully credible. Thus, comparisons of economic performance across policy regimes reflect differences in the long-run stochastic behavior of the economy and not the short-run transition costs. Number of replications To begin a stochastic simulation, residuals are drawn for one period from a normal distribution with mean zero and the estimated historical variance-covariance matrix. The model is solved in 1989Q1 by using these residuals and the fixed lags and exogenous variables. tions are computed by the Fair-Taylor algorithm. assumed to be z ero. The future expecta- Future residuals are The stochastic solution for 1989Q1 is then used for the necessary lags in solving 1989Q2. In solving 1989Q2, a new draw of residuals is taken from their estimated distribution, but future residuals are again assumed to be zero. This process is repeated for twenty quarters, thus completing one stochastic replication over the baseline Gagnon and Tryon period. Twenty stochastic replications are conducted for each policy 4 rule, for a total of 400 draws of the residuals. In order to make more accurate comparisons across policy rules, we repeated the same sequence of stochastic shocks for each rule. Somewhat to our surprise we found that differences in the computed RMSDs across replications for the same rule were two orders of magnitude greater than differences in the RMSDs across policy rules for the same replication. To test whether the RMSD under one rule differed significantly from the RMSD under another rule, we computed the difference between the two RMSDs for each replication. Assuming that these differences are normally dis- tributed, we were able to test the null hypothesis that their mean value is zero, i.e. that there is no difference in the stability of the given variable under either policy rule. Because of the adjustment lags present in many equations in MX3, the random shocks tend to have persistent effects. Within a given replication, as shocks are drawn in successive periods their effects are combined with the gradually declining effects of earlier shocks. Since each replication begins without the effects of any lagged shocks, the RMSD of most variables increases over the first few years of a replication. In order to focus on the long-run stability of the variables, we computed RMSDs for only the last two years (eight quarters) of each five-year replication. There were four replications for which we were unable to obtain a solution in at least one period for at least one policy rule. In order to obtain 20 complete replications with identical shocks under all rules, we simulated the model for a total of 24 sets of stochastic shocks. Difficulty in solving the model for a particular set of shocks is clearly not independent of the nature of those shocks. more likely to lead to solution problems. Unusually large shocks are Thus, our results may be biased by the exclusion of those replications that could not be solved under all policy rules. 4. These replications were conducted with TROLL 13.1 software using the stochastic simulator package. Each replication requires about 40 minutes of processing (CPU) time on an Amdahl 5850. 5. The RMSDs of a few variables continued to increase mildly over the fourth year of the replications, but we did not believe that the increase was strong enough to warrant discarding an additional year of data. - 4- Gagnon and Tryon The number of replications run for each rule was determined primarily by limits on available computer time. In order to gauge the significance of our results, we report confidence intervals and t-tests calculated on the assumption that the MX3 model is approximately lognormal. This assumption was tested using 100 replications of a single period stochastic disturbance; the deviations from baseline for all variables of interest were checked for normality using the Jarque-fiera test statistic. We were unable to reject the null hypothesis of normality at the 952 confidence level for all variables except consumption and the real exchange rate. To economize on computation time, the Fair-Taylor algorithm was allowed only one type-III iteration over a forecast horizon of twenty quarters. The type-II convergence criterion used was 0.02 percent. In most cases type-II convergence was achieved, but sometimes the solution stops at the iteration limit of 100. A series of test solutions indicated that these restrictions allow reasonably accurate results. MONETARY BASE AND NOMINAL INCOME TARGETTING We begin with two simple alternatives, monetary base and nominal income targetting: (1) RSt - RS* - 1.5 [log(MBt) (2) RSt - RS*t - 1.5 [log(GDPVt) - log(tfB*)] - log(GDPV^)] where RS is the short-term interest rate in decimal form at an annual rate; MB is the monetary base; and GDPV is nominal GDP. Asterisks denote target values. In rule (1) the monetary authorities target the monetary base; in rule (2) the monetary authorities target nominal GDP. Unlike in our previous work, in this paper we implement the monetary policy reaction functions only for the United States: in the other regions monetary policy is assumed to hold the monetary base fixed (on its baseline path). Columns 1 and 2 of Table 1 summarize the results of the stochastic simulations for these rules. The table shows the root-mean-square deviation (RMSD) of each variable from its baseline path over the period 1992:1-1993:4 averaged over the 20 replications. - 5- Below each RMSD is a 90 Gagnon and Tryon percent confidence interval for the true RMSD based on a sample of 20 independent observations. PGDP is the GDP deflator; CAB is the current account balance divided by nominal GDP; REFW is a weighted real exchange rate; RL is the long-term interest rate; C is real private consumption; and MBR is the monetary base deflated by the consumption deflator. The variables are measured in logarithms, except for the interest rates and the ratio of the current account balance to nominal GDP, which are decimal fractions. We also report statistics for the first differences (quarterly growth rates) of some of these variables. Column 1 of Table 2 shows the mean values of the differences in RMSDs for these two policy rules. An asterisk denotes that the difference is significantly different from zero at the 10 percent level; two asterisks denote significance at the 5 percent level. The most striking aspect of these results is that the monetary base and nominal income rules are very similar. As expected, the RMSD of nominal GDP is lower with nominal income targetting (by 0.9 percentage points), and the RMSD of the monetary base is lower when it is targetted (by 0.4 percentage points). (The monetary base rule does not require that the money base target be met exactly, so there is partial accomodation of money demand shocks in this case.) In each case the RMSD of real GDP from baseline is about 6 percentage points, while the RMSD in the growth rate of real GDP is around 2 percentage points. The variability of the growth rate is significantly lower for the nominal income rule, but the magnitude of the difference is slight (0.2 percentage points). The variability of the price level and inflation is essentially the same under both regimes. Thus, the reduction in variability of nominal income is not passed through 6. The confidence intervals are computed under the assumption that the deviations of each variable from baseline are normally distributed. For normal deviations the sample variance follows a chi-square distribution. Although the sample contains 160 observations, we allowed for only 20 degrees of freedom in computing our confidence intervals because the deviations within each of the 20 replications were highly autocorrelated. Our intervals are therefore larger than true 90 percent confidence intervals. 7. The discussion of our results focuses on the RMSDs for output, prices, and interest rates. The real exchange rate and the current account balance are presented as macro variables of general interest. Consumption and real money balances are included because they represent an alternative pair of variables that the monetary authority might wish to stabilize. Generally speaking, consumption variability is highly correlated with output variability. The variability of the real monetary base is less easy to characterize. - 6- Gagnon and Tryon to its components (real output and prices); instead, this regime effectively exploits offsetting variations in the components to meet the target for nominal income. There is a significant difference in the variability of the interest rate: under nominal income targetting the RMSD of the short-term interest rate is higher by 117 basis points and the variability of its first difference is 175 basis points higher. This increased variability is passed through to the long-term interest rate and the real exchange rate. o These results are consistent with our earlier findings. A priori, one might expect nominal income targetting to stabilize prices and output better than monetary base targetting, because (log) nominal income is simply an equal-weighted average of the price level and output in logarithms. The similarity of nominal income and monetary base targetting implies that money demand shocks in MX3 are not too large relative to other disturbances. Since money demand in MX3 is roughly proportional to nominal income, and the adjustment lag is relatively short, it is perhaps not surprising that these two rules have similar effects on prices and output. INTEREST RATE SMOOTHING An important feature of feedback rules of the form used in (1) and (2) is that they can lead to "instrument instability," i.e., substantial variation in short-term interest rates from one period to the next. This is not necessarily a theoretical problem, since the interest rate need not enter directly into private agents' utility or the monetary authority's objective function. However, excessive interest rate (or exchange rate) volatility is sometimes viewed as undesirable in and of itself. To ad- dress this question, we consider a variation of the nominal income rule 8. These results are also robust to changes in the feedback coefficient by a factor of two or three. Gagnon and Tryon (1992) show that increasing the feedback coefficient reduces the RMSD of the target variable and increases the RMSD of the short-term interest rate. A higher feedback coefficient on nominal income does not reduce the RMSD of the price level or output separately. - 7- Gagnon and Tryon that smooths fluctuations in the short-term interest rate. The rule is calibrated so that a persistent deviation in nominal GDP will provide the same interest rate response as rule (2) in the long run, but not in the short run: (3) RSt - RS*t - 0.8 ks t .2 - ^ . j ] + (1-0.8) 1.5 [logiGDPVJ - log(GDPV*)J The results are shown in column 3 of Table 1 and columns 2 and 3 of Table 2. The addition of interest rate smoothing to the nominal income rule reduces the RMSD of the change in the short-term interest rate by 200 basis points. This reduction is not at the expense of any importanc in- crease in volatility in the targets; the RMSD of nominal income rises by only 0.2 percentage points; the changes in the variability of real output and prices are of the same order of magnitude. Nominal income targetting combined with interest rate smoothing produces almost exactly the same results as the monetary base rule, as shown in column 2 of Table 2. The ability of interest rate smoothing to dampen fluctuations in interest rates without substantial increases in the RMSDs of other variables is noteworthy. We believe that this result is due to two basic properties of the MX3 model. First, adjustment lags are quite large throughout the model, implying that the short-run response of the model to monetary policy is much less than the long-run response. Second, the be- havioral equations are forward-looking, so that future monetary policy has a strong impact on current behavior. If the interest smoothing rule is credible, agents believe that a sustained upward shock to nominal income will cause the monetary authority to initiate a series of increases in the short-term interest rate. These expected future increases in the interest 9. Alternatively, this rule could be motivated by a desire to include lagged information in the target. In a model with adjustment lags it is optimal in principle to react to both current and lagged shocks. However, it is infeasible to compute the optimal response pattern to lagged information in a model of this size. 10. This result appears to be robust with respect to the parameters of the smoothing rule. We performed a limited number of trials with different feedback coefficients on the lagged interest rate and nominal GDP; we found that increasing either coefficient always yielded a smaller RMSD of the associated variable at the expense of the other variable. The RMSDs of real output and prices were unaffected in these trials. - 8- Gagnon and Tryon rate will dampen current consumption and investment more than if agents were not forward-looking. OBSERVATION ERRORS Rules (4) and (5) incorporate observation error into rules (2) and (3). These rules are motivated by the fact that the monetary authority cannot accurately observe current nominal GDP. Many components of nominal GDP are observable contemporaneously, but many others are measured only with a lag. We postulate that the monetary authority uses the contemporaneously available indicators of nominal GDP to make an unbiased estimate of current nominal GDP. The contemporaneous indicators may include--but are not limited to--asset prices, commodity prices, interest rate spreads, and in-house surveys of business activity. The error in estimated nominal GDP is captured by c, which is assumed to be normally distributed with zero mean and no autocorrelation. (4) RSt - RS*t - 1.5 [log(GDPVc) + et - logiCDPV^)] (5) RSt - RS*t - 0.8 [*Stml - *S*.2] + (1-0.8) 1.5 [log(GDPVt) + € t - log(GDPV*)J To estimate the magnitude of the monetary authority's observation error, we calculated the difference between the consensus forecast for current quarter nominal GDP growth published in Blue Chip Indicators Economic 12 and BEA's final estimate of nominal GDP growth. The standard deviation of the error (in logarithms) is 0.0068, or 2/3 of a percent. The results are shown in the fourth and fifth columns of Tables 1 and 2. The addition of observation error degrades the ability of the authority to control the target variable, and without interest rate smoothing, the RMSD of nominal income increases by 0.4 percentage points. 11. We believe that the interaction between the private sector and the monetary authority should be modeled symmetrically: if the private sector can respond simultaneously to innovations in the monetary instrument, then the monetary authority should be allowed to react simultaneously to innovations in private variables. We conjecture that dynamic instability of macroeconomic models under some policy rules may be due specifications that do not allow the monetary authority to respond to any contemporaneous information. 12. The forecasts were published at the beginning of the third month of the quarter and were based on information collected and analyzed during the second month. The sample period was 1980:1 through 1988:4. - 9- Gagnon and Tryon The RMSD of the short-term interest rate rises by about 70 basis points, and the RMSD of the real exchange rate also rises, by about 1.5 percentage points. However, none of these differences is statistically significant. With interest rate smoothing, there is virtually no difference between the nominal income rules with and without observation error. This is because with smoothing, the authorities do not respond nearly as much to temporary shocks, and the impact of observation errors is correspondingly reduced. The existence of observation error thus strengthens the case for interest rate smoothing. As in the standard signal extraction problem, the monetary authority's optimal response to an innovation in the target variable is reduced by the presence of noise in the observation. Because the effect of an observation error is much less persistent than the effect of a structural disturbance, a partially delayed response of monetary policy helps to filter out the effect of noise on the policy instrument. REAL PRICE CONTRACTS In another paper presented at this conference, Fuhrer and Moore argue that U.S. macroeconomic time series are better modeled with staggered price contracts in real, as opposed to nominal, terms. In particular, Fuhrer and Moore provide evidence that the U.S. inflation rate is much more persistent in the face of shocks than can be explained by staggered nominal price contracts. We wanted to explore the implications of real price con- tracts in the MX3 model. We were especially interested to see whether our conclusions about nominal income targetting with and without interest rate smoothing are robust to this alternate specification of the model. The real contracting model is described in equations (6)-(8). (6) log(PCTPt) - ajlogCXp + a2logUrt_2) + c ^ l o g U ^ ) + c ^ l o g U ^ ) (7) log(Vt) - o1log(Xt/PGDPt) + + a3log(Xt_2/PGDPt_2) (8) <*2log(Xtml/PGDPtl) + ailog{Xt3/PGDPt3) logU t /PGDP t ) - c^logd^) + « 2 l o g ( V l ) + a 3log(V2) + a 4 log(y t+3 ) + 7(ajlog(CUt) + * 2 log<a/ t+I ) + a3log<0/t+2) + a4log(07t+3)) The GDP deflator, PGDP, is a geometric average of the contract prices, X, that are still in effect in the current period. - 10 - Equation (6) Gagnon and Tryon implies that the longest contract price lasts for four quarters. The coefficients a,- a, sum to unity and equal the proportion of contracts outstanding that were negotiated at times t, t-1, t-2, and t-3, respectively. V is the average real contract price currently in effect. Equation (8) states that the current real contract price depends on the expected future real contract price as well as the expected future level of capacity utilization, CU. The parameter 7 reflects the sensitivity of the real contract price to excess demand. We tried to run stochastic simulations of the model using the coefficient values estimated by Fuhrer and Moore, however, we were unable to complete a single replication with those coefficient values. We were able to complete 20 replications using the coefficients in MX3's nominal price contract equations. The problem appears to be associated with the coeffi- cient -y, which is nearly two orders of magnitude larger in MX3 than in Fuhrer and Moore. The results are displayed in columns 6 and 7 of Tables 1 and 2. The RMSDs of all variables are larger with real price contracts than with nominal price contracts. Although the difference is sometimes quite large, it is never significant. Because of our uncertainty about the ap- propriate coefficients to use, we choose not to focus on the effect of real price contracts per se. We are interested, however, in the comparison of policy rules with and without interest rate smoothing when contracts are written in real terms. Because inflation is more persistent with real price contracts, we were concerned that interest rate smoothing might prove to be destabilizing under real contracts even though it is not destabilizing under nominal contracts. Column 7 of Table 2 demonstrates that this concern appears unwarranted. Only the nominal monetary base shows any large increase in variability, and that increase is not statistically significant. actually declines. Moreover, the RMSD of the real monetary base The short-term interest rate exhibits a large and sig- nificant decrease in variability under interest rate smoothing, just as it did with nominal price contracts. LONG-TERM INTEREST RATES It is of some interest to understand the implications of various monetary policy rules for the behavior of long-term interest rates. If long-term interest rates are determined simply by expectations of future short-term interest rates, we should find that long-term rates are less variable than - 11 - Gagnon and Tryon short-term rates. This conclusion follows from that fact that future shocks are expected to equal zero and that the model is expected to gradually return to baseline in the absence of shocks. In order for long- term rates to be more variable than short-term rates, the model would have to allow for permanent shocks to the inflation rate or to the real interest rate. (Alternatively, we could incorporate an ad hoc risk premium in the long-term interest rate.) The basic MX3 model includes only a one-period interest rate. For this paper, the model was modified to define a long-term interest rate, modeled as an exponentially declining weighted sum of expected future short-term interest rates: CO (9) RL - (1-7) 2 7 1 E [USt+1] - (l-7> RSt + 7 E t [ ^ t + I l i-0 The weights were chosen to approximate a ten-year bond (7 - 0.975, at a quarterly rate). The long-term interest rate does not enter into any other equation in the model, and there is no stochastic term in this definition. The results for the long-term interest rate are shown in the last two rows of both tables. As expected from the experimental design, the variability of the long-term rate is in all cases substantially less than the short-term rate. CONCLUSIONS This paper contains three main findings. First, in the MX3 model the variabilities of prices and output are roughly equal whether the monetary authority targets the monetary base or nominal income. Second, interest rate smoothing provides a significant reduction in interest rate variability with almost no increase in the variability of prices and output, and this conclusion is not affected by a modification of the model that increases the persistence of the inflation rate. Third, it appears that random errors in the observation of a nominal income target may not significantly increase the variability of prices and output. 13. The consumption and investment equations depend on the long-term interest rate implicitly as a function of expected future short-term rates. - 12 - Gagnon and Tryon The result that monetary base and nominal income targetting are bi^adly equivalent is robust within the framework of our model, but may not be entirely conclusive. A priori, there are reasons to prefer nominal income targetting, because nominal income is closer to the ultimate goals of policy than is the money supply. This paper does not resolve the issue of the optimal target variable(s) for monetary policy, nor does it derive optimal coefficients for the rules considered here. Calculation of op- timal rules would be prohibitively expensive with our model, and the conclusions would be particularly sensitive to specification and estimation errors in the model. We believe that our other two findings are robust to the specification of the model and policy rules. The ability of the monetary authority to smooth interest rates to a significant extent without destabilizing other variables depends mainly on the existence of forward-looking agents and adjustment costs in economic activity. The result that observation error does not significantly affect the outcomes under different rules depends on adjustment lags and on the relatively small size of the observation error. Since we were able to measure the observation error directly, we have a high degree of confidence in our finding that it does not destabilize real output and prices significantly. - 13 - Gagnon and Tryon 1. Summary of Stochastic Simulations of Policy Rules Rule 1: Variables m Rule 2: GDPV Rule 3: Rule 4: RS & GDPV GDPV & e GDP 0.0730 (0.058-0.098) 0.074 (0.060-0.100) 0.075 (0.060-0.102) 0.069 (0.055-0.094) AGDP 0.022 (0.018-0.030) 0.020 (0.016-0.027) 0.022 (0.017-0.029) 0.021 (0.017-0.028) PGDP 0.087 (0.069-0.117) 0.086 (0.069-0.117) 0.084 (0.067-0.113) 0.087 (0.069-0.117) APGDP 0.014 (0.011-0.020) 0.014 (0.011-0.019) 0.014 (0.011-0.018) 0.014 (0.011-0.019) GDPV 0.037 (0.030-0.050) 0.027 (0.021-0.036) 0.029 (0.023-0.039) 0.035 (0.028-0.048) MB 0.020 (0.016-0.030) 0.023 (0.019-0.032) 0.031 (0.025-0.042) 0.030 (0.024-0.041) CAB 0.004 (0.003-0.005) 0.003 (0.003-0.005) 0.003 (0.003-0.005) 0.004 (0.003-0.005) RS 0.030 (0.023-0.040) 0.040 (0.032-0.054) 0.026 (0.020-0.035) 0.053 (0.042-0.071) ARS 0.010 (0.008-0.013) 0.028 (0.022-0.038) 0.007 (0.006-0.010) 0.034 (0.027-0.046) RL 0.004 (0.003-0.006) 0.005 (0.004-0.007) 0.004 (0.003-0.006) 0.007 (0.006-0.010) ARL 0.002 (0.001-0.002) 0.002 (0.002-0.003) 0.001 (0.001-0.002) 0.003 (0.003-0.005) RERW 0.130 (0.103-0.175) 0.130 (0.104-0.177) 0.130 (0.104-0.176) 0.157 (0.125-0.213) ARERW 0.058 (0.046-0.079) 0.063 (0.050-0.085) 0.060 (0.048-0.082) 0.072 (0.057-0.097) C 0.061 (0.049-0.083) 0.062 (0.050-0.084) 0.623 (0.050-0.084) 0.067 (0.054-0.091) AC 0.016 (0.012-0.021) 0.015 (0.012-0.021) 0.015 (0.012-0.021) 0.016 (0.013-0.022) MBR 0.077 (0.061-0.104) 0.087 (0.069-0.118) 0.078 (0.062-0.106) 0.090 (0.072-0.122) AMBR 0.013 (0.010-0.018) 0.015 (0.012-0.020) 0.013 (0.011-0.018') 0.016 (0.013-0.022) - 14 - Gagnon and Tryon Summary of Stochastic Simulations of Policy Rules (cont'd) Root-mean-squared deviation from baseline Rule 5: Rule 6: Rule 7: Variables GDPV & RS & e GDPV & FM GDP 0.075 (0.060-0.101) 0.134 (0.107-0.182) 0.139 (0.111-0.188) AGDP 0.022 (0.017-0.029) 0.025 (0.020-0.034) 0.028 (0.022-0.037) PGDP 0.840 (0.067-0.113) 0.161 (0.128-0.218) 0.156 (0.124-0.211) APGDP 0.014 (0.011-0.018) 0.026 (0.021-0.036) 0.027 (0.021-0.036) GDPV 0.029 (0.023-0.039) 0.047 (0.038-0.064) 0.063 (0.050-0.085) MB 0.031 (0.024-0.041) 0.035 (0.028-0.048) 0.058 (0.047-0.079) CAB 0.003 (0.002-0.005) 0.004 (0.003-0.006) 0.004 (0.003-0.005) RS 0.025 (0.020-0.034) 0.071 (0.056-0.096) 0.056 (0.045-0.076) ARS 0.008 (0.006-0.010) 0.031 (0.025-0.042) 0.013 (0.011-0.018) RL 0.004 (0.003-0.006) 0.010 (0.008-0.013) 0.010 (0.008-0.013) ARL 0.001 (0.001-0.002) 0.003 (0.002-0.004) 0.003 (0.002-0.003) RERW 0.132 (0.110-0.178) 0.153 (0.122-0.207) 0.164 (0.131-0.222) ARERW 0.061 (0.048-0.082) 0.064 (0.051-0.087) 0.062 (0.050-0.084) C 0.062 (0.049-0.083) 0.101 (0.080-0.136) 0.101 (0.081-0.137) AC 0.016 (0.012-0.021) 0.020 (0.016-0.027) 0.021 (0.017-0.028) MBR 0.076 (0.061-0.104) 0.157 (0.125-0.212) 0.141 (0.112-0.191) AMBR 0.013 (0.010-0.018') 0.027 (0.021-0.036') 0.024 (0.019-0 .OSS') - 15 - GDPV & RS & FM Gagnon and Tryon 2. Differences In RMSDs across Policy Rules GDP -0.0011 -0.0022 -0.0011 0.0026 AGDP 0.0018* 0.0004 -0.0014* -0.0006 PGUP 0.0009 0.0029 0.0020 -0.0001 APGDP 0.0004 0.0006 0.0002 -0.0001 GDPV 0.0092 0.0074 -0.0018 -0.0039 -0.0036 -0.0106 -0.0070 -0 0042 0.0001 0.0001 -0.0001 -0.0002 RS -0.0117 0.0026 0.0143* -0.0067 ARS -0.0175** 0.0025 0.0200** -0.0059 RL -0.0008 -0.0002 0.0006 -0.0009 ARL -0.0007** -0.0001 0.0006** -0.0003 RERW -0.0016 -0.0005 0.0011 -0.0148 ARERW -0.0049* -0.0024 0.0026 -0.0057 C -0.0007 -0.0008 -0.0001 -0.0020 0.0003 0.0001 -0.0003 -0.0008 MBR -0.0086 -0.0001 0.0085 -0.0010 AMBR -0.0017 -0.0002 0.0015 -0.0012 MB CAB AC * significant at 10 percent level, ** significant at 5 percent level. - 16 - Gagnon and Tryon 2. Differences In BMSDs across Policy Rules (cont'd) Variables Rule 3-Rule 5 Rule 2-Rule 6 Rule 6-Rule 7 GDP 0.0003 -0.0497 -0.0044 AGDP 0.0001 -0.0047 -0.0021 PGDP 0.0001 -0.0626 0.0031 LPGDP -0.0000 -0.0104 0.0000 GDPV -0.0000 -0.0172 -0.0089 MB 0.0003 -0.0093 -0.0213 CAB 0.0000 -0.0008 0.0004 RS 0.0002 -0.0259 0.0171* -0.0004 -0.0028 0.0182** RL 0.0002 -0.0035 0.0009 LRL 0.0002 -0.0006 0.0005 RERW -0.0011 -0.0169 -0.0045 LRERU -0.0003 -0.0008 0.0020 C 0.0006 -0.0320 0.0000 LC 0.0000 -0.0037 -0.0007 HBR 0.0014 -0.0586 0.0148 -0.0099 0.0024 &RS MBR . o.oooi * significant at 10 percent level. ** significant at 5 percent level. - 17 - Gagnon and Tryon APPENDIX: Some Detailed Results Tables Al through A5 present more detail on the results of our stochastic simulations for a small subset of the policy rules. The top panel of Table Al shows the sample autocorrelations of the deviations of several variables from their baseline values under nominal income targetting (Rule 2). All of the variables have a high degree of autocorrelation in the levels, but none of them have highly autocorrelated growth rates (first differences). This autocorrelation is due to the presence of adjustment lags in most of the model's equations. The autocorrelation is par- ticularly high for the price level, real money balances, and output. The bottom panel of Table Al presents statistics relating to the sample distribution of the deviations. These statistics were computed 14 from 100 replications of a one-quarter solution. Of the variables tested, only the weighted real exchange rate and consumption reject our null hypothesis of normality. Table A2 displays the RMSDs for each quarter, computed across 20 replications. For the levels of the variables, the RMSDs tend to increase over successive quarters. In most cases the RMSD appears to stabilize after 12 quarters, but there is still some slight increase in the RMSDs for real output and the price level after 12 quarters. For the dif- ferences of the variables, the RMSDs appear quite stable over time. The average RMSDs differ from those presented in Table 1 because they were computed using observations from all 20 quarters and 20 replications, rather than the last eight quarters of 20 replications. Table A3 contains the RMSDs for each replication, computed across 20 quarters. This table shows how different the results were for each set of stochastic shocks. Tables A4 and A5 show that the differences in RMSDs across rules are quite small, despite the large differences in RMSDs across quarters and across replications that are documented in Tables A2 and A3. 14. The model was solved for only one quarter because the persistence of deviations in the model implies that the distribution of a variable in a given quarter is dependent on the number of quarters that have been simulated stochastically prior to the given quarter. Hence, if we had included observations from different quarters we would have been sampling from populations with different distributions. As the number of stochastically simulated quarters increases, the distributions of the observed deviations should approach a stationary distribution. - 18 - Gagnon and Tryon Al. Properties of Simulated Deviations from Baseline Rule 2: GDPV First through Fifth Order Autocorrelation Variables GDP LGDP PGDP LPGDP GDPV MB CAB RS ARS RL ARL RERW LRERW C AC MBR AMBR 1 0.976 -0.161 0.997 0.862 0.618 0.617 0.867 0.618 0.230 0.798 -0.108 0.775 0.269 0.964 0.015 0.997 0.592 2 4 3 0.962 0.201 0.991 0.704 0.265 0.062 0.721 0.265 -0.247 0.708 -0.076 0.506 -0.042 0.925 -0.125 0.992 0.432 5 0.963 0.023 0.979 0.650 0.315 -0.274 0.591 0.315 0.130 0.848 -0.399 0.504 -0.181 0.943 -0.322 0.988 0.478 0.966 0.094 0.983 0.640 0.240 -0.337 0.605 0.240 0.076 0.722 0.215 0.325 -0.059 0.943 -0.050 0.989 0.346 0.950 0.189 0.979 0.723 0.396 -0.409 0.487 0.396 0.077 0.730 0.116 0.595 -0.186 0.911 -0.042 0.986 0.579 Test of Normality' Variables GDP PGDP MB RS RL CAB RERW C MBR Statistic P-value 2.64 2.69 3.23 2.93 1.70 1.51 8.16* 9.39* 0.28 (.733) (.740) (.801) (.769) (.573) (.530) (.983) (.991) (.128) Skewness Excess Kurtosis -.396 -.136 .410 -.383 -.027 -.229 .697 -.745 .114 * Reject normality at 5 percent significance level. Autocorrelations computed from 20 replications of 20 quarters. Normality tests based on 100 replications of one quarter. - 19 - -.075 -.756 -.320 -.342 -.637 -.391 .120 .191 .120 Gagnon and Tryon A2. RMSD by Quarter Across 20 Replications Rule 1: MA GDP AGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR GDP AGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR 89 1 89 2 89 3 89 4 90 1 90 2 90 3 90 4 0.0179 0.0185 0.0046 0.0068 0.0186 0.0037 0.0019 0.0055 0.0086 0.0012 0.0014 0.0513 0.0552 0.0134 0.0134 0.0071 0.0055 0.0154 0.0185 0.0105 0.0068 0.0121 0.0058 0.0019 0.0086 0.0086 0.0013 0.0014 0.0552 0.0552 0.0123 0.0134 0.0083 0.0055 0.0234 0.01670.0195 0.0101 0.0181 0.0085 0.0021 0.0127 0.0097 0.0019 0.0015 0.0870 0.0606 0.0233 0.0170 0.0149 0.0094 0.0375 0.0204 0.0291 0.0106 0.0343 0.0125 0.0028 0.0188 0.0107 0.0023 0.0017 0.1141 0.0590 0.0356 0.0179 0.0222 0.0096 0.0280 0.0207 0.0384 0.0104 0.0289 0.0132 0.0020 0.0198 0.0086 0.0019 0.0015 0.1102 0.0642 0.0338 0.0145 0.0278 0.0105 0.0359 0.0192 0.0455 0.0095 0.0265 0.0150 0.0020 0.0225 0.0083 0.0015 0.0014 0.0921 0.0545 0.0341 0.0145 0.0326 0.0095 0.0442 0.0160 0.0495 0.0100 0.0301 0.0137 0.0028 0.0206 0.0085 0.0022 0.0019 0.1075 0.0565 0.0423 0.0150 0.0396 0.0089 0.0487 0.0254 0.0531 0.0119 0.0286 0.0154 0.0032 0.0231 0.0110 0.0029 0.0018 0.1085 0.0698 0.0449 0.0183 0.0435 0.0095 91 1 91 2 91 3 91 4 92 1 92 2 92 3 92 4 0.0443 0.0235 0.0577 0.0125 0.0314 0.0157 0.0035 0.0235 0.0100 0.0025 0.0013 0.1312 0.0722 0.0434 0.0143 0.0479 0.0104 0.0536 0.0236 0.0638 0.0119 0.0336 0.0173 0.0030 0.0260 0.0096 0.0024 0.0017 0.1363 0.0562 0.0491 0.0183 0.0545 0.0115 0.0582 0.0244 0.0690 0.0109 0.0338 0.0162 0.0030 0.0243 p.0075 0.0028 0.0015 0.1342 0.0586 0.0529 0.0190 0.0597 0.0119 0.060C 0.0194 0.0736 0.0104 0.0313 0.0157 0.0024 0.0236 0.0092 0.0026 0.0016 0.1449 0.0448 0.0541 0.0122 0.0643 0.0079 0.0666 0.0197 0.0757 0.0113 0.0319 0.0151 0.0027 0.0226 0.0090 0.0035 0.0017 0.1335 0.0486 0.0596 0.0154 0.0684 0.0098 0.0656 0.0233 0.0773 0.0131 0.0310 0.0147 0.0029 0.0220 0.0103 0.0037 0.0015 0.1438 0.0509 0.0580 0.0126 0.0704 0.0113 0.0687 0.0208 0.0793 0.0144 0.0353 0.0165 0.0030 0.0247 0.0107 0.0043 0.0012 0.1200 0.0627 0.0587 0.0148 0.0731 0.0133 0.0671 0.0282 0.0816 0.0139 0.0316 0.0198 0.0028 0.0297 0.0090 0.0042 0.0014 0.1297 0.0508 0.0589 0.0153 0.0736 0.0103 Gagnon and Tryon A2. RMSD by Quarter Across 20 Replications (cont'd) Rule 1: MA GDP AGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR 93 1 93 2 93 3 93 4 Avg 0.0688 0.0142 0.0847 0.0154 0.0366 0.0209 0.0034 0.0314 0.0085 0.0045 0.0014 0.1340 0.0542 0.0570 0.0145 0.0728 0.0138 0.0714 0.0259 0.0908 0.0160 0.0391 0.0219 0.0044 0.0328 0.0110 0.0038 0.0020 0.1331 0.0678 0.0563 0.0178 0.0762 0.0172 0.0819 0.0222 0.0963 0.0138 0.0417 0.0222 0.0045 0.0333 0.0085 0.0039 0.0017 0.1280 0.0623 0.0674 0.0175 0.0839 0.0151 0.0875 0.0181 0.1032 0.0154 0.0453 0.0231 0.0039 0.0347 0.0112 0.0043 0.0014 0.1044 0.0637 0.0707 0.0155 0.0909 0.0137 0.0561 0.0212 0.0662 0.0120 0.0319 0.0161 0.0030 0.0242 0.0095 0.0031 0.0016 0.1178 0.0588 0.0490 0.0157 0.0574 0.0111 Gagnon and Tryon A3. RMSD by Replication Across 20 Quarters Rule 1: MA GDP AGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR GDP AGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR TR1 TR2 TR3 TR4 TR5 TR6 TR7 TR8 0.0436 0.0173 0.0694 0.0081 0.0347 0.0205 0.0034 0.0307 0.0072 0.0028 0.0017 0.1824 0.0518 0.0285 0.0118 0.0465 0.0087 0.0302 0.0195 0.0355 0.0094 0.0176 0.0115 0.0018 0.0172 0.0084 0.0028 0.0016 0.0729 0.0432 0.0271 0.0139 0.0338 0.0095 0.0573 0.0221 0.0545 0.0098 0.0235 0.0126 0.0024 0.0189 0.0096 0.0033 0.0015 0.0575 0.0471 0.0526 0.0149 0.0547 0.0112 0.0291 0.0163 0.0470 0.0120 0.0316 0.0136 0.0018 0.0204 0.0100 0.0017 0.0014 0.0835 0.0512 0.0198 0.0155 0.0298 0.0097 0.0490 0.0233 0.0618 0.0107 0.0295 0.0106 0.0026 0.0159 0.0079 0.0023 0.0015 0.0911 0.0571 0.0569 0.0178 0.0641 0.0101 0.0285 0.0167 0.0381 0.0121 0.0212 0.0143 0.0032 0.0214 0.0081 0.0030 0.0016 0.0814 0.0552 0.0326 0.0136 0.0402 0.0110 0.1074 0.0194 0.1342 0.0117 0.0358 0.0188 0.0019 0.0282 0.0071 0.0026 0.0013 0.1150 0.0594 0.0912 0.0143 0.1144 0.0114 0.0314 0.0217 0.0107 0.0060 0.0259 0.0077 0.0021 0.0115 0.0062 0.0030 0.0019 0.1379 0.0485 0.0304 0.0118 0.0168 0.0063 TR 9 TR 10 TR 11 TR 12 TR 13 TR 14 TR 15 TR 16 0.0239 0.0244 0.0375 0.0112 0.0401 0.0135 0.0050 0.0202 0.0102 0.0022 0.0014 0.1354 0.0655 0.0408 0.0133 0.0373 0.0110 0.0538 0.0210 0.0727 0.0089 0.0333 0.0093 0.0043 0.0140 0.0107 0.0025 0.0018 0.1181 0.0732 0.0713 0.0186 0.0755 0.0087 0.0809 0.0183 0.0780 0.0110 0.0254 0.0145 0.0020 0.0218 0.0083 0.0036 0.0016 0.1447 0.0633 0.0631 0.0131 0.0697 0.0099 0.0447 0.0192 0.0404 0.0129 0.0310 0.0134 0.0033 0.0201 0.0114 0.0027 0.0015 0.0761 0.0559 0.0262 0.0112 0.0331 0.0080 0.0614 0.0144 0.0714 0.0179 0.0311 0.0173 0.0016 0.0260 0.0117 0.0048 0.0016 0.0689 0.0473 0.0415 0.0161 0.0591 0.0118 0.0346 0.0332 0.0309 0.0064 0.0374 0.0087 0.0032 0.0131 0.0091 0.0022 0.0015 0.2531 0.0794 0.0311 0.0184 0.0238 0.0097 0.0195 0.0224 0.0275 0.0108 0.0248 0.0127 0.0020 0.0191 0.0101 0.0022 0.0015 0.1003 0.0657 0.0244 0.0182 0.0230 0.0106 0.0606 0.0197 0.0754 0.0123 0.0257 0.0179 0.0016 0.0268 0.0090 0.0020 0.0013 0.0724 0.0571 0.0523 0.0145 0.0634 0.0127 Gagnon and Tryon A3. RMSD by Replication Across 20 Quarters (cont'd) Rule 1: MA GDP AGDP PGDP ^PGDP CDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR TR 17 TR 18 TR 19 TR 20 0.0258 0.0191 0.0290 0.0108 0.0319 0.0137 0.0026 0.0206 0.0114 0.0022 0.0016 0.0965 0.0585 0.0303 0.0157 0.0190 0.0086 0.0590 0.0236 0.0833 0.0201 0.0537 0.0298 0.0031 0.0447 0.0127 0.0045 0.0013 0.0832 0.0741 0.0479 0.0211 0.0579 0.0182 0.0665 0.0277 0.0701 0.0164 0.0243 0.0227 0.0052 0.0341 0.0082 0.0029 0.0016 0.0910 0.0621 0.0561 0.0182 0.0576 0.0164 0.1033 0.0167 0.1162 0.0124 0.0398 0.0217 0.0028 0.0326 0.0093 0.0053 0.0017 0.1186 0.0433 0.0774 0.0174 0.1046 0.0121 Avg 0.0561 0.0212 0.0662 0.0120 0.0319 0.0161 0.0030 0.0242 0.0095 0.0031 0.0016 0.1178 0.0588 0.0490 0.0157 0.0574 0.0111 Gagnon and Tryon A4. Difference in RMSD Between Rules 1 and 2 Across Replications GDP LGDP PGUP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR GDP LGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR 89 1 89 2 89 3 89 4 90 1 90 2 0.0024 0.0018 0.0001 0.0001 0.0027 -0.0079 0.0002 -0.0184 -0.0132 -0.0007 -0.0008 -0.0065 -0.0067 0.0006 0.0003 -0.0022 -0.0006 0.0012 0.0018 0.0003 0.0001 0.0023 -0.0088 0.0002 -0.0060 -0.0132 -0.0004 -0.0008 -0.0030 -0.0067 0.0004 0.0003 -0.0037 -0.0006 0.0020 0.0025 0.00030.0002 0.0037 -0.0114 0.0003 -0.0089 -0.0121 -0.0007 -0.0008 -0.0051 -0.0062 0.0009 0.0007 0.0013 0.0003 0.0044 0.0031 0.0004 0.0001 0.0063 -0.0170 0.0006 -0.0232 -0.0179 -0.0013 -0.0007 -0.0098 -0.0066 0.0020 0.0011 0.0020 -0.0009 0.0010 0.0019 0.0004 0.0001 0.0052 -0.0095 0.0002 -0.0158 -0.0207 -0.0008 -0.0010 -0.0045 -0.0066 0.0019 0.0004 0.0009 -0.0024 0.0004 0.0018 0.0003 0.0001 0.0049 -0.0074 -0.0000 -0.0099 -0.0156 -0.0005 -0.0007 -0.0018 -0.0056 0.0014 0.0004 -0.0010 -0.0006 91 1 91 2 91 3 91 4 92 1 92 2 92 3 92 4 -0.0002 0.0021 0.0003 0.0004 0.0058 -0.0059 0.0004 -0.0148 -0.0221 -0.0008 -0.0008 -0.0056 -0.0044 0.0005 0.0005 -0.0029 -0.0019 -0.0001 0.0026 0.0004 0.0004 0.0071 -0.0049 0.0003 -0.0138 -0.0208 -0.0010 -0.0009 -0.0038 -0.0085 -0.0001 0.0008 -0.0036 -0.0029 0.0002 0.0024 0.0005 0.0003 0.0085 -0.0098 0.0004 -0.0136 -0.0245 -0.0012 -0.0008 -0.0050 -0.0078 0.0000 0.0006 -0.0046 -0.0015 -0.0004 0.0014 0.0004 0.0001 0.0086 -0.0068 0.0000 -0.0104 -0.0153 -0.0007 -0.0006 -0.0001 -0.0052 -0.0002 0.0004 -0.0054 -0.0025 0.0002 0.0020 0.0001 0.0003 0.0084 -0.0052 0.0002 -0.0127 -0.0157 -0.0008 -0.0006 0.0005 -0.0063 -0.0002 0.0005 -0.0052 -0.0023 -0.0007 0.0019 -0.0002 0.0003 0.0084 -0.0040 0.0001 -0.0119 -0.0198 -0.0006 -0.0005 -0.0004 -0.0050 -0.0004 0.0004 -0.0072 -0.0019 -0.0008 0.0018 -0.0002 0.0005 0.0082 0.0003 -0.0002 -0.0160 -0.0206 -0.0009 -0.0005 -0.0007 -0.0043 -0.0008 0.0001 -0.0093 -0.0031 -0.0017 0.0026 -0.0000 0.0005 0.0077 -0.0010 -0.0000 -0.0062 -0.0204 -0.0005 -0.0010 0.0001 -0.0054 -0.0011 0.0005 -0.0111 0.0012 90 "* 0.0014 0.0019 0.0002 0.0002 0.0053 -0.0105 0.0002 -0.0165 -0.0152 -0.0008 -0.0007 -O.OO'-l -0.0048 0.0012 0.0006 -0.0011 -0.0023 90 4 0.0016 0.0032 0.0002 0.0004 0.0058 -0.0083 0.0003 -0.0111 -0.0235 -0.0011 -0.0011 -0.0053 -0.0099 0.0010 0.0012 0.0003 -0.0007 Gagnon and Tryon A4. Difference in RKSD Betveen Rules 1 and 2 Across Replications (cont'd) GDP LGDP PGDP LPGDP GDPV MB CAB RS &RS RL ARL RERW ARERW C AC MBR AMBR 93 1 93 2 93 3 93 4 -0.0017 0.0011 0.0002 0.0005 0.0094 -0.0014 0.0002 -0.0094 -0.0133 -0.0009 -0.0006 -0.0006 -0.0029 -0.0011 0.0003 -0.0120 -0.0019 -0.0024 0.0022 0.0006 0.0004 0.0115 -0.0046 0.0006 -0.0086 -0.0203 -0.0009 -0.0008 -0.0025 -0.0078 -0.0016 0.0005 -0.0140 -0.0007 -0.0015 0.0020 0.0007 0.0003 0.0132 -0.0062 0.0003 -0.0094 -0.0210 -0.0012 -0.0008 -0.0049 -0.0046 -0.0015 0.0002 -0.0118 -0.0018 -0.0018 0.0009 0.0008 0.0003 0.0137 -0.0070 0.0003 -0.0127 -0.0124 -0.0011 -0.0003 -0.0036 -0.0036 -0.0015 -0.0001 -0.0111 -0.0033 - 25 - Avg -0.0003 0.0021 0.0003 0.0003 0.0077 -0.0066 0.0002 -0.0121 -0.0181 -0.0008 -0.0008 -0.0031 -0.0059 -0.0002 0.0005 -0.0064 -0.0016 Gagnon and Tryon A5. Difference in RMSD Between Rules 1 and 2 Across Quarters TR1 GDP LGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR TR3 TR4 TR5 TR6 -0.0034 -0.0008 0.0020 -0.0007 -0.0011 -0.0008 0.0017 0.0021 0.0025 0.0015 0.0027 0.0020 0.0036 0.0001 -0.0012 -0.0012 0.0032 0.0018 0.0004 0.0004 0.0002 0.0002 0.0006 0.0005 0.0086 0.0029 0.0050 0.0053 0.0078 0.0044 0.0065 -0.0101 -0.0045 -0.0079 -0.0208 -0.0084 -0.0003 0.0005 0.0001 -0.0001 0.0008 0.0002 -0.0085 -0.0049 -0.0087 -0.0191 -0.0166 -0.0038 -0.0160 -0.0185 -0.0186 -0.0118 -0.0220 -0.0137 -0.0008 -0.0004 -0.0008 -0.0010 -0.0005 -0.0005 -0.0005 -0.0008 -0.0007 -0.0007 -0.0008 -0.0009 -0.0044 -0.0026 -0.0045 -0.0080 0.0002 -0.0033 -0.0044 -0.0053 -0.0053 -0.0047 -0.0079 -0.0055 -0.0018 -0.0006 0.0009 0.0004 -0.0012 0.0000 0.0005 0.0002 0.0006 0.0004 0.0008 0.0003 -0.0128 -0.0077 0.0021 -0.0169 0.0019 0.0007 -0.0006 -0.0006 -0.0026 -0.0059 -0.0039 -0.0015 TR 9 GDP AGDP PGDP APGDP GDPV MB CAB RS ARS RL ARL RERW ARERW C AC MBR AMBR TR2 0.0014 0.0025 0.0054 0.0007 0.0142 -0.0047 0.0005 -0.0186 -0.0221 -0.0014 -0.0008 0.0045 -0.0062 -0.0029 0.0006 -0.0081 -0.0035 TR 10 TR 11 -0.0055 0.0027 0.0022 0.0010 0.0052 -0.0006 0.0005 0.0002 0.0115 0.0053 -0.0199 0.0038 0.0006 -0.0001 -0.0187 -0.0085 -0.0176 -0.0146 -0.0002 -0.0007 -0.0007 -0.0004 -0.0007 0.0034 -0.0052 -0.0040 -0.0030 0.0014 0.0005 0.0003 -0.0088 0.0031 -0.0019 0.0002 TR7 TR8 -0.0047 0.0017 0.0001 0.0001 0.0072 -0.0087 0.0004 -0.0147 -0.0168 -0.0004 -0.0006 -0.0031 -0.0038 -0.0015 0.0006 -0.0117 -0.0025 0.0062 0.0020 -0.0007 0.0003 0.0091 -0.0140 0.0000 -0.0138 -0.0195 -0.0009 -0.0007 -0.0038 -0.0034 0.0034 0.0005 0.0044 -0.0005 TR 12 TR 13 TR 14 TR 15 TR 16 0.0035 0.0018 0.0001 0.0006 0.0074 -0.0114 0.0005 -0.0152 -0.0184 -0.0010 -0.0009 -0.0051 -0.0067 0.0015 0.0001 0.0047 -0.0048 0.0000 0.0007 -0.0052 0.0005 0.0053 -0.0046 -0.0004 -0.0128 -0.0115 -0.0010 -0.0004 -0.0075 -0.0052 -0.0001 0.0003 -0.0232 -0.0027 0.0033 0.0040 0.0044 0.0003 0.0085 -0.0118 0.0005 -0.0303 -0.0360 -0.0014 -0.0014 -0.0076 -0.0116 0.0022 0.0014 -0.0035 -0.0008 0.0001 0.0028 0.0006 0.0003 0.0048 -0.0011 0.0002 -0.0109 -0.0215 -0.0009 -0.0010 -0.0046 -0.0087 -0.0001 0.0006 -0.0052 -0.0007 -0.0026 0.0014 0.0002 0.0002 0.0044 -0.0020 0.0002 -0.0052 -0.0152 -0.0007 -0.0006 -0.0014 -0.0019 -0.0012 0.0002 -0.0011 -0.0022 Gagnon and Tryon A5. Difference in RHSD Between Rules 1 and 2 Across Quarters (cont'd) TR 17 GDP LGDP PGDP LPGDP GDPV MB CAB RS LRS RL ARL RERW ARERW C AC MBR AMBR 0.0025 0.0019 0.0014 0.0001 0.0083 0.0008 0.0003 -0.0148 -0.0144 -0.0010 -0.0008 -0.0051 -0.0061 0.0028 0.0007 -0.0117 0.0003 TR 18 TR 19 TR 20 Avg 0.0001 0.0027 -0.0014 -0.0003 0.0014 0.0011 0.0027 0.0021 0.0034 -0.0021 -0.0037 0.0003 0.0003 0.0005 -0.0004 0.0002 0.0176 0.0077 0.0006 0.0080 -0.0033 -0.0133 0.0080 -0.0066 0.0002 0.0004 -0.0002 0.0002 -0.0095 -0.0016 -0.0151 -0.0121 -0.0131 -0.0242 -0.0089 -0.0181 -0.0014 -0.0003 -0.0012 -0.0008 -0.0005 -0.0011 -0.0006 -0.0008 -0.0075 -0.0025 -0.0029 -0.0031 -0.0071 -0.0083 -0.0048 -0.0059 0.0001 0.0012 -0.0002 -0.0002 0.0008 0.0002 0.0006 0.0005 -0.0084 0.0143 -0.0207 -0.0064 -0.0003 0.0032 -0.0013 -0.0016 - 27 - Gagnon and Tryon REFERENCES Blue Chip Economic Indicators (Alexandria, VA: Capitol Publications) various issues, 1980-1989. Fuhrer, Jeff, and George Moore "Inflation Persistence," in this conference series, 1992. Gagnon, Joseph E. "A Forward-Looking Multicountry Model for Policy Analysis: MX3," Economic and Financial Computing 1, 1991, pp. 311-61. Gagnon, Joseph E., and Ralph W. Tryon "Stochastic Behavior of the World Economy under Alternative Policy Regimes," International Finance Discussion Papers No. 428, March 1992. - 28 - COMMENTS ON PRICE AND OUTPUT STABILITY UNDER ALTERNATIVE MONETARY POLICY RULES Satyajit Chatterjee1 The goal of this paper is to use a medium sized econometric model of the U.S., West Germany, and Japan, to evaluate the performance of the U.S. economy under alternative monetary policy rules. This model has been developed by the first author and discussed in detail in Gagnon (1991). Two key aspects of the model are that expectation formation is forward looking and approximately rational (in the sense that it approximately corresponds to the predictions of the model) and that the long run properties of the model match those of a standard real neoclassical growth model. These features make the model attractive for monetary policy analysis. Adherence to rational expectations means that the policy analysis exercise is immune from the Lucas critique. The fact that the long run properties of the model are those of a real neoclassical growth model means that the short run dynamics (for which monetary factors matter) do not extrapolate into bizarre long run behavior. In addition, the behavioral equations of the model, while not grounded in explicit optimization, are carefully motivated. The reader gets a sense of the kind of structure (in terms of preferences, technology and market opportunities) that would generate these decision rules. In what follows I will not comment any further on the structure of the model. What I will focus on instead are the different monetary policy rules considered in the paper and the manner in which they Economist, Federal Reserve Bank of Philadelphia. Chatterjee are evaluated. Let me begin with the latter point. The authors "evaluate" policy ruies in terms of their effect on output, price and interest rate variability. The presumption is that a rule that generates more variability is inferior to one that generates less. However, a more appealing way to evaluate policy rules is to determine how they affect the variability of utility. From the discussion in Gagnon (1991), it would appear that the authors have in mind a situation where consumption goods and real money balances are the only arguments in people's utility function. Therefore, it is the variability in these quantities that ought to matter. Indeed, since the authors actually estimate the parameters of the utility function they can evaluate the different policy rules directly in terms of expected utility. My sense is that the ranking of rules would be significantly affected by this choice of metric. In particular, rules that smooth interest rates might generate greater variability in real money balances and hence be less desirable. I turn now to the different policy rules evaluated in the paper. Presumably, the ultimate object of interest here is the character of the optimal monetary policy rule given the structure of the model. However, due to its complexity it is computationally infeasible (but not impossible) to calculate the optimal policy rule for a given criterion function. Instead, the authors provide us with the operating characteristics for a collection of reasonable looking rules. While this is understandable, we are left nevertheless with an uncomfortable imbalance in the paper: while a lot of care has gone into modelling individual decision rules as resembling the result of some sort of optimization exercise, no attempt is made to model the monetary policy rule as resembling the result of some sort of an optimization exercise on part of the monetary authority. 2 Chatterjee Consequently, 1 find it difficult to get interested in the operating characteristics of the economy under any of these rules. What could be done to alleviate this problem? One possibility, which I find personally attractive is to pose the optimal monetary policy question in a model for which it is computationally feasible to obtain an answer. I am thinking about fairly abstract general equilibrium monetary models like the representative agent cash-in-advance models of the type popularized by Lucas (Lucas (1984), Lucas and Stokey (1987)). Recently, researchers (Cooley and Hansen (1989)) have used calibrated versions of this model to obtain answers to question like: what happens to the operating characteristics of the economy if the monetary growth rate is raised from 3% to 6%? It is not too difficult to extend this kind of analysis to compute optimal feedback rules. We would then have a numerical candidate rule which we know to be optimal for a simpler economy. It would then be of interest to see how this rule performs when it is used in an econometric model of the kind that the authors have estimated and which incorporates real world frictions absent from the simpler model. To summarize, I find the model estimated by the authors to be reasonable and certainly worth taking seriously. My sole concern has to do with the manner in which the model is used. I would liked to have seen different policies ranked according to expected utility (or failing that, at least in terms of variability of consumption and real money balances). I would also liked to have seen some attempt at studying the operating characteristics of an approximately optimal monetary policy rule. 3 Chatterjee REFERENCES Cooley Thomas F., Hansen Gary D., "The Inflation Tax in a Real Business Cycle Model." American Economic Review, vol. 79, (September 1989), pp. 733-48. Gagnon, Joseph E., "A Forward Looking Multicountry Model for Policy Analysis: MX3." Economic and Financial Computing, vol 1, (1991), pp. 311-61. Lucas, Robert E., Jr., "Money in a Theory of Finance." Rochester Conference Series on Public Policy, Carnegie vol. 21, (1984), pp. 9-46. ..-«.-». f Stokey Nancy L. , "Money and Interest in a Cash-in-Advance Economy." Econometrica, vol. 55, (May 1987), pp. 491-514. 4 MONETARY POLICY EXPERIMENTS IN A STOCHASTIC OVERLAPPING GENERATIONS MODEL OF THE TERM STRUCTURE Steven Russell1 In recent years economists have begun to experiment with the construction of dynamic, stochastic general equilibrium models designed to confront the data provided by macroeconomic time series — models that can explain, or help explain, the relationships between and within various series that constitute the "stylized facts" of the business cycle. The exercise of data confrontation seems to take place in two steps. First, an investigator specifies a model, and chooses its parameters in a way that seems empirically plausible. This step is sometimes called "calibration." Next, the model is simulated, and the properties of the artificial time series it generates are compared to those of actual time series data, paying special attention to the particular stylized facts emphasized by the investigator. In practice there is usually some (and often a great deal of) interaction between first and second steps: parameter values are very often chosen with an eye towards producing artificial data with the desired properties. The model of choice for these sorts of exercises has been the representative agent, infinite-horizon capital accumulation model, augmented by positing stochastic variation in technological productivity. The augmented model has become known as the "real business cycle" (RBC) model. RBC modeling has provided valuable insights into the nature 1 Federal Reserve Bank of St. Louis. research assistance. Lynn Dietrich provided Russell and sources of the business cycle, and is clearly a growth industry among macroeconomists. One common criticism of RBC models is that they emphasize real sources of cyclical variation at the expense of sources that are monetary in nature. Some RBC modelers have attempted to respond to this criticism by introducing money into the model and examining the effects of various assumptions about monetary policy. Unfortunately, the model does not seem well suited to this purpose. Since it is devoid of the sorts of exchange frictions that are necessary to provide a "natural" role for money, money demand must be induced via ad hoc devices such as placing real balances in the utility function or imposing cash-in-advance constraints. One characteristic finding is that monetary policy is entirely neutral in the long run, and relatively ineffectual even in the short run. Results of this sort have led critics to allege that RBC practitioners introduce money into their models only in order to demonstrate its unimportance. One problem with "monetary" RBC models that has attracted a good deal of attention has been their inability to produce liquidity effects — their inability, that is, to generate nominal interest rates that decline in response to monetary injections. Recently, investigators such as Fuerst (1992) and Christiano and Eichenbaum (1992) have succeeded in constructing RBC models that produce liquidity effects. While this is certainly a very interesting development, a skeptic might view the scope and intricacy of the assumptions these investigators must make in order to achieve such effects as a testament to the limitations of the RBC model as a framework for monetary analysis. The most popular dynamic general equilibrium alternative to the representative-agent, infinite horizon -2- Russell model is the overlapping generations (OLG) model. The OLG model has been, and for the most part remains, the model of choice for theorists interested in monetary issues. However, OLG modelers have rarely attempted to confront business cycle data in anything like the sense that RBC modelers attempt to do so. The principal reason for this is probably the "time horizon question." RBC models can be calibrated, by appropriate choice of the representative agent's rate of time preference, so that a "period" seems to represent an interval of a quarter or a year, and in particular so that business-cycle-like variation occurs over intervals that are short relative to the decision horizon of the agent. This cannot be done with conventional OLG models, since the agents that populate them live for only two or three periods. The relative length of agents' decision horizons may not be the only (or best) criterion according to which the empirical plausibility of a model can be evaluated, however. Another criterion is whether the model is flexible enough to capture the features of the economy that most economists consider critical to understanding the phenomena under study. When this criterion is applied, overlapping generations models compare quite favorably to RBC models. The generational heterogeneity of the population of agents creates the potential for an active money market, and active primary and secondary markets for government securities — active in the sense that the agents in the model actually trade these objects with other agents. In addition, it is relatively easy to introduce the sorts of intragenerational heterogeneity necessary to produce active private credit markets. In OLG models, monetary injections can easily take the form of open market purchases, rather than the "helicopter drops" favored by RBC modelers. Partly as a result, it is relatively easy to use these models to analyze -3- Russell the interaction between fiscal and monetary policy: indeed, realistic descriptions of open market operations virtually require explicit consideration of this interaction. Perhaps the most important difference between OLG and RBC models is that the former pr