The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Intermediate Targets and Indicators for Monetary Policy A Critical Survey Federal Reserve Bank of New York Intermediate Targets and Indicators for Monetary Policy A Critical Survey Federal Reserve Bank of New York CONTENTS Foreword £. Gerald Corrigan i Introduction Richard G. Davis 1 Possible Roles for the Monetary Base Ann-Marie Meulendyke 20 Monetary Aggregates as Intermediate Targets John Wenninger 67 Liquid Asset Measures as Intermediate Targets and Indicators for Monetary Policy Gabriel S.P. de Kock and Lawrence J. Radecki 109 A Review of Credit Measures as a Policy Variable Lawrence J. Radecki 183 Targeting Nominal GNP Spence Hilton and Vivek Moorthy 232 Interest Rates as Targets and Indicators for Monetary Policy Charles Steindel 274 Commodity Prices as Intermediate Targets and Information Variables Spence Hilton 305 Possible Roles of the Yield Curve in Monetary Policy Arturo Estrella and Gikas Hardouvelis 339 The Use of Dollar Exchange Rates as Targets or Indicators for U.S. Monetary Policy Charles Pigott and Christopher Rude 363 Optimal Monetary Policy Design: Rules vs. Discretion Again A. Steven Englander 421 A Review of Federal Reserve Policy Targets and Operating Guides in Recent Decades Ann-Marie Meulendyke 452 FOREWORD For nearly two decades after the Federal Reserve-Treasury Accord of 1951, U.S. monetary policy was viewed, both by its practitioners and by others, as mainly a matter of "leaning against the wind"-of tightening money market conditions when inflationary pressures threatened and easing them when the economy weakened. The 1960s, however, saw a gradual revival of interest in the possible role of the money stock in cyclical fluctuations and, especially, in the longer run behavior of prices. At the same time, the economy was experiencing a gradual, though irregular, acceleration in the trend rate of inflation, a development that was to reach its climax in the late 1970s and early 1980s. Against this backround, the Federal Open Market Committee began to experiment more systematically with internal objectives for money and credit growth beginning around 1970. This was done in the belief that such targets could provide a better focus for the potential impact of policy on the economy and, in particular, on the longer run trend of inflation. Subsequently, a 1975 congressional resolution called on the Federal Reserve to report formally to Congress on its "plans and objectives" with respect to growth in money and credit. This requirement was reaffirmed and refined in the Humphrey-Hawkins legislation of 1978. Thus came into being a set of "intermediate targets" for monetary policy—intermediate, that is, between instruments such as open market operations and the discount rate and the broad objectives of policy with respect to economic performance. The potential usefulness of such "intermediate targets" seemed-and continues to seemsubstantial. At least in principle, targets for growth infinancialaggregates such as money and broad measures of credit can provide an appropriate way of structuring the intermediate and longer run strategy of monetary policy, improving both the process of internal decision making and the communication of policy objectives to the Congress and to the public at large. Longer term objectives for money and credit growth can have an especially important role to play in ensuring that shorter term decisions are consistent over the long run with stability in the value of the nation's money, an objective that is uniquely a concern of the central bank. The decade of the 1980s just passed, however, proved to be a particularly difficult period for the effective use of money and credit targets. A major wave of financial innovation and deregulation and a sharp reduction in the rate of inflation and the accompanying decline in interest rates, together with I other factors not completely understood, have produced major changes in the relationship of the various measures of money and credit to economic activity. The Federal Reserve has found it necessary to adapt to these changing conditions by reducing or eliminating the use of some financial measures as intermediate targets while increasing emphasis on others and, more generally, by relying importantly on an eclectic and judgmental approach to policy making. Moreover, given the apparently impaired usefulness of the more traditional measures of money and credit as policy targets and indicators, some have sought to find new measures, such as commodity prices and the yield curve, to serve as guides for policy. In these circumstances, it seemed useful to us at the Federal Reserve Bank of New York to review the evidence on the potential value of a range of financial measures as intermediate targets and/or indicators. The results of this review are contained in the studies that make up the present volume. The papers survey the vast literature that has grown up around these various measures while undertaking new research where that seemed to be needed. We hope that the results, as collected in one volume, will prove useful to all those concerned with issues of monetary policy. Not surprisingly, given the experience of the 1980s, each of the measures examined in these studies is found to have serious limitations. Thus some of the measures are clearly unsuitable as intermediate targets, if only because they cannot be readily controlled with existing monetary instruments. Moreover, in varying degrees, all the measures prove to have only an imperfect relation to the broader economic measures policy seeks to influence. Despite the problems that have emerged with the traditional money and credit measures in the 1980s, however, the Federal Reserve remains committed—not only as a matter of law but also as a matter of philosophy—to the formulation of money and credit objectives that will serve as a general guide to its longer run intentions. Paths for the downward adjustment of the longer run trend of money and credit growth are an essential element in any coherent long-run strategy for eliminating inflation as a significant factor in our economic life. The experience of the 1980s provides an important reminder, however, that such an approach to monetary policy must retain a great deal of flexibility, including the need to look at a wide variety of financial and nonfinancial indicators in framing judgments about policy. In any case, the Federal Reserve will undoubtedly continue to study the evolving performance of money and credit measures in an effort to find which of them may prove most suitable for framing the strategy of monetary policy in the conditions of the 1990s. E. Gerald Corrigan July 1990 ii INTRODUCTION Richard G. Davis Over the years, a broad array of mainly financial variables has been proposed for use in formulating and implementing monetary policy. This collection of papers examines the potential value of these various measures as intermediate targets and/or indicators of monetary policy. It includes a review of the Federal Reserve's evolving approach to the use of policy targets and operating guides in the postwar period. It also contains an analysis of the recent academic literature on the theory of policy rules that is relevant to the potential usefulness of intermediate targets. Systematic analysis of monetary tactics and strategy in light of the relationships among policy instruments, a broad array of monetary and financial variables and measures of economic performance, began to expand rapidly in the late 1950s. Over the subsequent decades the subject has generated a large body of literature. One early source of motivation for this work was monetarist criticisms of the Federal Reserve's post-Accord procedures. In these procedures, the behavior of the money stock played, at most, only a limited role. Another impetus to the literature on monetary tactics was progress in modeling the markets for reserves and money. This work provided far greater analytical and quantitative detail on the connection between Federal Reserve actions and the response of the reserve and money aggregates than had previously been available. Continuing controversy over the appropriate role of money stock targets sustained and intensified interest in the question of intermediate targets and their implementation in the 1960s and 1970s. Interest in the subject was especially intense in the period after the October 1979 announcement of a change in operating procedures designed to improve the implementation of targets for the monetary aggregates. By the early 1980s, signs of an emerging breakdown in the existing relationships of the money measures to the economy generated suggestions that money stock targets be augmented or replaced by broad measures of liquid assets and credit. As the extent of the shift in the relationship of all these various measures to GNP became more apparent, however, research interest in their use as intermediate targets or indicators waned and their role in policy making diminished. More recently, as discussed in the relevant papers in this volume, some interest has been expressed by economists and policy makers in possible roles for nominal GNP and/or for market measures such as l commodity prices, the yield curve, and foreign exchange rates as policy targets and/or indicators. In general, however, confidence that there exist financial measures that can replace in part or in whole a basically judgmental, pragmatic, and eclectic approach to policy seems currently (1990) at a rather low ebb. Virtually without exception, the results reported in this volume support such a skeptical attitude. Nevertheless, as argued below, the issue is far from closed. Indeed, interest in the problem of devising and implementing "intermediate" guides for policy is likely to prove a hardy perennial in the years ahead. I. Some Terminology One product of the debate on these issues has been the development of a useful and reasonably settled vocabulary to discuss them. One can imagine a spectrum of economic measures that has, at one end, the "ultimate targets" of monetary policy. These almost always include the price level and real output and sometimes also include the behavior of the balance of payments and the foreign exchange value of the dollar. At the other end of this spectrum are the "instruments" of monetary policy. These include open market operations, the discount rate, and in earlier periods, required reserve ratios and Regulation Q ceilings on deposit interest rates. Just one step along the spectrum beyond these instruments are "operating targets," measures that can be controlled with a rather high degree of precision through manipulation of the policy instruments. Potential operating targets include measures such as nonborrowed reserves, the nonborrowed monetary base, and short-term money market rates, most notably the federal funds rate. Borrowings from the discount window clearly also constituted a potential operating target under the system of lagged reserve accounting that prevailed between 1968 and 1984, since the Trading Desk could take required reserves as predetermined within any reserve averaging period. Even under the present system of approximately contemporaneous reserve accounting, most people would probably still want to count borrowings as a potential operating targetthough to achieve it in any given reserve maintenance period means that the Desk must correctly estimate required reserves in the current period as well as market factors supplying reserves and the levels of excess reserves. "Intermediate" measures, whether considered as "targets" or as "indicators," are variables that, as the term suggests, are intermediate between (1) the instruments and operating targets that are capable of rather tight control and (2) the ultimate target measures that can only be influenced indirectly. Measures of the money stock are perhaps the classic examples of such "intermediate" 2 variables, but as noted, the list includes other broad financial aggregates, such as credit extended to the nonfinancial sectors, as well as market measures, such as the foreign exchange rate, that are thought to be significantly influenced by movements in the operating targets. Some measures, as discussed in more detail in the appropriate papers, are a little harder to classify. Thus, for example, short-term interest rates are usually treated as operating targets but may also be treated as intermediate targets. Conversely, the monetary base is most often discussed as an intermediate measure but sometimes, more controversially, is viewed as a potential operating target. Nominal GNP is sometimes treated as a potential intermediate measure, at one step removed from its ultimate target components of prices and real output. The various intermediate measures may have the potential to serve as intermediate "targets" and/or as intermediate "indicators" of monetary policy. "Targets" are, obviously enough, objectives the Federal Reserve seeks to achieve over some time period with some degree of precision and under some particular set of circumstances. The concept acquired legislative status with the 1975 congressional resolution requiring the Federal Reserve to report on its "plans and objectives with respect to the growth of the monetary and credit aggregates over the coming year," language that was repeated in the Humphrey-Hawkins legislation of 1978. The concept of an "intermediate target" seems to imply that to qualify, a measure should be (a) reasonably subject to control by the Federal Reserve through adjustment of its operating targets and (b) reasonably closely (that is, predictably) related to the ultimate targets or, in practice, at least to nominal GNP. Consequently, the papers in this volume examine the various measures considered as potential intermediate targets from both points of view. The concept of intermediate measures as monetary "indicators" is a bit more complicated because it is sometimes taken to mean a measure of the stance of monetary policy and is sometimes interpreted as an indicator of current or future developments in the economy. In much of the earlier literature (early 1960s), the term was interpreted in the sense of indicators of the stance of monetary policy—that is, as measures that could provide, in some sense, an index of monetary "ease" or "restraint." The attempt to pin down such an index produced, and indeed continues to produce, a certain amount of confusion and ambiguity. Consider, for example, a situation in which the Federal Reserve is using interest rates as an operating target and has no intermediate target objective for the money stock. Suppose on entering a recession, the policy makers progressively lower their interest rate target, but, owing to the recession-induced decline in the demand for money, the money stock falls (probably along with a drop in total reserves). Measured in terms of intentionality, policy has clearly "eased," because the declining short-term rates are, at least in large part, the direct result of 3 policy decisions to ease. But if one believes that it is not intentionality but rather the impact of policy on the economy that matters, and if one also believes that this impact is best signaled by the money stock, then in this instance, the declining money stock indexes not an "easing" but a "tightening" of policy. This may be a terminological problem in the sense that one may want to talk about an indicator of policy intentions or an indicator of policy impact and these may not be the same thing. But the distinction between measures of intention and measures of impact, if they are in fact different, may also raise an econometric issue: how to decide which intermediate measure, if any, should be treated as "predetermined" for estimating purposes. In this example, the money stock or short-term rates? In any case, the recent technical literature has tended to focus on intermediate "indicators" (sometimes, in this context, also called "information variables") not as measures of the stance of policy, but as measures of the present or prospective state of the economy. This is the sense in which the term is generally used in the present volume. To be sure, there are places in the literature where the two senses of a monetary "indicator" are conflated. For example, a rise in commodity prices or a steepening of the yield curve may be taken as indicating both that the prospects are for rising inflation in the future and that policy has been "easy" or, perhaps, "too easy." Clearly, the main requirement for a good intermediate indicator of the state of the economy is that it be reliably (predictably) related to the current or prospective behavior of ultimate goals such as inflation and/or real output. In practice, statistical tests have often been couched in terms of the relationship of the measure to nominal GNP. A question arises whether a measure that has proved to be a good indicator in this sense can then be used as an intermediate target while still continuing to be a good indicator. It has sometimes been asserted that when a financial aggregate such as the money stock becomes an intermediate target, presumably chosen in part because of its good indicator properties, these properties will then be altered (for the worse?) by the very fact of its targeting by the authorities. This may or may not be a problem with respect to financial aggregates that are treated both as intermediate targets and indicators. It clearly could be a complicating issue, however, for such market measures as commodity prices, interest rates, and the foreign exchange rate. Knowledge in the market that the behavior of the measure is being used by the authorities to make policy decisions is very likely to alter that behavior. Partly for this reason, proponents of these latter measures have generally advocated them for only a single purpose: for example, commodity prices and the yield curve, simply as indicators; interest rates and the dollar, either as indicators or as operating or intermediate targets, but not both as indicators 4 and as targets. II. Is There a Case for Intermediate Targets? It is clear that a coherent monetary policy requires a decision on operating targets. It is equally clear that "indicator" measures providing advance information about the current or prospective state of the economy are, almost by definition, of value. The usefulness of intermediate monetary targets, however, has always been more controversial. No measure selected for such a role will ever be perfectly predictably related to the ultimate targets that matter. At least some uncertainty, some shortterm instability, and some longer term drift in the relationship of any intermediate target to final objectives seems inevitable. It has therefore been argued that the use of intermediate targets will result in suboptimal decisions. Policy makers will adjust their operating targets, not directly in terms of the settings most likely to achieve their ultimate objectives, but instead in terms of the settings most likely to achieve the intermediate target. According to this line of thought, intermediate measures such as the money supply may be useful, at best, as variables that may shed light on (1) the current state of the economy, perhaps because of more prompt reporting, or (2) the economy's prospective future state, because of leading indicator properties. On the other hand, their use as intermediate targets is likely only to produce poorer control over ultimate targets than if instruments were adjusted directly in terms of objectives for these latter targets. The logic of this criticism of intermediate measures as targets seems impeccable. Nevertheless, it misses the heart of the case for the use of such targets, a case that encompasses a much wider range of considerations. This broader case envisions a number of potential benefits from the use of intermediate targets. It has been argued, for example, that intermediate targets can usefully provide a means of communicating the central bank's intentions to the public. Moreover, such targets can provide a form of central bank accountability. The ultimate target measures may not be well suited for these various purposes. Thus, as discussed in the paper on nominal GNP targeting, there may be real problems in having an independent central bank set or announce goals for ultimate targets. Equally to the point, actual economic performance over any given period is subject to many important influences in addition to monetary policy. Hence it may be quite inappropriate to judge the success of this policy by the actual performance of the economy—that is, by the ultimate target measures-given the role of nonmonetary influences. By contrast, an intermediate target—a goal for the rate of money growth, for example—can 5 be judged in advance for its probable consistency with acceptable economic outcomes. Moreover, it can be used to judge, ex post, whether the central bank's day-to-day decisions have been appropriate to achieving its intermediate target objective. Moreover, the existence of an intermediate target, defined over a time period such as a year, can be useful to the central bank as an internal check on the appropriateness of the shorter term settings of its operating targets.1 But there are other fundamental arguments for the use of intermediate targets-provided suitable targets can be found. Thus it has generally been argued that over the long run, monetary policy can only affect nominal magnitudes. Its longer run influence over real growth, real interest rates, and employment mainly reflects its success or lack of success in achieving an environment in which economic decisions can be made with a minimum of concern and uncertainty about price level instability. If this view is correct, the appeal of intermediate targets in providing a "nominal anchor" for policy decisions is fairly clear. Such targets can provide, in principle, an indication that the longer run thrust of policy will be consistent with longer run goals for price behavior. In principle, at least, any one of the various monetary and credit aggregates could, if used as intermediate targets, provide this kind of "nominal anchor" for policy—as could nominal GNP. Another role that has been suggested for intermediate targets is in dealing with the potential conflict that may exist between short- and long-run optimal policy, an issue known as the "time consistency problem" in the academic literature and in more informal discussions as the "credibility problem." A conflict between short-run and long-run optimizing can arise from the fact that in the short run, the monetary authorities can probably engineer some extra real output, at least up to a point. They can only do this, however, through an expansionary policy that yields more inflation than is built into the public's expectations. According to widely accepted theory, increases in wages and prices that are more rapid than expected will "fool" the public into supplying more labor and goods under the mistaken impression that the higher wages and prices represent higher real rewards. In the short run, there may be pressure on the central bank to seek output gains through such "surprise" inflation. But once the public comes to recognize that the policy makers are operating in a way that accelerates inflation, the public will anticipate this acceleration. Put simply, the attempt to boost output through policies that create surprise inflation will be self-defeating. Over time, the public will catch on, and the higher inflation will no longer be a "surprise." Inflation that is anticipated will ^ e s e various arguments were cited in a speech, "The Contributions and Limitations of 'Monetary1 Analysis," given by Paul A. Volcker in September 1976 and most recently reprinted in the 75th Anniversary issue of the Federal Reserve Bank of New York's Quarterly Review, May 1989. 6 have no power to induce higher output. Thus, over the longer run, the effort to induce higher output through excessively stimulative policies will fail. Output will be no higher than it otherwise would have been—trending at its potential rate over time—but the rate of inflation will be higher. Thus on balance, stimulative policies that seem attractive period by period will, over the longer run, simply result in higher inflation without any output gains-a result desired by no one. An intermediate target, publicly announced and faithfully adhered to, could, in theory, avoid this kind of outcome. It could do so by effectively tying the hands of the authorities, preventing them from yielding to the temptation to seek short-run output gains in a process that over the longer run only guarantees higher inflation. Probably the best known prescription for using an intermediate target in this way is the constant money growth "rule" or, in some versions, money growth targets that settle down to such a rule after some period of accommodation to disequilibrium initial conditions. Of course a monetary growth rule also has potential disadvantages. Thus while it may ensure reasonable long-run price stability, it makes no provision for accommodating shocks—whether from supply or demand-and thus may achieve long-run price stability only at the expense of unsatisfactory shorter run outcomes for both output and prices. It might be possible to design a more complex monetary growth rule that allows money growth to adjust to such short-run disturbances, but in a predetermined way that still prevents the authorities from seeking short-run output gains at the expense of higher average inflation. However, monetary rules that embody such automatic response features may themselves create credibility problems—as discussed in the paper in this collection that reviews the "time consistency" literature. It has to be emphasized that all these various potential virtues of intermediate targets—improved accountability, improved communication with the public, provision of a nominal anchor, and prevention of short-run decisions that serve merely to raise inflation over the longer run—can be achieved only if suitable target measures exist. As noted earlier, "suitable" in this context means measures that are "sufficiently" controllable and "sufficiently" stable in their relation to the ultimate objectives. But this is not an all or nothing matter. No intermediate target will be perfectly controllable, even over a year. And no measure will be related in a perfectly predictable way to the ultimate targets. At least some slippage on both counts is inevitable. On the other hand, even if there is some slippage, the benefits derived from intermediate targeting may, over the longer run, outweigh the costs that arise as a result of this slippage. Clearly it is a matter of more or less-that is, a question of how much slippage can be expected from the use of intermediate targets, on the one hand, and how much one values their potential longer run benefits on the other. Typically, individuals most 7 concerned with long-run inflation results have tended to minimize the problems with intermediate targets, while those most concerned with the shorter run real output consequences have tended to worry most about these problems. in. Evaluating the Candidates Eight papers in this volume examine individual candidates or groups of candidates—for example, the multiple measures of money and credit—as potential targets and/or indicators of policy. While the papers differ somewhat in organization and emphasis, they all touch on certain common issues. These include (1) the theoretical basis for believing that the particular measures in question might be useful targets or indicators, (2) the statistical evidence for believing a relationship to ultimate targets exists and evidence for the stability of any such relationship, (3) issues of central bank control, and (4) the question whether the measure, even if not used as a formal target, might be useful in a subordinate role. For example, the paper on interest rates considers the possibility that even if interest rates make little sense as an intermediate target, upper and lower bounds for real short-term rates might nevertheless be useful as "constraints" on settings for an interest rate operating target. As noted earlier, the mostfrequentlyadvocated measures for intermediate targeting over the past three decades have been the various measures of the money stock and the monetary base and, more recently, liquid assets and various credit measures. The statistical results for these measures form a vast literature varying in method, sophistication, periods covered, and conclusions. This literature is summarized and evaluated in some detail in the individual papers in this volume. It may be useful here, however, to give some crude sense of the problems that developed for these measures in the 1980s. Charts 1 to 6 show the departure from trends of the GNP velocities of a number of potential intermediate target measures in the 1980s. These departures are clearly large in all cases—greatest for Ml, the monetary base, and nonfinancial credit; less for M2, M3, and liquid assets. These departures from past experience are fairly easy to explain in some cases. Thus the velocity of narrow money (and the monetary base) almost certainly fell because of declines in inflation, nominal interest rates, and hence the opportunity cost of holding these measures. Explanations in the case of the broader measures that internalize the effects of market interest rate movements seem less clear. In any case, the same pattern of major departures from earlier postwar relationships is evident in Table 1, showing regressions of growth in nominal GNP on current and lagged growth in these various financial measures. As the error measures suggest, equations estimated on data from 1960 to 8 Chart 1 Velocity of Monetary Base (GNF/Monetary Base) Level 4th quarter 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 Chart 2 Velocity of M1 (GNP/M1) Level 4th quarter 4.5 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 Chart 3 Velocity of M2 (GNP/M2) Level 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 Chart 4 Velocity of M3 (GNP/M3) Level 4th quarter i i i i i i i i i i i i i i > i 1970 10 1972 1974 1976 1978 1980 1982 1984 1986 1988 Chart 5 Velocity of L (GNP/L) Level 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 Chart 6 Velocity of Nonfinancial Debt (GNP/Debt) Level 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 Table 1 Summary Statistics from Reduced Form Equations 1960-H to 1979-IV Average Error Actual-Predicted for 1980-89 Average Absolute Error 2.05 -4.13 5.72 6.16 3.92 1.86 -0.38 3.55 4.48 0.23 3.94 1.86 -0.55 3.40 4.49 0.30 3.78 2.01 -0.97^' 3.67i' 4.54i' 0.29 3.80 2.08 -2.65i' 4.35i' 4.77i/ 0.27 3.85 1.92 -2.40 3.99 4.41 R2 Y = 2.49 + 1.18M1 (2.39)(6.29) Y = 1.43 + 0.85M2 (0.99)(5.19) Y = 1.57 + 0.75M3 (1.07)(5.07) Y = 0.37 + 0.93L (0.25)(5.69) Y = 1.00 + 0.87Debt (0.59)(4.54) Y = 2.72 + 0.98Base (2.34)(5.33) SEE 0.34 3.66 0.24 DW RMSE Notes: All equations regress the growth rate of nominal GNP on the current and four lagged growth rates of the financial aggregate. Figures in parentheses are "t" values. L represents the Federal Reserve Board's measure of liquid assets. V 1980-1 to 1989-m Table 2 Summary Statistics from Reduced Form Equations 1981-1 to 1989-IV R2 Y = 5.85 + 0.16M1 (4.17)(1.03) Y = 7.21 + 0.01M2 (2.88X-0.02) Y = 8.05 + 0.06M3 (3.16)(-0.21) Y = 9.46 + 0.21L (3.33X-0.21) Y = 10.80 + 0.30Debt (2.85)(-0.90) Y = 0.37 + .99Base (0.12)(2.56) SEE 0.06 3.67 1.36 0.02 3.76 1.31 0 3.79 1.33 0.06i' 3.69±; 1.38i7 0.07i' 3.661' 1.5 li' 0.16 3.48 1.51 DW Notes: All equations regress the growth rate of nominal GNP on the current and four lagged growth rates of thefinancialaggregate. Figures in parentheses are "t" values. L represents the Federal Reserve Board's measure of liquid assets. y 1981-1 to 1989-m. 12 1979 do a very poor job in estimating GNP growth in the 1980s. And as Table 2 shows, similar equations estimated over data from 1981 to 1989 have almost no explanatory power, with coefficients that are not significant (indeed usually negative!) for all measures except the monetary base. This kind of result makes clear the reasons for the growing disillusionment in recent years with the potential of these measures as intermediate targets or indicators. The other property required of a potential intermediate target (as opposed to indicator) is of course controllability, and that poses a different set of problems. Some of the broader measures, such as total liquid assets and aggregate credit, are clearly not closely related to Federal Reserve operating targets. They can perhaps only be controlled indirectly-that is, by first controlling GNP! The narrower measures such as Ml and M2 have clearly retained substantial interest rate sensitivity for horizons out to a year or so because many of their component own-rates respond only slowly to changes in market rates. As a result, growth in such measures may be rather sensitive to changes in interest rate operating targets. That very sensitivity is itself a problem, however. Thus it makes the GNP outcome associated with any successfully achieved growth rate target for the aggregate quite sensitive to shifts in the demand for goods and services~at least over periods out to a year or so. Use of such targets, therefore, either will imply a wide range of uncertainty about the GNP outcome associated with a given money growth rate or will make it seem desirable to define targets in terms of broad ranges of growth rates. This latter approach, however, clearly weakens the usefulness of such targets for many of the purposes they are designed to serve. The breakdown in earlier relationships between financial aggregates and nominal GNP in the 1980s has lead a number of academic writers to suggest that the velocity problem could be "solved" by targeting nominal GNP directly. Such an approach has some attractive features. A long-run nominal GNP growth rule set in light of the expected trend growth in real output would establish a "nominal anchor" for policy. Adhered to as a "rule," with or without automatic feedback mechanisms, it would solve the short run/long run inconsistency problem and would satisfy the other objectives of an intermediate target. In the short run, adherence to a nominal GNP target would automatically offset both the price and the real effects of demand shifts and would split the impact of supply shifts between real output and prices. Unfortunately, the nominal GNP approach appears to have equally large problems. First, it "solves" the velocity problem only on the assumption that there exists a way of accurately achieving a nominal GNP target with the means at the disposal of the central bank. Obviously the method of choice would not be through intermediate money targets, for that would simply reintroduce the 13 velocity problem. A different option would be to aim at GNP targets through constant resettings of an interest rate operating target. Clearly success with such an approach is far from assured. A second difficulty is that to get a handle on final objectives with a nominal GNP target, you need to have a predictable relationship between nominal GNP and output in the short run—that is, you need to be able to predict the price/output split resulting from a given GNP result, at least to the extent that you have short-run output objectives. But of course this problem is shared with other potential intermediate targets such as money growth rates. A third difficulty is that a fixed nominal GNP rule could, under quite plausible conditions examined in the GNP paper in this volume, generate problems of dynamic instability in the path of real output in the face of supply shocks and, apparently, in the face of prior misses in hitting the nominal GNP objective. Such problems can be avoided by resetting, on a discretionary basis, the nominal GNP target year by year. But this approach would create a very uncomfortable situation for a central bank that is "independent within the government." Year-by-year settings of GNP targets would come very close to setting year-by-year targets for real growth and inflation. Such a situation could well create pressures for precisely the kind of short-term optimization that produces the worst of all possible worlds over the long run-that is, it would seem to maximize the risks of creating the kind of "time consistency" problem cited earlier. Overall, it seems quite possible that as a practical matter, discretionary nominal GNP targeting could result in worse inflation outcomes than might exist in the absence of any intermediate target at all. In summary, the nominal GNP route appears, on closer examination, to be no panacea for the problems created by the velocity instabilities of the 1980s. The remaining measures examined in this volume, commodity prices, the yield curve, and the foreign exchange value of the dollar, have generally been proposed as intermediate indicators that might be used to guide settings of the operating targets, rather than as intermediate targets themselves. Because these markets are often regarded as efficient in incorporating relevant economic information, they could perhaps signal changes in the economic outlook very quickly. In the case of commodity prices, it seems likely that these prices could in fact be targeted through direct open market operations in commodity markets. Indeed, that is just what a "commodity standard," whether defined in terms of a basket of commodities or a single commodity such as gold, would involve. Instead of such operations, it has been suggested merely that commodity prices may represent sensitive advance indicators of changes in general inflation rates that can be used to signal the need to tighten or ease the conventional operating targets. The results surveyed in the paper on commodity prices included in this volume suggest that 14 movements in commodity price indexes do have a marginal contribution, but only a marginal contribution, to make in forecasting inflation. As leading indicators of turning points in broad movements in the overall inflation rate, commodity price measures, suitably averaged and smoothed, do have some predictive value. But they have also at times given false signals of turning points in the general inflation rate. In the case of correct signals, moreover, their lead times tend to be rather variable and there appears to be little relation between the magnitude of commodity price movements and the magnitudes of subsequent movements in overall inflation. On balance, it appears that commodity prices may be reasonable additions to the items the central bank "looks at" when it surveys the prospects for inflation. They do not, however, add much to more conventional methods of assessing the outlook for inflation. Another market measure that has been proposed as an indicator but not as a target of monetary policy is the yield curve. In this case, the question whether the measure is to be thought of as an indicator of the stance of policy or an indicator of the future course of the economy is somewhat ambiguous. And this ambiguity is directly related to the theoretical assumptions underlying the attention sometimes given to the yield curve for either or both of these roles. Belief in the indicator properties of the yield curve appears to rest on the expectations theory of the yield curve-the view that longer term rates should be regarded as (weighted) averages of the market's expectations of the future course of successive short rates. While this theory has considerable intuitive appeal, empirical tests of its validity over the years have produced mixed results. Even if the expectations theory of the yield curve is accepted as correct, moreover, the theoretical implications of particular yield curve configurations, as interpretations both of monetary policy and of prospective economic performance, appear to be ambiguous. This ambiguity stems in part from another widely accepted theoretical premise—that nominal interest rates reflect the sum of a real rate and an inflation premium that allows for the expected rate of inflation over the life of the instrument. Thus an upward rising yield curve, for example, could imply either that the market expects real short-term rates to rise in the future or that it expects the rate of inflation to rise. Against this background, the paper on the yield curve in this volume points out that it would be very hard for a central bank to interpret the significance of, for example, a steepening of the yield curve on the basis of theoretical considerations alone. Such a steepening could mean that the market expects inflation to accelerate, which the central bank could interpret as a need to tighten. Alternatively, it could mean that the market expects a rise in the productivity of capital and hence a rise in real short-term rates and an acceleration of real growth. This cause of a steepening in the yield 15 curve might or might not imply a need to change the settings of operating targets, depending on circumstances. A third possibility is that the steepening reflects a market judgment about the future of monetary policy itself-that is, that policy is expected to tighten, real short rates to rise, and therefore, quite possibly, real growth to slow. So a central bank looking at a change in the yield curve must try to sort out its possible meanings and then must decide what implications, if any, the change may have for policy. Despite these interpretive ambiguities at the theoretical level, the yield curve paper gives some fairly concrete results. It suggests, for example, that the Federal Reserve does have significant power to affect the yield curve by changing the federal funds rate as an operating measure. Since no one proposes the yield curve as a target, however, this is of rather limited significance. But the paper goes on to suggest that the yield curve, simply as an empirical matter, has proved to have significant forecasting value for both real output and inflation, even in the presence of other forecasting variables such as short-term interest rates, the leading indicators, and the consensus of economists' forecasts. One has to wonder, however, how this forecasting value might be affected if the yield curve were to become a major forecasting tool for the authorities and if the market were to become aware of such a development and were to respond accordingly. A kind of "two-person game" situation between the market and the authorities might greatly distort the behavior of the yield curve relative to what it would be in the absence of a belief in its indicator significance. Finally, the increasing sensitivity of the U.S. economy to international developments has led to growing interest in the use of the foreign exchange value of the dollar as a guide for U.S. monetary policy. However, the paper on this topic emphasizes that, because of important differences between exchange rates and more traditional variables, the systematic use of the dollar's value in U.S. monetary policy operations is likely to be highly problematic and would almost certainly raise considerations beyond those traditionally incorporated in U.S. policy deliberations. In particular, manipulation of policy instruments to regularly counter or "target" dollar movements could be destabilizing for the U.S. economy under a wide range of circumstances and could require a significant degree of international policy coordination. The paper does suggest that exchange rates can play a useful role as policy indicators but generally only under fairly limited circumstances. At times, for example, foreign exchange market conditions have proved helpful in gauging market perceptions and the likely reactions to policy changes. Beyond these circumstances, however, the evidence raises considerable doubts about the reliability of exchange rates as regular indicators of underlying inflation pressures or the monetary policy stance. Accordingly, on present knowledge, the case for upgrading the role of the 16 dollar in U.S. monetary policy formulation appears questionable. IV. A Future for Intermediate Targets? The cumulative effect of the papers included in this volume is to leave one impressed with the limitations of all the various measures, certainly as intermediate targets and, for the most part, even simply as indicators. But if all potential intermediate targets have problems, it is also important to recall the many ways in which policy conducted without any such target is itself less than satisfying. In practice, conducting policy without reference to intermediate targets means setting operating targets directly in line with changing assessments of the likely outcomes for the ultimate goal variables. Perhaps most often, this will mean adjusting some money market rate in line with changing projections of the future behavior, under assumed paths for such a rate, of prices and real output. The difficulties of this approach to policy making are numerous. Perhaps the most obvious problem is the need to assess correctly the future state of the economy under alternative assumptions about settings of the operating targets. Note that it is the future state of the economy that matters given the universally acknowledged existence of significant lags in the impact of policy on output and prices. While there is substantial evidence that experienced macro forecasters can improve significantly on naive extrapolative procedures in projecting the future, it is also clear that forecasting remains as much an art as a science. Macro forecasting normally reflects a blend of reliance on econometric models, interpretation of incoming information (both statistical and "anecdotal") on the current state of the economy, and the selective use of an array of leading indicator measures. While such forecasting is clearly useful—indeed, absolutely necessary given the lags of policy's impact—it is also obviously fallible. As a further complication, policy decisions must be based on multiple forecasts, implicit or explicit, that are conditional on multiple alternative settings of the operating variables under consideration. The well-known limitations on the ability to forecast raise the risk, moreover, that policy makers will find themselves putting undue weight on the current state of the economy despite the acknowledged importance of lags in the process. And of course absent intermediate targets, policymaking procedures do not readily lend themselves to an "objective," quantitative way of communicating the intentions of policy to the public or of evaluating its success after the fact. Even more serious, an approach that relies on setting operating targets in light of projected future economic outcomes fails to provide a "nominal anchor" for policy and does nothing to solve the conflict between period-by-period and long-run optimizing in policy making. 17 So we have a real tension here. On the one hand, intermediate targets, if suitable ones exist, have the potential for improving the overall performance of monetary policy, especially over the longer run. But, to repeat, "suitable" means not merely controllable, but sufficiently tightly related to ultimate goals that slippages can be ignored and thus the forecasting problem bypassed. The experience of the 1980s has left serious doubts that such "suitable" target measures do in fact exist. Faced with this tension, the Federal Reserve has in practice compromised. It has continued to set intermediate targets for money and credit aggregates—as, indeed, it is required to do by law—but it has defined these targets in terms of rather wide ranges (generally 4 percentage points for annual growth rates). Moreover, on occasion the Federal Reserve has felt free to allow even these wide ranges to be violated when it has appeared likely that the targets could be achieved only at the expense of inferior economic outcomes~or at least outcomes that are "inferior" within the one-year time horizon of the current targeting process. The target measures have been given more attention when at the top or bottom of their ranges, with particular attention focused on the behavior of M2. In summary, intermediate targets have continued to exist, but only as rather wide ranges and without any clearly defined means of connecting them with day-to-day or month-to-month operational decisions. Even under these circumstances, the extant intermediate targets have had some value in providing a "nominal anchor"—though one that tends to drag a bit—and have provided a means of connecting, if somewhat loosely, short- to intermediate-run objectives with the longer run objectives for price performance. Nevertheless, it is apparent that their usefulness for these purposes falls far short of what, at least in theory, could be provided by more formal adherence to a satisfactory intermediate target. The broad appeal of the intermediate target concept is such that interest in it seems bound to persist. Research on the subject has continued, both within and without the Federal Reserve System. In particular, some economists at the Federal Reserve Board have developed evidence to suggest that long-run M2 velocity may have retained enough stability to make M2 behavior a useful indicator of the longer run behavior of inflation. In the meanwhile, it is possible that after the major shocks to the monetary aggregates (and possibly also to broad credit measures) from financial innovation and deregulation in the 1980s, these aggregates may settle down to a pattern of behavior that, if changed from earlier decades, has nevertheless again become predictable enough to be useful. The future role of intermediate targets in the policy-making process is certainly likely to depend in part on such potential developments. But given the short-run slippages that would inevitably persist between intermediate targets and ultimate objectives even under the best of circumstances, the future role of 18 intermediate targets probably also depends on the weight that is given to the objective of long-run price stability. It is in the context of such an objective that the potential usefulness of intermediate targets is particularly clear. 19 POSSIBLE ROLES FOR THE MONETARY BASE Ann-Marie Meulendyke The concept of the monetary base has a long history. In essence, the monetary base consists of currency issued by the monetary authority, primarily the Federal Reserve System in this country, and reserve balances held with the monetary authority by depository institutions. The most commonly used measures of the monetary base make an adjustment for reserve requirement changes (see appendix for definitions). Several names have been applied to what is now most often called the monetary base. A number of writers have used the term "high-powered money."1 Gurley and Shaw (1960) introduced the term "outside money."2 These names for the monetary base, perhaps, are more suggestive of why the concept has played a role in policy discussions. The monetary base consists of monetary instruments that can function as reserves of the banking system and are obligations of the government or the central bank. It is readily observable with a short lag. Furthermore, many economists have argued that it should be feasible to control the monetary base because its components are on the Federal Reserve's balance sheet. In addition, because of regulatory and behavioral linkages between the monetary base and various monetary aggregates that have been proposed as intermediate targets, the base's behavior has the potential to affect broader monetary and economic variables. Three different roles have been proposed for the monetary base. The oldest use for the concept has been as an analytical device to explain relationships between gold flows and central bank actions and the behavior of money, credit5 and prices. The second suggested use of the monetary base is as an operating target of policy, most commonly as a means of achieving desired behavior of an intermediate target such as a monetary aggregate. The third type of proposed use of the monetary base is as an intermediate target or as an indicator to guide the central bank in its efforts to achieve its long-run goals of sustainable economic expansion and price stability, especially the latter. *See, for instance, Burgess (1936), Friedman (1959), and Lothian (1976). *They describe "outside money" as an obligation of the government (including the central bank) that is outside the control of the private sector. In contrast, "inside money" is as an obligation of the private sector. 20 The analysis of the possible roles for the monetary base has proceeded on more than one level. In concept, it has appeared in the context of models of money demand, business cycle determination, and the appropriate responsibilities of a central bank. The monetary base has had particular appeal to those economists who have pushed for limiting monetary expansion to promote price stability. Proponents of the quantity theory of money, the classical long-run neutrality of money, and rational expectations have all tended to give prominence to the notion of a long-run anchor for the price level. Many of them have supported the use of the monetary base in that role. Other analyses have focused on more narrowly technical issues associated with using the base in one or another of its potential roles. Topics considered have included the definition of the monetary base, details of its controllability, and institutional and regulatory questions. Overlapping the conceptual and technical sides of the issue have been the questions whether the monetary base can be exogenously determined and whether estimated relationships between the base and intermediate or ultimate policy goals would be sustained if serious efforts were made to control the base. Many of the debates are affected by the time period considered relevant for achieving the goals. In some cases, failure to clarify assumptions on that question has complicated developing an understanding of the issues. This paper examines the three proposed uses of the monetary base concept, considering both conceptual and practical aspects. It evaluates the components of the monetary base and the extent to which the base has the properties attributed to it. The paper then reviews the evidence from experiences with targeting the base or a related reserve measure. In summary, the monetary base concept appears to be useful as a descriptive tool. Considered as a possible operating target, the base would seem to present some problems if efforts were made to control it directly over time periods shorter than one to two quarters. Considered as an intermediate operating guide, the base may be of some value, especially as a warning signal of potential inflationary policies, but the generally weak relationships between the monetary base and ultimate policy goals suggest that it should be used cautiously. A review of experience in Switzerland and West Germany with the use of measures similar to the monetary base as intermediate targets suggests some problems but also a measure of success. I. The Monetary Base as a Descriptive Device Early in the twentieth century, economists were making analytical use of the monetary base concept, although not by name. In 1911, Irving Fisher ([1911] 1963) described the role of currency as 21 a circulating medium and as a major source of reserves to the banks. The other form of reserves was interbank deposits; small banks generally held balances at large banks. Fisher described changes in currency and reserves as the primary source of growth or decline in the quantity of money because, in normal times, the ratio of currency to deposits would be determined by payment practices and customs that did not change easily.3 Furthermore, Fisher argued that prudent banking practice and the high cost to a bank of being unable to meet withdrawals would determine the ratio of reserves to deposits within a fairly narrow range. Fisher anticipated that disturbances to the ratios would occur during periods of transition to new institutional arrangements. He also expected the ratios to shift gradually over the business cycle in response to sticky adjustment of prices and interest rates. Randolph Burgess (1936), an officer at the New York Federal Reserve's open market desk in the 1920s and 1930s, described in 1936 what then seemed to be a common view of the monetary base (which he called high-powered money) as follows: It may be said, from one point of view, that there are in any country two kinds of money, and for the sake of giving them names they may be called high-powered money and low-powered money. The central bank deals in high-powered money, the money which constitutes bank reserves. Historically, this high-powered money has been closely related to a country's basic reserves of gold and currency, though the specific form of this relationship shows wide variations under different banking laws. . . . When the amount of high-powered money increases, the amount of low-powered money tends to increase also, but in multiple relation to the high-powered money, (pp. 5-6) Burgess presented high- and low-powered money as descriptive concepts to help explain how the purchase or sale of gold or securities would lead to growth or shrinkage of commercial bank deposits. Provision of high-powered money by the central bank makes it possible for the banks to create more low-powered money according to what has now become a familiar textbook multiplier model.4 In that model, the creation of deposits is constrained by limits on the ratio of deposits to reserves. The ratio is partially determined by banking practices, as Fisher had suggested, but under the Federal Reserve, the relationship has also been importantly influenced by the reserve requirement ratios that are applied. The relationship further depends upon the ratio of currency to deposits. These relationships can be expressed in equation form as M=mB, where M is the monetary aggregate in question, B is the monetary base, and m is the multiplier linking them. In addition, M is 3 Fisher defined money as currency and coin, but this analytical work was based on such money plus bank deposits. 4 See, for instance, Mayer, Duesenbury, and Aliber (1987, pp. 161-70). 22 defined to equal C+D, where C is currency and D is deposits, and B is defined to equal C+R, where R is reserves in the form of vault cash or balances at the Federal Reserve. The multiplier can be expressed as a function of ratios linking deposits, reserves, and currency as follows: Let k=C/D and r=R/D. Then m=M/B=(C+D)/(C+R). Dividing the numerator and denominator by D and substituting gives m=(k+l)/(k+r). If underlying behavioral characteristics serve to make the ratios k and r stable, then the monetary base multiplier can be a useful device for relating central bank actions affecting reserves to the behavior of money. In that case, a change in the monetary base would be associated with a proportional change in money. If either of the ratios, k or r, shiftsfrequently,however, the multiplier will not be stable. Under these circumstances, money and the monetary base will not move closely together. Furthermore, the level of the reserve-deposit ratio, r, will have other effects. A low ratio implies that a low weight is given to deposits relative to currency in the monetary base, since only that fraction of deposits represented by reserves is counted. Many economists over the years have expressed unease over the implicit weighting scheme that gives relatively heavy weight to currency, because they believe that the behavior of deposits has a major role in influencing economic activity. Nonetheless, during long stretches of time, the two ratios have been sufficiently stable that many analysts have downplayed the worries about the low weight given to deposits. The most dramatic exception to that pattern of stability occurred in the 1930s, at the time when Burgess was writing. Bank failures in the early 1930s periodically raised the demand for currency and made the ratio of currency to deposits highly variable. Moreover, by the time Burgess published his book, excess reserves had risen to unusually high levels; the ratio of reserves to deposits was higher than it had previously been and rather variable. Hence, it is not surprising that Burgess did not consider the possibility of using high-powered money as a guide to policy. Henry Simons, writing in 1934 and observing the problems associated with fractional reserve banking, proposed that the variation in the multiplier could be eliminated if 100 percent reserve requirements were imposed. Under that arrangement, currency and deposits would receive equal weight and shifts between them would have no effect on the relationship between the monetary base and money. If the banks held 100 percent reserves against deposits, then r would presumably be around 1. The multiplier would be approximately (k+l)/(k+l) or 1 regardless of the ratio of currency to deposits. Simons recognized that his proposal was radical. Nonetheless, he believed it would be feasible to separate institutions providing deposits from those providing credit. Milton Friedman (1959) supported Simons' proposal although he believed it was necessary to 23 address its primary weakness. Issuing deposits subject to 100 percent reserve requirements would be costly. Because the banks could not use those deposits as a source of funds for making loans and investments, they would have to charge large fees on deposits to cover the overhead costs associated with them. Alternative instruments that were exempt from the 100 percent reserve requirement would have to fund credit extension. Consequently, those instruments could pay interest instead of incurring charges. Given the relative attractiveness of these alternative liabilities, they would probably evolve into something that closely resembled money. Broader forms of money would then consist of deposits subject to very high and very low reserve requirements, and the high and stable ratio of reserves to deposits that the proposal was designed to achieve would be lost. Friedman recommended paying interest on required reserves to mitigate the cost disadvantage associated with deposits subject to the high reserve requirement. Nonetheless, the instruments funding credit extension would have to be priced attractively, and some risk would remain that they would take on some of the characteristics of money. A less radical proposal for handling the problem of minimizing the variation of the ratio of reserves to deposits was to make reserve ratios uniform according to the nature of the deposit rather than the characteristics of the issuing institution. In the years since most of the literature on this subject was written, steps have been taken to make the ratios more uniform with respect to transaction deposits. However, time and savings deposits are subject to low requirements, and beginning in the 1980s, personal time and savings deposits have been exempt from requirements. For money definitions that include some of these deposits-M2 and M3—shifts between transaction and nontransaction deposits will affect the ratio of reserves to deposits and hence the monetary base multiplier. Whether the existence of low and differing reserve ratios for various types of deposits interferes with the usefulness of the monetary base in either of the roles proposed remains controversial. One proposal has been made to create a monetary base measure that includes artificial reserve ratios higher than those that actually prevail. To a limited extent, the "St. Louis" definition of the monetary base in effect does that since it uses a 12 percent reserve ratio against transactions deposits (see appendix) rather than the actual reserve ratio, which has recently been around 8 percent. The German concept of the central bank money stock (discussed below) also uses a higher than actual reserve ratio. In neither case was the step taken explicitly for the purpose of raising the weight on deposits. The results fell out of efforts to minimize distortions imposed by changes in the level and structure of reserve requirement ratios. 24 II. The Monetary Base as an Operating Target The case for using the monetary base as a tool in the policy process was made in the 1960s in a series of books and articles promoting the control of a monetary aggregate as a means of achieving the longer run monetary policy goals of price stability and an expanding economy. In this role, the monetary base is, in the terminology most commonly used in this literature, an operating target. The base would be controlled not for its own sake but in order to achieve desired behavior of another variable, usually a monetary aggregate. By the 1960s, when monetary aggregate targeting became a popular objective, sufficient stability had returned to the ratios of reserves to money (Chart 1) and currency to money (Chart 2) to make the arguments for such an approach to targeting plausible, although many analysts remained skeptical. (See the paper by John Wenninger in this volume.) Milton Friedman (1959) proposed targeting a monetary aggregate and suggested that manipulating high-powered money might be a reasonably effective way to attain the monetary targets. Karl Brunner and Alan Meltzer (1968) made a similar proposal and explored several aspects of the multiplier relationship. In several articles that appeared in the St. Louis Federal Reserve Review, Andersen and Jordan (1968a, 1968b) and Burger, Kalish, and Babb (1971) developed in some detail the suggestion to target the monetary base. The St. Louis articles presented multiplier relationships between the monetary base and Ml similar to those described above, although the relationship was more elaborately drawn.5 While the economists promoting control of the monetary base recognized the problems associated with the differential weights given to currency and deposits, they doubted that these difficulties would prove to be serious in practice. At the time they were writing, reserve requirements were high enough to discourage banks from holding large quantities of excess reserves. The economists believed that the presence of binding required reserve ratios would make the ratio of reserves to deposits relatively stable as long as adjustments were made whenever the Federal Reserve changed the specified ratios. Building on Fisher's arguments, they expected payment conventions and the absence of banking crises to provide stability to the ratio of currency to deposits. Empirical analysis of the data covering the 1950s and 1960s generally gave some support to their expectation. Burger, Kalish, and Babb presented such an examination and recommended a control procedure in 5 A variety of multipliers have been computed that make separate allowances for transactions deposits and time deposits. Johannes and Rasche (1979) built an elaborate multiplier model that has been regularly updated for Shadow Open Market Committee meetings. See also Rasche and Johannes (1987). 25 11 [• 10.5 h r ^ \y\ 10 H 9.5 i i Chart 1 Total Reserves as a Percentage of M1 8.5 \- i h | 7.5 l i i i l i i i i i i i l i n l u i l n i iiilnil i i i l i i i l i i i l i i i l i i i l i i i i i i i lui 1960 1965 1970 1975 i n l i i i h I I I I I I I M I " i l i i i l i i i l i n JJJ 1980 Note: Reserve data are seasonally adjusted and are adjusted for changes in reserve requirements. M1 values are seasonally adjusted. Ratios are plotted quarterly. Chart 2 Currency as a Percentage of M1 30 h 28 L 26 h • 24 h r • i 22 20 h I n I M I iiiiiniiiii . . . I M . I > H I I M I . I I I I I I I n I I I I I I M I I n I I n I 11 I n M I I n I I I I I I I I I I I I I I I I 11 I I I I I I I I I I I 1 I I,11 lI .In lI ,I 111I MI JI M . L M I M . I M J M . I M , 1955 1960 1965 1970 Note: M1 values are seasonally adjusted. Ratios are plotted quarterly. 1975 1980 1985 1989 which the monetary base would be targeted to achieve desired growth in Ml. They proposed estimating the multiplier from recent behavior of its constituent ratios. Then they compared their model forecasts of the multiplier with the actual values of the multipliers. They found the errors to be small in size and observed that they were not cumulative. They concluded that if the proposed monetary base targets were achieved and the multipliers were the same as those that actually occurred, then money growth would have deviated only slightly from a smooth path. From these results, they argued that following their procedures would produce a reduction in unwanted variation in money growth. The Burger, Kalish, and Babb proposal would have required the central bank to control the monetary base. Most of the supporters of the proposal argued that the central bank could control the monetary base since the base consisted of items on the central bank's balance sheet that could be observed with at most a one-day lag (Balbach 1981). Others attacked the proposal, suggesting that the consequences of trying to control the base would be undesirable because the observed multiplier relationships on which such proposals were based were estimated when the monetary base was, in fact, determined endogenously. The critics predicted that the relationships would break down if the base were actually controlled. To explain the issues involved, it is helpful to review the potential methods for controlling the monetary base. As the appendix indicates, the monetary base can be viewed as consisting of three elements: currency in circulation, total reserve balances of depository institutions at the Federal Reserve, and a factor that adjusts for the effects of changes in reserve requirement ratios. The adjustment factor does not introduce control problems, except possibly briefly at those few times that the Federal Reserve changes reserve requirement ratios. Currency, the biggest component of the base, has been issued by the Federal Reserve to depository institutions on demand. The Federal Reserve debits the reserve accounts of a depository institution when it issues the currency. Limiting currency issuance would be a sharp break with a tradition that extends back to the beginnings of the Federal Reserve. Most of those proposing control of the base did not contemplate a limitation on Federal Reserve issuance of currency, but instead advocated offsetting undesired currency movements with increases in or restrictions on the provision of total reserves through open market operations (Burger, Kalish, and Babb 1971). The Federal Reserve would have precise knowledge of the amount of currency it had issued, and therefore would know the size of offsetting adjustments in reserves needed as soon as any unwanted currency movements took place. 27 In practice, the process is not so simple. Total reserves, the factor that would need to be manipulated to achieve a desired level of the monetary base, consists of reserves provided by open market operations—nonborrowed reserves-and reserves obtained at the discount window—borrowed reserves.6 If, for example, the Federal Reserve attempted to control the monetary base by offsetting undesired expansion of currency through a reduction in nonborrowed reserves, depository institutions might borrow at the discount window to obtain the reserves lost, lifting total reserves back to the level where they stood before the reduction in nonborrowed reserves occurred. Consequently, it would not be possible to control the monetary base even though it would be possible to control what is often referred to as the nonborrowed base—the monetary base minus borrowed reserves. Whether depository institutions would in fact borrow at the discount window to offset the reduction in nonborrowed reserves would depend upon their demand for total reserves. Depository institutions hold total reserves in order to meet reserve requirements on average over a reserve maintenance period (which has been two weeks long since 1984, but was one week long between 1968 and 1984, when much of the discussion of controlling the base took place). Some depository institutions may also choose to hold reserves in excess of requirements if day-to-day flows through reserve accounts are large relative to their reserve requirements.7 The cost of close management of reserves to avoid being unexpectedly short may exceed the interest lost from holding some excess reserves. If the total reserve component of the monetary base target is to be achieved, a decline in nonborrowed reserves cannot be offset by an increase in borrowed reserves. A reduction in nonborrowed reserves can only result in an equal decline in total reserves if the lower level of nonborrowed reserves is consistent with existing demands for reserves to meet requirements—because excess reserves are larger than depository institutions desire-or if depository institutions adjust loans and deposits to reduce required reserves by the full amount of the reduction in nonborrowed reserves. In order to lower required reserves, deposits would have to fall by a multiple of the desired decline. Whether such sharp adjustments to deposits over a short time period are feasible has been technically, the Federal Reserve cannot, through use of open market operations, achieve desired nonborrowed reserve levels with precision because there are a number of factors on its balance sheet, such as Treasury cash and Federal Reserve float, that it does not control and can only observe after the fact. Most of the time, forecasts of those factors are sufficiently accurate that average errors in reserves for a reserve maintenance period are small. Occasionally, however, they are large. 7 In equation form: total reserves = nonborrowed reserves + borrowed reserves = required reserves + excess reserves. 28 extensively debated. Those favoring short-run control of the base generally argued that depository institutions could make quick adjustments to deposits if they were given the incentive to do so. They believed that any additional nonborrowed reserves provided would quickly be lent or invested, setting in motion a multiple expansion process that would raise deposits until the unwanted excess reserves had disappeared. In a parallel fashion, they believed that depository institutions would respond to a shortage of nonborrowed reserves by contracting loans and deposits until required reserves had shrunk to the point that they were consistent with existing supplies of nonborrowed reserves. They argued that regulations that weakened the money multiplier process were serious impediments to such adjustments. In particular, they cited the relatively low discount rates (or alternatively the variable spread between the discount rate and the federal funds rate) and lagged reserve requirements that interfered with banks' ability to adjust required reserves within a reserve maintenance period.8 (Between 1968 and 1984, there was a two-week lag between the reserve computation and maintenance periods. Since 1984, required reserves on transactions deposits have been quasi-contemporaneous, with a two-day lag at the end of the period.) Another impediment to monetary base control was the way the discount window functioned. Many of the supporters of base control suggested that the discount window be closed or that the discount rate be set high enough to ensure that borrowing would be costly. The latter option would not make the monetary base precisely controllable, since banks still could borrow reserves in excess of those consistent with the target, but the high cost of doing so would discourage borrowing and make the monetary base approximately obtainable. The severe restrictions on borrowing would be presumed to limit deviations between the monetary base and the nonborrowed base. Other analysts claimed that removing the impediments would not solve the problems of controlling the base. They argued that depository institutions could not easily make the large adjustments to loans and investments or to deposits needed to achieve the required reserve levels consistent with the desired monetary base. They reasoned, first, that each individual institution would have difficulty gauging whether an excess or shortage of reserves relative to its requirements resulted from a deliberate policy adjustment, from a poor distribution of reserves in the banking system—either among depository institutions or over the maintenance period~or from a Federal Reserve error in predicting reserve balances on the day. Only the first of those sources would induce depository institutions to consider taking the steps needed to change required reserves; changes in reserves from ^ornton (1982) discusses these arguments. 29 the latter two sources would be quickly reversed, and banks would attempt to maintain existing levels of reserves and deposits. Each individual depository institution would have trouble discovering the true reserve picture from observing its reserve levels because flows of reserves are huge relative to average reserve balances. Many large banks routinely experience reserve positions that within a day range from being overdrawn to being substantially in excess, even when overall reserve levels are consistent with requirements. Second, even when a depository institution correctly perceived a policy-based reserve shortage or excess, its most likely response would be to adjust its loan and deposit rates and fees gradually.9 There are costs to making price changes because they have to be disseminated and because they may complicate efforts to achieve good relations with customers. In addition, customers would not respond immediately to the changes in pricing unless the short-run interest elasticity of demand for deposits was very great. But most observers have perceived the short-run interest elasticity of deposits to be relatively modest. In that circumstance, sharp swings in interest rates would have to occur to achieve the necessary changes in deposits. A depository institution facing a reserve shortage or excess might try to pass it to another institution by adjusting its activity in the federal funds market. If all depository institutions did that simultaneously, the federal funds rate would show large short-run movements. Some observers argued that those large swings in rates would induce instability in the demand for money through a complex lagged adjustment mechanism. Thus, money would not respond as hoped to a reserve excess or shortage, but would exhibit a series of overshoots and reversals. Both the targeted monetary aggregate and interest rates would be subject to undesirable short-run volatility. To the extent that the supporters of strict monetary base targeting addressed this issue, they suggested that depository institutions would respond by building up their normal holdings of excess reserves. Once the institutions held enlarged levels of excess reserves, they could allow excesses to vary in the short run whenever required reserves and nonborrowed reserves were inconsistent. Those excess reserves would serve as a shock absorber to mitigate the short-term fluctuations in interest rates. If excess reserves became large and variable, those opposing monetary base targeting argued, the reserve-deposit ratio would become variable and the multiplier would cease to have the relatively stable relationship observed in the historical data generated when excess reserves were small and the 'Demand deposits cannot pay explicit interest. Depository institutions may pay interest implicitly by extending services for which they do not charge. Since rates on NOW accounts were decontrolled in 1986, interest paid on them has been adjusted very slowly if at all. 30 monetary base was not a target. If the multiplier were unstable, controlling the base would not achieve desired money targets.10 For the most part, the proponents of base targeting rejected these objections. They believed that the multiplier was structurally stable because of mandated reserve ratios and underlying customary behavior patterns and institutional structures. Hence, they anticipated that the ratios would either maintain or quickly return to their norms (Brunner and Meltzer 1983). If the ratio of currency to deposits is determined by the structure of payment practices, then currency will be induced to shift in the same direction as the change in deposits, taking some of the adjustment burden off reserves. One issue rarely mentioned in the debate that probably deserved more attention was the appropriate time period over which it was desirable to achieve control of the monetary base. Very short-run control of the monetary base was never advocated for its own sake. Instead, it was seen as a means to achieve a desired path for money and in turn for economic activity and prices. Many of the supporters of a monetary base approach to money control argued for short-run targets for two reasons. First, they were afraid that if deviations were permitted, the central bank would allow misses of sufficient duration to affect adversely economic activity and prices. Second, they believed close control was feasible at low cost.11 Their critics had a range of views with regard to the first point but, as indicated above, disagreed with the second. Many of the operational and stability concerns associated with monetary base targeting would be reduced if the time period for achieving the target were lengthened beyond the single reserve maintenance period cited in most of the control proposals. Presumably, the monetary base could be controlled with only modest errors over a one- to two-quarter horizon if nonborrowed reserves—or the federal funds rate—were manipulated to bring the monetary base back on track once evidence developed that it was deviating significantly from a desired path. Technical control should be easier than it is for Ml or M2 because the base does have a closer relationship to reserves and can be observed accurately with no lag. The more meaningful question is whether the monetary base can be controlled in a way that does not introduce short-run instability to interest rates and money demand. Without actual experience, it is difficult to answer that question. 10 Carlson (1981) describes the two viewpoints. "Thornton (1983), generally a supporter of monetary base targeting, recognized that very short-term control was not feasible because loan and investment decisions, under any reserve accounting scheme, would not be closely linked with current reserve levels. Thus, the adjustment would not occur instantly. He did not think that the delays would be a problem because banks would adjust over meaningful periods. His arguments on this point are not typical of the literature supporting base targeting. 31 Nonetheless, the longer the control horizon, the more likely the answer will be that control of the base without unacceptable instability in rates and money demand is indeed possible. Although most proposals concerning the monetary base made since the early 1980s have focused on its use as an intermediate target rather than as an operating target, one exception has been the work of McCallum (1987, 1988a, 1988b, 1988c). Influenced by the breakdown of the relationship between Ml and GNP in the 1980s, he suggested two modifications to the common monetary targeting proposal. First, he suggested that nominal GNP could serve as an intermediate target as long as it was chosen to ensure that inflation could not be rapid. Second, he advocated an adaptive policy rule that contained the means for the procedure to recover when underlying relationships between the base and GNP shifted as a result of deregulation or other developments. (See the paper by Spence Hilton and Vivek Moorthy in this volume.) McCallum suggested that the monetary base could serve as the operating instrument. He seemed to argue that the base could be controlled directly on a day-to-day basis, although he only advocated a quarterly average growth target. McCallum did not discuss his assertion that the monetary base could be controlled. His equation for quarterly target growth of the monetary base consists of three terms.12 The first is a constant, equivalent to the desired trend growth in nominal GNP-in his preferred specification, 3 percent at an annual rate. The second term subtracts from the constant the average increase in monetary base velocity over the previous four years. This term introduces a gradualresponseto changes in monetary base velocity so that the model will adapt to changing trends. The third term provides for a partial response to deviations of GNP from its constant growth target. McCallumrecommendeda 25 percent per quarter adjustment factor in response to such deviations. in. The Monetary Base as an Intermediate Target of Policy The monetary base has been proposed as an intermediate target in place of a traditional monetary aggregate in order to resolve the possible short-run control problems associated with such an aggregate or to treat the monetary base itself as a narrow monetary aggregate. The idea that the monetary base could be considered a monetary aggregate developed gradually. Lothian (1976) found that the monetary base showed a more consistentrelationshipwith net national product than did a broad monetary aggregate across groups of twenty-nine or forty countries (using two different 12 In equation form: bt - b^ = 0.00739 - (l/16)[xt_, - x ^ - bM + bM7] + .25 (x\x - xM), where b is the log of the monetary base and x is the log of GNP. An asterisk indicates a target value. 32 formulations). He posited that differential interest rates, inflation, and regulations such as interest rate ceilings and reserve requirements had a larger influence on the demand for deposits than they did on the demand for non-interest-bearing central bank monetary assets. Drawing on the econometric techniques used to evaluate monetary aggregates as potential intermediate targets, analysts have tried to test empirically the ability of the monetary base to serve in the role of intermediate target for the United States. The traditional approach has been to estimate a model to see how well it predicts the behavior of economic activity or prices. Although one may question whether using a model built to accommodate the observed characteristics of one monetary variable is appropriate for evaluating monetary variables with a different set of characteristics, the approach has been widely used. Differences in methodologies and assumptions have yielded conflicting results. A. Studies in Which GNP Is the Dependent Variable A number of authors have estimated "St. Louis-type" models, substituting the monetary base for Ml or M2. Models of this type were developed at the St. Louis Federal Reserve in the late 1960s by Andersen and Jordan and updated and modified in subsequent years. In these models, a measure of nominal income (generally GNP) is regressed on a measure of money and a measure of fiscal actions (often high employment expenditures). Initially the models were estimated in first difference form, although later the variables were typically specified as rates of change. Sometimes variables have been added to allow for the effects of strikes and other shocks, such as the two large increases in oil prices in the 1970s. A series of studies were made in the late 1970s and early 1980s that compared the models' performance when the monetary base was used as the monetary aggregate with the models' performance when Ml was the aggregate. The models were estimated with data covering the previous twenty to thirty years. The models generally achieved better fits for Ml than for the monetary base, although the differences ranged from trivial to very large depending on specification and time period. A few of the authors offered simulations after the estimation period that tended to provide predictions of similar quality for the monetary base and Ml. Andersen and Karnosky (1977) used a simplified version of the St. Louis equation from which they omitted the fiscal variable (because the sum of coefficients on the fiscal variable was usually insignificant). They estimated their model from 1952 to 1961, using quarterly data, and then extended it year by year through 1975. (They did not report their regression results.) In each case, they used the estimated model to simulate the next four quarters' GNP. They found that in the equations using Ml, the errors had modestly 33 lower variance than the errors in the equations using the monetary base. However, the equations using the monetary base achieved lower mean errors and mean absolute errors. The authors concluded that it would be worthwhile to consider using the monetary base as an intermediate target because its forecasting performance was only slightly worse than Ml, and it was easier to control. Davis (1979-80) ran similar simplified St. Louis model regressions, which also omitted the fiscal variable. His regressions used a sample period from 1961 to 1978 and two subperiods. He found significantly better fits for Ml than for the monetary base, using as measures of goodness of fit either R2 adjusted for degrees of freedom or the standard error of the estimate. (Table 1 contains highlights of reported results for this and other studies.) He also estimated similar equations using either the currency component or the reserve component of the monetary base. The limited relationship he found between the monetary base and GNP seemed to derive from the currency component, not the total reserve component. Cullison (1982) argued that the monetary base equation in Davis* formulation was misspecified because it contained the current value of the monetary base, which was largely endogenous over the sample period. He also criticized the choice of sample period, arguing that the starting and ending points should coincide with the same stage in the business cycle to allow for the cyclical pattern in the coefficients. His equations used 1959 as a starting point and ended in 1969, 1973, and 1979 (all cycle peaks). His results show much less of a gap between Ml and the monetary base, if judged by the values for adjusted R2. His two formulations, with and without the current value of the monetary base, give similar results. Gambs (1980) estimated somewhat different models from those used by Davis. He made his estimates over a longer time period (1953-78) and included a variable for high employment expenditures, an approach closer to the traditional St. Louis model. The monetary variables, especially the base, explained larger portions of the variation in GNP than they did in Davis9 studies, but Gambs still observed a sizable differential between Ml and the monetary base. Hafer (1984) made a study using a sample period running from 1960 to 1980. He used a more elaborate version of the St. Louis model that included high employment expenditures and special factor variables for oil price impacts, strikes, and wage and price controls. He allowed coefficients to vary over time. With GNP as the dependent variable, he achieved better overall results, presumably because additional variables were included. He found that the version using Ml showed only a slightly better fit than the version using the monetary base. Benjamin Friedman and Kuttner (1989), using data available in 1988, reestimated equations 34 Table 1 Nominal GNP Equations: Selected Model Results Davis (1979-80) and (1990) GNP, = a + I&nvi for i = 0.....4. 1961-1 to 1978-IV 1970-1 to 1978-IV 1961-1 to 1969-IV (Davis 1990) 1981-1 to 1989-IV Financial Measure R2 SEE R2 SEE R2 SEE 52 SEE Ml 0.31 2.96 0.23 2.25 0.19 3.64 0.06 3.67 M2 0.25 3.07 0.27 2.19 0.08 3.88 0.02 3.76 MB-B 0.08 3.41 0.01 2.56 -0.06 4.16 0.16 3.48 Total reserves -0.02 3.59 0.19 2.31 -0.11 4.26 - - Currency plus nonmember bank cash vault 0.11 3.35 0.13 2.40 0.01 4.03 - - Cullison (1982) GNP, = a + E^im,.! for i = 0....4. 1959-IV to 1969-IV 1959-IV to 1973-m 1959-IV to 1979-IV R2 SEE R2 SEE R2 SEE Ml 0.360 0.0066 0.364 0.0072 0.371 0.0078 MB-SL 0.230 0.0072 0.280 0.0076 0.290 0.0083 Financial Measure GNP, = a + Iib,.iMB,.i for i = 1,2. 1959-IV to 1969-IV 1959-IV to 1973-m 1959-IV to 1979-IV Financial Measure R2 SEE R2 SEE R2 SEE MB-SL 0.270 0.0070 0.270 0.0077 0.260 0.0085 35 Gambs (1980) GNP, = a + Zjbim,.! + I^E,., for i = 0.....4. 1953-1 to 1978-IV R2 SEE Ml 0.41 3.74 1.18(6.53) MB-SL 0.31 4.04 0.94(5.67) Financial Measure 5^b,., Hafer (1984) GNP, = a + p, XJwnn + fc I , . ^ + Z, for i j = 0.....4 1960-1 to 1970-11 1970-m to 1980-IV Financial Measure R2 SEE Ifo.; R2 SEE 2^),., Ml 0.479 2.320 1.142 (3.31) 0.469 2.992 1.361 (2.92) MB-SL 0.468 2.345 0.955 (3.42) 0.460 3.018 2.191 (2.99) Friedman and Kuttner (1989) GNP, = a + Ijbim,.! + Z^E,., for i = 1.....4. 1960-n to 1979-m 1960-n to 1986-IV 1970-1 to 1986-IV Financial Measure R2 R2 52 Ml 0.32 0.11 0.02 MB-B 0.23 0.10 0.02 M2 0.27 0.19 0.06 36 Board Staff (1988) GNP, = a + 2ibini,.i + EftE^ for i = 0.....8 (or less). 1961-n to 1979-IV R2 SEE Ml 0.21 3.4 MB-B 0.06 3.7 Ml-A 0.19 3.4 M2 0.18 3.5 Financial Measure Stone and Thornton (1988) GNP, = a + ZibiiiVi + IjdjGNP,.j 1961-1 to 1980-IV Financial Measure W for i = 0,...,4 and j = 1,2. 1961-1 to 1987-11 R* 1981-1 to 1987-H^ W Ml 0.356 0.159 0.348 MB-SL 0.230 0.156 0.422 MIA 0.334 0.142 0.468 M2 0.199 0.179 0.295 Total reserves 0.204 0.132 0.320 - For this sample period, i = 0,1,2 and j = 1. 37 Naive Model GNPt = a + 2jd,GNPt.j for j • 1,...,7 (first two columns) for j = 1,...,4 (third column) 1961-1 to 1980-IV 1961-1 to 1987-11 52 0.069 R2 0.109 1981-1 to 1987-11 R2 0.244 Notes: t-statistics are in parentheses. Notation changed from that of the author to make variables consistent. GNP m E MB-SL MB-B Z = = = annual rate of change in nominal GNP annual rate of change infinancialmeasure annual rate of change in high employment expenditures = annual rate of change of monetary base-St Louis measure = annual rate of change of monetary base-Board of Governors measure = energy and strike variables and price control dummies. similar to those of Gambs for 1960 to 1979. They found a similar differential in the performance of Ml and the monetary base, but generally weaker relationships to GNP, presumably because of the different sample period. In the early 1980s, the seemingly dependable relationships between the various monetary measures and GNP underwent substantial shifts. Two developments appeared to be primarily responsible. The first was the deregulation of interest rates applied to a variety of types of deposits. Such deregulation made it more attractive to hold wealth in money form and led to a downward shift in the velocity of money. The second development was the fundamental reduction in the trend rate of inflation, which also lowered velocity. It was suggested that deregulation would have less effect on the monetary base than on Ml, 38 since neither component of the base paid interest in the earlier years or in the 1980s.13 Nonetheless, the monetary base was somewhat affected by deregulation since induced changes in transactions deposits affected its total reserves portion. While the currency component of the base might remain stable, any instability in total reserves would complicate targeting and control. Presumably, the reduction in the trend rate of inflation, which increased the demand for money generally, would affect the demand for the monetary base as well as Ml and M2. In any case, as the 1980s unfolded, the data strongly suggested that structural shifts had in fact occurred. The shifts were most dramatic for Ml, as the charts of income velocity in the introductory essay in this volume illustrate, but they also occurred in the monetary base velocity relationship. It would be hardly surprising in the face of such patterns that extending the observed log-linear relationships underlying the St. Louis equations would produce unsatisfactory results. Nonetheless, a number of authors pursued such an approach. Friedman and Kuttner (1989) ran the same model cited above, extending the sample period to 1986. For sample periods running from 1960 to 1986 and 1970 to 1986, the proportions of the variation in GNP that were explained by either Ml or the monetary base were very low. Staff members at the Board of Governors (1988) estimated reduced-form equations that included a monetary variable and a measure of fiscal stimulus for the period from 1961 to 1979. The sum of the coefficients on the monetary variable was constrained to equal one to be consistent with long-run neutrality of money. The estimates using the monetary base were considerably poorer than those for Ml, Ml-A, or M2. The staff then simulated the models for the postsample period from 1980 to 1988. The simulations showed large errors for all measures.14 The monetary base showed bias, but the error statistics were better than those on the standard narrow money definitions and similar to those for M2 and a shift-adjusted version of Ml-A. Stone and Thornton (1988) used rates of change of various money measures and lagged values 13 Gambs (1980) presents the argument without endorsing it. The Shadow Open Market Committee recommended monetary base targeting, partly to deal with deregulation. 14 Summary statistics for Board staff simulations for 1980-1 to 1988-IH: Mean absolute error Root mean squared error Bias (mean error) SH=shift-adjusted Ml 6.1 7.4 -4.5 Base 4.8 5.6 -3.3 Ml-A 5.1 6.7 -.2 MI-ASH 4.5 5.7 -1.4 M1SH 6.1 7.5 -4.1 M2 4.3 5.5 -1.6 M2SH 3.9 5.1 -1.2 39 of GNP as independent variables and obtained results that were generally similar to those obtained by others over the initial sample period of 1961 through 1980. Extending the period of estimation to 1987 weakened the explanatory power but greatly reduced the differences among monetary measures. Furthermore, out-of-sample simulations from the first estimations, over the period from 1981 to 1986, showed considerable deterioration with respect to both Ml and the monetary base. Stone and Thornton went on to reestimate the models, using data generated during and after the structural shifts described above. When the authors reestimated the model with 1981 as the starting date (and reduced the lags to gain degrees of freedom), the values for adjusted R2 actually improved compared with those obtained for the earlier period, and the monetary base outperformed all of the other measures except MIA, to which it was a close second. They argued that the monetary base could appropriately play a role in the policy process. In the introduction to this volume, Davis offers updated versions of some of his earlier equations, now estimated over 1981-89. He achieves the same ordering as Stone and Thornton, with the base improving and Ml deteriorating relative to the earlier period. But the values for adjusted R2 and the coefficients are very different from those found by Stone and Thornton. While the time periods are not identical, the main source of the difference is the inclusion or exclusion of lagged values of nominal GNP with the independent variables. (The effects of alternative specifications are discussed below.) Milton Friedman (Darby et al. 1987) performed a slightly different exercise and reached a similar conclusion. He calculated eight-quarter moving standard deviations of the quarterly change in income velocities of Ml, M2, and the monetary base, both before and after the structural shift of the early 1980s. He found no increase in the size of the standard deviations of these measures in the 1980s, suggesting that velocity had not become more unstable. He claimed that monetary base velocity had the smallest standard deviation of the three measures, although the differences among them were not dramatic. (He presented his results in a graph. Visually, Ml appears to have deteriorated somewhat, while it is hard to tell about the other measures.) I repeated the calculations he described using data covering 1972-80 and 1982-89. The first time period is shorter than that examined by Friedman, who began his exercise with 1961. The second time period is longer; Friedman's sample ended in 1986. My calculations show some deterioration for the standard 40 deviations of Ml and M2 velocity and some improvement for the monetary base.15 B. Studies in Which Prices Are the Dependent Variable Another pair of studies (including one cited above) examined the impact of the choice of monetary variable on the rate of change in prices rather than on nominal income. Unfortunately, these studies only examined the relationships through the 1970s. Compared with the GNP equations over similar sample periods, the monetary base did relatively better, and in one of the studies it came out ahead of Ml. Hafer substituted the rate of change in prices for GNP in equations that included longer lags than his GNP equation. He observed a stronger relationship between the monetary variables and prices than between the monetary variables and GNP. Ml still outperformed the monetary base, but the difference narrowed (Table 2). Fama (1982) used a different approach from the others, and concluded that the monetary base was superior to Ml as a monetary measure. Fama developed a model based on Fisher's work and various rational expectations models in which the long-run quantity theory was assumed to hold. He posited that the rate of change in the demand for real money was positively related to the rate of change in anticipated real activity and negatively related to the nominal interest rate (assuming money does not pay interest). He then made the assumption that real economic activity is determined outside the money sector (in keeping with the long-run neutrality of money hypothesis). From these assumptions, he built a model to test the performance of alternative definitions of money. He ran a number of variants of the model, estimated from 1954 to 1976, first using annual data and then quarterly or monthly data. Inflation was expressed as a function of the monetary variable and nominal interest rates (both of which were expected to have positive coefficients) and as a function of actual and anticipated economic activity (expected to be negatively related). When the three-month Treasury bill rate proved insignificant, Fama dropped it. He used current money and money lagged one period. Initially, he tried money with a lead as well, on the assumption that it might serve as a proxy for the expected course of monetary policy, but dropped it when the 15 Standard deviations of eight-quarter moving averages of logs of velocity of various monetary measures: Mi Mi MB-B MB-SL 1972-80 0.02333 0.01388 0.01578 0.01663 1982-89 0.02587 0.01614 0.01238 0.01252 41 Table 2 Price Equations: Selected Model Results Hafer (1984) P, = Eibimt.i + Z, for i = 0.....20 1960-1 to 1970-n 1970-ffl to 1980-IV Financial Measure W SEE±' 2,b,.| R2 SEE^ 2,bt., Ml 0.737 0.896 0.942 0.681 1.278 0.972 MB-SL 0.671 1.002 0.848 0.504 1.593 0.749 Fama (1982) P, = a + £,b,nit, + EjCjA^ Financial Measure Economic Activity Measure Ml for i = 0,1. *? R2 SEE rr* rIP 0.79 0.0130 0.42 0.11 Ml rGNP 0.80 0.0126 0.50 MB-SL rIP 0.91 0.0083 MB-SL rGNP 0.95 DD rIP DD rGNP £.b, j - -0.03 0.92 -1,0,1 0.12 -0.07 0.90 -1,0 0.06 -0.08 -0.01 0.71 -1,0,1 0.0064 0.38 0.07 -0.12 0.72 -1,0 0.65 0.0167 0.52 0.22 0.02 0.85 -1,0,1 0.65 0.0166 0.55 0.21 0.00 0.82 -1,0 Notes: Notation changed from that of the author to make variables consistent. P m Z Ml MB-SL A rIP rGNP DD = = = = = = = = = annual rate of change in prices (Hafer: GNP deflator, Fama: CPI) annual rate of change in financial measure energy and strike variables and price control dummies annual rate of change in Ml annual rate of change in monetary base-St. Louis measure annual rate of change in economic activity measure annual rate of change in real industrial production annual rate of change in real gross national product annual rate of change in demand deposits. -;SEE %rk = standard error of the estimate. = residual autocorrelation coefficient (k = 1,2,3). 42 coefficients proved insignificant. He measured economic activity alternately with industrial production and real GNP. Current economic activity was included along with activity with a one-period lag and a one-period lead. The leading variables (which had highly significant negative coefficients) were justified as proxies for anticipated real economic activity. Expecting objections to the use of leading variables, he also proxied anticipated activity with a stock price index, which gave a poorer fit but had the same sign and did not substantially change the coefficients of the other variables. He ran the various models, alternately including the (St. Louis) monetary base, Ml, or demand deposits as the monetary variable. In each model variant, the base dominated both Ml and demand deposits as measured by adjusted R2, the standard error of the regression, and lagged residual autocorrelations. He reran some of the models using quarterly and monthly data and achieved the same ordering of results. Hence, he concluded that the monetary base was the appropriate monetary variable to follow to achieve a price goal. His model, which is complex, does not appear to have been updated to include the 1980s. McCallum cited Fama's work as justification for using the monetary base as a target variable. It is unfortunate that he did not offer an update of the price relationship. C. Empirical Estimates of GNP and Price Relationships To gain some additional insights into relationships using the rate of change in nominal GNP as the dependent variable in the 1980s, I estimated a variety of regressions similar in formulation to many of those reported here. Two time periods were used, 1972-80 and 1981-89. To allow for the observed structural shifts in the early 1980s, it was necessary to use relatively short time periods for this type of estimation. Hence, the results must be interpreted very cautiously. The exercises suggest some ratios arising from sensitivity to the specifications used. Including or excluding lagged GNP or the contemporaneous monetary variable made sometimes sizable differences in the results, as did specifications of the lag structure (Table 3). Results can be judged by values of adjusted R2 or by the plausibility of sums of coefficients (which for most of the monetary variables might be greater or less than one but which should not differ dramatically from one). In all specifications and under either definition of the monetary base, the base performed considerably better in the later period than in the earlier period. Indeed, it generally showed a better relationship to GNP in the later period than Ml showed in the earlier period. Ml was the only monetary variable tested that showed reasonable results in the earlier period. The deterioration in Ml as a predictor of GNP between the 1970s and the 1980s is dramatic if judged by sums of coefficients but modest if judged by error statistics. The M2 equations do not appear to be 43 Table 3 * A Summary of Selected Regressions Using Nominal GNP as the Dependent Variable Time Period of Estimate: 1972-80 Unconstrained Mon Variable with Lagged GNP; Contemp & Lagged Money Unconstrained Mon Variable with Lagged GNP; Lagged Money Only Poly Distr Lag Poly Distr Lag Not Tied; Not Tied; with Lagged GNP; with Lagged GNP; Contemp & Lagged Lagged Money Only Money Poly Distr Lag Poly Distr Lag Not Tied; Not Tied; No Lagged GNP; No Lagged GNP; Contemp & Lagged Lagged Money Only Money Poly Distr Lag Tied Far End; Poly Distr Lag Tied Far End; with Lagged GNP; with Lagged GNP; Contemp & Lagged Lagged Money Money Only Poly Distr Lag Tied Far End; No Lagged GNP; Contemp & Lagged Money Statistical Measure M1 AdjR2 St err regress Sum coef .38 3.61 1.23 .07 4.43 .62 .41 3.51 1.28 .15 4.22 .74 .40 3.55 .87 .11 4.33 .09 .37 3.62 1.33 .17 4.17 .78 .38 3.62 .95 M2 AdjR2 St err regress Sum coef .05 4.47 .58 .08 4.4 .56 .09 4.38 .56 .08 4.38 .55 .11 4.33 .49 .11 4.34 .44 .11 4.32 .57 .11 4.32 .56 .13 4.27 .50 St err regress Sum coef -.1 4.81 -.03 -.1 4.83 -.46 -.13 4.88 -.19 -.08 4.77 -.46 -.10 4.81 -.26 -.84 4.78 -.50 -.1 4.80 -.18 -.11 4.82 -.53 -.07 4.74 -.25 Total reserves AdjR2 St err regress Sum coef -.11 4.84 -.15 -.07 4.75 -.15 -.08 4.77 -.15 -.07 4.75 -.15 -.03 4.65 -.15 -.03 4.65 -.13 -.05 4.70 -.15 -.04 4.67 -.14 -.01 4.60 -.15 -.06 4.72 .02 -.09 4.8 -.67 -.02 4.64 .06 -.09 4.80 -.67 .03 4.52 .06 -.04 4.69 -.71 .01 4.57 .04 -.07 4.74 -.65 .06 4.45 .04 -.08 4.76 1.37 -.11 4.84 .41 -.04 4.68 1.39 -.11 4.84 .42 -.01 -.06 4.73 -.01 4.60 1.37 -.08 4.76 .44 .02 4.54 1.09 Currency AdjR2 Mon base—Board AdjR2 St err regress Sum coef Mon base—St Louis AdjR2 St err regress Sum coef Note: Polynomial distributed lags are third order. 4.60 1.15 .20 Table 3 - Continued Time Period of Estimate: 1981-89 Unconstrained Unconstrained Mon Variable Mon Variable with Lagged with Lagged GNP; Lagged GNP; Contemp and Lagged Money Money Only Poly Distr Lag Poly Distr Lag Not Tied; Not Tied; with Lagged GNP; with Lagged GNP; Contemp & Lagged Lagged Money Only Money Poly Distr Lag Poly Distr Lag Not Tied; Not Tied; No Lagged GNP; No Lagged GNP; Contemp & Lagged Lagged Money Money Only Poly Distr Lag Poly Distr Lag Tied Far End; Tied Far End; with Lagged GNP; with Lagged GNP; Contemp & Lagged Lagged Money Money Only Poly Distr Lag Tied Far End; No Lagged GNP; Contemp & Lagged Money Statistical Measure M1 AdjR2 St err regress Sum coef .31 3.15 .13 .29 3.19 .21 .37 3.00 .11 .31 3.13 .23 .23 3.31 .04 .10 3.58 .22 .21 3.37 .03 .19 3.40 .22 3.46 0 M2 AdjR2 St err regress Sum coef .23 3.34 .31 .20 3.39 .44 .25 3.27 .31 .20 3.37 .43 .12 3.54 .32 .05 3.69 .47 .13 3.53 .25 .19 3.40 .41 .05 3.69 .26 Currency AdjR2 St err regress Sum coef .43 2.87 1.62 .44 2.84 1.66 .41 2.89 1.71 .43 2.85 1.66 .43 2.85 2.00 .43 2.84 2.16 .32 3.11 1.77 .45 2.81 1.64 .36 3.02 1.86 St err regress Sum coef .19 3.41 -.04 .09 3.61 -.02 .12 3.54 -.05 .09 3.60 -.01 .05 3.70 -.11 -.05 3.87 -.07 .14 3.51 -.05 .06 3.67 -.01 .06 3.68 -.11 Mon base—Board AdjR2 St err regress Sum coef .31 3.15 .85 .31 3.16 .94 .21 3.35 .90 .31 3.14 .91 .17 3.44 1.01 .17 3.45 1.25 .17 3.45 .78 .30 3.16 .99 .13 3.53 .92 Mon base—St Louis AdjR2 St err regress Sum coef .36 3.02 .71 .36 3.04 .81 .36 3.02 .76 .35 3.04 .80 .30 3.17 .89 .22 3.34 1.11 .24 3.30 .70 .37 3.00 .78 .24 3.31 .81 Total reserves AdjR2 t* .16 Chart 3 Currency as a Percentage of Nominal GNP 4.8 4.6 h 4.4 © a. 4.2 h 4 h 38 1970 1975 1980 1985 1S Note: Nominal GNP and currency values are seasonally adjusted. Ratios are plotted quarterly. well specified in either period. They are poorer predictors than other work on M2 might suggest16 When currency was placed in the equations, the values for adjusted R2 were much higher in the later period than in the earlier period-when there was no apparent relationship. During the 1970s, currency represented an expanding share of Ml (Chart 2) but declined relative to GNP (Chart 3). In the 1980s, currency's share of Ml showed no trend, while currency as a percent of GNP rose modestly. The estimated coefficients for currency seemed implausibly high in the later period. The equations did not pick up a significant relationship between total reserves and GNP in either period. The behavior of total reserves was different in the two decades, with total reserves falling sharply as a share of Ml in the 1970s and more slowly in the 1980s (Chart 1). Movements in total reserves responded to both the behavior of reservable deposits and the shifts in the reserve ratios arising from sources other than 16 See Wenninger (1990). 46 formal changes in requirements. Using the monetary base, which combines currency and total reserves, in the equation produced more plausible coefficients than using either currency or total reserves separately. These results give some support to the notion that the monetary base might have a role to play as an intermediate target or at least as an indicator. The important role of currency in the relatively good results achieved for the base in the 1980s means that one would want to know more about why currency behaved as it did and whether it was likely to continue to do so. When prices were used as the dependent variable in empirical work covering the 1980s, the short estimation period was even more of a handicap than it was for nominal GNP. Results obtained for the earlier years suggested that the monetary variables affected prices with lags that were on average significantly longer than those observed for nominal GNP. Furthermore, the strong results obtained by Fama and Hafer were achieved with relatively complex models incorporating rather specific views of the relationships among economic variables and would be hard to replicate. Nonetheless, while recognizing that the exercise was subject to many pitfalls, I estimated several regressions with rates of price change as the dependent variable.17 Prices were defined alternately with the GNP deflator or the consumer price index (CPI). The two price measures produced some differences but a generally similar ordering of results. Table 4 presents selected results for regressions that used the implicit deflator as the dependent variable. Most of the equations also included lagged values of real GNP. Those without lagged real GNP generally had insignificant values for adjusted R2 and less plausible coefficients. (When real GNP was included, the monetary coefficients could be expected to be modestly lower than one.) With the notable exception of currency, all of the monetary measures showed considerably poorer results for the 1980s than they did for the 1970s if judged by adjusted R2. For the monetary base, the coefficients were more plausible in the 1980s than in the 1970s, although very low when real GNP was omitted. They nonetheless were more plausible than those on currency. The results for the 1980s were disappointing, although the short time period and the number of changes in money demand occurring in the early part of the decade were probably contributory factors. More broadly, one could ask how much light any of these exercises, those with nominal GNP or those with a price index as the dependent variable, shed on the issues they are supposed to address, namely, the relative merits of these measures as intermediate targets of monetary policy. The basic 17 To gain a few more observations, 1980 was chosen over 1981 as a starting date in these regressions, although the earlier date may have introduced more distortions from the transition period. 47 Table 4 00 A Summary of Selected Regressions Using the Implicit Deflator as the Dependent Variable Time Period of Estimate: 1972-80 Unconstrained Mon Variable with Contemp & Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Tied Far End; with Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Tied Far End; with Lagged Real GNP; Lagged Money Only Poly Distr Lag Not Tied; with Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Not Tied; with Lagged Real GNP; Lagged Money Only Poly Distr Lag Not Tied; No Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Not Tied; No Lagged Real GNP; Lagged Money Only Statistical Measure Ml AdjR2 St err regress Sum coeff .38 1.88 .78 .35 1.92 .68 .35 1.93 .65 .39 1.86 .76 .41 1.83 .77 .32 1.97 .53 .33 1.95 .53 M2 AdjR2 St err regress Sum coeff .52 1.65 -.08 .58 1.54 -.08 .49 1.71 -.29 .59 1.54 -.01 .54 1.62 -.10 .54 1.62 -.36 .50 1.69 -.39 St err regress Sum coeff .34 1.94 .90 .38 1.87 .63 .36 1.91 .84 .36 1.91 .64 .34 1.94 .84 .26 2.06 .92 .23 2.09 1.17 Total reserves AdjR2 St err regress Sum coeff .17 2.17 .02 .19 2.15 .06 .20 2.13 .07 .16 2.19 .06 .17 2.17 .07 .02 2.36 .13 .05 2.32 .13 .57 1.57 1.64 .62 1.48 .59 1.53 1.92 .58 1.55 1.94 .48 1.97 .60 1.50 1.96 1.72 2.45 .46 1.75 2.42 .56 1.58 1.92 .56 1.58 2.24 .56 1.58 2.22 .54 1.61 2.24 1.61 2.22 .29 2.01 2.75 .30 1.99 2.73 Currency AdjR2 Mon base—Board AdjR2 St err regress Sum coeff Mon base—St Louis AdjR2 St err regress Sum coeff Note: Polynomial distributed lags are third order. .54 Table 4 - Continued Time Period of Estimate: 1980-89 Unconstrained Mon Variable with Contemp & Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Tied Far End; with Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Tied Far End; with Lagged Real GNP; Lagged Money Only Poly Distr Lag Not Tied; with Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Not Tied; with Lagged Real GNP; Lagged Money Only Poly Distr Lag Not Tied; No Lagged Real GNP; Contemp & Lagged Money Poly Distr Lag Not Tied; No Lagged Real GNP; Lagged Money Only Statistical Measure M1 AdjR2 St err regress Sum coeff -.06 2.71 -.17 .10 2.49 -.11 .07 2.54 -.19 .08 2.53 -.11 .04 2.58 -.17 -.03 2.67 -.26 -.02 2.66 -.28 M2 AdjR2 St err regress Sum coeff .25 2.74 -.01 .05 2.56 -.36 .08 2.52 -.06 .03 2.60 -.04 .05 2.56 -.06 -.04 2.69 -.21 -.03 2.67 -.13 Currency AdjR2 St err regress Sum coeff .76 1.56 2.44 .64 1.58 2.32 .63 1.60 2.24 .65 1.55 2.40 .65 1.56 2.34 .42 2.00 2.19 .43 2.00 2.23 St err regress Sum coeff .46 2.32 -.06 .31 2.18 -.03 .32 2.17 -.04 .29 2.22 -.03 .30 2.20 -.04 .12 2.47 -.02 .09 2.51 -.04 Mon base—Board AdjR2 St err regress Sum coeff .15 2.43 .70 .23 2.31 1.12 .15 2.43 .90 .21 2.35 1.11 .12 2.46 .91 -.08 2.74 .45 -.08 2.74 .40 .39 2.46 .54 .21 2.34 .77 .12 2.47 .67 .19 2.38 .79 .10 2.50 .70 -.08 2.74 .21 -.09 2.74 Total reserves AdjR2 Mon base—St Louis AdjR2 St err regress Sum coeff s .21 model from which all of them were drawn was built to capture the behavioral characteristics that had been observed for Ml in the 1950s and 1960s. At the time, none of the components of Ml paid explicit interest, and interest sensitivity of the demand for money did not appear to be very large. The relationships have all changed in recent years, with short-run interest rate sensitivity increasing considerably, especially for Ml. M2 has been sensitive to relative interest rate moves all along, although the patterns have changed as deregulation of rates has spread. Furthermore, since each of the monetary measures has different characteristics, it is doubtful that the same model would prove optimal for each of the monetary variables. Thus, there is no reason to believe that models developed in the 1960s to explain Ml behavior would reveal the correct relationship or even the correct ordering among monetary variables over the 1980s. It is possible that better relationships between the aggregates and the target variables could be found than any of these equations have discovered. IV. Does the Concept of the Monetary Base Make Sense? Benjamin Friedman (1988), commenting on McCallum's article (1988a), questioned the logic of using the monetary base as either an indicator or as an intermediate target. He had two basic criticisms. First, a large portion of the monetary base consists of currency, and the demand for currency is not well understood. He noted that currency outstanding as of the end of 1986 amounted to almost $900 for every U.S. citizen, far in excess of what surveys suggest most people hold.18 Some currency is used for illegal transactions that are not captured by U.S. GNP statistics. A portion is used for transactions in countries where the local currency is not stable or is not freely convertible. In addition, he simulated McCallum's equations for 1963-87, substituting alternately currency and total reserves for the monetary base. On the basis of root mean squared errors, currency outperformed the base while total reserves produced very large errors. Friedman did not try to explain the poor relationship between total reserves and economic activity. One possibility was that the ratio of total reserves to M2, the aggregate with the best track record in the 1980s, might have been unstable. As Chart 4 shows, however, the ratio of total reserves to M2 was reasonably steady. Indeed, in the 1980s, the ratio of total reserves to M2 showed less quarterly variation than the ratio of total reserves to Ml (Chart 1), even though a large share of M2 deposits were not subject to reserve requirements. Friedman's poor results for total reserves may have reflected variations in the ratios of currency to deposits and reserves to deposits that were not picked 18 The figure was around $900 million (not seasonally adjusted) at the end of 1989. 50 Chart 4 Currency and Total Reserves as Shares of M2 c 2 5 Q> Total reserves as a share of M2 I , ., I ,, , I , , ,I i •,, i , , , 11 1 1 1 1 1 1 _L , I , , , I . . . I , • , I 1970 1975 1980 Note: Reserve data are seasonally adjusted and are adjusted for changes in reserve requirements. M1 and currency values are seasonally adjusted. Ratios are plotted quarterly. 1 1985 1989 up by the model. Some formulations of the St. Louis equations reported in the preceding section give slightly better results for total reserves than Friedman found (although they are still poor). Thus, on the basis of limited evidence, it is hard to make a choice between currency and the monetary base. The validity of either currency or the monetary base as money measures was debated implicitly in the early twentieth century. The question was asked, Should deposits be considered "money," or should money be defined to consist only of gold and other government obligations? Irving Fisher and other economists argued persuasively that deposits performed essentially the same payment and accounting services as currency and coin. Therefore, ignoring them would lead to a serious understatement of existing monetary services. The subject was mostly left to rest until the 1970s. Lothian (1976) presented a rationale for a narrow money measure such as the monetary base or currency. He argued that all financial assets provided a mix of money- and bond-type services. 51 Currency and non-interest-bearing reserves have little opportunity to provide bond-type services. Deposits, on the other hand, may pay interest either explicitly or implicitly, and the institutions accepting them may offer bond-type services to depositors. In an environment with variable regulations and inflation, the mix of money- and bond-type services provided by deposits may vary more than the mix provided by the non-interest-bearing monetary base. Consequently, Lothian concluded that the base was likely to be a better proxy for monetary services when deposit characteristics differed (over time or across countries). Lothian's cross-country empirical results were very similar for the monetary base and currency. A reasonably strong relationship between currency and economic activity might be plausible. Currency is convenient for carrying out certain classes of transactions. If those transactions represent a relatively steady share of economic activity, then a close relationship might be expected. One might anticipate, however, that currency would perform a gradually declining share of monetary services whenever the number and sophistication of substitute means of payment grew. During the 1970s and early 1980s, rapid inflation stimulated efforts to find substitutes for non-interest-bearing currency. The pattern over that period did show a decline in currency relative to GNP (Chart 3), although currency was a growing share of Ml. Countering that tendency in more recent years has been slower average inflation and the increased ease of obtaining currency from automated teller machines, developments which have raised the value of currency as a provider of payment services. The irregular but rising pattern for currency use during much of the 1980s may in part reflect those factors. Nonetheless, Friedman's criticisms of currency as a potential intermediate target deserve attention. The use of U.S. currency for transactions that are not included in U.S. GNP statistics because they are illegal or take place in other countries would seem to be a serious problem only if the proportion of currency used for such transactions was variable. Limited direct information about variability over long periods of time is provided by two Board staff studies (Avery et al. 1986,1987). The Board made surveys of cash holdings for transactions purposes by American adults in 1984 and 1986. Both surveys could account for about 11 to 12 percent of the cash outstanding. The 1987 article reported that the staff discovered a survey of cash holdings undertaken in 1944. That survey "accounted for a proportion of circulation currency that was remarkably consistent with the 1984 and 1986 data" (Avery et al. 1987, p. 191). Considering the numerous changes in payment practices and the rise of the illegal drug trade between the 1944 survey and the two more recent surveys, the similarity is surprising. In 1944, currency represented a share of Ml similar to its share in the mid 1980s, but it constituted a much larger share of GNP-on the order of 10 percent. (Of course 1944 was also a year 52 of turmoil, and people in the war zones would most likely have sought to hold U.S. currency.) In any case, over short time periods U.S. currency can undergo significant unpredictable variations when developments abroad provoke large movements of currency outside the United States. If one should wish to make currency the centerpiece of policy, one confronts the problem of how to control it. As long as currency is provided on demand, it is not directly controllable. Control would have to be achieved indirectly by influencing demand. Variables that act upon the growth of nominal economic activity, such as real interest rates and reserve provision, would influence currency over time. The authorities, therefore, would have a greater chance of success if they used currency as an intermediate indicator rather than as a target. They could use a procedure that adjusts for trend changes and large shipments of currency abroad. V. Experiences with Monetary Base Targets The United States has no direct experience with monetary base targeting, but the nonborrowed reserve targeting procedures employed between October 1979 and the autumn of 1982 have some features in common with a nonborrowed base operating target. In addition, Switzerland and West Germany have used measures similar to the monetary base in a role closer to an intermediate target than an operating target. A. Reserve Targeting in the United States Under the procedures followed between 1979 and 1982, the primary operating target was nonborrowed reserves. The procedures focused on intervals three weeks to five weeks in length, thus encompassing more than a single one-week reserve maintenance period. By averaging reserves over several weeks, the procedure avoided forcing offsets to very short-run deviations of reserves from the desired path—deviations which were thought likely to be self-reversing. In many ways, the procedure resembled one that could be used to target the nonborrowed base. No underlying growth rate objective was set for nonborrowed reserves (or any other reserve measure) over quarterly or annual time periods. Instead, the staff derived the targets for nonborrowed reserves by estimating the volume of total reserves believed to be consistent with desired Ml. Under the procedure, the staff made estimates of currency and of reserve ratios expected to be associated with desired Ml, two steps which were also recommended for developing an operating target for the monetary base. The nonborrowed reserve operating objective was derived from the total reserve target by subtracting an amount of borrowing believed to be consistent with interest rate levels that in turn would be consistent with the 53 desired money growth. (In practice, the FOMC set the initial allowance for borrowing at a level close to that most recently in effect. It occasionally raised or lowered the allowance modestly if money growth had been more or less rapid than desired.) The operating targets were for nonborrowed reserves rather than total reserves because, as indicated earlier, existing institutional arrangements did not permit the Federal Reserve to control total reserves~or the monetary base-even if it had wanted to do so. The degree of similarity between targeting a nonborrowed reserve or nonborrowed base measure and a total reserve or total base measure depends upon the way banks respond to the reserve shortages or excesses. As noted in the earlier discussion, differing views on that response mechanism lie at the heart of the debate about the controllability of the monetary base. Limited information from the 1979-82 experience suggested that the process of adjusting Ml deposits to changes in reserve availability began fairly quickly but continued over a period of several months. Thus, the nonborrowed reserve targeting procedure produced reserve and money adjustments in the desired direction within a reasonable time period, but the adjustments were not by any means instantaneous. Even though the procedure allowed for some modest smoothing of the federal funds rate, the funds rate varied considerably more, on a day-to-day and week-to-week basis, than it had before the procedures were introduced (when deliberate actions had been taken to smooth the rate day to day). Some of the rate variation may have reflected specific features of the institutional reserve accounting structure, such as the length of the reserve maintenance period, the monetary authority's reaction when depository institutions failed to meet requirements, the presence of lagged reserve requirements, and the rules of access to the discount window. But much of this rate variation probably stemmed from the lags that must inevitably be part of the banks9 adjustment process in a system involving many institutions, large flows, and a complex set of motivations for bank pricing policies. The other notable feature of the period was that Ml growth slowed on average and came closer to its desired targets over extended periods of time, but it experienced considerable variability from month to month and even quarter to quarter-more than it had before. The variability is not completely understood, but the lags in the banks' response system could very easily have produced a damped oscillatory response in Ml to the changes in reserve pressures. Some of that variation might have been reduced through modifications to the targeting procedures, but there are limits to how much reduction could have been achieved in the face of delays in the adjustments undertaken by the commercial banks. It appears that the nonborrowed reserve targeting procedure followed was able to bring Ml 54 back on target when it veered off and to prevent an off-target trend. Nonetheless, it did introduce a substantial amount of short-run variability to both Ml and interest rates. Monetary base targets abroad For practical evidence using the monetary base as an intermediate target, one must look outside the United States. This section reviews the experiences of Switzerland and of West Germany, two countries that have made considerable use of monetary base targets. The Swiss National Bank (SNB) began targeting the monetary base in 1979, after having targeted Ml from 1973 to 1978. As described by Kohli and Rich (1986) and Rich (1987), the SNB considers the monetary base to be an intermediate target. The base is believed to be reasonably well linked over time with price stability and approximately controllable within a year's time frame through adjustments to reserves. The SNB sets annual target rates, which are chosen to be consistent with moving as close to its long-run goal of price stability as is considered feasible. The main restraint is that with heavy dependence on international trade, exchange market conditions must also be considered. The SNB does not view short-run manipulation of output and employment as appropriate activities for the central bank. It attempts to achieve price stability in a way that causes as little fluctuation in output as possible. Rich and Kohli reported that the SNB judged its experience with monetary base targeting to have been largely successful, with a couple of notable exceptions. The SNB had its first problem with monetary base targeting in 1979. A strong exchange rate and trade pressures meant that the SNB was unable to shield the country successfully from the inflationary impulse of that year's oil price shock. It temporarily set aside its monetary base target that year, permitting an overshoot, but then accepted shortfalls in the next two years that offset the overshoot. Then in 1987, strength of the Swiss franc in the exchange markets and concerns that such strength was damaging the export sector led the SNB to try to discourage franc appreciation by pursuing a somewhat more expansionary policy than it would have preferred. During the 1980s, the SNB ran into some problems with shifting money demand (Belognia 1988). In contrast to most other western countries, Switzerland experienced a significant fall in the demand for Ml early in the decade, apparently in several steps, and a more modest reduction in the demand for the monetary base. As the SNB observed the shifts, it lowered its targets for the monetary base. Because of these shifts, the SNB increased the attention paid to money market rates in order to obtain early indications of possible money demand shifts. In 1988, the SNB introduced revised liquidity requirements and a new interbank payments system. Both factors acted to lower the demand 55 for Ml. In response, the SNB deliberately undershot its base target in that year and again in 1989. Undershooting was also deemed consistent with handling the inflationary impulse that had developed in response to the earlier accommodative posture. Targets for the monetary base apply only to the year as a whole. The growth of the base is monitored during the year. Deviations from the target are tolerated if special factors suggest that they are likely to be self-reversing. Otherwise the degree of reserve provision is adjusted to ensure that the base will be brought back on target. Rich and Kohli reported that the SNB had the ability to control the monetary base within what had been judged to be acceptable limits. In most years the actual figure was within 1 percentage point of the target, and often it was closer. A miss of 1 percentage point or less is not believed to do serious damage to the drive for price stability. The exceptions have been years when a deliberate decision was made to miss the target because of an observed special factor. The SNB achieves the targets by adjusting reserve availability through temporary repurchase operations against foreign currencies (known as swaps) and by placing limits on the commercial banks' access to the discount and Lombard windows. The banks are required to report in advance any planned large takedowns of discount or Lombard credit, giving the SNB room to offset the reserve effects. This restriction in borrowing access has led the banks to hold more than minimal levels of excess reserves. It has also led to occasional wide fluctuations in the overnight money market rates on the last day of the month, when reserve requirements must be met, although these fluctuations have little or no effect on longer term rates. Variations in excess reserves arise in the adjustment process, since banks cannot instantaneously adjust their deposits when the base level is changed. The SNB has judged monetary base targeting to have been largely successful, both because unplanned extended misses have been small and because average price inflation has been considerably lower than in other countries. Control of the base was credited with limiting the potential for inflation. Others have challenged that view, suggesting that the determination to control inflation and a conservative fiscal policy established a climate where almost any technique could have worked (Wenninger 1983). Although the policy climate was obviously helpful, monetary base targeting has served as a visible sign of the central bank's efforts to contain inflation. Consequently, monetary base targets in Switzerland appear to have played a useful part in bringing about the desired results. Like Switzerland, West Germany appears to provide an example of relatively successful use of the monetary base as a policy target. The West German Bundesbank targeted a measure called the central bank money stock (CBM) from 1975 through 1987, setting annual target ranges. The measure, 56 still in use today as an indicator, closely resembles the monetary base. CBM consists of currency plus compulsory cash reserves held by the banking system at constant January 1974 reserve ratios, and thus is essentially a monetary base measure adjusted for reserve requirement changes. Articles by Bundesbank staff raised objections to equating CBM with the monetary base (Dudler 1980, 1982). The articles indicated that Bundesbank policy makers disliked identifying CBM with the base for two reasons. First, they wanted to distance their approach from that of using CBM as an operating target; when CBM targeting began, most of the discussion of the monetary base in the literature focused on direct short-run control. Second, they viewed CBM as a proxy for a broad money measure, M3. M3 consists of currency plus sight deposits and time deposits of less than four years and deposits at statutory notice. Throughout the time the Bundesbank targeted CBM, data were available much more promptly for CBM than for M3. Most of the time, the two measures tracked each other's behavior fairly closely. The Bundesbank manipulated interest rates when it observed CBM going off track, and although it did not set a formal target for M3, it assumed that efforts to bring CBM into line with the target would also lead to appropriate behavior for M3 (Deutsche Bundesbank 1985). In 1988, the Bundesbank changed its monetary target to M3 (Deutsche Bundesbank 1988). In making the change, it again pointed out the similarities between longer run movements of M3 and CBM. It indicated that both aggregates had shown fairly stable positive relationships to the growth of nominal production potential and negative relationships to interest rates. Movements in both measures led movements in economic activity. The change in the target variable was made because CBM had been tending to move in an exaggerated fashion, overstating swings in money growth rates in preceding months. At the time, currency was expanding disproportionately because interest rate levels were low. Because CBM's movements were often dominated by fluctuations in currency, CBM was responding too much, in the Bundesbank's view, to movements in interest rates. In other words, CBM was not serving as a good proxy for M3. The Bundesbank also cited some past episodes of divergence, but reported that those episodes had generally not lasted overly long. Another motive for changing the target was some discomfort with the abstract nature of the reserve component of CBM. Actual required reserve ratios were very different (and generally lower) in 1988 than they had been in 1974. Furthermore, several components of broad money were not subject to minimum reserve requirements. Nonetheless, the data delays that had previously led the Bundesbank to target CBM rather than M3 continued after 1988, and for this reason the Bundesbank retained CBM as an early indicator of changes in M3 growth. The Bundesbank over the years took its targets, whether for CBM or M3, seriously. It was 57 often successful in achieving them, but it also experienced a number of misses both planned and unplanned.19 In choosing its annual monetary target, the Bundesbank made estimates of the growth rate of the target variable likely to be consistent with desired nominal GNP. It took account of what it considered to be unavoidable inflation, likely prospects for real economic expansion, and forecasts of velocity. Consequently, it varied the targeted growth rate range for CBM significantly from year to year as its estimates of appropriate economic activity and velocity changed. Sometimes it chose to miss a target during the year because of conflicts with desired developments in the foreign exchange markets or because of unanticipated shifts in the behavior of velocity. The Bundesbank was generally pleased with this flexible use of monetary targets. Like Switzerland, West Germany generally achieved inflation rates below those in most other industrialized countries. It is of course true that the anti-inflation policy in West Germany, as in Switzerland, had strong internal support that greatly increased the likelihood it would succeed. Nonetheless, the CBM targets made a contribution to achieving the price objective. Appendix: Definitions of Monetary Base and Reserve Measures The Monetary Base There are a number of possible approaches to defining the monetary base. Two definitions are widely available, one prepared by the St. Louis Federal Reserve, which pioneered the empirical work, and the other prepared by the Board of Governors. The base concept can be developed from either the sources side or the uses side of the balance sheet. The results should be equivalent, although details of timing and treatment of seasonal adjustment issues can introduce slight differences. The St. Louis staff has approached the construction of the base conceptually from the sources side. It describes the base as consisting of Federal Reserve credit-holdings of securities in the portfolio, loans by the discount window, and other balance sheet items, as well as the gold stock, Special Drawing Rights, and Treasury currency. Several categories of liabilities are subtracted, namely, Treasury and foreign deposits at the Federal Reserve, Treasury holdings of currency, and certain miscellaneous items. When staff members actually construct what they call the source base, 19 See articles entitled "The Economic Scene in the Federal Republic of Germany" in various issues of the Monthly Report of the Deutsche Bundesbank. 58 however, they define it as currency in the hands of the public plus required reserves plus excess reserves. In other words, they define the source base in terms of its uses, although they treat vault cash contemporaneously rather than in lagged form. For dates before 1984, when contemporaneous reserve requirements (CRR) were introduced, the staff has allowed for the lag structure on reserve requirements. The Board staff defines the base in terms of its uses, as consisting of currency and total reserves held by depository institutions. More specifically, its definition of the base before adjustment for reserve requirement changes consists of (1) total reserves, (2) required clearing balances and adjustments to compensate for float at Federal Reserve Banks, (3) the currency component of the money stock less the amount of thrift institutions' vault cash holdings normally included in the currency component of the money stock, and (4) the excess of current vault cash over the amount applied to satisfy current reserve requirements at insitutions not having required reserve balances. For dates after the introduction of CRR in February 1984, the Board measures currency and vault cash over the two-week computation period ending Monday, two weeks and two days before the two-week reserve maintenance period begins. Before CRR, all components of the monetary base other than excess reserves were seasonally adjusted as a whole, rather than by component, and excess reserves were added without seasonal adjustment. Since CRR, the seasonally adjusted series has consisted of seasonally adjusted total reserves, which include excess reserves on a not seasonally adjusted basis, plus the seasonally adjusted currency component of the money stock plus the remaining items seasonally adjusted as a whole. The major difference between the two definitions arises from the adjustment procedures to handle changes in reserve requirement ratios. The Board's "break adjustment" method allows for discontinuities arising from regulatory changes in reserve requirements by taking the new ratios, calculated for different classes of institutions, and applying them to the actual historical deposit series for each banking group. The annual indexing of low reserve tranches is applied to the reserve ratios as if the change were phased in over the whole year instead of being applied in the first maintenance period of each year. It includes the 3 percent requirement on nonpersonal time and savings deposits but ignores the requirement on Eurodollar deposits. The St. Louis reserve adjustment magnitude (RAM) employs a different approach. Instead of using current reserve ratios to define the monetary base, the RAM uses hypothetical reserve requirements. For years since 1980, the RAM has set the reserve ratio on transactions deposits equal to 12 percent, the marginal reserve requirement on such deposits. (The actual average reserve ratio 59 has recently been about 8 percent because there are zero and 3 percent reserve tranches.) The RAM measure assumes a zero reserve requirement for other types of deposits, even though some of them are actually subject to a 3 percent ratio. For years before 1980, it uses a ratio equal to member bank deposits and only applies the ratio to such deposits. It applies separate ratios for time and savings deposits. Interim techniques were used between 1980 and 1987 when new reserve requirements were being phased in. For a more detailed description of the techniques of reserve adjustment, see Gilbert (1980, 1984, 1987). Studies have suggested that it makes an important difference whether the base is adjusted for reserve requirement changes (Haslag and Hein 1989). However, the different techniques of adjusting for reserve ratios are of limited importance over most time periods. The St. Louis base measure is higher than the Board base since it uses a higher reserve ratio, and it thus gives more weight to deposits relative to currency. For brief time periods, the movements of the two bases often differ. However, the differences in rates of growth are generally slight for periods of a quarter or more.20 Other Measures Total reserves consist of reserve balances with Federal Reserve Banks, excluding required clearing balances and adjustments to compensate for float, plus vault cash held during the lagged computation period by institutions having required reserve balances at Federal Reserve banks, plus the amount of vault cash equal to required reserves during the maintenance period at institutions having no required reserve balances. Nonborrowed reserves consist of total reserves less those reserves borrowed by depository institutions from the Federal Reserve's discount window. Borrowed reserves consist of those reserves acquired from the Federal Reserve's discount window. For some purposes, extended credit borrowing by banks in financial difficulty is treated as part of nonborrowed reserves. Required reserves consist of reserves that depository institutions must hold against deposits. Requirements are based upon deposit levels during various computation periods according to the type of deposit and the size of the institution. Excess reserves consist of total reserves not needed to meet reserve requirements. ^ o n e of the studies cited in the text reported results for both base measures in the regressions. In the equations I ran, the measures of significance differed somewhat for the two base measures, although the sums of the coefficients were generally similar. 60 Bibliography Andersen, Leonall C, and Jerry L. Jordan. 1968a. "Monetary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabilization." Federal Reserve Bank of St. Louis Review, November, pp. 11-24. . 1968b. "The Monetary Base-Explanation and Analytical Use." Federal Reserve Bank of St. Louis Review, August, pp. 7-11. Andersen, Leonall C, and Denis S. Karnosky. 1977. "Some Considerations in the Use of Monetary Aggregates for the Implementation of Monetary Policy." Federal Reserve Bank of St. Louis Review, September, pp. 2-7. Avery, Robert B., George E. EUiehausen, Arthur B. Kennickell, and Paul A. Spindt. 1986. "The Use of Cash and Transactions Accounts by American Families." Federal Reserve Bulletin, vol. 72, no. 2 (February), pp. 87-108. . 1987. "Changes in the Use of Transactions Accounts and Cash from 1984 to 1986." Federal Reserve Bulletin, vol. 73, no. 3 (March), pp. 179-96. Balbach, Anatol B. 1981. "How Controllable is Money Growth." Federal Reserve Bank of St. Louis Review, April, pp. 3-12. Bank for International Settlements, Monetary and Economic Department. 1980. The Monetary Base Approach to Monetary Control. September. Banque Nationale Suisse. 1988. "Monetary Policy in 1989." Bulletin Trimestriel, no. 4 (December), pp. 271-72. . 1989. "The Monetary Policy of the Swiss National Bank in 1989 and 1990." Bulletin Trimestriel, no. 4 (December), pp. 291-92. Belongia, Michael. 1988. "Stability of Swiss Money Demand: Evidence for 1982-1987." Banque Nationale Suisse Bulletin Trimestriel, no. 1 (March), pp. 68-74. Board of Governors of the Federal Reserve System. 1988. The Monetary Base. Staff Study, December. Brunner, Karl. 1961. "Some Major Problems in Monetary Theory." American Economic Review, vol. 51, no. 2 (May), pp. 47-56. . 1980. "The Control of Monetary Aggregates." Controlling Monetary Aggregates in, pp. 1-65. Federal Reserve Bank of Boston, October. 61 Brunner, Karl, and Allan H. Meltzer. 1964. "Studies on Money and Monetary Policy-Some Further Investigations of Demand and Supply Functions for Money." Journal of Finance, vol. 19, no. 2 (May), pp. 240-83. . 1968. "Liquidity Traps for Money, Bank Credit, and Interest Rates." Journal of Political Economy, January-February, pp. 1-37. . 1981. "Time Deposits in the Brunner-Meltzer Model of Asset Markets." Journal of Monetary Economics, vol. 7, no. 1, pp. 129-39. . 1983. "Strategies and Tactics for Monetary Control." Carnegie-Rochester Conference Series on Public Policy, no. 18, pp. 59-104. Brunner, Karl, Alex Cukierman, and Allan H. Meltzer. 1983. "Money and Economic Activity, Inventories, and Business Cycles." Journal of Monetary Economics, vol. 2. Burger, Albert E., Lionel Kalish HI, and Christopher T. Babb. 1971. "Money Stock Control and Its Implications for Monetary Policy." Federal Reserve Bank of St. Louis Review, October, pp. 6-22. Burgess, Randolph. 1936. The Federal Reserve Banks and the Money Market. New York: Harper and Brothers. Cagan, Phillip. 1982. "The Choice among Monetary Aggregates as Targets and Guides for Monetary Policy." Journal of Money, Credit, and Banking, vol. 14, no. 4 (November), pp. 551-686. Carlson, John B. 1981. "The Monetary Base, the Economy, and Monetary Policy." Federal Reserve Bank of Cleveland Economic Review, Spring, pp. 2-13. Christiano, Lawrence J. 1986. "Money and the U.S. Economy in the 1980Y. A Break from the Past?" Federal Reserve Bank of Minneapolis Quarterly Review, Summer, pp. 2-13. CuUison, William E. 1982. "Money, the Monetary Base, and Nominal GNP." Federal Reserve Bank of Richmond Economic Review, May-June, pp. 3-13. Darby, Michael R., William Poole, David E. Lindsey, Milton Friedman, and Michael J. Bazdarich. 1987. "Recent Behavior of the Velocity of Money." Contemporary Policy Issues, vol. 5, no. 1 (January), pp. 1-33. Davis, Richard G. 1979-80. "The Monetary Base as an Intermediate Target for Monetary Policy." Federal Reserve Bank of New York Quarterly Review, Winter, pp. 1-10. . 1980. "The Monetary Base as an Intermediate Target for Monetary Policy in the United States." The Monetary Base Approach to Money Control, pp. 214-29. Basle: Bank for International Settlements, Monetary and Economic Department, September. Deutsche Bundesbank. 1985. "The Longer-Term Trend and Control of the Money Stock." Monthly Report of the Deutsche Bundesbank, vol. 37, no. 1 (January), pp. 13-26. 62 . 1988. "Methodological Notes on the Monetary Target Variable M3." Monthly Report of the Deutsche Bundesbank, vol. 40, no. 3 (March), pp. 18-21. Dudler, Hermann-Josef. 1980. "Money-Market Management, Supply of Bank Reserves and Control of the Central-Bank Money Stock." The Monetary Base Approach to Money Control pp. 52-77. Basle: Bank for International Settlements, Monetary and Economic Department, September. . 1982. "The Implementation of Monetary Objectives in Germany-Open Market Operations and Credit Facilities." Central Bank Views on Monetary Targeting, pp. 18-29. Federal Reserve Bank of New York, May. Fama, Eugene F. 1982. "Inflation, Output and Money." Journal of Business, vol. 55, no. 2, pp. 201-31. Fellner, William. 1982. "Criteria for Useful Targeting. Money versus the Base and Other Variables." Journal of Money, Credit, and Banking, vol. 14, no. 4 (November), pp. 641-60. KisHyr, Irving N. [1911] 1963. The Purchasing Power of Money. Reprint. New York: Augustus M. Friedman, Benjamin M. 1988. "Conducting Monetary Policy by Controlling Currency Plus Noise-A Comment." Carnegie-Rochester Conference Series on Public Policy, no. 29. Friedman, Benjamin M., and Kenneth N. Kuttner. 1989. "Money, Income, and Prices After the 1980s." National Bureau of Economic Research, Working Paper Series, no. 2852. February. Friedman, Milton. 1959. A Program for Monetary Stability. Millar Lectures, no. 3. Fordham University Press. Friedman, Milton, and David Meiselman. 1963. "The Relative Stability of Monetary Velocity and the Multiplier in the United States, 1897-1958." In Commission on Money and Credit, Stabilization Policies. New Jersey: Prentice Hall, pp. 165-268. Gambs, Carl M. 1980. "Federal Reserve Intermediate Targets: Money or the Monetary Base?" Federal Reserve Bank of Kansas City Economic Review, January, pp. 3-15. Garfinkel, Michelle R., and Daniel L. Thornton. 1989. "The Link Between Ml and the Monetary Base in the 1980s." Federal Reserve Bank of St. Louis Review, September-October, pp. 35-52. Gilbert, R. Alton. 1980. "Revision of the St. Louis Federal Reserve's Adjusted Monetary Base." Federal Reserve Bank of St. Louis Review, December, pp. 3-10. . 1984. "Calculating the adjusted Monetary Base under Contemporaneous Reserve Requirements." Federal Reserve Bank of St. Louis Review, February 1984, pp. 27-32. . 1987. "A Revision in the Monetary Base." Federal Reserve Bank of St. Louis Review, August-September, pp. 24-29. 63 Goodhart, Charles A., M.D.K.W. Foot, and A.C. Hotson. 1979. "Monetary Base Control." Bank of England Bulletin. June, pp. 149-59. Gurley, John G., and Edward S. Shaw. Money in a Theory of Finance. Washington, D.C.: Brookings Institution, 1960. Hafer, R.W. 1984. "The Monetary Base or Ml? Results from a Small Macromodel." Temple University Journal of Economics and Business, vol. 36, no. 1 (February), pp. 85-93. Hafer, R.W., Scott E. Hein, and Clemens J. Kool. 1983. "Forecasting the Money Multiplier: Implications for Money Stock Control and Economic Activity." Federal Reserve Bank of St. Louis Review, October, pp. 22-33. Hall, Thomas E. 1988. "McCallum's Base Growth Rule: Results for the United States, West Germany, Japan, and Canada." U.S. Department of State, Bureau of Economic and Business Affairs, Planning and Economic Analysis Staff, Working Paper Series. December. Haslag, Joseph H., and Scott E. Hein. 1988. "Evidence on the Two Monetary Base Measures and Economic Activity." Federal Reserve Bank of Dallas, Research Paper. December. . 1989. "Reserve Requirements, the Money Base, and Economic Activity." Federal Reserve Bank of Dallas Economic Review, March, pp. 1-15. Johannes, James M., and Robert H. Rasche. 1979. "Predicting the Money Multiplier." Journal of Monetary Economics, vol. 5, pp. 301-25. Kohli, Ulrich, and Georg Rich. 1986. "Monetary Control, The Swiss Experience." Cato Journal, vol. 5, no. 3 (Winter), pp. 911-26. Lindsey, David E. 1980. "Control of the Monetary Base and Reserves - The U.S. Case." The Monetary Base Approach to Money Control, pp. 205-13. Basle: Bank for International Settlements, Monetary and Economic Department, September. Lindsey, David E., et al. 1981. "Monetary Control Experience Under the New Operating Procedure." New Monetary Control Procedures-Federal Reserve Staff Study, vol. 2, February. Lothian, James R. 1976. "The Demand for High-Powered Money." The American Economic Review, vol. 66, no. 1 (March), pp. 56-68. Mayer, Thomas, James S. Duesenberry, and Robert Z. Aliber. 1987. Money, Banking and the Economy. 3d ed. W.W. Norton and Company. McCallum, Bennett T. 1987. "The Case for Rules in the Conduct of Monetary Policy: A Concrete Example." Federal Reserve Bank of Richmond Economic Review, September-October, pp. 10-18. . 1988a. "Postwar Developments in Business Cycle Theory: A Moderately Classical Perspective." Journal of Money, Credit, and Banking, vol. 20, no. 3 (August), pp. 459-71. 64 . 1988b. "Robustness Properties of a Rule for Monetary Policy." Carnegie-Rochester Conference Series on Public Policy, no. 29, pp. 173-204. . 1988c. "Targets, Indicators and Instruments of Monetary Policy." American Enterprise Institute Conference on Monetary Policy in an Era of Change, October. Osborne, Dale K. 1986. "Velocities of Ml, and the Money Base: A Correction of Standard Formulas." Federal Reserve Bank of Dallas Economic Review, January, pp. 10-24. Poole, William. 1988. "Lessons on Monetary Policy from the 1980s." The Journal of Economic Perspectives, vol. 2, no. 3 (Summer), pp. 51-72. Rich, Georg. 1987. "Swiss and United States Monetary Policy: Has Monetarism Failed?" Federal Reserve Bank of Richmond Economic Review, vol. 73, no. 3 (May-June), pp. 3-16. Rich, Georg, and Kurt Schiltknecht. 1980. "Targeting the Monetary Base~The Swiss Approach." The Monetary Base Approach to Money Control, pp. 150-70. Basle: Bank for International Settlements, Monetary and Economic Department, September. Roberds, William. 1989. "Money and the Economy: Puzzles from the 1980's Experience." Federal Reserve Bank of Atlanta Economic Review, September-October, pp. 20-35. Robertson, D.H. 1948. Money. 4th ed. Cambridge: Cambridge University Press. Schiltknecht, Kurt. 1982. "Switzerland~The Pursuit of Monetary Objectives." Central Bank Views on Monetary Targeting, pp. 72-79. Federal Reserve Bank of New York, May. . 1983. "Targeting the Monetary Base: The Swiss Case." Improving Money Stock Control: Problems, Solutions, and Consequences, chap. 2, pp. 67-96. Boston: Kluwer-Nijhoff. Simons, Henry C. 1948. Economic Policy for a Free Society. Chicago: University of Chicago Press. Stone, Courtenay C , and Daniel L. Thornton. 1988. "Selected Aggregates as Intermediate Policy Targets: Some Initial Evidence." Federal Reserve Bank of St. Louis Research Paper, February. Tatom, John. 1980. "Issues in Measuring an Adjusted Monetary Base." Federal Reserve Bank of St. Louis Review, December, pp. 11-29. Thornton, Daniel L. 1982. "Simple Analytics of the Money Supply Process and Monetary Control." Federal Reserve Bank of St. Louis Review, October, pp. 22-39. Tinsley, Peter A., et al. 1981. "Money Market Impacts of Alternative Operating Procedures." New Monetary Control Procedures-Federal Reserve Staff Study, vol. 2 (February). Von Hagen, Jurgen. 1988. "Alternative Operating Regimes for Money Stock Control in West Germany: An Empirical Evaluation." Weltwirtschaftiches Archiv, vol. 124, pp. 89-107. 65 Wenninger, John. 1983. Discussion of a paper by Kurt Schiltknecht. Improving Money Stock Control: Problems, Solutions, and Consequences, chap. 2, pp. 91-96. Boston: Kluwer-Nijhoff. Wood, John H. 1984. "The Search for a Monetary Policy Rule in an Uncertain World." Federal Reserve Bank of Dallas Economic Review, September. 66 MONETARY AGGREGATES AS INTERMEDIATE TARGETS John Wenninger The Federal Reserve began to use monetary aggregates as intermediate targets in the mid1970s. Increasing recognition of the drawbacks to relying on interest rates and the growing need to formulate a long-run strategy for containing inflationary pressures were the main reasons for adopting monetary targets. The experience gained in using the monetary aggregates in the policy process was in many respects the precipitating force behind much of the work that has been undertaken over the years on the general topic of using intermediate variables to attain ultimate policy goals. And the difficulties encountered in using monetary targets provided much of the impetus to search for alternative intermediate variables for policy such as total credit and liquid assets initially and commodity prices and the yield curve later on. This paper reviews the recent experience with using monetary aggregates as intermediate targets. It begins with a brief overview of the improvements in the policy process that were expected from formulating policy with monetary targets and the many practical problems that developed in using the monetary targets. The second section presents some of the theoretical arguments for and against using money as a target. A simple IS-LM model is used to illustrate the conditions under which money would be a good target variable as well as the conditions under which money could be readily controlled using the monetary base (as opposed to using interest rates). The third section of the paper provides some perspective on the stability of the money-income and the base-money relationships by presenting simple extrapolations of velocity and money multiplier trends. These simple techniques show that prior to the mid-1970s a strong empirical case for using monetary targets could be made, particularly for Ml. Both velocity and the money multiplier seemed to follow stable trends. Since the mid-1970s, however, the velocity link has become considerably less stable, and there is also some evidence that the relationship between Ml and the base has deteriorated as well. For the broader aggregates, while the multiplier relationships were not as tight as those for Ml even before the mid-1970s, the changes in velocity relationships have not been as pronounced as those for Ml. These results suggest that the broader aggregates are currently better suited for policy 67 purposes. The fourth section confirms these conclusions using money-income and base-money reduced form equations. The final section explores the reasons behind the recent instability in Ml relative to economic activity by focusing on the effects of innovation and deregulation on the public's demand for money. The way banks adjust the rates on various types of deposits as market rates change in a deregulated environment appears to have made the use of Ml as an intermediate variable more problematic than the use of the broader aggregates. The first appendix to this paper describes the breakdown in the relationship between narrow money and inflation. This appendix also contains a discussion of recent research on using M2 in combination with the level of unused capacity in the economy as a long-run leading indicator of inflation (often referred to as p*). The second appendix examines the operating procedures the Federal Reserve has used over the years in attempting to keep money on target. I. Circumstances Leading to the Adoption of Monetary Targets By the mid-1970s, the problems with using nominal interest rates as policy guides when inflationary expectations were shifting became increasingly apparent. Increases in inflation and inflationary expectations could cause real rates to fall even while nominal rates held steady or rose. Although real rates were difficult to measure, they were generally accepted as the important variables for spending decisions. It was also recognized that interest rates could give the wrong impression about the posture of monetary policy over time if interest rates were changing because of a shift in aggregate demand. To indicate more clearly whether policy had eased or tightened, policy makers should pay attention both to the movements in interest rates and to measures of bank reserves and/or the money stock. Around that same time, it also appeared increasingly important to establish a longer run strategy for lowering inflation.1 Control of the money supply, because of its presumed long-run relationship to inflation, was viewed as a key variable to attain price stability over time. In addition, monetary targets came to be seen by many as an approximate means to communicate longer run policy objectives to Congress and the general public, to establish central bank credibility and thus lower inflationary *For a discussion of the monetary aggregates as practical guides for policy, see Paul Volcker, "The Role of Monetary Targets in an Age of Inflation," Journal of Monetary Economics, vol. 4, no. 2 (April 1978), pp. 329-39. 68 expectations, and ultimately to hold the central bank accountable for its policies. Finally, monetary targets could provide a more enduring standard against which shorter run policy decisions could be evaluated over time. Monetary targets might also expedite policy decisions because some empirical work had suggested that movements in money led movements in nominal GNP and inflation by several months. Overall, many analysts believed that monetary targets could provide a nominal anchor for policy, both for internal operational purposes and for external communications. Monetary targets, however, were far from a panacea for policy. Establishing monetary targets also required dealing with many practical problems. These problems included: —choosing the monetary aggregate or aggregates to be targeted and the appropriate definitions of the monetary aggregates during a period of innovation and deregulation; —establishing the time period over which targets should be set, howfrequentlythey should be reset, how much they could reasonably be changed from year to year, and whether the previous target value for money or the actual value should be used as the base for the target in the next period (base drift); —deciding how quickly a deviation in money from target should be corrected to avoid overshooting the target and needing to reverse policy; —determining whether reserves or short-term interest rates would be better instruments to attain monetary targets; -exercising judgment in interpreting deviations in money from target as economically significant or insignificant; and deciding whether the policy process should focus only on money or on a broad range of indicators to ensure that money was giving the correct signal; —deciding how to identify and then deal with shifts in the trend of velocity and the demand for money, particularly when inflation and inflationary expectations dropped sharply following a period of monetary restraint, thereby changing the relative demands for financial and real assets. II. Theoretical Considerations in Using Money as a Policy Guide The theoretical reasons for and against using monetary aggregates as intermediate targets were debated long before many of the practical problems mentioned above became apparent. Readers already familiar with these arguments can skip to Section in where the empirical results begin. At a theoretical level, two models that represent extreme and opposite possibilities can be used to illustrate the circumstances under which the money supply would or would not be useful as an 69 intermediate target. In the first model, money plays a key role as an intermediate target. Causation flows from an exogenously determined supply of high-powered money (reserves and currency) to the money stock, with a stable or predictable money multiplier serving as the link. The money stock, in turn, determines the level of nominal spending, this time with a stable, predictable velocity providing the link. The observed correlation between money and income occurs because money induces changes in nominal spending. In addition, since the demand for money is viewed as being stable and relatively insensitive to changes in interest rates, changes in exogenous expenditures will not cause income to deviate from the desired path. Hence, managing the monetary base in such a way that money grows at a moderate rate would also achieve a noninflationary path for nominal spending. In other words, monetary control is the key to preventing inflation in the longer run. In the other model, monetary aggregates as intermediate targets or indicators would have a much reduced role, if any. In this model, causation runs from shifts in exogenous spending to changes in total nominal spending and then to increases or decreases in the quantity of money demanded. The changes in the quantity of money demanded in turn lead to changes in the demand for the monetary base, changes which are passively accommodated by the central bank. In other words, money is endogenously rather than exogenously determined, and the correlations between money and income and between money and the monetary base result from changes in spending that induce changes in the demand for money and the demand for the base. Even if the central bank attempted to restrain the monetary base and therefore money, the high interest rate sensitivity of the demand for money and the instability in the money demand function would make control of the money stock as an intermediate target a difficult exercise. In this model, money might serve as an early indicator of changes in spending (only because money data are available before GNP data), but there would be little value in controlling money because keeping money on a steady path would not prevent changes in autonomous expenditures from affecting GNP.2 In this model, inflation would not occur so much because of changes in the money supply, but rather because of changes in autonomous expenditures that pushed aggregate spending beyond the economy's capacity to produce. In other words, a sensible fiscal policy 2 For good textbook presentations of alternative approaches to stabilization policy, see Thomas Mayer, James Duesenberry, and Robert Aliber, Money. Banking and the Economy (New York: W.W. Norton and Company, 1987); and Lawrence Ritter and William Silber, Principles of Money. Banking and Financial Markets (New York: Basic Books, Inc., 1989). 70 Table 1 Basic IS-LM Modef (1)Y =-cr + X (2)M = -ar + bY + Z (3)M = R/m + dr + K M = narrow money supply r = the interest rate Y = income Z = money demand shift X = autonomous expenditures R = monetary base K = money supply shift m = money multiplier a,b,c,d = structural parameters Reduced Form Multipliers JL JL + r = M = + c a+d+bc a+d+bc a+d a+d+bc 1 m(a+d+bc) Y = 1 a+d+bc 1 a+d+bc b a+d+bc a+bc m(a+d+bc) + a+bc a+d+bc a+d+bc bd a+d+bc c m(a+d+bc) + X -'Prices are assumed constant in the short run so that nominal and real values are the same. would be the key to price stability.3 In practice, however, causation in the economy does not flow in just one direction, as portrayed in these extreme models. The basic IS-LM model can be used to illustrate some of the complexities facing policy makers in deciding whether to use money as an intermediate guide. The first table shows the basic IS-LM model and the derivation of the reduced form multipliers. 3 For overviews of the theoretical arguments for intermediate targets, see Edward Kane, "Selecting Monetary Targets in a Changing Financial Environment," Money Policy Issues in the 1980s (Proceedings of a symposium sponsored by the Federal Reserve Bank of Kansas City, August 1982), pp. 181-206; Benjamin Friedman, "The Value of Intermediate Targets in Implementing Monetary Policy," Price Stability and Public Policy (Proceedings of a symposium sponsored by the Federal Reserve Bank of Kansas City, August 1984), pp. 169-91; and Richard G. Davis, "Monetary Aggregates and the Use of 'Intermediate Targets' in Monetary Policy," New Monetary Control Procedures, vol. 1, Board of Governors of the Federal Reserve System Staff Study, February 1981, pp. 1-44. 71 For the Federal Reserve to control money and GNP with a great degree of precision, it would be necessary to know the structural parameters of the model (a, b, c, and d) and the source and size of other exogenous shocks to the economy (K, Z, and X) in order to offset their impacts on money and GNP. If the Federal Reserve cannot anticipate and offset these shocks, both money and GNP will be affected. For shifts in the expenditure function (X) and shifts in the supply of money function (K), the impacts on money and GNP will have the same signs. Therefore, by looking at money, the Federal Reserve can at least discern whether it needs to contract or expand the supply of reserves to stabilize GNP. However, for money demand shifts (Z), the multipliers have opposite signs and the Federal Reserve could destabilize GNP by attempting to keep money on target when the demand for money changes. Finally, even if the demand for money were stable and the Federal Reserve only needed to respond to shifts in the expenditure function (X), stabilizing money would not necessarily stabilize GNP entirely. Only in the special case when the interest elasticities of the supply of and demand for money (a and d) were equal to zero would keeping money on target ensure that the desired performance of GNP would be attained. In this case, neither money nor GNP would be affected as a result of shifts in the expenditures function (X). In the other extreme case, however, when the interest elasticity of the demand for money (a) would be very large, shifts in autonomous expenditure would have small effects on money but relatively large impacts on GNP. Money might therefore stay quite close to target even while GNP deviated substantially. Clearly, even a very simple model underscores the need for caution in using money as an intermediate target. Unanticipated shifts in the demand for money, of course, appear to be the most serious difficulty that could arise. In this case, a reduction (increase) in the demand for money, without any change in the supply of money, could create inflationary (recessionary) pressures. In addition, a very large money demand interest elasticity could also be a significant drawback to monetary targeting. An increase (decrease) in autonomous expenditures could create inflationary (recessionary) pressures, by pushing GNP growth beyond (below) its potential, without appreciably more (less) rapid money growth.4 4 For more background, see William Poole, "Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Model," Quarterly Journal of Economics, May 1970, pp. 197-216; John Wenninger, "Financial Innovation, A Complex Problem Even in a Simple Framework," Federal Reserve Bank of New York Quarterly Review, Summer 1984, pp. 1-8; and James Tobin, "Monetary Policy: Rules, Targets, and Shocks," Journal of Money, Credit, and Banking, November 1983, pp. 506-18. 72 In general, the following conditions would favor monetary targeting when the monetary base is the policy instrument: — low interest rate elasticity in the supply of money function (d) and small values of shifts in a the supply of money (K). These conditions would tighten the linkage between the base and the money stock. The ratio of the money stock to the base (the money multiplier) could be used to determine how much monetary base the Federal Reserve should supply to attain its monetary objectives. — low interest rate elasticity in the money demand function (a) and small values of shifts in a the demand for money (Z). Under these circumstances, the money stock would largely determine the level of income, and any shifts in the expenditures function (X) would mainly affect interest rates. Money and income would move together in a predictable way (stable velocity), and velocity could be used to calculate the target path for money consistent with a desired GNP objective.5 III. Predictability of Velocity and Multiplier Trends This section uses extrapolations of velocity and money multiplier trends to illustrate the predictability of these trends with straightforward and transparent techniques. The analysis reveals that for Ml, in particular, the predictability of these monetary relationships has deteriorated considerably. The recent breakdown in the money-income (GNP-M1) relationship has been well documented in numerous other studies.6 To illustrate the timing and relative magnitudes of this breakdown, the 5 It is, of course, also possible to control the money stock using the interest rate. In this case, the reduced form for money is M=(-a-bc)r+bX+Z. When M is controlled using the interest rate, shifts in money demand (Z) and autonomous expenditures (X) would have larger impacts on M than when the base is used, but shifts in the supply of money function (K) would not have any effect. Hence, the choice of instrument—the base or the interest rate—would depend in part on the relative sizes of the shift parameters in the IS-LM model. 6 For overviews of the breakdown from the monetarist and nonmonetarist perspectives, see Benjamin Friedman, "Lessons on Monetary Policy in the 1980s," The Journal of Economic Perspectives, Summer 1988, pp. 51-72; and William Poole, "Monetary Policy Lessons of Recent Inflation and Disinflation," The Journal of Economic Perspectives, Summer 1988, pp. 73-100. Some analysts, however, do not agree that a breakdown in the relationship of money and GNP occurred in the 1980s. They find that an alternative definition of Ml that excludes NOW accounts has tracked (continued...) 73 second table shows the annual errors in forecasting the growth of Ml velocity (as well as the velocities of M2 and M3) as an average of the previous ten years. This approach allows changes in the trend of velocity to be gradually reflected in the forecast without putting too much weight on any given year. While any such naive approach to forecasting velocity is somewhat arbitrary, it is sufficient to show the relative stability of these velocity trends over time. Indeed, monetary targets appealed to policy makers because the techniques used to project velocity could be simple and could easily be explained to Congress and the public. That is, elaborate econometric models were not required for selecting the monetary targets and making them understood as long as the econometric evidence suggested a low interest rate elasticity and stability in the demand for money function. From the early 1960s to the mid-1970s, the average error in forecasting Mi's velocity was zero, while the root mean squared error (RMSE) was about 2 percentage points. Errors of this size were interpreted as meaning that the monetary authorities should use a range about 3 percentage points wide rather than setting a single numerical target for Ml. From the mid-1970s to 1988, however, the average error was -1.5 percentage points and the RMSE more than doubled to 4.4 percentage points.7 6 (...continued) economic activity fairly well. For more detail, see Michael Darby, Angelo Masarco, and Michael Marlow, "The Empirical Reliability of Monetary Aggregates as Indicators," Economic Inquiry, October 1989, pp. 555-85. For a wide range of views on the behavior of velocity in the 1980s, see Michael Darby, William Poole, David Lindsey, and Milton Friedman, "Recent Behavior of the Velocity of Money," Contemporary Policy Issues, vol. 5, no. 1 (January 1987), pp. 1-33. 7 Unusual movements in the actual Ml velocity statistics relative to expected velocity occurred in two subperiods of the interval from the mid-1970s to the late 1980s. The first was 1972 to 1978, when velocity (except for one year) was unusually strong, most likely as a result of increased emphasis on cash management as nominal rates rose to extremely high levels. For more detail, see Stephen Goldfeld, "The Case of the Missing Money," Brookings Papers on Economic Activity, 3:1976, pp. 683-739; and Thomas Simpson and Richard Porter, "Some Issues Involving the Definition and Interpretation of the Monetary Aggregates," Controlling Money Stock HI, Federal Reserve Bank of Boston, October 1980, pp. 161-233. Other analysts argued that the demand for money had been stable but the function had not been properly specified. See, for example, Michael Hamburger, "Behavior of the Money Stock: Is There a Puzzle?" Journal of Monetary Economics. July 1977, pp. 265-88. The second period of instability was from 1982 to 1986, when velocity declined sharply. This decline probably stemmed from the sharp drop in interest rates at a time when the deregulation of deposit interest rates had made the demand for Ml more interest sensitive, at least in the shorter run. For more detail, see John Wenninger, "Responsiveness of Interest Rate Spreads and Deposit Flows to Changes in Market Rates," Federal Reserve Bank of New York Quarterly Review, Autumn 1986, (continued...) 74 In the mid-1980s, the extremely large errors prompted the Federal Reserve to drop its target range for Ml. The velocities of the broad aggregates in the earlier time period were not as predictable as Mi's velocity, with RMSEs 1/2 to 1 percentage point greater than for Ml. However, the velocities of the broader aggregates did not become as unpredictable as Mi's in the later period. For M2, the RMSE increased from 2.5 to 3.3 percentage points, whereas for M3 the velocity forecast errors declined slightly to 2.5 percentage points. This change in the relative predictability of the velocities of Ml and the broader aggregates was a major factor in persuading the Federal Reserve to place greater emphasis on the broad aggregates during the 1980s and to cease setting targets for Ml. 8 Table 3 contains some additional information on velocity growth since 1961. The left side shows maximum and minimum values for velocity growth on an annual basis, while the right side shows the maximum and minimum values on a ten-year moving average basis. Whether calculated on a year-to-year basis or on a ten-year moving average basis, Mi's velocity looks the least stable. On a ten-year moving average basis, M3's velocity looks the most stable, while on a year-to-year basis M2's velocity appears the most stable. Again, when looked at in this light, the broad aggregates appear to be better suited for policy purposes than Ml. Not only are the RMSEs smaller on average, but they also have narrower ranges of extreme values. Next, we turn to the stability over time of the relationship between the money supply and the monetary base.9 Some analysts have raised serious questions about using the base as a policy 7 (...continued) pp. 1-10; and Courtney Stone and Daniel Thornton, "Solving the 1980s' Velocity Puzzle: A Progress Report," Federal Reserve Bank of St. Louis Review. August-September 1987, pp. 5-23. Also see references cited in footnote 4. ^ e r e was also evidence that the demand functions for the broad aggregates had been more stable than the demand for Ml. For more detail, see John Wenninger, "Money Demand—Some Long-Run Properties," Federal Reserve Bank of New York Quarterly Review. Spring 1988, pp. 23-40. *We use the St. Louis adjusted monetary base (adjusted to neutralize the effects of changes in reserve requirements) because it has been used extensively in earlier empirical work and is available for a long time period. For a detailed review of the relationship between Ml and the base focusing on the effects of deregulation, see Michelle R. Garfinkel and Daniel Thornton, "The Link Between Ml and the Monetary Base in the 1980s," Federal Reserve Bank of St. Louis Review. September-October 1989, pp. 35-52. 75 Table 2 Errors in Predicting Velocity (Using Simple Velocity Trends) GNP/M1 GNP/M3 -10.8 1.8 2.7 0.6 -1.4 -0.8 -0.9 3.2 3.9 -2.5 2.3 2.9 -1.2 -3.7 -0.3 4.5 1.0 -2.0 -3.4 0.9 6.2 0.6 0.5 -0.5 -6.4 -2.1 1.1 -1.9 -4.0 4.5 1.5 -0.2 -1.9 -1.5 -1.6 2.6 4.0 -2.9 2.1 6.1 -3.7 -4.1 -0.7 1.0 -0.8 1.7 -0.8 0.4 3.3 -0.4 1.4 -2.0 -5.7 1.4 -0.8 -0.2 -2.9 4.1 1.6 1961 to 1988 Mean = RMSE = -0.7 3.5 0.1 2.9 0.0 2.7 1961 to 1974 Mean = RMSE = 0.0 2.1 0.5 2.5 -0.1 3.0 1975 to 1988 Mean = RMSE = -1.5 4.4 -0.4 3.3 0.1 2.5 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1.7 1.0 -0.4 -1.6 2.8 1.8 -3.9 -1.0 0.1 -3.2 -0.3 0.7 3.2 -0.5 2.5 0.6 0.6 2.9 -2.0 -1.1 0.2 -8.9 GNP/M2 -2.8 0.8 -7.2 Table shows errors resulting when each year's velocity growth (fourth quarter to fourth quarter) was predicted as the average growth rate over the previous ten years. 76 Table 3 Range of Values for Velocity Growth (1961 to 1988, in Percent) Yearly (Fourth Quarter to Fourth Ouartert Minimum Maximum Range Ten-Year Moving Average Maximum Range Minimum Ml -9.4 6.2 15.4 0.0 3.8 3.8 M2 -5.4 4.3 9.7 -0.9 1.0 1.9 M3 -6.1 5.3 11.4 -0.3 -1.3 1.0 instrument.10 Over the years, other reserve aggregates such as total reserves, reserves on private deposits (RPDs), nonborrowed reserves, and the nonborrowed base have been studied. In addition, some analysts who accept the notion that it would be desirable to control money have left open the question whether it should be done through a reserve aggregate or through a short-term interest rate. When controlled with an interest rate variable, the money demand function often has been viewed as the reduced form equation for money.11 Whether some reserve measure or interest rate should be used in practice depends upon the properties of the money demand and supply functions, that is, their interest rate sensitivity and relative stability. The Federal Reserve has at different times used both short-term rates and reserves aggregates in its attempt to control money (Appendix 2 examines this topic in more detail). The empirical work in this paper focuses only on the monetary base because other papers in this volume will deal with alternative reserve measures and interest rates in the policy 10 Benjamin Friedman, "Conducting Monetary Policy by Controlling Currency Plus Noise," Carnegie-Rochester Conference Series on Public Policy, no. 29 (1988), pp. 205-12. The main problem with the base is that it consists of 75 percent currency and 25 percent reserves, and the Federal Reserve has usually provided an elastic currency. It is not clear that the Federal Reserve could control the entire base as an instrument simply by adjusting the relatively small reserves portion. 1 Barnes L. Pierce and Thomas D. Thomson, "Some Issues in Controlling the Stock of Money," in Controlling Monetary Aggregates II: The Implementation (Proceedings of a conference sponsored by the Federal Reserve Bank of Boston), September 1972, pp. 115-36. See also Richard G. Davis and Frederick C. Schadrack, "Forecasting the Monetary Aggregates with Reduced Form Equations," Federal Reserve Bank of New York, Monetary Aggregates and Monetary Policy, October 1974, pp. 60-71. 77 process. Tables 4 and 5, similar in format to Tables 2 and 3, respectively, assess the predictability of the money multiplier (ratio of a monetary aggregate to the monetary base) using the naive model of a tenyear moving average as the projection of the growth rate of the multiplier for each year. Like velocity, the money multiplier would be a much more useful policy tool if relatively simple techniques could be used to predict its performance, not only in terms of clearly expressing the intent of policy but also in terms of attaining monetary objectives. Not all those who would advocate using the monetary base would agree with such an approach, but some have pointed out the relative ease with which the trend in the multiplier might be predicted.12 In Tables 4 and 5, the annual data (fourth quarter to fourth quarter) are used to correspond to thefirameworktypically used for setting the longer run monetary targets (see footnote 12 for references on short-term multiplier forecasts). Table 4 shows that the Ml multiplier was the most predictable over the 1961 to 1974 period, with a RMSE (1.1 percentage points) about half as large as the RMSEs for the M2 and M3 multipliers. However, over the 1974 to 1988 period the RMSE for the Ml multiplier more than doubled and the RMSEs for the multipliers of all three aggregates were in the 2.0 to 2.S percentage point range in the later period. Hence,