The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Open Market Policies and Operating Procedures-Staff Studies BOARD OF GOVERNORS OF THE FEDERAL RESERVE SYSTEM Library of Congress Catalog Card Number 72-611769 C o p ie s o f th is p a m p h le t m a y b e o b ta in e d fr o m P u b lic a tio n s S e r v ic e s , D iv is io n o f A d m in istr a tiv e S e r v ic e s , B o a r d o f G o v ern ors o f th e F e d e r a l R e s e r v e S y s te m , W a s h in g to n , D .C . 20 5 5 1 . T h e p r ice is $ 2 .0 0 p e r c o p y ; in q u a n titie s o f 10 o r m o re s e n t to o n e a d d r e ss, $ 1 .7 5 e a c h . R e m itta n c e s h o u ld b e m a d e p a y a b le to th e o r d e r o f t h e B o a r d o f G o v e r n o r s o f th e F e d e r a l R e s e r v e S y ste m in a fo rm c o lle c tib le a t p a r in U .S . c u r r en cy . (S ta m p s a n d c o u p o n s n o t a c c e p te d .) P U B L I S H E D I N J U L Y 1971 The staff studies included in this book were prepared as part of a study of the policies and operating procedures im plicit in the policy directives of the Fed eral Open Market Committee. These stu dies represent the views of the individ ual authors and should not be attributed to either the Federal Open Market Com mittee or individual members of that Committee. The role of monetary aggregates and of money market conditions in the decision-making process of the FOMC and in the day-to-day conduct of open market operations is described in an article, “Monetary Aggregates and Money Market Conditions in Open Market Policy,” which appeared in the Federal Reserve B u lletin for February 1971. This article is reprinted here as an appendix. THE FOMC DIRECTIVE AS STRUCTURED IN THE LATE 1960’s: THEORY AND APPRAISAL Stephen H. Axilrod SHORT-RUN TARGETS FOR OPEN MARKET OPERATIONS Richard G. Davis SELECTION OF A MONETARY AGGREGATE FOR USE IN THE FOMC DIRECTIVE Leonall C. Andersen DETERMINING THE OPTIMUM MONETARY INSTRUMENT VARIABLE John Kareken, Thomas Muench, Thomas Supel, and Neil Wallace THE TRADE-OFF BETWEEN SHORTAND LONG-TERM POLICY GOALS James L. Pierce TACTICS AND STRATEGY IN MONETARY POLICY Benjamin M. Friedman RULES-OF-THUMB FOR GUIDING MONETARY POLICY William Poole APPENDIX: MONETARY AGGREGATES AND MONEY MARKET CONDITIONS IN OPEN MARKET POLICY Stephen H. Axilrod by Stephen H. Axilrod THE FOMC DIRECTIVE AS STRUCTURED IN THE LATE 1960’s: THEORY AND APPRAISAL CONTENTS 3 INTRODUCTION 3 3 5 8 STRUCTURE OF THE DIRECTIVE Nature of first paragraph Operational elements in the second paragraph Role for Manager’s judgment 9 FUNCTION OF MONEY MARKET CONDITIONS AS AN OPERATING GUIDE Day-to-day role of free reserves and the Fed eral funds rate Money market conditions in relation to bank deposits Money market conditions in relation to over-all interest rates Evaluation of the need for a money market conditions guide 10 11 12 14 17 17 19 21 23 POSSIBLE RELATIONSHIP BETWEEN THE STRUCTURE OF THE DIRECTIVE AND A THEORY OF MONETARY POLICY FORMULATION Formulation of longer-run projections Role of money market conditions and proviso in relation to longer-run projections Errors and uncertainties considered Money market conditions: policy target aspects 24 RECAPITULATION AND CONCLUDING REMARKS 28 APPENDIX: An Empirical View of “Even Keel” FOMC DIRECTIVE IN LATE 1960‘s INTRODUCTION This paper attempts to lay out and appraise the workings of, and a possible theory for, the structure in the latter part of the 1960’s of the Federal Open Market Committee’s directive to the Manager of the Open Market Account. An effort is made to indicate and evaluate the practice of open market policy as it flowed from the structure of the directive. The paper also attempts to outline one theoretical ration ale for the directive’s structure and to indicate the nature of the flow of economic information, including projections, that appears to be re quired to satisfy such a theoretical under pinning for the directive. STRUCTURE OF THE DIRECTIVE For many years the FOMC directive has con tained two paragraphs. The first paragraph is a statement about the economy and the general goals of monetary policy, while the second contains operating guides for the Account Manager covering the interval between Open Market Committee meetings. In the recent past this interval has generally been 3 or 4 weeks. The nature of the information and instructions in these two paragraphs has changed over the years. In this section, the paragraphs as they were formulated in the late 1960’s will be de scribed and evaluated.1 NATURE OF FIRST PARAGRAPH. The first paragraph of the directive typically con tained statements about over-all economic ac tivity, prices, various financial flows—particu N o t e . —The author is Associate Director, Division of Research and Statistics, Board of Governors of the Federal Reserve System. 1 The wording of the directive issued on Aug. 12, 1969, is as follows: The information reviewed at this meeting indicates that expansion in real economic activ ity slowed somewhat in the first half of 1969 and some further moderation is projected. Sub stantial upward pressures on prices and costs are persisting. Most market interest rates recently have receded slightly from their earlier highs. In July the money supply expanded as U.S. Govern ment deposits decreased further; bank credit larly bank credit and money—and interest rates. Generally, only the statement about over-all economic activity had a future cast to it. But the time horizon for this future was often rather indefinite. Sometimes the wording has been such that the reader would think it referred to no more than a quarter ahead, or to the quarter in process. An example of such wording would be “economic activity appears to be slowing.” On the other hand, at times statements simply noted that economic activity is projected to slow. In such cases the time ho rizon appears more indefinite. To understand the magnitude and timing of the future projections of economic activity that provide a basis for FOMC decisions and for the instructions given to the Account Manager, it is necessary to look outside the directive it self. For the public, the policy record that is published at the same time as the directive (both with about a 3-month lag) contains a general indication as to the direction and mag nitude of the gross national product, but the references are qualitative and not necessarily consistent as to time periods mentioned. The declined on average, after adjusting for an in crease in assets sold to affiliates and to customers with bank guarantees. The run-off of large-denomination CD’s which began in mid-December continued without abatement in July, and there apparently were net outflows from consumer-type time and savings accounts at banks and nonbank thrift institutions combined. The over-all balance of payments deficit on the liquidity basis re mained very large in July; the balance on the official settlements basis was still in surplus in the first half of the month but subsequently shifted toward deficit as U.S. banks’ borrowings of Euro-dollars leveled off. Foreign exchange markets appear initially to be adjusting in an orderly fashion to the announced devaluation of the French franc. In light of the foregoing de velopments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to the reduction of inflationary pres sures, with a view to encouraging sustainable economic growth and attaining reasonable equili brium in the country’s balance of payments. To implement this policy, System open market operations until the next meeting of the Com mittee shall be conducted with a view to main taining the prevailing firm conditions in money and short-term credit markets; provided, how ever, that operations shall be modified if bank credit appears to be deviating significantly from current projections or if pressures arise in con nection with foreign exchange developments or with bank regulatory changes. FOMC itself has available to it specific, dated projections presented by the staff at each meet ing. The staff projections are in considerable detail with specific numbers and generally with a time horizon of about a year. Thus, the economic analysis behind FOMC instructions to the Account Manager in the directive cannot be understood by reference only to the first paragraph of the directive; it requires other documentation. Some might argue that there is no reason for the first para graph to express the full scope of the eco nomic and financial analysis that lies behind the specific operating instruction of the second paragraph. However that may be, the main point here is that the structure and meaning of the directive that is issued cannot be under stood in itself but must be considered in rela tion to the information gathering, economic analysis, and policy discussion that are integral to the FOMC meeting. In that respect, it should, of course, be pointed out that the staff material and projections may not give a cor rect impression of the views of the members of the FOMC. Their outlook for the future has often been different from the staff’s, and a thorough understanding of their views and the relation of these views to the directive requires access to the minutes of the meeting, although a brief summary of the policy discussion is contained in the policy record that is published along with the directive. The final sentence of the first paragraph of the directive states the goals of monetary pol icy as they relate to the balance of payments, economic growth, and inflation. From time to time the structure of this sentence is rearranged so as to give particular emphasis to the bal ance of payments, or to price stability, or to the need to encourage economic growth, as may be appropriate. This rearrangement can then be taken to represent a general statement of the Committee’s over-all priorities with re spect to the ultimate goals of policy. There is no explicit mention of potential trade-offs among various competing goals, however, in the final sentence of the first para graph. The order in which the goals are pre sented may give some indication of priorities attached to particular goals by the Committee, but there is nothing to indicate that the Com mittee is considering the sacrifice of a degree of attainment of one goal in order to obtain a greater degree of attainment for another goal. In fact, it may be an overstatement to suggest that rearrangements of the wording of this sen tence indicate explicit consideration by the Committee of the trade-off problem. It is more likely that rearrangement should be interpreted as indicating that the Committee is moving, for example, toward an emphasis on combating in flation rather than encouraging growth. But whether the Committee believes it can have both some desired level of economic growth and a desired degree of price stability over some given time period is certainly not made clear in the general statement of goals. The indefinite nature of the time horizon of the first paragraph and its very general state ment of goals make its connection with the op eration elements of the second paragraph rather tenuous. The second paragraph refers explicitly to how the Account Manager should operate in the market over the interval be tween Committee meetings. Presumably, these operations would be consistent with the desires of the Committee with respect to the economy and the balance of payments as expressed in the last sentence of the first paragraph. But how these two paragraphs relate to each other is not made clear in the directive itself, or in the policy record accompanying the directive. In order to relate them it would seem neces sary to analyze the relationship of: (1) the op erating variables that the Manager works with, (2) the financial flows and over-all interest rates that result from these operations, and (3) re lated effects over time on economic activity, prices, and the balance of payments. Thus, while clear for the first paragraph of the direc tive, it is even clearer for the whole structure of the directive as constructed in the late 1960’s— not to mention other periods— that the directive cannot be analyzed independently of the total FOMC DIRECTIVE IN LATE 1960’s flow of material and projections given to the FOMC, and of the nature of the discussion undertaken by the FOMC in relation to this material. In brief, the procedures of the FOMC and the directive are inseparable.2 Before discussing the relationships among the day-to-day operational variables in the sec ond paragraph of the directive, aggregate mon etary flows, over-all interest rates, and longerrun projections of the economy, it is desirable to describe the constituent elements of the sec ond paragraph that affect the Manager’s opera tions. These are by no means clear, of course, as expressed in the second paragraph of the directive, but they are fairly clear to those present at FOMC meetings and with access to the full FOMC documentation. OPERATIONAL ELEMENTS IN THE SECOND PARAGRAPH. The second para graph of the FOMC directive generally has asked the Manager to maintain— or ease, or seek tauter, as the case may be—money market conditions. Sometimes the term “money market conditions” has been expanded to “money and short-term credit market conditions.” In addi tion, in the last 4 years of the 1960’s the sec ond paragraph included a proviso clause, which noted that the money market conditions should be attained provided bank credit was not deviating significantly from projections. More over, the second paragraph contains, when appropriate, references to “even keel” around periods of Treasury financings. And finally, this paragraph has made references to possible modifications of operations in cases of liquidity crises or similar emergencies, such as excep tionally large outflows of funds from banks or thrift institutions at interest-crediting periods or potential domestic market reactions to foreign exchange market developments. The Manager also appears to have a continuing authority to 2 This paper will not, however, discuss the content of FOMC discussions and the nature of the go-around among FOMC members, but will rather con centrate on the economic issues germane to the theo retical basis of the directive— thereby implicitly discussing the types of decisions that would appear to require FOMC discussion. avert disorderly market conditions; just how such conditions are defined is unclear, but they are generally taken to mean a drying-up of trading in securities and large and cumulative downward price movements for which no end seems in sight. Money and other short-term market condi tions. The money and other short-term market conditions referred to in the second paragraph include principally the Federal funds rate, bor rowings by member banks, and net free or net borrowed reserves. At times, the rate on 3month Treasury bills has been included in this constellation. The words “other short-term market conditions” have generally been taken to indicate inclusion of the 3-month bill rate, although that rate has also at times been something of a factor in operations even with out such specific wording. The emphasis placed on the bill rate has varied considerably with monetary and economic conditions. For instance, in the early 1960’s when it was thought that international flows of funds were responsive to relations between short-term market rates here and abroad, much attention was paid to the 3-month bill rate in opera tions. Also, the 3-month bill rate was a partic ularly important operating variable when it and the whole bill rate structure were hovering around Regulation Q ceilings, and the Com mittee did not wish to encourage either a large expansion in bank credit that might be associ ated with a decline in the bill rate or a large contraction that might be associated with a rise in the bill rate. A constellation of money market conditions, rather than a single indicator, has been relied on for operating purposes because changes in reserve distribution and other temporary mar ket factors may result in divergent tendencies in any one of the money market conditions, and such a divergent tendency might be offset through manipulation of other conditions in order to maintain an over-all degree of ease or tightness in the money market. For example, when reserves are distributed in favor of lead time went on, and nondeposit funds became relatively important sources of bank credit, the bank credit referred to became the total of member bank deposits plus the average for the month of weekly data on liabilities to branches abroad, and then finally plus the average of borrowings through commercial paper issued by bank-related affiliates. A theory behind the proviso clause will be discussed in an ensuing section of the paper, including some discussion of what particular aggregates might best be included in such a clause. The proviso was generally a two-way pro viso. That is, the Manager was directed to ease money market conditions a little if bank credit were falling short of expectations and was di rected to tighten them a little if bank credit were rising above expectations. Sometimes, however, the proviso was expressed so that its effect was only one way. For example, if the Committee were particularly anxious to avoid a sharp rise in bank credit, it might have di rected the Manager to alter money market conditions only if bank credit were rising above projections. For the most part, projections of bank credit provided by the staff were for only 1 month ahead, although on occasion figures for a slightly longer time period were presented. The monthly projections were based on recent trends in deposit data, knowledge as to likely Treasury financing activity, expectations as to the effects of market interest rates on time de posits given Regulation Q ceilings at banks, and a view as to the intensity of loan demands in light of the outlook for GNP. Either the projections have assumed no change in money market conditions, or if a change in the sec ond paragraph was to be considered by the FOMC, then they have been based on some what tighter or easier money market conditions, as the case might be. The Committee generally, but not always, accepted the staff projections as the appropriate quantities for the proviso clause. 3 Loans and investments of the large weekly re porting banks are available weekly in fairly detailed There was nothing in the directive to indi categories, while estimates for loans and investments cate when the proviso would be put into effect at all commercial banks are available only for the last Wednesday of the month. —whether it would be after 1 week of devia ing money center banks the Federal funds rate will often decline, and this would appear to be an indication of easing in the market unless member bank borrowings are permitted to rise and net borrowed reserves deepen. On the other hand, when reserves and funds move away from money center banks, the Federal funds rate will tend to rise because these major money center banks appear to be more willing than other banks to borrow and to pay higher rates for day-to-day money in the market. In such a case, if member bank borrowings are not per mitted to decline somewhat, the over-all money market will appear to tighten. There are limits within which these trade-offs can take place, and the range of trade-offs represents the over all constellation of money market conditions that have been the day-to-day operating guide for the Account Manager. The operating emphasis on money market conditions has meant that the directive was es sentially accommodative, in the sense that market demands for credit and money would be accommodated at a given Federal funds rate or level of net borrowed or net free reserves. Some constraint on the degree of accommoda tion was instituted by the proviso clause, but in practice this represented a rather minor ele ment of constraint, in part because the Com mittee was willing to tolerate wide swings in bank credit and in part because the proviso clause was not in application taken as a strong target of policy. Bank credit proviso. The proviso clause in the directive during the latter part of the 1960’s was for the most part related to bank credit, although in its early days required reserves were used (and on one or two occasions money supply was noted along with bank credit). The bank credit referred to was origi nally a proxy for daily-average bank credit as measured by total member bank deposits, a se ries for which daily figures are available.3 As FOMC DIRECTIVE IN LATE 1960’s tion from projections or 2 or 3. Nor was there anything in the directive to indicate how much of a change in money market conditions the Manager should seek in light of a deviation of bank credit from projections. Much of the time the word “significantly” appeared in the proviso in relation to deviations from projec tions, and this would appear to indicate that the deviation would have to be relatively large, with the dimension having to be gleaned by the Account Manager from Committee discus sion. No large change in money market condi tions was ever undertaken by using the proviso clause. Only small shadings were undertaken, no matter how large the deviation of bank credit from projections, with the FOMC recon sidering its whole stance at the next Commit tee meeting. When it was used, the proviso clause was generally not taken as a target, or at least not as a strong target, because the Ac count Manager was not directed to alter mar ginal reserves and money market conditions as need be to attain the specified bank credit range. Even keel. The words “even keel” have re ferred to the operations of the Federal Reserve Open Market Account around periods of Treas ury financings. As the appendix notes, “in practical terms ‘even keel* has meant that, for a period encompassing the announcement and settlement dates of a large new security offer ing or refunding by the Treasury, the Federal Reserve has not made new monetary policy decisions that would impede the orderly mar keting of Treasury securities and significantly increase risks of market disruption from sharp changes in market attitudes in the course of a financing.” The past timing of even keel and its effect on interest rates and monetary aggregates are discussed in some detail in the appendix and will not be repeated here. However, two points should be highlighted. One is that there have been rather marked fluctuations in both dayto-day interest rates and longer-term interest rates during even-keel periods, as well as fluc tuations in member bank borrowings and net reserves; but in spite of such fluctuations, the trend of the narrow money market measures has not generally changed during even-keel pe riods. As a second point, it should be noted that during even-keel periods the money sup ply and bank credit have often risen relative to their trend and that they have not always com pletely dropped back after even keel. If any general conclusion about even keel can be drawn, it may be that in such periods the Fed eral Reserve has permitted somewhat more ex pansion in monetary aggregates than it might otherwise have done in order to keep interest rate fluctuations more damped than they other wise would be. But whether such a conclusion should be at tributed to even keel, as such, is a question. Since the FOMC directive has been essentially an accommodative directive, and regardless of whether the System maintained even keel, very lumpy credit demands, such as the Treasury’s, would have been associated with an enlarged expansion in bank credit and money. The major impact of even keel has been that the System refrained from changing its constella tion of money market conditions in a period of Treasury financings, whereas it would not re frain from doing so in periods of particular corporate or State and local government financ ings. The reasons for refraining with respect to Treasury financings are the very large size of such financings and the extreme sensitivity of the markets as a whole to the receptions given these financings. Moreover, should such financ ings fail, the System would be under extreme pressure to take up the slack since the Treas ury generally requires the money either to roll over maturing debt or to finance committed expenditures. Liquidity, emergency, and other provisions. As noted above, one use of liquidity and emer gency provisions in the second paragraph of the directive has been to guard against market disruption in case of very large and unex pected net outflows of funds from banks and savings or other types of financial institutions. While these net outflows would often make funds available to the securities markets, they could raise the threat that the institutions would not be able to meet commitments and, therefore, that confidence in the institutions, and perhaps in financial markets generally, might be dissipated—with undesirable reper cussions on the economy itself. Certain kinds of liquidity and emergency provisions have also been used at times when foreign exchange markets have been in flux, and large outflows of funds from the dollar were in prospect that would have exerted strong and undesired up ward pressure on the interest rate structure in this country. Finally, it might be noted that the second paragraph of the directive has at times given the Manager authority with respect to adjust ing operations to take account of changes in the discount rate or reserve requirements when it seemed relatively certain that such changes were about to take place. Exactly how he should adjust operations is, of course, not spelled out in the directive. But some guidance has been given through Committee discussion or through staff analysis. Nevertheless, in this respect as in others, there is a role, although circumscribed, for the Manager’s judgment. ROLE FOR MANAGER’S JUDGMENT. The Manager’s judgment as to what money market conditions to seek has been circum scribed in recent years through greater statisti cal specification by the Committee. The staff has presented projections of bank credit, as noted earlier, and also detailed projections of member bank deposits, the money supply, time deposits, nondeposit sources of funds, and in terest rates generally, on the assumption of un changed money market conditions or, as an al ternative, either tightened or eased conditions. Ranges have been given both for money mar ket conditions and for the projected monetary aggregates. Needless to say, not all members of the FOMC would accept staff specifications as their own. Thus some members might pre scribe a slightly different range for the Federal funds rate, even for a directive for unchanged money market conditions. And some members might be more willing than others to see bank credit expand above, or move below, projec tions. Given the multiplicity of variables and the sometimes conflicting desires of various Com mittee members, the Manager has had consid erable scope to play off one variable against another as consistent with his sense of the de sires of the majority of the FOMC so long as at least some key variables remain within spec ified ranges. The problem of compromising among objectives, is made more difficult be cause not all Committee members necessarily discuss the same variables, so the Manager cannot be sure of the wishes of those members who have not expressed themselves with re spect to, say, the Federal funds rate or the 3month bill rate. Finally, it might be noted that the Manager seems to have had some capacity marginally to alter money market conditions if credit markets more generally were being buf feted by unusual conditions or if the public’s view of System monetary policy seemed to be changing undesirably—with market expecta tions developing that policy was either tighter or easier than the FOMC desired— as a result of a published series of money market statistics or operations deviating significantly from pre vious trends or actions. While a good deal of specification is pre sented by the staff and while the various Com mittee members themselves often specify numerically what they hope to see happen, de velopments often turn out differently from projections. This, of course, has been less likely to happen with the narrow money mar ket conditions— such as the Federal funds rate and net borrowed reserves—since these have been the principal operating variables the Manager sought to attain; and it has been much more likely to happen with bill rates, longer-term interest rates, bank credit, and money supply. In large part, of course, unex pected developments are the result of errors in specifying the relationship between money market conditions and monetary aggregates, or it may be that the levels of economic activity and credit demands are stronger or weaker FOMC DIRECTIVE IN LATE 1960’s than assumed for purposes of making the projection. But whatever the reason for the difference between the projected and the actual outcome with respect to interest rates and monetary ag gregates, or even with respect to narrow money market conditions, some outcomes are acceptable to the FOMC even though un specified as a possibility. For example, a greater than expected rise in interest rates, as com pared with projections, may turn out to be acceptable to the FOMC if this occurs at a time when demands in the economy are turn ing out to be larger than anticipated. In fact, the FOMC may often have told the Manager not to offset a market-generated tendency for interest rates to rise, or to fall. Whether the multiplicity of short-run tar gets means that the Manager has had more scope for judgment than if he had only a sin gle target is an open question. If the single target were net borrowed reserves, it would be clear that the Manager would have almost no scope for judgment, because net borrowed re serves are one of the more certainly attainable objectives within the constellation of short-run targets. However, if the single target were a rate of increase in the money supply, the Man ager might have to exercise a very considerable degree of judgment because he would likely be faced with sharp day-to-day variations in de posits and hence would have to make almost continuous judgments as to whether he should tighten money market conditions or ease them in the particular statement week in order to make sure that over the month, the quarter, or whatever the relevant period, he would attain the desired money supply target. While the degree of judgment required of the Manager need not be a principal factor in determining FOMC operating targets, the at tainability of targets with a reasonable degree of accuracy should probably be a criterion. What types of targets are so attainable, and over what time periods, are not within the pur view of this paper. The only point that might be added here is that emphasis on money mar ket conditions in the second paragraph of the directive has reflected in part a sense by the FOMC that such conditions represented an attainable target, one to which the Manager could be held accountable, and one that might minimize his scope for judgment in day-to-day operations. Other targets too might be feasible —and perhaps more desirable for economic reasons—but they would require more day-today judgmental decisions by the Manager since the target (for example, money supply or bank credit) might be one or two steps removed in terms of availability of statistics from the dayto-day flow of bank reserve adjustment data and money market information. Such targets might be attainable, but they would require that the FOMC provide the Manager with more day-to-day— or more importantly more week-to-week—freedom in operations and might also require greater tolerance for errors, given existing institutional arrangements (such as the structure of reserve requirements). FUNCTION OF MONEY MARKET CONDITIONS AS AN OPERATING GUIDE As an operating guide, money market condi tions have given the Manager a rather specific means by which he could determine whether or not to inject or absorb reserves. The net borrowed reserve position of member banks is relatively easy to meet within a week, particu larly since required reserves are given as a result of lagged reserve accounting, and the Federal funds rate is available every day. In addition to providing the Manager with a tar get that he can achieve and thus one to which he can be held accountable, the money market conditions target permits market demands to influence money, bank credit, and reserves, as has been earlier noted. In that sense it permits, among other things, the market to make its own seasonal adjustment of the money supply and related items. At the same time, of course, nonseasonal changes in demand would also be accommo dated. Whether such accommodation is desira ble has been one of the critical issues over the years in the FOMC’s method of operation, since it raises the danger of providing or ab sorbing bank reserves, credit, and money in a procyclical fashion. DAY-TO-DAY ROLE OF FREE RE SERVES AND THE FEDERAL FUNDS RATE. This section will analyze in detail the day-to-day operating function of free reserves and other money market conditions, principally the Federal funds rate. The net reserve position and the Federal funds rate are basic elements of money market conditions influencing the Manager’s day-to-day decisions as to whether to buy or sell securities. In the framework of the directive of the late 1960’s, it is his task to supply or absorb reserves in response to market demands under given money market conditions. The Federal funds rate—the rate banks charge for selling excess reserves to other banks, usu ally on an overnight basis—is one of the most sensitive measures of the demand for or the supply of reserves. While shifts in the distribu tion of reserves among major banks, or between major money market and country banks, affect this rate, a persisting tendency for the rate to rise from previous levels indicates a greater desire for reserves relative to supply than in earlier periods, and vice versa. The Federal funds rate generally bears a consistent, and relatively stable, relationship to the net free or net borrowed reserves position of member banks, although there can be week-to-week fluctuations between the two measures as a result of reserve distribution problems or unusual Treasury and other short term financing demands in the market. There can also be a longer-run shift in the relation ship—for example, the Federal funds rate may rise relative to net borrowed reserves if bank deposit drains cumulate and bank liquidity be comes increasingly strained, thereby increasing banks’ demands for Federal funds borrowings (and assuming their effective demand for bor rowing at the discount window is restricted by Federal Reserve rationing). In day-to-day op erations the Federal funds rate and net re serves have been considered jointly, while rec ognizing the necessity of some give-and-take in maintaining an over-all unchanged state of ease or tightness for the money market (as suming the FOMC voted for an unchanged state of money market conditions). The net reserve position of member banks is measured by the difference between their ex cess reserves and their borrowings. For pur poses of understanding the relation of free re serves to System operations, however, it is better to look at such reserves as the difference between nonborrowed reserves (the reserves that can be supplied through open market op erations) and required reserves (the result of joint decisions by banks and the public affect ing the level and distribution of deposits, at given interest rates). If the FOMC voted to keep money market conditions unchanged, the Account Manager would assume that the net reserve position of banks should remain about where it was in previous weeks. In his operations the amount of reserves he supplied or absorbed through the market would depend on other sources of nonborrowed reserves and on required reserves during the statement week. Thus, the Desk has to have at hand projections of float, the Treas ury balance at the Federal Reserve, currency in circulation outside the banking system, gold flows, and foreign drawings or repayments on Federal Reserve swap lines, all of which are factors other than his own operations that affect nonborrowed reserves and that are for the most part outside his control. In addition, the Desk would need to have for the current statement week estimates of the amount and distribution of deposits by type of deposit and class of bank in order to obtain a measure of required reserves. Under the lagged reserve scheme put into effect in Sep tember 1968, required reserves in a current statement week are based on deposits 2 weeks earlier, and thus the Desk knows with cer tainty what required reserves will be in the FOMC DIRECTIVE IN LATE 1960’s current week. But the System had operated with a money market conditions target (with or without a proviso) for a great many years before adopting the lagged reserve provision, and the theory of using money market condi tions as an operating guide is little different with or without lags—although the timing of the effects of operations on key financial variables might be affected by the presence of lags. MONEY MARKET CONDITIONS IN RELATION TO BANK DEPOSITS. Over the very short-run period of a bank reserve statement week, bank deposits are probably determined mainly by credit demands on banks and by bank investment policies, given money market conditions and, more generally, the level and structure of interest rates. As in dividual banks enter a new statement week, they are confronted with particular supply and demand conditions. On the supply side, they are faced with a set of fund availabilities given to them and about which they can do little (U.S. Government and private demand depos its, which in large part are beyond their influ ence in the short run)4 and costs (reserve requirements; rates on Federal funds on Euro dollars, and on CD’s and other time depos its if available under Regulation Q; and so forth) that influence their willingness to obtain additional funds and affect their loan terms 4 There are obvious exceptions to the statement that pertain to both private and Government demand deposits, but some of these in reality apply to banks’ lending or borrowing policy rather than demand de posit flows as such. For example, banks can obtain U.S. Government deposits at times by bidding for Treasury bills offered with payment through credit to tax and loan accounts. But to an individual bank this is a temporary source of funds, which it considers on the same basis as Federal funds. The Federal funds rate represents the opportunity cost to the bank that influences the price at which it bids for the tax and loan balance. Such U.S. Government deposits are probably more appropriately considered as Federal funds in contrast to, say, normal seasonal deposit flows. Similarly, policy with respect to compensating balances may be changed by banks in the short run, but this is probably better considered as a factor in loan terms and conditions. and portfolio policies. On the demand side, banks have formulated portfolio policies and they are faced with demands for loans, reflect ing the underlying demand for goods and serv ices and given the costs to borrowers of var ious alternative methods of financing, including banks’ own loan rates and terms. Through in teraction of these supply and demand forces, a certain volume of credit will be extended by banks and a volume of deposits will be gener ated. A similar short-run process takes place re gardless of whether reserves are lagged. A bank’s willingness to extend loans or to com pete for time deposits, even under a lagged scheme, will be limited by its seasonal pattern of demand deposit flows and by the cost to it of obtaining reserves in the Federal funds mar ket, including particularly expected deposit flows and costs of Federal funds 2 weeks hence when reserve requirements on the current week’s deposits have to be met. It must be as sumed under existing procedures that the dis count window is not a permanent source of re serve supply and that it can provide funds to individual banks only for short and infrequent periods when their reserve calculations go as tray. While the general theory of operating with a money market conditions guide is the same when reserves are lagged as when they are not, there may be some difference in timing of bank response to System operations. For ex ample, if the System is tightening under an un lagged scheme, it is possible for the banking system to adjust to a smaller increase in non borrowed reserves by selling assets to the public and reducing required reserves in the current week. Under a lagged scheme, the banking system cannot reduce required reserves in the current week, but that does not mean that banks need necessarily avoid preparing for the tightening of conditions in the current week. Clearly, they may still sell assets to the public in the current week—thereby reducing deposits currently and required reserves 2 weeks from now. However that may be— and the charac teristic of bank reactions to changes in reserve availability within short-run periods is an area where further empirical research is much needed—in this paper it is assumed that bank deposits in the very short run, such as a state ment week, are not much affected in practice by System operations within that period, and that the operating option for the System is whether to supply the necessary required re serves through the discount window or by providing nonborrowed reserves. If money market conditions are kept un changed, the System through open market op erations will supply or absorb enough nonborrowed reserves—given the other factors affecting nonborrowed reserves—to keep the net reserve position of banks and member bank borrowings (the most volatile element in the net reserve position under current circum stances, with excess reserves generally at mini mal levels) at around their previous levels. And apart from reserve distribution problems, the Federal funds rate would generally also show little net change. Because projections of non-System factors affecting nonborrowed reserves are uncertain (and in the days before the lag, projections of required reserves too were uncertain), the be havior of the Federal funds rate in the course of a statement week helps provide a clue as to whether the staff projections of net borrowed reserves and factors affecting such reserves are correct. For instance, if staff projections show that net borrowed reserves early in the state ment week are deeper than those prevailing in earlier weeks (and thus would require System reserve-supplying operations under an un changed policy), while at the same time the Federal funds rate is opening lower than in previous weeks, the Manager might consider holding off on any reserve-supplying opera tions in the expectation that there were in fact more reserves available than the projection for net borrowed reserves indicated. This might then turn out to be the case when the next day’s figures became available because, say, float was running higher than was allowed for or than was normal for that particular time of the year. The interplay between statistical projections and the Federal funds rate is a val uable source of information to the Account Manager. If the FOMC voted to tighten money mar ket conditions, the Account Manager would conduct his operations so as to force banks to borrow more at the discount window than they had in earlier weeks, assuming excess reserves are at minimal levels. As banks find that they are forced more into the discount window, they also find fewer reserves available relative to demand in the Federal funds market (both being aspects of a reduced supply of nonbor rowed reserves by the System) and the Fed eral funds rate tends to rise. Banks will also begin to undertake portfolio adjustments, such as selling Treasury bills, particularly if they think the tighter conditions are likely to per sist; they will begin to alter offering rates on CD’s and Euro-dollars; and they will begin to change loan terms and conditions. These changes soon begin to show up in the rate of growth of bank deposits and credit. For exam ple, slower growth than otherwise in deposits may develop over a period of weeks as indi vidual banks begin selling securities to the nonbank public as part of the adaptation to tighter money market conditions. MONEY MARKET CONDITIONS IN RELATION TO OVER-ALL INTEREST RATES. While following a money market conditions target essentially has meant that the System would accommodate whatever market demands for money and deposits developed at a given Federal funds rate and bank net re serve position, this did not necessarily mean that the System could be construed as stabiliz ing interest rates other than the overnight money rate. Interest rates broadly conceived will probably tend to fluctuate less in the short run under an accommodative monetary policy than they might otherwise. But still there are likely to be rather wide swings, and also trend movements, in interest rates on obligations maturing in 2 or 3 months and longer as FOMC DIRECTIVE IN LATE 1960’s a result of shifts in credit demand or market expectations, if money market conditions re main unchanged. Experience in the latter half of 1969 is evidence in this respect, although the markedly slower rate of growth in the money supply that developed simultaneously would also be consistent with the hypothesis that an unwillingness on the part of the Sys tem to accommodate completely demands for money—however that unwillingness came about—was an important causative factor in the increase in interest rates. A number of factors can account for the over-all variability in interest rates under an unchanged money market conditions target. One, of course, is expectations. An increase in inflationary anticipations, for example, will increase the interest rate premium demanded by investors and will make borrowers more willing to pay it. Similarly, an abatement of inflationary expectations will have the reverse effect. Expectational effects on interest rates can also develop out of shifting attitudes with re spect to fiscal and monetary policies. Anticipa tions of a fiscal surplus, and of course the ac tual development of one, may lead to declines in interest rates on both short- and long-term Treasury securities as dealers become more willing to position securities currently in antici pation of a relative scarcity of securities later or in recognition of a shortage in the process of developing. Similarly, a pervasive attitude that the monetary authority may at some time in the fu ture begin to ease money market conditions is likely to bring interest rates down currently as investors attempt to acquire large amounts of high-yielding securities. In the bill market, such a phenomenon may be associated with declines in both 3- and 6-month bill rates, but often a relatively greater decline develops in the 6-month bill rate—reflecting the greater likelihood that short-term rates will be lower in the longer-term future than over the very near term. Expectations of a tightening in money market conditions will have the reverse effect. But if expectations of a shift in money market conditions prove unfounded, interest rates are likely to revert to previous levels. A more permanent effect on interest rates, however, can develop as money market condi tions remain unchanged over a sustained pe riod because of a cumulating tightness that de velops on banks. For instance, if member bank borrowings from the Federal Reserve remain at, say, around $1 billion for a number of months, many banks will have sought funds at the discount window a number of times. Given the attitude of the Federal Reserve that such borrowing should be only occasional and pri marily for unforeseen reserve adjustment con tingencies, the reluctance of banks to borrow will tend to increase with the number of times they have previously borrowed. Thus, as a given degree of pressure on bank reserve posi tions is sustained, banks will increasingly sell Treasury bills, reduce purchases of municipal securities, and make other adjustments that re duce the likelihood of their having to come to the discount window. These adjustments will add to upward pressures on interest rates. Such a process tends to be intensified in pe riods when Regulation Q ceilings are at unrealistically low levels and banks are forced to adjust portfolio policies and loan terms because of large losses of time deposits. Interest rates tend to rise under such circumstances partly because the banks appear to be more efficient investors than are the large number of individ uals and corporations. But in addition, it is likely that the structure of interest rates may be affected—with long-term interest rates rising relative to short-term rates—as those with drawing funds from banks, such as corpora tions, invest largely in short-term market in struments, while banks react not only by selling Treasury bills but also by reducing ac quisitions of long-term State and local govern ment securities and by stiffening lending terms, which may force some business borrowers into the open market, including the capital market, for funds. Finally, over-all interest rates may vary, given money market conditions, along with changes in basic credit demands, which may re flect changes in the trend of GNP. A weaken ing of demands for bank credit will reduce the need for banks to undertake liquidity and portfolio adjustments and will contribute to a lowering of market interest rates in general. Changes in demands on bond markets—pre dicted in part on, say, changing needs to fi nance business capital outlays—will also affect long-term interest rates while money market conditions remain unchanged. However, in these, as in other instances of changing credit demands, the extent of the change in interest rates will be influenced by expectations and will also be limited by the accommodative posture of the Federal Reserve—that is, by the ex tent to which the Federal Reserves does or does not permit money market conditions to change. In general, as credit demands weaken, the accommodative monetary policy at given money market conditions will be consistent with interest rate declines, but the extent of decline in the short run will be limited by Sys tem actions leading to unchanged, rather than to easing, day-to-day financing rates and member bank indebtedness at the Federal Reserve. Similarly, as credit demands strengthen, inter est rates generally will rise, but the degree of rise in the short run will be limited by System actions maintaining day-to-day financing costs at previous levels rather than letting them rise and making it more expensive for dealers to underwrite the securities that are issued and more expensive for individual banks to accom modate loan demands through marginal bor rowing in the Federal funds market. EVALUATION OF THE NEED FOR A MONEY MARKET CONDITIONS GUIDE. One of the chief advantages of operating with money market conditions as a guide would ap pear to be the automatic seasonal adjustment that is provided for bank reserves and money. For instance, the drain on bank reserves from outflows of currency to the public around the Thanksgiving and Christmas holiday periods and the greater transactions need for demand deposits are not permitted to tighten the money markets, since the System provides offsetting reserves to the banks through open market operations. The resulting increase in the money supply, as it recurs regularly, would be represented as no more than seasonal in the money supply statistics. In addition, other tem porary demands are provided for, even though they may not recur year after year and thereby qualify as seasonal demands. An example would be a one-time speed-up in corporate tax payments. The desirability of stabilizing money market conditions in order to provide an automatic short-run accommodation to banks’ changing demands for reserves may, of course, be open to question. One reason for operating in that way is that banks have not had automatic ac cess to the discount window. If there were such access, and assuming that the discount rate were continuously in touch with market rates, member bank borrowings—rather than nonborrowed reserves—might be permitted to fluctuate for seasonal reasons. But apart from that possibility, the theory behind the directive has appeared to imply the desirability of pro viding seasonal and other temporary accom modation to the market on the grounds that the market cannot be completely relied on to arbitrage out, through the interest rate mecha nism, the shifting seasonal demands for credit and money. It seems unlikely, for instance, that the market would fully anticipate tax-pe riod needs for credit at times of seasonal slack and thereby avoid severe crunches in credit markets at tax dates. Of course, one might argue that the market’s learning process is rapid and that it would not take more than one or two tax dates before the market did learn to borrow in advance, when short-term interest rates would be tending to be lower. While there is something to be said for ac commodating seasonal and temporary market demands in the System’s day-to-day opera tions, there are also dangers. The chief danger is that if economic activity is advancing faster FOMC DIRECTIVE IN LATE 1960’s than expected, there is likely to develop over the short run a larger expansion of bank credit and money than is desired for seasonal, tem porary, or longer-run growth reasons. On the other hand, if the economy is weakening, the System is likely to find itself in a position of absorbing more reserves over the short run than it may wish to when taking into account the sustainable growth needs of the economy. This condition might be corrected, of course, either by strict adherence to the pro viso (that is, by making it more of a target) or by adjusting the money market conditions tar get when the FOMC again meets. But in very weak or very strong economic situations, small adjustments in money market conditions— and experience shows that in the past the FOMC has moved in small steps with respect to money market conditions—may not be suffi cient to achieve over-all financial conditions consistent with desired economic activity. A focus on money market conditions, therefore, and a concern with stability of money market conditions tend to limit the System’s ability to control monetary aggregates and to effect the desired associated changes in over-all credit conditions and interest rates. While money market conditions have gener ally been considered to be an operating and merely instrumental target, they have been moved infrequently enough and slowly enough that, for all practical purposes, they as sumed the aspects of a goal of policy. The sta bility of the money market has clearly been a short-run goal, but often the desire not to have sharp shifts in money market conditions has appeared to be a longer-run goal, in that the System in the past has appeared reluctant to change money market conditions by more than small, gradual amounts. Such short-run and longer-run goals for the money market can often interfere with the attainment of the long er-run interest rate, bank credit, and money objectives of policy— all of which appear to be more closely related to economic activity than are money market conditions themselves. It is not without reason that the System pays such close attention to the money market and its operations. Many of the reasons have been discussed earlier in this paper. In particu lar, the use of such a target for enabling the System to provide for the seasonal and tempo rary reserve needs of the economy has been noted. In addition, at least the theoretical con sistency between money market conditions and longer-run policy goals will be sketched out in a subsequent discussion of how the System might attain credit conditions and monetary flows consistent with a desired GNP, while op erating day to day on money market condi tions, through an interlocking set of short- and long-run projections of financial and real flows. But perhaps the chief reason why the FOMC has focused on the money market in its operations has been the feeling that such a focus would lead to less interest rate fluctua tion and less danger of liquidity crises than would a focus on a monetary aggregate. The history of central banking, and particularly the genesis of the Federal Reserve System, has had as one of its main themes the need to have an institution that will be able to avert old-fashioned financial panics by providing a source of ultimate liquidity to the economy. Thus, the state of the central money market— where liquidity pressures focus—has histori cally been a main concern of the Federal Re serve. Perhaps partly explainable as an outgrowth of such a tradition, it would appear that the structure of the directive in the late 1960’s, not to mention earlier years, was con sistent with a belief by the FOMC that wide fluctuations in interest rates over the short run are more likely than short-run swings in the money supply or bank credit to cause destabil izing disturbances in the behavior of borrowers and lenders, who rely to a great extent on the interest rate structure as a source of informa tion about current and prospective credit and possibly economic conditions. The sharp rise in both short- and long-term interest rates over the latter half of 1969 cer tainly raised questions, however, as to how much stability in interest rates is produced by a focus on narrow money market conditions. Setting aside the question of whether one should stabilize interest rates at all in the short run, it might be pointed out that more stability could be introduced into the interest rate structure, if that were desirable, by encourag ing offsetting fluctuations in the Federal funds rate. That is, a tendency for bill or other inter est rates to rise could be offset by forcing the Federal funds rate down, and vice versa. This might be desirable, depending on economic prospects, but there is the danger that such a policy would simply increase the likelihood of providing reserves procyclically. For example, if people expected interest rates to rise, an ef fort by the System to lower the Federal funds rate and to provide more nonborrowed re serves in order to prevent such a rise would result in an even larger short-run rise in the money supply than would otherwise be the case. And this might over the longer run fore stall a rise in market interest rates if the greater expansion in money should lead to in flationary expectations. While changes in money market conditions to offset fluctuations in over-all interest rates are not desirable in a period when the econ omy is either strengthening undesirably or weakening undesirably, it may be desirable to permit money market conditions to move in such a way as to reinforce over-all interest rate movements. That is, the money market itself might be permitted to tighten as other interest rates rise, or to ease off as other interest rates decline. But if the money market is permitted to tighten sharply, there is a danger that the tightening might affect the solvency of dealers in securities who may have exposed positions and may rely on the money market for financ ing. Thus, an excessive tightening of the money market over the short run could lead to some failures of underwriters and to an associ ated weakening of confidence generally. While there is reason for the System to as sure a degree of stability in the money market, more fluctuation in money market conditions than has been permitted seems to have desira ble aspects. An emphasis on money market conditions apparently leads many market par ticipants to view a change in money market conditions as signaling a change in policy. If the money market were permitted to fluctuate more, this view might be eroded. To the extent that that happened, the System’s flexibility in attaining targets for interest rates more gener ally, reserves, or other monetary aggregates would be enhanced. A greater fluctuation in money market vari ables, once the market had become accus tomed to such fluctuation, would not appear in and of itself to affect credit conditions that af fect spending. As the Federal funds rate fluctuates up and down, banks are unlikely to change loan and investment policies, and deal ers in securities are unlikely to become signifi cantly more or less aggressive in bidding for a position in securities. But a clear trend in money market conditions toward either the tight or easy side would, as it has in the past, have an effect on over-all credit conditions. If the money market were permitted to fluc tuate more, this might make it possible for the System to carry out an open market policy with less short-run variability in the money supply, bank reserves, bank deposits, and possibly even interest rates generally. But whether it is better policy to minimize short-run variability in the money supply or short-run variability in money market conditions is a much debated question. If the System were to move to a monetary aggregate target for the short run, the effect on money markets would depend on how the value of the aggregate was chosen. The System could choose, for example, to expand bank credit in accommodation of Treasury financing demands in a current month just as it would under a fixed money market conditions target. If the staff projected that bank credit would expand at a 15 per cent annual rate in a month with fixed money market conditions, FOMC DIRECTIVE IN LATE 1960’s given the Treasury financing and past seasonals, and the Committee accepted the 15 per cent as a suitable target for the month, then it is likely that money market conditions, assum ing the staff is correct, would remain relatively stable within the month and would show little change from the previous month. In practice, however, if an aggregate were taken as a primary target, the money market would be likely to fluctuate more than in the past because the Manager would have to move rapidly to attain the aggregate target if the projections appeared to be wrong. But an ag gregate target over a 1-month period is not likely to be considered except as a part of a desired longer-term trend. And as it became clear to the market what the longer-term trend appeared to be, some of the short-run varia tion in money market conditions might tend to moderate as borrowers and lenders became more efficient in discounting the future. POSSIBLE RELATIONSHIP BETWEEN THE STRUCTURE OF THE DIRECTIVE AND A THEORY OF MONETARY POLICY FORMULATION The second paragraph of the directive is es sentially an instruction to the Manager on how to operate in the open market during the inter val between Committee meetings. In that sense the second paragraph need not be interpreted as representing monetary policy, if monetary policy as it influences financial markets is to be judged by such key variables as over-all credit conditions, interest rates, the availability of funds to the mortgage market, the money supply, and the liquidity positions of banks, other financial institutions, corporations, and individuals. All of these key financial variables can change while the operating phrases in the second paragraph of the directive remain un changed, at least as the directive was struc tured in the latter part of the 1960’s. It takes only a cursory reading of history to point out such periods, but the one that comes to mind most quickly is the period from the spring to the end of 1969, when there was a sharp tight ening in what almost anyone would call mone tary policy—whether judged by interest rates, money supply, or liquidity—without any ac companying change in the second paragraph of the directive. Since money market conditions themselves are not a key variable affecting spending, the theory of using money market conditions, with a proviso clause, as day-to-day operating vari ables in the directive can be explained by not ing one possibility of how the second para graph of the directive of the late 1960’s might relate to projections for key financial variables that affect the economy and to projections of economic activity itself. It should first be pointed out that the view of these interrelation ships to be presented here represents a theory that it is not clear that all, or even most, mem bers of the FOMC held, particularly as the theory pertains to the role of the proviso. Nev ertheless, it is a theory that is generally con sistent with the type of information presented by the staff to the FOMC, although as will be brought out in the concluding section of the paper, there are gaps between theory and practice. Some of these gaps may reflect the fact that the FOMC itself did not accept or did not follow the theory, and some may be because the detailed information and interrela tionships required by the theory simply were not ascertainable with a reasonably small mar gin of uncertainty, given the state of economic knowledge. FORMULATION OF LONGER-RUN PROJECTIONS. The staff ordinarily presents to the FOMC longer-term projections of devel opments in the economy, with certain assump tions as to monetary policy. These assumptions have been expressed in various ways at various times; for example, at times they have been expressed in terms of a particular bill rate, at other times in terms of a growth in bank credit, and at times in terms of growth in total reserves. Most frequently, perhaps, the policy assumptions are stated as a collection of finan cial flow and interest rate variables that are believed to be mutually consistent.5 Basic to the formulation and operations of monetary policy is a long-run forecast of how the economy is likely to develop over a period of, say, 1 year. For the purposes of this analy sis, the techniques of such forecasts—the alter natives and problems of which have been under intensive debate among economists for some time now—will not be discussed. Within the structure of a long-run forecast of economic activity—meaning GNP in both real and nomi nal terms—there would be contained a time path of economic activity. The units of time could be as small as one would like, but the state of economic data and the art of forecast ing suggest one quarter as a reasonably short division of time for projections of real eco nomic activity and associated financial flows. While the quarterly pattern of projections within the context of a longer-run projection may be satisfactory for policy formulation by the FOMC, it seems clear that even shorterrun projections, at least of certain key financial variables, are needed for the operations of pol icy in the open market in order to verify that policy is on the track of the longer-run projec tion, assuming that attainment of the latter projection represents a goal of policy. But before discussing the projections needed for day-to-day open market operations, it is necessary, first, to consider in a little detail the assumptions behind the longer-run forecast of real activity and financial flows, since this fore cast is presumed to provide the ultimate guide line for operations. One basis for a longer-run forecast would be an assumption of no change in over-all credit conditions as currently pre 5 The following few paragraphs on the formulation of longer-run projections and their relation to operat ing guides are based for the most part on a paper, “Notes on Monetary Policy Formulation and Opera tions,” written by the author for another occasion. vailing, or changes in credit conditions could be posited if required to lead to a desired GNP. One reason for using an assumption about credit conditions is that most of the links thus far found between financial condi tions and categories of spending appear to be from the credit side—interest rates and credit availability—rather than from the asset side— money supply and so forth. A forecast could also be constructed on the assumption of no change in the rate of money supply growth from, say, a growth of the previous several months on average. And, of course, assump tions about credit conditions imply a particular money supply growth, and vice versa. But for purposes of presenting a theory consistent with the directive of the late 1960’s, it will be as sumed that projections of GNP are based on credit market assumptions. An assumption of unchanged credit condi tions from those prevailing in the recent past might not be inconsistent with some fluctuation or movement of nominal interest rates, but it would not be consistent with such large varia tions as to change the willingness of borrowers to undertake credit-financed spending from what had been anticipated at the time of the forecast. Real economic activity also depends, of course, on past financial market conditions as they have come to influence spending in the quarters ahead. Finally, fiscal policy, wage and price pressures, and exogenous shocks to the system—such as technological changes, unfore seen defense emergencies, and sudden surges of consumer optimism or pessimism—all influ ence the forecast of economic activity. For the forecast level of GNP to be real ized, a certain pattern of financial flows would be required, given current and past credit con ditions. This pattern would reflect the credit demands of businesses, individuals, the U.S. Treasury, and State and local governments. The financing of these demands, given a level and structure of interest rates, would imply a distribution of financial assets held by consum ers and others that would in effect serve as a source of funds for the borrowers. Thus, the FOMC DIRECTIVE IN LATE 1960’s money supply, time deposits, savings and loan shares, and so forth fall out of the projection; and so does the need for aggregate bank re serves. If the pattern of real economic activity in the projection is satisfactory to the monetary authority, then in view of how the projections were made, there will be no need for monetary policy to be changed—in the sense that there is no need for open market operations to be directed toward achieving firmer or easier over-all credit conditions. But that does not mean that there would be no short-run varia tions in rates of growth in bank reserves and the money supply, given the lumpiness of var ious types of demands from both the U.S. Government and businesses, as required to be consistent with the longer-run financial and credit flows necessary to achieve the desired level of growth of economic activity. If some other pattern of change in real eco nomic activity were desired by the FOMC, a consistent projection of real economic activity and financial flows could, of course, also be worked out, with the effects of past monetary policies imposing a restraint on how soon a more desired economic goal might be achieved or on how large a wrench might be required in the financial system to attain it. ROLE OF MONEY MARKET CONDI TIONS AND PROVISO IN RELATION TO LONGER-RUN PROJECTIONS. A structure of interest rates and pattern of financial flows consistent with the credit and money demands generated by the desired level of economic ac tivity can be attained by using money market conditions as a day-to-day guide for open mar ket operations as described in the earlier sec tion of the paper, provided the relationship among money market conditions, financial and monetary flows, over-all credit conditions, and desired GNP can be reasonably well predicted. In this context, day-to-day open market opera tions conducted in terms of money market con ditions can be said to be free of the sin of money market myopia. But they can only be said to be so if there is no hesitancy in reset ting the money market conditions guide when it appears that over-all credit conditions are becoming tighter, or easier, than desired. What all this amounts to is that money market con ditions have little meaning for policy in and of themselves and that they acquire meaning only as they lead to changes in financial vari ables that affect spending. Needless to say, however, there can be many slippages between the specification of the set of money market conditions and the ensu ing financial developments that more directly affect GNP (as well as reflect GNP), just as there can be large miscalculations as to the basic state of aggregate demands in the econ omy or of the degree of fiscal stimulus and re straint. Because of these slippages and because money market conditions in themselves do not include variables that directly affect spending, it would appear that such conditions would have to be varied frequently as errors in speci fication between money market conditions and variables that affect spending become apparent or as errors in projections of aggregate demand become apparent. In practice, therefore, one would on theoretical grounds expect rather fre quent changes in both the directive and in projections. One way of hedging against the possibility that given money market conditions are lead ing to a policy that condones undesirable eco nomic developments is to make short-run fore casts for time units of less than one quarter—such as for the months within the quarter—for certain key banking and mone tary variables, such as total reserves, nonbor rowed reserves, money supply, and time de posits, that are immediately responsive to open market operations. In other words, money market condition targets can be set in the ex pectation that they will lead to a certain growth of bank credit, money, and reserves over a particular 1-month period, which repre sents an interval roughly reflective of the time between FOMC meetings. And the growth rate in such variables over that month—as well as the successive monthly projections —would be consistent with the quarterly growth rates that are implicit in the credit con ditions leading to the GNP forecast—provided all the elements were put together consistently, that is, with correct analysis of the relation ships between real economic activity and credit conditions (taking due account of the distinc tion between nominal and real interest rates in judging the appropriateness of credit condi tions), between credit conditions and the pub lic’s preferences for assets, and between finan cial flows this month and next month. That such relationships can be predicted with accu racy represents, of course, a very heroic as sumption, but this paper is discussing theory as much as reality. The proviso clause in the directives of the late 1960’s can be interpreted as using total member bank deposits subject to reserves— called the bank credit proxy—as a variable for testing the consistency between money market conditions and projected developments in the real economy. If the successive weekly and monthly observations of this variable were ris ing faster than projected, the assumption would be that GNP was stronger than ex pected. If this variable were weaker than pro jected, the assumption would be that GNP was weaker. On this theory that the proviso clause is the link between the day-to-day money market conditions target and the ultimate GNP goal, two principal criteria for the variable to be in cluded in the proviso clause could be reasona bly posited: one would be its responsiveness to GNP, and the other would be the ready availa bility of data on a daily basis so that they could be taken into account in the course of operations. Still another criterion might be the controllability of the variable through open market operations; but this criterion becomes more important to the degree that the proviso is considered more as a target to be attained rather than as an indicator of GNP trends. And the proviso may have certain target as pects because under particular conditions— such as inflation—the FOMC might wish to put more stress on attaining the specified ag gregates if it felt relatively more uncertain about appropriate credit conditions because of inability to evaluate the impact of inflationary expectations on interest rates. The ambiguities in the concept of the proviso—whether it is a target or an indicator of whether GNP and as sociated credit demands are behaving as ex pected—are discussed in somewhat more detail later in this section. Whether total member bank deposits meet the first criterion of being related to aggregate economic demands in a consistent manner is a testable proposition. On a priori grounds, one might think that the money supply would be a better variable in this respect, since the income elasticity of money probably dominates the in terest rate elasticity of money. Total member bank deposits, on the other hand, include a time deposit component that is highly elastic with respect to interest rates and probably less elastic with respect to income. In its short-run forecasts of total member bank deposits, the staff does attempt to esti mate the extent to which time and savings de posits, as well as demand deposits, will be af fected by the level of market interest rates expected to accompany a given level of money market conditions. Thus, an expected amount of so-called intermediation or disintermediation is included in the forecast. For purposes of the proviso clause, the assumption could then be made that if the projection of total member bank deposits is wrong, it is wrong not be cause of errors in forecasting intermediation or disintermediation, but because the assumption about aggregate demands is wrong. It is ob vious, however, that the staff may also miscal culate the income elasticity of total member bank deposits, even if its forecast of GNP is correct. The monthly projections of monetary aggre gates provided to the FOMC may be thought of as the link between day-to-day money mar ket conditions and real economic activity. This link depends on a degree of detailed knowl edge about the functioning of the economy and FOMC DIRECTIVE IN LATE 1960’s about interrelationships between real and finan cial variables and among financial variables that is barely attainable by the human mind, and is certainly not at hand at the moment. Thus, at best, the directive may be said to have been working with a very imperfect mechanism, but a mechanism—that is, a pro viso clause—which was probably better than no such mechanism at all, for it may give correct signals in periods when there are large devia tions in GNP as compared with projections. Before the problems of errors and uncer tainties implicit in such a theory and practice of the directive are discussed in somewhat more detail, the ambiguities in the role of the proviso clause in practice need to be brought out. Many apparently have considered that the proviso clause represented a target for policy, not an indicator of whether money conditions were set in such a way as to achieve a desired GNP. Those who have considered the proviso as a target, therefore, have been concerned about whether it measures bank credit prop erly, if that is taken as a goal of policy. It may have been concern with the target aspect of the proviso that led the FOMC to add to total member bank deposits the funds obtained abroad through Euro-dollars and obtained do mestically through nondeposit sources when specifying the ranges for the proviso. But if the proviso is taken purely in its indicator role— that is, its role as reflecting transactions or credit demands in the economy—it is not clear that it needs to be a comprehensive measure of bank credit. In this sense, the use of the term “bank credit proxy” may have led to consider ably more conceptual confusion than is necessary. The theoretical bases for considering the proviso clause as a target as compared with considering it as an indicator of whether the relationship between money market conditions and evolving GNP is about as expected would appear to be quite different. Taking it as a tar get, one would have to argue that the proviso clause should contain a flow variable readily controllable by the Federal Reserve and most likely to lead to desired GNP in the future, given the lags in monetary policy. Moreover, one would probably also have to argue that the proviso clause should be the principal op erating instruction. However, taking the pro viso clause as an indicator of GNP (not as an indicator of monetary policy in this context, it should be stressed), one might argue that it need only contain a flow variable that is highly income-sensitive and that is readily available. It is not immediately apparent that, insofar as monetary aggregates are concerned, a target variable and indicator-of-currentGNP variable need be one and the same, though this is an empirical question basically. But there does seem to be some uncertainty in the FOMC directive as to which type of vari able has been sought. ERRORS AND UNCERTAINTIES CON SIDERED. As the previous section has at tempted to make clear, there is considerable scope for error in the relationship between the operating targets in the second paragraph of the directive and the ultimate goal of policy— a satisfactory performance of the economy in terms of activity, prices, and the balance of payments. Errors in projections of GNP and in prices, since they are given in framing monthly and quarterly financial projections, can ob viously lead to errors in the directive variables given to the Manager. In addition, GNP might be correctly projected, but the staff might err in its evaluation of the relationship between current financial flows and the given GNP. Finally, there may simply be random variations, or noise, affecting monthly esti mates of monetary flows. One result of ran dom events or noise as a source of misestimation would be that, if money market conditions were given in the directive, bank credit might turn out to be stronger or weaker than pro jected, but still not be inconsistent with the de sired GNP. Nevertheless, the deviation of bank credit from projections might trigger the pro viso clause and set up a chain of events that would lead to an undesired GNP. The possibil ity of this sort of error is one of the reasons why the proviso clause was generally not trig gered except in cases of large deviations from projections, and that when triggered, it led to only very minor changes in money market conditions. There are potential sources of error that would affect operations, of course, regardless of whether the directive was couched in terms of some monetary aggregate rather than money market conditions, or whether the clauses in the directive were reversed—that is, with an aggregate in the principal clause and money market conditions in the proviso. But the sources of error might differ somewhat. With some sort of monetary aggregate target—such as the money supply—there would be some built-in protection against underevaluating the effect of inflationary expectations on nominal interest rates and thereby choosing a wrong in terest rate target when using a market condi tions guide. On the other hand, a money sup ply target might very well be set wrongly—say, too low—in relation to liquidity demands, with the result that credit conditions become too tight to achieve desired GNP. In general, linkages between financial varia bles and economic activity, as well as among financial variables, including money market conditions, are—despite two decades of empir ical research—still subject to considerable un certainty. As a result, any form of directive by the FOMC is likely to involve the risk of error and thus of poor policy after the fact, though presumably economic research will lead us to a point where it will be possible to specify op erating variables that at least minimize the po tential deleterious effect on the economy of mistakes in projecting relationships among economic and financial variables. Whether such operating variables would encompass monetary aggregates, interest rates, or some combination of the two is not within the pur view of this paper. The potential sources of errors are the result of uncertainties as to linkages between and among financial and economic variables, as well as the unpredictability of exogenous shocks to the economy, such as wars, techno logical breakthroughs, and erratic changes in consumer buying sentiment. There is uncer tainty as to which financial variables affect economic activity—for example, it is not clear whether or what type of rationing occurs in the economy when there is a shortage of credit relative to demand, or whether the balancing of demand and supply is accomplished com pletely through interest rates. It is not clear what the lags are between changes in financial variables and changes in economic activity. And it is not clear how strong a change in fi nancial variables is required to obtain a given effect on economic activity—that is, whether the money supply should rise or fall 2 or 4 per cent, or whether interest rates should fall or rise 2 or 4 percentage points. In addition to these uncertainties as to link ages, there are uncertainties as to how much variability should be permitted in key financial variables over the short run. One of the prem ises underlying the form of the FOMC direc tive in the latter part of the 1960’s was that it is better to keep money market conditions stable over the short run, while permitting more short-run variability in such items as the money supply, longer-term interest rates, and even Treasury bill rates. A decision to stabilize money market condi tions would appear to assume that this will lead to fewer mistakes with respect to other fi nancial variables that more directly affect the desired volume of economic activity than would a decision to stabilize a longer-run in terest rate or a money supply variable itself. In other words, the directive of the late 1960’s seemed to assume that the greater variability in member bank borrowings and the Federal funds rate that might result from specification of a money supply or total reserve target would be more harmful to the economy— given the prevailing state of uncertainty as to what should be the level, rate of change, and value of key financial variables—than would stability of money market conditions. The rea son would have to be that a money market FOMC DIRECTIVE IN LATE 1960’s conditions target gives maximum scope for per mitting market demands to determine financial flows and for permitting expectations to deter mine movements in interest rates away from the basic relationship to the Federal funds rate. These may be determinations that the FOMC felt it could not make directly, at least in the short run, because in the current state of knowledge it could not know the linkages; or because it believed that the demand for money is inherently unstable; or because, out of concern with the potential for liquidity crises, it placed higher value on money market stability in the short run than on predeter mined levels or rates of change in other varia bles. Perhaps at the risk of reading more into the framing of the directive than was in the minds of the framers, it would appear that uncertain ties as to linkages between financial variables and economic activity, and uncertainties as to the ability to determine the short-run demand for money and bank reserves, were important factors behind the choice of money market conditions as the principal operating target. In addition, it is likely that money market condi tions can be thought of as bearing a closer and more predictable relationship to over-all credit conditions and liquidity positions of banks and other key lending institutions. This may be a reason why those who adhere to a view that credit conditions—rather than changes in the public’s holdings of financial assets, particu larly money—determine spending may feel more comfortable with the money market con ditions target. But for those who hold such a theory, it is difficult to understand why it would not be better to specify some particular interest rate, constellation of interest rates, or desired reduction or enhancement of liquidity for banks, as a target instead. However that may be, the uncertainties faced by the policy-makers, together with the need to provide the Manager with an attaina ble target, provided them with a reason for ad hering to money market conditions as a shortrun operating guide for the System Account Manager, while at the same time keeping an eye on other financial variables that bear more direct relations to spending and to GNP in the formulation of policy. MONEY MARKET CONDITIONS: POL ICY TARGET ASPECTS. While the staff s presentations and projections of GNP and financial variables in both the short run and the long run do give members of the FOMC an idea of what is likely to happen to key variables under given money market condi tions, there is still the danger that a directive couched primarily in terms of money market conditions will lead to unexpected and unde sired changes in variables that are more di rectly reflective of the impact of monetary pol icy on GNP. This can happen not only because of errors in staff projections but also because money market conditions themselves can come to be taken as an objective of pol icy. Money market conditions can become an objective of policy partly because the need for a stable money market in the short run is over stressed. But it can happen in part because a continued stable money market comes to be viewed by the market as an objective of pol icy. When this occurs, the System often tends to get locked in, because it feels that any change in money market conditions will be in terpreted as a change in policy and, therefore, lead to overreactions by market participants and others. This is particularly true in periods, such as 1969, when abatement of inflationary psychology appeared to be the ultimate aim of monetary policy. With that aim, there seemed to be the fear that any change in money mar ket conditions would be interpreted itself as signaling a change in policy and thus would fuel inflationary psychology. Whatever the relation in particular periods among money market conditions, over-all credit conditions, and the money supply, it does seem clear that concentration on money market conditions in the operating paragraph of the directive has led both the Committee and the market at times to interpret these con ditions as policy itself. If an operating directive or so large an easing (or tightening) of credit were phrased in terms of some monetary ag conditions as might be necessary to achieve its gregate, or even in terms of over-all credit economic goals. Thus, there may have been a conditions, the Manager might have more dif conflict between the attitude toward money ficulty in operating but there would tend to be market conditions and what is necessary to less confusion between operating variables and achieve changes in financial variables that the financial conditions that are the goals of more directly affect changes in the public’s policy. Such a directive might also lead to spending propensities. more fluctuation in money market conditions —but that would come to be considered nor mal. However, it is difficult to predict how money market conditions would react over the longer run to such a recasting of the directive, RECAPITULATION AND since the market itself might find ways of sta CONCLUDING REMARKS bilizing itself as borrowers and lenders come to discount the future more accurately. This paper has attempted to indicate how In sum, the second paragraph of the direc the construction of the second paragraph of tive would appear to have had only a tenuous the FOMC directive, as it was in the late relationship to monetary policy as most econo 1960’s, related to the flow of money and depos mists perceive such policy. That relationship its and of interest rates broadly conceived in has depended on staff projections of the rela the practice of open market operations. It also tionship between money market conditions and attempted to present one theory—though ad other financial variables. These projections are mittedly one that might not be generally held generally made known in summary form to the or acted upon by the FOMC—as to how the public when the policy records are released money market conditions operating guide in after a 3-month lag. Unless money market the second paragraph, in conjunction with the conditions themselves change, many in the proviso clause, could be fitted into a nexus of market do not consider that monetary policy financial and nonfinancial projections of the has changed, and it is not completely clear economy and related to financial variables that that this view has not also been held by many more directly affect spending decisions. It was members of the FOMC. not the task of the paper to determine if an The focus on money market conditions has other theory—for example, one that put more in practice tended to prevent the Committee stress on monetary aggregates both in opera from adjusting these conditions rapidly. tions and in their role in economic forecasting Changes in money market conditions, when —would improve the functioning and posture they have been undertaken, have been under of monetary policy. But the paper has pointed taken gradually. Another reason for gradual out the great uncertainties present in the eco changes, apart from concern with the money nomic and financial relationships that would market as such, has been the uncertainty of the have to be projected both over short and over System as to effects of its actions or as to their longer periods of time to satisfy the theoretical desirability. This resulted in a directive that basis presented here for the FOMC directive specified attainment of slight, modest, or mod of the late 1960’s. Uncertainties, though per erate changes in money market conditions. But haps of not exactly the same sort, would also because of this unwillingness to move money plague other conceivable forms of a directive. market conditions rapidly at times, the System While the general problem would appear to may also have been put in the posture of not J be one of finding a form for the directive that being able to encourage so rapid an accelera would minimize the potential for errors in pol tion (or decelaration) in money supply growth icy, it does not appear that the directive of the FOMC DIRECTIVE IN LATE 1960’s late 1960’s, even on its own terms, quite lived tions will, perforce, be appropriate. Such a up to the theory that has been constructed for theory does not imply that monetary policy it here. There were, in other words, gaps be- y ' would stabilize either interest rates broadly tween theory and practice. Some of these gaps conceived or a rate of change in some monemay have occurred because the theory re tary aggregate. It does imply, however, that quired more knowledge or explanation than over the short run money demands would be was, or conceivably could have been, pro accommodated at any given Federal funds duced; some, because the FOMC simply oper rate, and to that extent policy operations ated on another theory or theories; and some would tend to moderate fluctuations in other because money market conditions in practice interest rates, although such rates would still took on aspects of a target role instead of be affected by changes in expectations and shifts in credit demand. playing only an instrumental role in policy. (3 i Under such a theory, economic and as The following points recapitulate the high sociated financial projections are required for lights of the paper and offer some conclusions: 1. Neither the first nor the second para several quarters ahead, as are short-run projec tions—for, say, a month—of key monetary graph of the FOMC’s directive to the Ac flows, such as bank credit and the money sup count Manager, nor the relation between the ply. The short-run projections can be used to two paragraphs, has been completely under indicate whether the money market conditions standable when the directive is considered by fixed for the interval between FOMC meetings itself, or perhaps even when it is taken in con are leading to the flows of bank credit and junction with the simultaneously published money that were projected over the longer run policy record. It can be best understood as an to be consistent with desired GNP, given credit aspect of the whole procedure at FOMC meet conditions, and interest rates. To the degree ings, including the economic information and that the short-run flows are showing changes projections presented and the discussion of greater or less than projected, the presumption policy by the members of the FOMC as ulti is that GNP, or aggregate demand, is stronger mately revealed in the minutes published for the or weaker than projected. In this view, the meeting. Within this context a theory for the proviso clause in the second paragraph serves directive might be constructed, particularly a as an indicator of aggregate demand, which theory that relates the operating instructions would suggest that the variable included in the of the second paragraph to the economic fore clause should be one that is dominated more cast and objectives that are noted, however by income elasticity than by interest rate elas vaguely, in the first paragraph. ticity. This may be an argument for using C£*)One theory for using money market money supply rather than bank credit, al conditions—essentially the net free or net bor though the staff projection of bank credit rowed reserve position of member banks and would have already allowed for the interest the Federal funds rate—as a day-to-day oper elasticity of bank deposits, particularly time ating guide for the Account Manager would be deposits. that such conditions bear a predictable relation 4. There are many gaps between theory to over-all credit conditions, that over-all and practice. The most obvious is that even if credit conditions (including interest rate struc such a theory provided a proper basis for pol ture, bank liquidity, and so forth) can be set icy, the requisite economic knowledge of inter so as to influence economic activity in a de relationships among financial variables and be sired direction or toward a desired level, and tween financial variables and real economic that the flow of bank reserves, bank credit, activity might not exist to permit the attach and money expected to result from the money ment of a high degree of probability to the nec market conditions and desired credit condi- essary projections. It is probably recognition of the uncertainties about the state of economic knowledge—not to mention the sharp and unexplained swings noticeable in daily and week-to-week deposit and reserve data—that led the FOMC to require implementation of the proviso clause for the most part only when deviations from projections were “significant” and that led to only very minor variations in money market conditions when the proviso clause was implemented. 5. Another gap between theory, at least as presented here, and practice is that the proviso may have been considered to serve partially as a target for monetary policy rather than as an indicator of the relationship between money market conditions and GNP. Viewing the pro viso clause as a target may help to explain why it focused on bank credit (which those who start from credit conditions may believe to be a reasonable short-run target related to spending)—without here discussing whether the change in bank credit was really a desirable flow target (as compared with other possibilities such as the change in aggregate reserves or money supply). But if the proviso were taken as a target, it does not appear to have been a very high-priority one, since experience shows that money market conditions were not varied rapidly enough or to the extent necessary to keep bank credit within proviso limitations when it tended to move significantly outside those limitations. 6. The desire of the FOMC to minimize short-term variability in money market condi tions, as well as the relatively small changes in such conditions that were undertaken when the money market target was shifted, suggests that in themselves money market conditions were to some degree a target of policy, rather than being merely instrumental variables through which the interest rate and financial flow, and ultimately economic, objectives of policy are attained. While the relation between free re serves and over-all credit conditions might be predictable— at least judgmentally if not econ- ometrically—experience, especially in 1969, appears to indicate that the relation is not con sistent—that is, with fixed free reserves, credit conditions can and will change. Thus, mini mizing fluctuations in money market condi tions and changing such conditions only grad ually over the longer run would represent yet another gap between theory as presented here and practice— unless, of course, the FOMC willingly accepts the changes in over-all credit conditions (not to mention inflows of mone tary aggregates) that accompany an un changed, or only gradually changing, level of free reserves. 7. As a short-run target, money market conditions have the advantage of permitting the market to make decisions about the appropri ate short-run flows of bank credit and money. But as is well known, so accommodative a monetary policy might lead the System also to provide larger or smaller amounts of reserves, credit, and money than are consistent with de sired economic objectives if credit demands turn out to be stronger or weaker than pro jected, or, expressed in another way, if bank’s demand for free reserves turn out to be weaker or stronger than expected. The proviso, of course, has represented something of a hedge against such undesired short-run developments. But if the economy weakens or strengthens considerably more than expected, the proviso is a weak hedge unless the FOMC is willing, either when it meets or in the interval between meetings, to move money market conditions, or permit them to be moved, rapidly enough to offset the changing impact on reserves of demand forces. For example, both the money supply and interest rates may be declining be cause demand is weakening; to turn the econ omy around under such circumstances may re quire a sharp easing of money market conditions (for example, a sharp short-run de cline in member bank borrowings, assuming they are already high) if the Federal Reserve is to do more than merely permit a built-in flexibility of over-all interest rates to brake the FOMC DIRECTIVE IN LATE 1960’s decline in economic activity and is to encour age an expansion of economic activity, credit, and money. 8. The target aspect of money market conditions inhibits the flexibility of monetary policy when these aspects become so ingrained in market thinking that the System is reluc tant to move for fear that any move will be overinterpreted. When combatting inflationary psychology is taken as a primary goal of pol icy, for instance, it becomes difficult to permit an easing in money market conditions because this might be taken as signaling an unwilling ness of the System to persist in its efforts to reduce inflationary expectations. 9. The short-run stability of money mar ket conditions and the gradualness of any longer-run change in money market conditions tempt one to the conclusion—be theory what it may—that a basic reason for couching the second paragraph in such terms was prag matic. Given uncertainties as to the proper levels or changes in money supply, bank credit, or interest rates, money market conditions represented objectives that were readily attain able, that kept the scope for the Account Man ager’s judgment within reasonable bounds, and to which the Account Manager could be readily held accountable* 10. Within the context of the theory pre sented in this paper, or almost any other theory of how monetary policy works, it would appear desirable to permit more short-run vari ability in money market conditions and to move such conditions more rapidly or frequently over the longer run in carrying out open market op erations. This would, at a minimum, reduce the market’s focus on these conditions and thus in crease the flexibility of the FOMC in attaining targets for reserve or monetary aggregates, if it so wished, or even in attaining credit condition objectives in relation to shifting demands and GNP by reinforcing, or offsetting, a tightening or an easing of trends in credit terms and con ditions when it appeared desirable to do so. 11. There are probably some limits to the flexibility that could be permitted in money market conditions, although the degree of limi tation is both a conjectural proposition and an empirical question on which precious little evi dence is available. Such limitations would ap pear to apply more to the tightening side than to the easing side. When interest rates are ris ing, a considerable tightening of the money market might have undesirable repercussions on such sensitive market participants as secu rities dealers, who might be faced with the pros pect of failures if carrying costs rose sharply relative to the return on their pre-existing se curity holdings, and might thereby lead to fi nancial crises that would affect confidence gen erally. Even on the easing side, a sharp easing of money market conditions could lead to an overly large build-up in speculative positions in securities, which might force the System to provide more reserves and money than it would otherwise want to, or be faced with considerable market confusion and churning if the market were forced to liquidate these posi tions over the short run. But in the absence of much recent experi ence with a monetary system in which rela tively wide fluctuations in money market conditions were permitted,0 it is obviously hard to tell how the market would react in such a different environment. If human nature is any guide, there will be periods of market prob lems, including undue speculation, no matter what the system by which monetary policy works. Some concern with money market con ditions might reduce this problem, but the con tribution of money market stability might not be commensurate with the key role of such conditions in the directives of the late 1960’s and of earlier years, and with the System’s ap parent unwillingness to change such conditions except by small degrees. 12. Perhaps the chief general conclusion to be drawn from the rather lengthy analysis 6 Very wide fluctuations may not develop if the market discounts the future properly. of this paper is that a workable theory for the directive of the late 1960’s might be con structed, but that in practice there are gaps— some of which may have been unavoidable given the state of economic knowledge—be tween theory and practice. Among the require ments for bridging theory and practice are a more certain knowledge of relations among fi nancial variables and between these and eco nomic activity, a sophisticated and consistent meshing of short- and long-run projections, and a willingness to reduce the target aspects of money market conditions and regard them solely as instrumental variables capable of short-run fluctuation and more rapid trend movement. APPENDIX: An Empirical View of “Even Keel” whole are highly sensitive to the reception of Treasury financings because o f the sheer size of offerings, the involvem ent of the U .S. Govern ment’s credit, and the key role o f the Govern ment securities market in liquidity and portfolio adjustments o f investors. Even keel should be sharply distinguished from the old pre-1951 policy o f pegging interest rates on U.S. Government securities. The even-keel policy does not provide any assurance that partic ular interest rates on new or outstanding Treas ury issues will be maintained. Rather, the evenkeel approach only helps to sm ooth the process of marketing several billion dollars o f Treasury issues (even more in the case o f advance or pre refundings). It provides those who help under write Treasury issues (such as banks and non N o t e — The views expressed in this appendix do bank U .S. Governm ent securities dealers) with a not necessarily represent those of the Federal Re short period of time in w hich market forces serve System. Parts of this paper are drawn from a rather than new m onetary policy decisions are previously unpublished paper on the subject prepared the main factors affecting interest rates. Those by the author and Joseph E. Bums. 1 Discussion of the even-keel policy has usually who make markets in U .S. G overnm ent securities been focused on its relation to tightening actions. But are by no means assured o f stable interest rates in practice the policy also influences the timing of on the new issues, but they do have som e time to easing actions. For instance, a discount rate reduc contact customers with n o more than a normal tion in the middle of a Treasury financing period market risk on their temporary holdings o f secu may be avoided because it might encourage undue speculative activity. rities. The words “even keel” refer to the policy pur sued by the Federal Reserve in relation to Treas ury financings. In practical terms even keel has meant that, for a period encompassing the an nouncement and settlement dates of a large new security offering or refunding by the Treasury, the Federal Reserve has not made new monetary policy decisions (as contained in announcements from the Board of Governors or as specified in the second paragraph of the policy directives of the Federal Open Market Committee) that would impede the orderly marketing of Treasury securi ties and significantly increase risks o f market dis ruption from sharp changes in market attitudes in the course o f a financing.1 Financial markets as a FOMC DIRECTIVE IN LATE 1960’s Because o f the relatively limited nature o f the Federal R eserve’s even-keel commitment, the def inition of the comm itm ent in terms o f financial variables is to a degree equivocal. The timing of even keel, the behavior o f interest rates and other monetary variables, and the extent o f Federal R e serve open market operations depend in large part on the type o f market and market psychol ogy that develops in anticipation or in the wake o f the Treasury financing involved. The purpose o f this paper is to review the behavior of key fi nancial variables during the 3 years 1966-68 in an effort to determine how much variation or stability they show during even-keel periods in comparison w ith other periods. This empirical approach is designed to shed some light on the variations in financial variables that have been tolerated under the constraint of even keel. But the results are necessarily limited by inability to quantify market attitudes, changes in which w ill influence the tolerance with which the market views differing degrees o f variations in interest rates, reserves, and related measures. The results are also limited in part by the “crude” nature o f the empirical analysis o f the paper, w hich consists o f charting time series for the relevant variables and o f scanning these series for differences in behavior. W hile such an ap proach has obvious limitations, its advantage is that even-keel periods can be easily viewed in re lation to longer-term trends and turning points in such trends. M oreover, fluctuations o f a variable within an even-keel period are also discernible. And questions as to the exact dating o f even-keel periods can be m inimized since the charts would indicate the direction o f change if 1 or 2 weeks were added to, or subtracted from, the beginning or end o f even-keel periods. the t im in g of even keel and O F T R E A SU R Y ISSU E. The policy directives o f the Federal Open Market Committee provide a basis for dating even-keel periods and for relating them to the type o f Treasury security offering. Such directives during even-keel periods would refer to Treasury financings as a factor to be taken into account in the conduct o f open market operations. Generally the directive would also stipulate that operations should be directed to maintenance o f prevailing m oney market con ditions. But it is also possible that the operations could be directed toward tightening or easing. type This could occur, for example, if the directive were written for a policy period that begins fairly well in advance of the anticipated Treasury financing announcement, thereby permitting some adjustment in policy prior to the financing period. Or this could occur to permit some shading to ward restraint or ease depending on the develop ing market attitude toward the financing, includ ing the speed with which the financing is distributed in the secondary market and the ex tent to which the market is tending to discount potential Federal Reserve action in advance. The time span of, and money market stability during, even keel has varied in the past with the nature o f the Treasury financing, with the market environment, and with the urgency behind the need for a monetary policy change. For purposes of this study, the interval from a week before the announcement of terms to a week after settle ment date has been taken as the basic unit of time for an even-keel period, but shortened when necessary to be consistent with the dating of FOMC directives referring to Treasury financ ings. The various relevant dates that bear on even keel are shown in Table 1. In practice, even keel might extend somewhat beyond 1 week after settlement date if an espe cially large volume o f new securities were left over hanging the market, whereas if the new offering were small or well distributed even keel might end at settlement date. And the period might not begin until 2 or 3 days before announcement date, depending on market conditions as they af fect the Treasury’s ability to appraise pricing of the new issues. On balance, the basic unit o f time for even keel in this study probably tends to err on the generous side. Even keel has been applied quite consistently to coupon issue financings, which are generally large in size. A period o f 2 to 3 weeks normally elapses between announcement o f the offering and payment. The Treasury sets the price and coupon rate when the offering is announced; a few days later books are open and the public places its orders; and a week and a half or more passes before payment or settlement date on the new issue. In contrast to offerings of coupon issues, the even-keel constraint has not been regularly a fea ture o f FOMC directives around Treasury bill financing periods. When it has been, the period ings varied between $3 Vi billion and %AVi billion in size. There are a number o f reasons for keeping the even-keel period short in relation to bill financ ings and for applying it less rigorously, if at all. First, the bill is auctioned, so there is less need to hold markets stable between announce ment date and auction date; in a coupon financ ing, on the other hand, the new issue is priced by the Treasury at announcement in the expectation that market attitudes will not shift significantly in the interval (typically 5 days in recent financ ings) until the books are open. Second, the risk of price fluctuation to holders o f bills, which ma- has generally been shorter than for coupon issues, although it has also overlapped a coupon issue period and thereby lengthened the time when even keel has been applied in consecutive weeks. Even keel has been noted in directives at times when bill issues for cash have been large and/or when short-term markets have been likely to be under particular strain. During the 3 years 1966-68, there were three instances in which the even-keel constraint was noted in the directive in relation to Treasury bill financings raising net new cash, out of 12 such financings in the period (other than simply additions to the regular weekly or monthly bill actions). The three financ TABLE 1: Treasury Financings During Even-Keel Periods D escription o f offering D a tes related to even keel B ooks opened Settlem ent date A ttrition or a llo tm e n t ratio D irective date A n nouncem ent date 1 2 /1 4 /6 5 1/11 1 /5 1 /1 0 1/19 Cash 1/11 2 /8 1 /2 6 1 / 3 1 - 2 /2 2 /1 5 1 3 .7 4 /1 2 5 /1 0 4 /2 7 51 2 -5 /A 5 /1 5 Rights (incl. pre refunding) Rights 2 .5 18 m o 7 /2 6 7 /2 7 8 /1 -8 /3 8 /1 5 8 .1 1 yr 4 yr 9 m o 1 0 /4 11/1 1 0 /5 10/11 1 0 /1 8 Rights (incl. pre refunding) TA 3 .5 185 d a y 247 day 11/1 1 0 /2 7 H /1 1 1 /1 5 Cash 3 .2 1 yr 3 m o 5 yr . 30(A L ) . 10(A L ) 1 /1 0 2 /7 1 /2 5 1 /3 0 2 /1 5 Cash 3 .9 1 yr 3 m o 5 yr . 10(A L ) .07CAL) 5 /2 4 /2 6 5 /1 -5 /3 5 /1 5 9 .1 1 yr 3 m o 5 yr .1 9 (AT) 4 .0 Type A m o u n t1 (billions o f dollars) M aturity 1966 1 .5 . 14(A L ) 10 m o 18 m o 4 yr 9 m o 1 7 . 4 / . 17(A T) ,4 6 (A T ) * 4 .3 /.2 0 ( A T ) 1967 6 /2 0 6 /2 8 7 /5 7 /1 1 R ights (incl. pre refunding) TA 7 /1 8 7 /2 6 7 /3 1 8 /1 5 Cash 3 .8 255 day 286 day 1 yr 3 m o .3 5 (A L ) 8 /1 5 8 /1 7 8 /2 2 8 /3 0 Cash 2 .6 3 yr 5 . 38(A L ) 10 /3 9 /2 2 1 0 /3 1 0 /9 TA 4 .4 1 0 /2 4 1 0 /2 5 1 0 /3 0 1 1 /1 5 Cash 4 .8 196 day 259 day 1 yr 3 m o 7yr 2 /6 1/31 2 /8 5/1 5/1 7/3 1 2 /5 2 /1 3 5 /6 5 /8 8 /5 2 /1 5 2/2 1 5 /1 5 5 /1 5 8 /1 5 R ights Cash R ights Cash Cash 1 2 .0 4 .1 3 .9 3 .2 5 .1 7 yr 15 m o 7 yr 15 m o 6 yr .2 8 (A T ) .3 9 (A L ) .3 0 (A T ) .2 8 (A L ) . 18(A L) 1 0 /2 3 1 0 /2 8 1 1 /1 5 Rights 5 .5 18 m o 6 yr ,3 3 (A T ) rno _ .3 6 (A L ) 1968 4 /3 0 7 /1 6 1 0 /8 1 0 /2 9 1 O ffered to the public. 2 A m o u n t exchanged in pre-refundings in b illio n s o f dollars. A L allo tm e n t ratio; A T attrition ratio; T A ta x anticipation bill. FOMC DIRECTIVE IN LATE 1960’s ture in a year or less, is smaller than to holders o f intermediate-term or long-term coupon issues. And third, the time span between auction and payment for bills is generally about 1 week, while for coupon issues 10 to 14 days usually elapse be tween subscription and payment dates; this is a technical matter, but presumably it reflects the shorter period normally required to distribute a new bill issue as compared with a longer-term obligation. E V E N KEEL A N D IN TER EST RATES. Interest rates have shown a relatively large amount o f movement during even-keel periods. M ovements o f interest rates are shown in Chart 1, with even-keel time spans represented by the shaded areas. It is not without interest that the even-keel periods defined as noted above take up roughly 40 per cent of the 36 months shown. Norm al quarterly refundings themselves would lead to even keel for about one-quarter o f the year, with the actual result being a little more or a little less depending on market conditions and also the requirements of monetary policy. W hen the Treasury raises cash, or undertakes advance refundings outside the regular quarterly refunding period, monetary policy is affected at rather more frequent intervals. Day-to-day m oney rates. Short- and long-term interest rates show different patterns of move ment during even-keel periods and also differ in relation to their behavior outside such periods. Day-to-day rates, such as the Federal funds and dealer loan rates, sometimes fluctuate rather sharply within an even-keel period, just as they do in other periods. For instance, the Federal funds rate fluctuates in response to week-to-week shifts in the distribution o f reserves between country and city banks. However, these rates gen erally do not show either an upward or down ward trend in even-keel periods. Trend move ments in such rates— that is, a clear upward or downward tendency persisting for some weeks— generally occur in the periods between even keel. W hile an absence of trend movements in dayto-day m oney rates is a characteristic of evenkeel periods, there have been a few exceptions during the period under review. In even-keel peri ods during the winter and spring of 1966, direc tives sought some reduction in reserve availabil ity, while taking into account forthcoming or current Treasury financings. These directives cov ered the mid-February and mid-M ay refundings. Federal funds and dealer loan rates did not in any event show a rising trend in the first o f these FIGURE 1 INTEREST RATES in c i. t . a . in c i. t . a . Per cent t.a . 6.5 5.5 4.5 3.5 6.5 5.5 4.5 MAR. JUNE 1966 SEPT. DEC. MAR. JUNE SEPT. 1967 Shaded areas indicate periods of even keel. T.A. tax anticipation bill. DEC. MAR. JUNE 1968 SEPT. DEC. periods, but in the even-keel period covering from about the third week in April to the third week in May, an upward trend in Federal funds and dealer loan rates was in practice permitted to develop. Because the A pril-M ay period illustrates a modest tightening o f policy during even keel, it is worthwhile to note the results o f the financing and market factors bearing on it. The financing involved was a $2.5 billion rights exchange (in terms o f public holdings) involving an offering of a single 18-month note. The attrition rate for this offering was very large— 46 per cent— the highest attrition rate by far in the period covered. Of course, A pril-M ay 1966 was a period of sharply rising loan demands in credit markets, so the unfavorable reception might be partly attributed to cash needs o f commercial banks and other holders o f the maturing issue. In addition, the market was disappointed at that time by a fading of hopes for a program for fiscal restraint. Fi nally, the offering was priced to have a yield ad vantage o f 10 to 12 basis points over the out standing market, which represents only a normal yield spread between new offerings and outstand ing issues of a comparable maturity. All in all, there appear to be a variety of market factors ac counting for the poor reception o f the issue, but tightening of monetary policy, as expressed by money market conditions, and expectations of further tightening certainly contributed. Bill rates. Treasury bill rates, as indicated by the yield on the 3-month bill, tend to display roughly the same kind of behavior— both in terms o f fluctuation and trend— during an evenkeel period as is characteristic of the span of sur rounding weeks and months. In 1965, a year not shown on the chart, bill rates— not to mention other rates— showed little movement in or outside even-keel periods. In the 1966-68 period, how ever, bill rates moved relatively widely both in and outside even-keel periods. As examples of cyclical-trend movements in bill rates in even-keel periods during the 1966-68 period, there were upward movements in the rate during the late July-late August 1966 period and in the M ay 1968 period; there were downward m ovem ents in the late January-late February 1967 period and in the late A pril-M ay 1967 pe riod. It is likely that the more evident trend movem ent in the 3-month bill rate, as compared with day-to-day m oney rates, in even-keel periods reflects the role o f expectations in determining in terest rates. With a 3-month horizon, investors in 3-month bills are more likely to be influenced by what monetary policy— and also other factors such as debt management and business credit de mands— may be expected to do in the period ahead. Consequently, even-keel policies w ould be come correspondingly less important in influenc ing these interest rates during the weeks in which even keel is in effect. Longer-term rates. Longer-term rates, as typi fied by the yields on 3- to 5-year Governm ent securities and on such securities maturing in over 10 years, would also tend to be less influenced than day-to-day m oney rates by current monetary policy, and longer-term rates do show trend movements both in and outside even-keel periods. They have both risen and fallen in even-keel pe riods, the direction being generally consistent with the over-all tendency o f surrounding peri ods. Rate movements appear to have generally been larger in magnitude outside even-keel pe riods, but this is by no means always the case. For instance, there was a very sharp rise in the yield on intermediate-term Governm ents in the mid-July-late August period o f 1966. T his was a relatively large refunding, including a pre-refund ing, that zeroed in on the intermediate-term cou pon area. Moreover, the financing took place in a period when financial market pressures were building to a peak; and certain tightening m one tary policy measures, including increases in re serve requirements announced in late June and mid-August, were put into effect quite close to the refunding period. With respect to open mar ket operations, the FOMC directive on July 26 indicated an even-keel stance and no change in money market conditions. While even keel was technically in effect in this financing, the sharp rise of interest rates in the maturity area containing one o f the new is sues offered in the refunding reflects the general expectation o f the time that financial markets were facing a credit crunch. This expectation, in turn, was partly a reflection o f the monetary pol icy actions that appeared to be in train before the even-keel period, and in prospect afterwards. Thus, a technical even-keel condition did not forestall a tightening of financial markets; nor was it accompanied, at that time, by any expan FOMC DIRECTIVE IN LATE 1960’s show less cyclical or trend m ovem ent than the 3month bill rate and longer-term market rates in even-keel periods, but they do fluctuate widely and occasionally do m ove persistently in one direction. Free reserves showed downward movements in the February and May 1966 periods, for example, when the FOM C was tightening in terms of re serve availability, while taking account o f Treas ury financing. On the other hand, free reserves rose, and member bank borrowings declined, in the even-keel period of O ctober-N ovem ber 1966, beginning the trend m ovem ent in those variables that lasted until the spring o f 1967. In 1968, net borrowed reserves deepened, and member bank borrowings rose, during the even-keel period in February. The FOM C direc tive o f February 6, 1968, sought to maintain firm conditions in the m oney market, but permitted operations to be modified to the extent permitted by the Treasury financing if bank credit appeared to be expanding as rapidly as projected. The ex pansion o f bank credit in that period apparently was sufficiently large to lead to some diminution in the extent to which reserves were supplied by open market operations (that is, through nonbor rowed reserves) relative to demand. sion in the monetary base (member bank reserve balances plus currency held by banks and the public), bank credit, or the money supply. Sharp downward movements in longer-term in terest rates began in the middle o f the M ay 1968 even-keel period and continued until the August period. Brightening prospects for fiscal restraint legislation contributed to the turnaround. And the decline was sustained by an accomm odative open market policy, as indicated by the mid-June and mid-July directives. These directives stipulated that open market operations should accom m odate tendencies for short-term rates to decline (in mid-June) and for less firm m oney market condi tions to develop (in m id-July). The mid-July directive took cognizance o f the forthcom ing A u gust refunding in the operating paragraph. But the mid-June directive did not take note o f an early July $4 billion tax bill financing, as the market atmosphere o f the time clearly posed no marketing problem for even a very large bill financing for cash. M A R G IN A L RESERVE M E A SU R E S. Free reserves and member bank borrowings, shown in Chart 2, behave somewhat the same in even-keel periods as does the cost of 1-day m oney— that is, Federal funds and dealer loan rates. T hey tend to FIG UR E 2 MARGINAL RESERVE MEASURES Incl. T.A. 1966 S h ad ed 1967 areas in d ic a te p e r io d s o f e v e n k e e l. T .A . t a x a n tic ip a tio n b ill. Billions of dollars T.A. 1968 M O N ETA R Y A G G R EG A TES. The relation between even keel and monetary aggregates (monetary base, bank credit proxy, and money supply) is both highly complex and erratic. As shown in Chart 3, it is difficult to perceive signif icant differences in behavior o f the monetary base in even-keel periods as compared with surround ing periods. In the summer and fall of 1966, the monetary base showed virtually no growth in or outside even-keel periods. Beginning in late 1966, the monetary base began to expand, and a more or less steady expansion persisted for the ensuing 2 years, with the rise in even-keel periods seem ingly little different from the rise outside such periods. It is true that in October of 1967 there was a relatively sharp increase in the monetary base during an even-keel period, as was also the case in N ovem ber 1968. The October 1967 period comprises a $4Vi billion tax offering. The re lationship to even keel was less direct than with an ordinary even-keel constraint. The sec ond paragraph o f the directive o f October 3, 1967, noted that operations should be directed to maintaining prevailing conditions in the money market with a proviso that operations should be modified to the extent permitted by Treasury financing to moderate any apparent tendency for bank credit to expand significantly more than currently expected. Apparently bank credit (as measured on a proxy basis weekly by total mem ber bank deposits) did not rise significantly more than expected, although the increase in the period was quite sharp as shown in Chart 4. Growth o f bank credit did slow in subsequent weeks. While the monetary base appears to show rela tively little difference in behavior in even-keel as compared with other periods, there are somewhat more frequent occurrences o f differential behav ior for bank credit and m oney supply measures (weekly figures on a daily-average basis).2 The February 1968 coupon financing was an instance of accelerated bank credit growth in an even-keel period. This financing was a combination “rights” exchange and cash financing, with the cash part settled a week later than the exchange. A bout $4 2 Technically, differences in behavior among bank credit, money supply, and the monetary base may be explained by changes in deposit mix or in deposit distribution between country and city banks. But the monetary base series comes from a source different from that for the credit and money supply series, and the seasonal factors could Also be inconsistent. FIGURE 3 MONETARY BASE Billions o f dollars, seasonally adjusted In c l, T.A , 1966 1967 1968 M o n e ta r y b a se: m em b e r ban k d e p o sits a t R eserv e B a n k s an d cu rren cy h eld b y b a n k s and th e n o n b a n k p u b lic . S h a d ed a reas in d ic a te p e r io d s o f ev en k e e l. T .A . ta x a n tic ip a tio n b ill. FOMC DIRECTIVE IN LATE 1960’s the Treasury ($4 billion in tax bills and the re mainder in coupon issues). On the other hand, through the summer and early fall o f 1966, bank credit showed no tendency to expand— even keel or not— despite about $8 billion of net new cash raised by the Treasury, practically all through new bill issues. In this period, banks were unable to compete effectively for time deposits. The money supply, too, showed more rapid growth at times in even-keel periods than in surrounding periods. A number of periods where this seems the case may be cited— February 1967; May 1967; May 1968; and October -N ovem ber 1968. It is not simple to develop an explanation for this phenomenon. One might hy pothesize that the process of exchanging securi ties, or issuing new securities, at times leads to enlarged holdings of cash balances as investors prepare for and consummate payments— either cash payments directly to the Treasury, or pay ments to other investors and underwriters for buying “rights” or in secondary market distribu tion of the new issues. Some confirmation o f that explanation might come from noting that money supply growth slowed or contracted following each of the even-keel periods noted above. CONCLUSIONS. 1. Even keel has been ap plied consistently to coupon issue financings. billion of new money was raised in the financing. The large net new cash demand made the financ ing similar in effect on bank credit to the tax bill financing noted above. There was, however, a contraction in outstanding bank credit for some weeks subsequent to the Treasury financing. Bank credit also appeared to show an acceler ated expansion in the^O ctober-Novem ber 1968 even-keel period. The mid-November financing did raise about $2 billion of new money. The ac celerated rate of credit expansion continued into December, sustained by issuance o f a $2 billion tax bill by the Treasury for payment in early D e cember— a financing that was not even keeled in the sense o f recognition in FOM C directives. It would appear that even keel is often associ ated with accelerated bank credit expansion in periods when even keel is applied to financings that raise large amounts o f net new cash and when at the same time market interest rates are low enough relative to Regulation Q ceilings that individual banks do not feel constrained in their ability to obtain time deposits and thus in their capacity to invest in U.S. Government securities as well as to make loans. In the long even-keel period in the summer of 1967, there was an ac celerated bank credit expansion, w hich help ed finance about $6Vi billion o f new cash raised by FIGURE 4 BANK CREDIT AND MONEY STOCK JW £ 1966 SEPT. DEC. MAR. JUNE 1967 Shaded areas indicate periods of even keel. T.A. tax anticipation bill. SEPT. DEC. MAR. JUKE 1968 With respect to bill financings, even keel has been applied in large financings, but only in cer tain market situations, and has been generally ig nored in small financings. 2. There is nothing in the material analyzed to suggest that even keel is necessarily a fixed pe riod or that it excludes some shading of policy toward restraint or ease. 3. Even keel has been consistent with vary ing movements of bank credit, money supply, and interest rates. If any variable were to be taken as an objective indicator of even keel, at least as it has unfolded in recent experience, one would select the cost of 1-day money, and assign marginal reserves to a secondary, but important, role. These are the variables most in the minds of market participants and also the ones that show the least trend movement during even-keel periods (after allowing for normal day-to-day or weekto-week fluctuations)—although even here market participants would tend to recognize that financ ing demands related to the distribution of newly offered Treasury securities would themselves tend to exert upward pressure on day-to-day money rates. 4. There have been fairly wide day-to-day fluctuations in money market variables during even-keel periods, and there have also even been some trend movements reflecting efforts by the FOMC to tighten or ease while taking account of Treasury financings. At times, this has been ac complished while not changing the attitudes of market participants because trend movements have been disguised for a few weeks by the large fluctuations that market participants are used to or because they have encompassed only a small portion of an even-keel period as defined for pur poses of this analysis. 5. While the wide variations in behavior of the variables examined suggests that the even-keel commitment is flexible not only in terms of tim ing but also in terms of credit conditions, any sharp movements permitted in day-to-day money market conditions, or even under some circum stances in interest rates, are likely over the short run to risk an unsuccessful Treasury refunding in the sense of an unexpectedly large attrition or high allotment ratio. 6. Bill rates and intermediate- and long-term rates are influenced by changes in the supply of securities and by expectations as well as by mon etary policy. Thus, it is not surprising that bill rates and other yields show movements independ ent of even keel. However, it may be that their movements during financings would be more ex aggerated without the even-keel constraint. But whether the trend of interest rates over a rela tively long period would be any different without even keel is quite another, and an unresolved, issue. 7. The behavior of monetary aggregates in even-keel periods has not been consistent. But when they have diverged from their behavior out side even-keel periods, it has been in the direc tion of relatively greater expansion, though often offset by slower growth or contraction in subse quent weeks. The relatively greater expansion, when it occurs, may not be a function of even keel, however. It may more basically be a func tion of the way monetary policy is conducted— with or without even keel. In general, monetary policy attempts to encourage credit conditions in the economy consistent with sustainable economic growth. The credit conditions sought by the Fed eral Reserve influence the interest rates the Treas ury has to offer on its securities and the type of buyer—for example, bank or nonbank—attracted to these securities. Treasury credit demands, like such demands from businesses or consumers, tend to fall in part on banks, who may either buy Treasury securities or help finance those who do. And money supply may also expand as an as pect of the financing and distribution process. Thus, credit demands or refinancings by the U.S. Government at times have led to an accelerated expansion in bank credit or money. But the ex tent to which this occurs will be affected by the existing tautness or ease of credit markets as in fluenced by monetary policy; in 1966, for in stance, net cash borrowing by the Treasury did not lead to expansion in bank credit or money. In any event, the significance of any accelerated expansion of monetary variables in even-keel pe riods—as in other periods—cannot be assessed without evaluating the credit conditions with which they are associated and the appropriateness of these conditions to the economic goals being sought. by Richard G. Davis SHORT-RUN TARGETS FOR OPEN MARKET OPERATIONS CONTENTS 39 INTRODUCTION 41 OFFSETTING ACTIONS NEEDED TO HIT TARGET VARIABLES AT THE DESIRED LEVELS A classification scheme for targets in terms of needed offsetting actions Targets in Group “A” Targets in Group “B” Targets in Groups “C” and “D” Summary 41 41 42 44 44 45 46 46 46 48 48 54 56 56 58 61 MULTIWEEK STRATEGIES FOR HITTING PARTLY ENDOGENOUS TARGETS Choice of variables to be used as a weekly target Translating the monthly target into appropriate values of the weekly target Error-response mechanisms TRANSLATING MONTHLY TARGETS INTO AVERAGE MONTHLY VALUES OF WEEKLY TARGETS Regression techniques Implications of the regression results THE EFFECTS OF QUANTITY TARGETS ON MONEY MARKET STABILITY The nature of the problem An experiment Evaluation 66 67 SOME MIXED STRATEGIES— BLENDING MONEY MARKET AND QUANTITY CONSIDERATIONS IN FRAMING TARGETS Pure money market strategies Pure quantity strategies The credit proxy proviso clause Using money market targets as a tactical device in a quantities-oriented strategy Quantity targets with money market modifiers A money market proviso 68 SOME GENERAL CONCLUSIONS 62 62 63 64 65 SHORT-RUN TARGETS INTRODUCTION This paper examines several types of targets that could be used by the Federal Open Mar ket Committee to guide the actions of the Manager during the interval between meetings. The paper considers a number of different monetary, banking, and money market meas ures that might be used as target variables, as well as strategies that could be used to aim the variable chosen at the target value selected by the Committee. The paper starts from the premise that whatever may be the ultimate goals of the FOMC and whatever criteria (or indicators) it may use to judge the impact of its decisions, the Committee must give instructions to the Manager to determine his actions between meetings. For various reasons, these instruc tions will not be couched in terms of goal vari ables such as the gross national product or the rate of change of the price deflator. These in structions may or may not be couched in terms of the same variables the Committee uses to measure the impact of its actions. For example, the Manager may be instructed to hold free reserves within a given range, making free reserves the target variable, and at the same time the Committee may use levels of free reserves to define degrees of policy tightness and ease. Thus the same variable, free reserves in this case, may serve both as “target” and as “indicator”—to use the Brunner-Meltzer termi nology.1 Alternatively, the instructions to the Manager might be couched in terms of free reserves, but the Committee might judge tight ness and ease in terms of some other variable such as the monetary growth rate—that is, target and indicator may be distinct. N o t e .—The author, who is Adviser to the Federal Reserve Bank of New York, is indebted to Paul Meek for many useful conversations on the subjects dis cussed in this paper. He would also like to thank Susan K. Skinner for excellent research assist ance. The author assumes sole responsibility for the views expressed. 1 See, for example, Karl Brunner and Allan Meltzer, “The Meaning of Monetary Indicators,” in G. Horwich, ed., Monetary Process and Policy (Homewood, 111.: R. D. Irwin, 1967), pp. 187-217. In any case, the target variable or variables will almost certainly come out of a familiar, if rather long, list of banking and money market measures—free reserves, nonborrowed re serves, the money supply, and so forth. A tar get may take many forms. Thus, for example, it might be stated in terms of a single value, or in terms of a range of values. The target might be a single variable or it might be several vari ables—provided, of course, that the values or ranges chosen for these variables were compa tible during the period in question. The pri mary target variable might be made subject to a side condition stated in terms of some other variable—as in the case of the so-called proviso clause.2 The target might be stated in terms of an explicit or implicit weighted average of several variables. For example, the famous notion of money market “tone” as a target may be thought of as a weighted average of several measures of marginal reserves and money mar ket rates—with the weights left unquantified, though hopefully reasonably well understood in any given historical situation. Much of the attention of this paper is de voted to ways in which the so-called “aggrega tive” or “quantity” variables (such as money supply and bank credit) could be used as tar gets, the accuracy with which such targets could be hit, the cost of aiming at such targets in terms of money market stability, and the ways in which the use of such targets could be reconciled with an acceptable minimum of market stability. These subjects have been cho sen for the major share of attention because they seem to be the ones of greatest interest at the moment. On the other hand, relatively little is said about money market targets as such. Experi ence has already taught us much about such targets and there seems little new to be added in a general way. Another subject that is treated only tangentially is the relative mer its as between different quantity (or aggrega 2 For a discussion of the proviso clause see pp. 64-65. tive) targets: Is M t a better target than M 2? Is M 2 a better target than bank credit? Most of the interesting problems that arise in trying to answer such questions turn out really to be questions as to which variable is the more eco nomically meaningful, which variable is the better measure of the impact of the System on the economy, and which variable better char acterizes the tightness or ease of policy. Since the paper is concerned only with the “target” properties of these variables as such, questions of this kind are considered off limits. While the paper makes no attempt to come up with a specific recommendation from among the various possible targets, the writer should perhaps admit to a feeling that for var ious reasons, the System is likely in the future to judge its performance more on the behavior of quantities—monetary and bank credit growth rates—than it has in the past and less on the basis of money market conditions. By the same token, the degree of control provided over some of these quantitative growth rates seems likely to become a more weighty consid eration in the choice of FOMC procedures in the future than it has been in the past— even if this means some reduction in the ability to sta bilize money market conditions. The format of the paper is as follows: The first section looks at the various possible tar gets from the point of view of the noncon trolled variables that must be offset to hit these targets. Out of this analysis it picks the varia bles that appear to be operationally feasible as week-to-week objectives. The second section considers the main elements of multiweek strategies, strategies through which targets not feasible as week-to-week objectives may never theless be hit over the longer period between FOMC meetings. The third section considers ways in which targeted values of such multi week targets (for example, the monetary growth rate) might be translated into appro priate average weekly values of the variables used as week-by-week targets (such as non borrowed or free reserves). In addition, this section attempts to provide some assessment of the accuracy with which variables such as the monetary or bank credit growth rates could be manipulated over periods as short as a month or a quarter if such manipulation became the deliberate and sole objective of open market operations. The fourth section tries to determine how much buffeting the money market might suffer if the use of money market targets were re placed by procedures that aimed directly at the monetary aggregates without special regard for their market consequences. The fifth section offers some “mixed” strategies through which control over monetary aggregates might be blended with concern for a reasonable degree of money market stability. The final section of fers some brief general comments on the re sults of the paper. For convenience, a list of the variables con sidered singly or in combination as possible targets, and the symbols used to represent them, follows ; Ru Rt Re Rb Rf Rr Bu Bt C Dp Dt Mx m2 BC FR rd n Member bank nonborrowed re serves Member bank total reserves Member bank excess reserves Member bank borrowed reserves Member bank free reserves Member bank required reserves Nonborrowed monetary base Total monetary base Currency in hands of nonbank public Private demand deposits adjusted Treasury deposits Private demand deposits adjusted plus currency Private deposits adjusted plus currency Total bank credit Bank credit proxy (variously de fined) Federal funds rate Discount rate Treasury bill rate SHORT-RUN TARGETS A CLASSIFICATION SCHEME FOR TARGETS IN TERMS OF NEEDED OFF SETTING ACTIONS. Table 1 summarizes the noncontrolled items that must be offset to fix the target variable listed in the left-hand col umn at whatever level or range has been speci fied by the Committee. The targets listed have been divided into four groups, A, B, C, and D, according to the different sorts of problems in volved in hitting specific values of each target within a given reserve-averaging period—that is, one week. TARGETS IN GROUP “A.” The first group of items, the A group, is the easiest to hit. The noncontrolled factors that must be offset to hit targets in this group are entirely exogenous with respect to open market opera tions. Hence there will be no feedbacks, no “simultaneity problem” involved. The first two items, nonborrowed reserves and the nonbor rowed monetary base, can be hit on a weekly basis with as much (but no more) accuracy as the noncontrolled operating transactions, such as float, can be forecast. The second two tar gets, nonborrowed reserves less reserves re quired behind Treasury deposits and the non borrowed base less such required reserves, obviously require, in addition to the operating transactions, correct forecasting of the behav ior of Treasury deposits at commercial banks. OFFSETTING ACTIONS NEEDED TO HIT TARGET VARIABLES AT THE DESIRED LEVELS The only variable under the complete and di rect control of the Open Market Account Manager is the size and composition of the Account’s portfolio. Nevertheless, for obvious reasons, the portfolio has in recent times sel dom if ever been suggested as a target varia ble. Thus every variable that has been sug gested is determined in part by actions not under the direct control of the Manager. Some of these actions are taken by the Treasury, some by the banks, some by the public, and some, even, are Acts of God! Thus the first question in determining the feasibility of hit ting any particular target variable is: “What noncontrolled items are involved?” Since hit ting the target means adjusting the portfolio to offset the effects of these noncontrolled items on the target variable, the next consideration may well prove to be: “How well can the movements in the noncontrolled items be pre dicted?” or alternatively, “Can the movements of the noncontrolled items be known soon enough so that the appropriate offsetting ac tions can be carried out during the time period —day, week, or month—during which the tar get variable is to be hit?” TABLE 1: Target Variables and Noncontrolled Items to be Offset T arget A - Rn-Rr T reas,. Bm-RT T r e a s ... B ............ B* R*-Rr T r e a s ... JB*-Rr T r e a s ... f A fi.............. .. c j[ M t ........... D i O perating transaction s Excess reserves B orrow ed reserves X X X X X irff....................... M em ber bank reserves required against— T im e Treasury Private dem and N et interbank X X X X X D em a n d dep o sits at n o n m em ber ba n k s T im e M arket d e p o sits dem ands a t n o n - fo r variou s m em ber instrubank s m en ts X X X X X X X X X X X X T o ta l required reserves X X X X X X X X X X X X X X X X X X X X X X 41 Free reserves, the final item in the A group, require the offsetting of both operating trans actions and changes in required reserves. Since the advent of lagged reserve accounting, how ever, required reserves are known at the begin ning of each week and hence present no additional forecasting problem. Prior to lagged reserve accounting, of course, required reserves were not strictly pre determined. However, the working assumption of the System seems to have been that re quired reserves were, in fact, largely independ ent of the level of nonborrowed reserves in the same week. Put differently, the banking sys tem’s purchase of assets was assumed to be rather unresponsive to the state of money mar ket conditions within the same week. Under the circumstances, required reserves would have been determined largely by conditions in previous weeks and therefore would have been essentially exogenous insofar as the current week was concerned. Hence, to the extent that this is true, the introduction of lagged reserve accounting has not changed the situation in any fundamental way, although it has reduced the number of exogenous variables that the System must forecast in any given week. TARGETS IN GROUP “B ” Success in hitting targeted values of variables in the B group involves offsetting items whose move ments are themselves functionally related to the volume of open market operations under taken within any given week. As a result, at tempts to aim directly at variables in the B group require not only projections of exogen ous movements in operating transactions but also some attempt to estimate the feedback ef fects of open market transactions on other var iables entering into the target. For the varia bles included in the B group, the crux of the feedback problem lies in the long-observed in terdependence between the volume of nonbor rowed reserves supplied through open market operations and the volume of borrowed re serves supplied through the discount window. To analyze the problem and the possibilities for hitting targets in the B group, it is useful to analyze the market for reserves as that mar ket exists within a given reserve-averaging pe riod. In the general case, it can be assumed that bank demands for both borrowed and ex cess reserves are interest rate sensitive—per haps the Federal funds rate is especially rele vant. Required reserves are predetermined by lagged reserve accounting as noted. Nonbor rowed reserves are an exogenous variable, be ing determined by market factors and System operations. In the linear case, the system of equations describing the reserve market is therefore simply as follows (ignoring constant terms): ( 1) Rb = (2) Re = b(r/f) + et (3) Rt ^ Rb + Ru (4) R t s Re + Rr where eh and ee are random variables repre senting the random components of the de mands for borrowed and excess reserves, re spectively. If the System supplies reserves, its actions will tend to raise nonborrowed reserves and lower interest rates. Excess reserves will tend to rise and borrowed reserves will tend to fall. The Trading Desk can predict the effects of its actions on these magnitudes and on total re serves only if it has some notions about the elasticities of demand for borrowed and excess reserves—in addition, of course, to predictions about the behavior of operating factors. Some features of the situation are brought out more clearly by solving equations 1 to 4 for the reduced form equation for total re serves. r^se*-r ^ w ’ + As the equation indicates, errors in predicting operating factors and stochastic elements in the demands for excess and borrowed reserves would remain as sources of error even if the SHORT-RUN TARGETS Desk knew the elasticities of demand for ex cess and borrowed reserves. Since a, the inter est rate coefficient of the demand for borrowed reserves, is equal to or greater than zero, and since bf the interest rate coefficient of the de mand for excess reserves, is equal to or less than zero, b/(b — a), the response of total re serves to an increase in nonborrowed reserves, will be nonnegative. However, the response of total reserves in a given statement week to an injection of nonbor rowed reserves in the same statement week may approach zero under two circumstances. First, it will approach zero for very large val ues of a—that is, when demands for borrowed reserves are very sensitive to the interest rate effects induced by increases or decreases in nonborrowed reserves. In the limit, an increase in nonborrowed reserves will be exactly offset by the repayment of borrowings without any perceptible fall in interest rates. Thus in this case, an increase in nonborrowed reserves has no effect on total reserves until the point is reached where borrowings are reduced to zero. Further increases in nonborrowed reserves be yond this point, however, increase total re serves by the same amount, with the interest rate falling enough to absorb the entire in crease into excess reserves. Second, b/(b — a) will approach zero as b approaches zero—that is, when demands for excess reserves are very msensitive to interest rates. In the limit, the demand for excess re serves reduces to a random variable (presuma bly with positive mean), and the reduced form equation 5 for total reserves becomes (5a) Rt = Rr + ee In this case, as in the first case, an increase in nonborrowed reserves will, in general, have no effect on total reserves within the same state ment week. An increase in nonborrowed re serves tends to push down interest rates. But since, by assumption, this fall in rates pro duces no rise in the demand for excess re serves, the fall in rates must proceed far enough to induce banks to repay borrowed re serves by the full amount of the increase in nonborrowed reserves. (As in the first case, this process is obviously limited by the fact that borrowings cannot fall below zero.) How likely is it that one or the other of these two extreme cases will prevail and that, as a result, the Desk will have little or no in fluence on total reserves within a given state ment week by its actions within that week? The first possibility, in which the elasticity of demand for borrowed reserves approaches in finity, can be ruled out. Indeed, all of the ar gument has been over whether borrowings show any substantial responsiveness to interest rates. The second case, however, in which the interest elasticity of the demand for excess re serves approaches zero, seems to have some real significance. Market experts apparently believe, for example, that when excess reserves are down to virtually frictional levels, as at present (fall 1969), demands for excess re serves may be quite insensitive to rate fluctua tions within the normal range, so that weekto-week fluctuations in excess reserves have to be treated as essentially random. The following example (summarized in Table 2) was suggested by one market ob server. Suppose, as at the present writing, that net borrowed reserves are around $1.0 billion in a given week. Suppose that in the following week required reserves rise by $200 million. If the System supplies $200 million in nonbor rowed reserves, free reserves will obviously re main unchanged. Abstracting from random shifts in the demand for excess and borrowed reserves, the Federal funds rate would also be unchanged, as would both excess and bor rowed reserves. Total reserves would rise by the $200 million increase in nonborrowed TABLE 2: Example of Change in Reserves—Tabular Summary (where bf(b — a) — 0.2) t?.. tA - 1 2, R esu ltin g change A ssum ed change am ple num ber R equired reserves +200 +200 N on borrow ed reserves +200 + 100 T otal Free E xcess B orrow ed reserves reserves reserves reserves 0 -2 0 0 +80 0 -1 0 0 +200 + 180 (and required) reserves. Now if, on the other hand, the System were to supply only $100 million in nonborrowed reserves, this market observer felt that borrowed reserves might rise by about $80 million instead of remaining un changed, and that excess reserves might fall by about $20 million instead of remaining un changed. In effect, this observer is estimating that b /(b — a) is only about 0.2, owing to a very low elasticity of demand for excess re serves. If this estimate is correct, a $100 mil lion reduction in the rate at which the System supplies nonborrowed reserves produces only a $20 million reduction in the rate at which total reserves grow. Certainly a value of 0.2 for b / (b — a) is by no means the same as a zero value, and it by no means implies a complete inability to influ ence total reserves within the statement week. It may be low enough, however, to reduce greatly the practical ability of the Desk to con trol total reserves on a week-by-week basis. Thus a $100 million cutback in the rate at which nonborrowed reserves are supplied might have a significant influence on the Fed eral funds rate while reducing excess (and total) reserves by only $20 million, a very small effect relative to the random component of week-to-week fluctuations in excess re serves. The implication of these estimates, if correct, is that the Desk might be able to exert a clear and substantial influence over total re serves within the statement week only, if at all, by inducing rather violent fluctuations in bor rowings and in the Federal funds rate. More over, attempts to hold the growth of total reserves below the predetermined rate of growth in required reserves very quickly run into the absolute limitation that excess reserves cannot be negative. TARGETS IN GROUPS “C” AND “D ” The variables in group C present problems similar to those in B. In principle, however, the problems are more severe because a larger number of functional interdependencies are in volved. For example, direct aim at the narrow money supply on a week-by-week basis would require not only predictions of operating fac tors and knowledge of the demand schedules of excess and borrowed reserves, but also knowledge of the demand schedules for all the major deposit liabilities as well—obviously an impractical requirement. From this point of view Mi, M2, the proxy, and bank credit are on a par since they all require knowledge of the same relationships—though arranged in differ ent ways. Finally, the group D variables—the money market rates—present a different kind of prob lem considered as week-by-week target varia bles. In principle, their successful use as tar gets would require a complete model of the money market. In practice, however, a tolera ble accuracy can be achieved by taking advan tage of the reasonably close relationship be tween money market rates and free reserves and borrowings. Thus the general strategy for a given week could be laid out by the level of free reserves thought to be compatible with the targeted level of, say, the Federal funds rate. Specific daily adjustments in this general strat egy could then be made in response to the emergence of rates in the market that deviates from targeted values. SUMMARY. The results of this survey of targets can be summarized as follows: The variables in the A group— nonborrowed re serves, the nonborrowed base, these two varia bles less reserves required behind Treasury de posits, and free reserves—can all be used as weekly targets subject only to errors in pre dicting operating transactions and, where rele vant, required reserves behind Treasury depos its. Variables in the B group—borrowings, total reserves, the total base, and these last two items less required reserves behind Treas ury deposits— can theoretically be used as weekly targets provided the Desk has at least some crude knowledge of the interest rate elas ticities of demand for excess and borrowed re serves (that is, b /(b - a ) ) . As with the A group, errors in hitting targets in the B group will be subject to errors in hitting operating transactions and also to errors in estimating SHORT-RUN TARGETS the demand elasticities of excess and borrowed reserves and/or stochastic shifts in the demand schedules for these quantities. Since it is diffi cult to imagine why errors in hitting operating factors should be systematically offset by er rors in judging the demands for excess and borrowed reserves, variables in the B group will be significantly harder to hit on a weekby-week basis than variables in the A group. Moreover, under certain assumptions about the demand elasticities of excess and borrowed re serves, targets in the B group, such as total re serves, may not be feasible at all as weekly targets. Thus it is often argued that the interest elasticity of demand for excess reserves is so low under present conditions that the Desk’s operations within a week have only a marginal influence on total reserves within that week— at least within tolerable limits of interest rate fluctuations—and that actual control would not be possible. Variables in the C group, such as the money supply and bank credit, could be aimed at directly on a week-by-week basis only with presently unavailable knowledge of numerous demand and supply schedules. The D group variables, including money market rate targets, present much greater problems in theory, as noted, than they do in practice when considered as weekly targets. The difficulty, and perhaps the impossibility, under present conditions of aiming directly at total reserves, the total base, and the various money and credit variables on a weekly basis does not mean that these variables cannot be used as targets to be approached indirectly over somewhat longer periods. It does, how ever, mean that week-by-week targets must be chosen from among those variables that can be used for this purpose, with the weekly setting of these targets picked according to some strategy designed to hit the basic target varia ble on the average over some longer period. Since, as indicated in the introduction, the “target problem” is the problem of the instruc tions to be given to the Manager by the Com mittee, the logical period in this context appears to be the period between FOMC meetings, presumably about a month. The next section considers the rough outlines of a multiweek control strategy for hitting target variables not suitable for use on a week-by-week basis. MULTIWEEK STRATEGIES FOR HITTING PARTLY ENDOGENOUS TARGETS If System actions to fix nonborrowed or free reserves within a given week have only a small influence on the behavior of variables such as the money supply or bank credit within that week relative to random or otherwise hard-topredict influences, these variables cannot be used as operationally meaningful targets gov erning Desk decisions on a week-by-week basis. To put it more concretely, the decision to increase nonborrowed reserves by $50 mil lion in a particular week, given a forecast of operating factors, implies a concrete decision about open market operations. The injunction to increase the money supply by $50 million in a particular week, by contrast, is probably almost empty of concrete implications for open market operations given the state of our knowledge at present and for the foreseeable future. Nevertheless, while variables in the C group such as the money supply (and proba bly variables in the B group such as total re serves) are not controllable on a week-byweek basis and are therefore not suitable as week-by-week targets, it is clear that the weekly settings of targets in the A group, such as nonborrowed and free reserves, do influence the behavior of the broader variables over the somewhat longer run. The question arises as to what kinds of strategies are available for using this influence to hit targets that are not under the direct con trol of the Desk over the longer period of about a month between FOMC meetings by setting week-by-week targets for variables that can be directly hit—subject only to errors in offsetting operating transactions and other purely stochastic matters. Any strategy would appear to consist of three basic elements: (1) the choice of a weekly target; (2) a procedure for translating the value of the monthly target into week-by-week values of the weekly target; and (3) a set of rules for responding to "misses” within the period between FOMC meetings. (Presumably responses in subsequent periods to misses over the entire period are a matter for the FOMC to decide at its meeting and therefore beyond the scope of this paper.) Assume that the FOMC issues its instructions in terms of a desired seasonally adjusted growth rate over the month for one of the broader magnitudes such as the money supply. This of course can immediately be translated into a seasonally wnadjusted level of the monthly target for the month in question. CHOICE OF VARIABLES TO BE USED AS A WEEKLY TARGET. As sug gested earlier, the first question in carrying out the Committee’s instructions is the choice of variables to be used as a weekly target. The previous analysis has argued that in practice, the choices probably narrow down to nonbor rowed reserves, free reserves, borrowings, and money market rates. It should be noted, how ever, that in any given week, the choice be tween free reserves and nonborrowed reserves is not really a choice at all. Given required re serves, a decision about a target level of nonborrowed reserves is simultaneously a deci sion about free reserves—and vice versa. There may be a tactical difference, though. For example, it might prove better to make decisions based on known past relationships between rates of growth of nonborrowed re serves and the monthly target variable rather than on known past relationships between lev els of free reserves and the rate of growth of the monthly target variable. TRANSLATING THE MONTHLY TAR GET INTO APPROPRIATE VALUES OF THE WEEKLY TARGET. Once a decision has been made as to which of the possible weekly target variables will be used, the next step is to translate the monthly target deter mined by the FOMC into appropriate values of the weekly target. For example, the pre liminary judgment might be reached that a 4 per cent targeted annual rate of growth in Mx for the month of October could be achieved by about a 5 per cent annual rate of growth in nonborrowed reserves. Such a judgment might be reached by means of a re gression equation or a “hand” method based on the projectionist’s “feel” for the probable behavior of the money supply under various assumed rates of growth in nonborrowed re serves. (The next section of this paper pre sents the results of a number of regression equations relating monthly values of directly controllable variables such as nonborrowed re serves to monthly values of targets from groups B and C—such as total reserves, the money supply, and bank credit.) Whatever method is used, the resulting monthly rate of growth in nonborrowed reserves could be translated into preliminary weekly targets for nonborrowed reserves simply by making the appropriate extrapolation from the level of such reserves in the last week of the previous month. The situation is illustrated in Figure 1. The top panel shows a weekly pattern of M x consistent with the targeted growth rate for the month designated “October.” The line AB in the bottom panel of the diagram shows the pre liminary weekly targets for nonborrowed re serves given the targeted behavior of M ERROR-RESPONSE MECHANISMS. The third and final broad element in a strategy consists of error-response mechanisms. Errors, or “misses,” are of course inevitable. They are of two basic types. First, weekly targets may be missed—in the case of free and nonbor rowed reserves because of misses in predicting operating factors; and in the case of borrowed reserves and money market rate targets, for other reasons as well. Second, even if the ac tual values of the week-by-week control varia bles are right on target, the expected path of the broader monthly target may not result. SHORT-RUN TARGETS The Manager might decide to respond only to the first type of error, errors in hitting the weekly target. In that case, successful hitting of the targeted path for nonborrowed reserves would imply that he would continue to move along the given path (AB in the diagram) even though M x was not responding as expected. This would be a plausible approach if slip pages between the money supply and nonbor rowed reserves could be assumed to be random FIGUR E 1 over time so that they could be expected to cancel out when averaged over a period of weeks. Even if the Manager does respond only to errors of the first kind—in this case errors in hitting the weekly targets for nonborrowed reserves—there are still a number of possible types of error-response open to him. Some of the possibilities are illustrated in Figure 1, where point H is assumed to be the tar geted level of nonborrowed reserves in the first week of October. Suppose the actual level falls short of this, say to point K. This might call for a new and steeper path of targeted values for the remaining 3 weeks such as KFD. Al ternatively, the Manager might try to offset all the effect on the monthly average of the first week’s miss in the second week. In that case he would aim for point I in the second week. If it were successfully hit, he would revert to his original path in the final 2 weeks—that is, points J and B. Presumably there are many other possibilities. Instead of responding only to errors in the weekly target, the Manager could also respond to errors in the monthly target. Thus, for ex ample, if all were going well, the level of M x reached in the first week would be point C in the upper panel of the diagram. Whether be cause of errors in hitting the weekly target or slippages between the weekly target and the value of the monthly target expected to be as sociated with it, Mi might fall short of this level in the first week of October, say to point G. Such an error might then call for resetting the target path of growth for nonborrowed re serves for the remaining 3 weeks of the month in an attempt to compensate for this “miss” of total reserves in the first week. In summary, if the FOMC wishes to aim for some target that cannot be aimed for directly on a week-to-week basis, a target that must in stead be approached indirectly over a period of weeks, decisions concerning the following must be made: (1) A week-by-week target variable must be chosen. (2) Some rule must be found for translating the FOMCs monthly target into weekly values for the week-by-week target. (3) Rules for responding to inevitable misses, whether in the weekly targets themselves or in the expected relationship between weekly tar get values and the monthly target, must be de vised. These three elements must be faced whether the monthly target be M u as assumed here, bank credit, total reserves, or what have you. TRANSLATING MONTHLY TARGETS INTO AVERAGE MONTHLY VALUES OF WEEKLY TARGETS As noted in the previous section, a complete strategy for hitting monthly deposit and bank credit targets involves translating monthly tar get values into appropriate values for one of the operational weekly targets. The translation procedure might itself be part of the instruc tions given by the FOMC to the Manager. However, it is more likely that the determina tion of such a procedure would be regarded as a technical problem—one that required weekby-week or even day-by-day flexibility and one that would best be solved by the operating per sonnel on the spot. REGRESSION TECHNIQUES, One way in which monthly deposit or credit targets could be translated into average monthly val ues of a weekly target such as nonborrowed reserves is by use of a regression equation re lating the FOMC’s monthly target to monthlyaverage values of the weekly target together with whatever lagged variables and seasonal dummies seem useful. Given such an equation, the average monthly value of the weekly target variable needed to achieve the FOMC’s de posit or credit objective could be calculated from its coefficient in the equation. Moreover, the pattern of forecast errors observed from applying the equation to data both within and without the equation’s sample period could be used as an indication of the accuracy with which monthly deposit and credit variables could be controlled by manipulating the aver age monthly value of the week-by-week target. There are, to be sure, a number of dangers involved in interpreting such equations. First, the historical tendency of the System to ac commodate demands for reserves in the proc ess of stabilizing money market conditions may introduce a simultaneous equations bias into the estimation of such equations. They may not be true reduced forms, and the coeffi cient of nonborrowed reserves may not give an unbiased estimate of the impact on deposits and credit of a given deliberate change in non borrowed reserves. Second, use of the error terms from such equations to evaluate the ac curacy with which the dependent variable could be controlled assumes perfect control of the independent variable. The current exogen ous variables included in the various equations presented below are, alternatively, nonbor rowed reserves, total reserves, and these two measures less required reserves behind Trea sury deposits. For reasons already discussed in detail, none of these variables can be set by the actions of the Desk without error, and it may be virtually impossible to set total re serves on a week-by-week basis under a wide range of circumstances. A third problem with these equations is that they make no allowance for the time distribution within a period of changes in reserve measures. Thus, for exam ple, they tacitly assume that the expected effect of raising the daily-average level of nonbor rowed reserves in a month by $100 million is the same whether the change is spread out evenly over the period or concentrated entirely in the final day. Despite these difficulties, as well as some other limitations to be mentioned later, such regression equations still seem to have a clear relevance to the problem at hand, and a large number of such equations are presented in the accompanying tables. In Tables 3 and 4, Parts A and B, it is assumed that nonborrowed re serves are to be the week-by-week target varia ble. Consequently the tables show a number of equations in which percentage changes in var ious potential monthly targets from groups B and C are regressed on current monthly per centage changes in nonborrowed reserves and in some lagged nonborrowed and total reserve changes. Part A of Table 3 shows equations for nonseasonally adjusted data, both with and without seasonal dummies. Table 4, Part A, repeats Table 3, Part A, with required reserves behind Treasury deposits subtracted from the various reserve measures. Again the results are reported with and without seasonal dummies. Part B of Tables 3 and 4 repeats Part A, this SHORT-RUN TARGETS time with seasonally adjusted data. All equa tions were estimated over the 1965-68 period. The standard errors are reported as percentage changes arithmetically blown up to annual rates. (The justification for annualization is simply that annual rates of change are the com mon measure in terms of which these growth rates are usually expressed.) In a large number of cases, the R 2’s of the equations are impressively high, but the stand ard errors are discouragingly large. For exam ple, by using seasonal dummies, the R 2’s of percentage changes in the deposit components of M 1 and M 2 are 0.95 and 0.88, respectively, but the standard errors amount to annual rates of 6.4 and 4.4 per cent (Table 3, Part A). If current movements in Treasury deposits are correctly allowed for, these standard errors drop somewhat, to 5.9 and 3.9 per cent, re spectively (Table 4, Part A ). As Part B of these tables shows, somewhat better results in terms of standard errors are obtained when seasonally adjusted data are used. However, it is clear that none of these standard errors are small in terms of the ranges of growth rates normally thought of as spanning the gap be tween “tight” and “easy” monetary policy. For example, the smallest standard error in the de mand deposit component of the money supply, 4.5 per cent, may be compared with the 2 to 6 TABLE 3: Current-Period “ Exogenous” Variable: N on borrow ed Reserves Regressions of Monetary Aggregates on Reserve Aggregates D e p e n d en t variable C on stant %A NBR % A N B R -l N B R -2 % A T R -1 % A T R -2 D .W . SE E Sim ple R 3 of % ANBR A. B ased on m onthly data without seasonal adjustm ent, 19 6 5 -6 8 W ithou t season al dum m ies: % A T R .......... ............................................. .3 5 8 2 % A D D ......................................................., .2 4 4 8 % A T D & D D ........................................... .5 9 9 8 .4 3 2 0 W ith season al dum m ies: % A T R .......................................................... .7 7 7 4 % A D D ......................................................... 2 .5 5 2 7 % A T D & D D ........................................... 1 .1 5 3 2 1 .0 8 7 0 .553 (5 .9 ) .9 7 7 ( 6 .1 ) .429 (5 .8 ) .525 (1 0 .2 ) .4 9 0 ( 3 .8 ) - .0 8 1 (-0 .4 ) .1 0 8 ( 1 .1 ) .0 4 2 ( 0 .6 ) .091 (0 .6 ) - .0 6 2 ( -0 .3 ) - .0 1 6 ( -0 .1 ) - .0 3 5 ( -0 .4 ) - .6 3 5 ( -4 .0 ) - .2 3 4 ( -0 .9 ) - .1 6 8 (-1 .3 ) .0 0 0 (0 .0 ) - .2 4 9 (-1 .6 ) - .1 0 4 ( -0 .4 ) - .0 2 2 (-0 .2 ) .0 4 3 ( 0 .5 ) .68 1 9 .6 4 4 0 .5 8 8 2 .5391 .56 1 3 .5091 .7553 .7 2 6 2 .2 0 8 (1 .2 ) .1 2 8 (1 .1 ) .1 5 5 ( 1 .9 ) .3 1 6 (3 .8 ) .5 4 3 ( 2 .7 ) - .0 9 3 (-0 .7 ) .1 2 6 (1 -4 ) .2 0 2 ( 2 .2 ) .2 7 5 ( 1 .2 ) .2 7 9 ( 1 .9 ) .1 6 4 ( 1 .6 ) .0 5 3 ( 0 .5 ) - .5 9 7 (-3 .4 ) - .1 8 3 (-1 .5 ) - .0 9 9 ( -1 .2 ) .0 2 6 ( 0 .3 ) - .2 4 7 ( -1 .4 ) - .0 8 1 (-0 .7 ) .0 1 3 ( 0 .2 ) .0 6 4 ( 0 .8 ) .7 5 6 2 .6 3 0 3 .9 5 1 7 .9 2 6 8 .8 8 3 5 .8 2 3 4 .8 6 4 5 .7 9 4 5 1 .8 2 9 .3 5 .5 3 5 2 .1 8 1 5 .9 5 .547 2 .1 0 7 .3 5 .5 3 5 1 .3 4 5 .1 4 .7 5 0 1 .7 8 9 .5 3 1 .5 0 6 .3 6 1 .1 8 4 .4 1 1 .2 7 4 .4 5 1 .8 7 3 .1 8 B . B ased on m onthly data seasonally adjusted, 1 9 6 5 -6 8 % A T R .......................................................... .0 6 6 4 % A D D ......................................................... .2 3 8 6 % A T D & D D ............................................ .5 4 6 8 .3 2 8 4 .7 5 8 (1 0 .2 ) .2 9 7 (2 .4 ) .2 3 5 ( 2 .9 ) .4 6 5 ( 5 .8 ) - .1 2 7 ( -1 .0 ) - .1 9 8 ( -0 .9 ) .1 1 9 ( 0 .8 ) .091 ( 0 .6 ) - .1 6 7 (-1 .2 ) - .0 2 8 (-0 .1 ) .1 3 7 ( 0 .9 ) .0 3 0 ( 0 .2 ) .2 2 7 ( 1 .5 ) .1 3 4 ( 0 .6 ) - .0 7 6 ( -0 .5 ) .1 0 8 (0 .7 ) .2 2 6 ( 1 .5 ) .1 3 6 ( 0 .5 ) - .0 1 3 ( -0 .1 ) .0 4 0 ( 0 .2 ) .7721 .7 4 5 0 .1 4 6 3 .0 4 4 7 .3 8 2 6 .3 9 1 0 .6 4 0 8 .5 9 8 0 1 .7 4 5 .2 0 1 .4 0 3 .5 0 1 .5 8 3 .4 3 C. Based on quarterly data seasonally adjusted, 1 9 6 0 -6 7 % A T R .......................................................... , - . 4 8 2 3 % A D D ......................................................... - .4 4 1 2 % A T D & D D ............................................ .4 5 2 8 .1 8 2 5 .899 (1 0 .7 ) .4 3 2 ( 2 .7 ) .5 4 6 ( 3 .6 ) .7 7 0 ( 5 .6 ) - .2 4 2 ( -1 .9 ) .0 3 6 ( 0 .1 ) - .1 1 7 ( -0 .5 ) - .2 8 1 ( -1 .3 ) - .1 0 5 ( -1 .1 ) - .0 3 9 ( -0 .2 ) — .2 4 5 (-1 .5 ) - .2 3 8 ( -1 .6 ) ,5 1 6 ( 3 .1 ) .2 7 0 ( 0 .9 ) .6 0 6 ( 2 .0 ) .6 9 6 ( 2 .6 ) N o t e .— Standard errors o f estim ates (SEE) are a t an n u al rates, "t” valu es are in parentheses. .3 9 2 ( 3 .2 ) .4 1 3 ( 1 .8 ) .617 ( 2 .8 ) .5 8 5 ( 2 .9 ) .8 9 2 8 .8 7 2 2 .5 6 2 3 .4781 .6 5 7 4 .5 9 1 5 .7 2 4 6 .6 7 1 6 1 .6 3 1 .2 2 1 .4 5 2 .3 0 1 .2 5 2 .2 1 1 .1 6 2 .0 1 per cent range of growth rates for Mx some times cited as the prudent limits of tight and easy money. A more graphic impression of the size of the errors can be obtained from Figures 2-5 showing the time series of residuals from certain of the equations presented in Tables 3 and 4, Parts A and B. The equations have been used to “predict” developments in the first half of 1969. As the figures indicate, the results were quite poor in most cases. Perhaps part of the difficulty could stem from struc tural changes produced by the inauguration of lagged reserve accounting in the fall of 1968. For the sake of completeness, a second set of tables, Tables 5 and 6, paralleling Tables 3 and 4, Parts A and B, has been prepared that assume, in effect, that total reserves rather than nonborrowed reserves are the variable to be used as the week-by-week target. Again, the control problem with regard to total re serves should be kept clearly in mind in evalu ating these equations. In any event, the lesson of Tables 5 and 6 is essentially the same as the lesson of Tables 3 and 4. The standard er rors of estimate (expressed, as before, in terms of annual rates) are still quite large relative to TABLE 4: Current-Period “ Exogenous” Variable: N onborrow ed Reserves A d ju ste d f o r R eserves R eq u ired A gainst U .S . G overn m en t D eposits Regressions of Monetary Aggregates on Reserve Aggregates Adjusted for Reserves Required Against U.S. Government Deposits D epend en t variable C onstant % A NBR* % A N B R -1 * % A N B R -2 * % A T R -1 * %A T R -2 * R 2 SEE Sim p le R* o f % anbr 1 .8 5 9 .7 1 .621 1 .4 6 9 .8 9 .8 2 8 1 .4 2 4 .9 4 .7 9 7 1 .7 3 6 .3 7 .5 9 7 1 .8 0 9 .8 5 1 .2 9 5 .9 3 D .W . A. B ased on m onthly data without seasonal adjustm ent, 19 6 5 -6 8 W ithout season al dum m ies: % A T R * ................................................ .38 5 0 % A D D .................................................. .0 1 4 6 % A T D & D D .................................... .5 1 9 2 % A P rox y ............................................. .4 5 9 4 W ith season al dum m ies: % AT R * ................................................ .9441 % A D D .................................................. 2 .2 4 1 4 % A T D & D D .................................... .8 9 5 6 % A P roxy............................................. 1 .0 9 9 3 .601 (6 .8 ) 1 .1 4 5 (1 2 .7 ) .4 9 5 ( 1 1 .0 ) .411 (7 .1 ) .495 (3 .8 ) “-.1 7 7 ( -1 .3 ) .0 6 0 (0 .9 ) .0 6 5 (0 .8 ) .099 (0 .7 ) - .0 3 7 ( -0 .3 ) .0 0 0 (0 .0 ) - .0 6 6 (-0 .7 ) - .6 8 0 (-4 .2 ) .048 (0 .3 ) — .0 6 5 ( -0 .8 ) .0 2 3 ( 0 .2 ) - .2 8 2 (-1 .8 ) .0 0 4 (0 .0 ) .0 0 2 ( 0 .0 ) .0 4 6 ( 0 .4 ) .7449 .71 4 5 .8 4 1 6 .8228 .8 0 1 5 .7 779 .6241 .5 7 9 3 .2 2 4 (1 .2 ) .348 (3 .0 ) .2 7 0 ( 3 .6 ) .223 (2 .5 ) .5 1 0 (2 .4 ) - .0 8 1 ( —0 .6 ) .1 6 3 (1 .9 ) .3 1 3 (3 .1 ) .1 2 5 (0 .5 ) .1 3 8 (1 .0 ) .0 8 8 ( 1 .0 ) .0 6 4 (0 .6 ) — .6 2 4 ( -3 .5 ) - .0 6 0 (-0 .6 ) - .0 3 0 ( -0 .4 ) .0 3 8 (0 .5 ) - .2 8 5 (-1 .6 ) - .0 4 4 (-0 .4 ) .0 3 4 ( 0 .5 ) .0 7 2 (0 .9 ) .80 6 3 .70 6 3 .95 7 9 .9 3 6 2 .9 0 9 3 .8 6 2 4 .8 4 9 0 .7 7 1 0 1 .1 0 3 .8 9 1 .4 0 4 .7 0 .7 1 7 8 .6 8 4 2 .3 629 .2 8 7 0 .5 1 6 9 .4 5 9 4 .5 1 9 6 .4 6 2 4 1 .7 9 3 .1 1 1 .8 7 4 .5 0 1 .3 6 3 .0 9 1 .5 6 3 .9 7 .9103 .8 9 3 0 .6 9 7 8 .6 397 .7 3 6 4 .6 8 5 7 .6 6 3 7 .5 9 9 0 1 .5 7 1 .0 5 1 .7 4 1 .9 1 1 .0 9 1 .9 3 1 .6 1 2 .2 2 B. Based on monthly data seasonally adjusted, 1965-68 % AT R * ................... ........................................... 1123 % A D D ..................... ........................................... 0771 % AT D & D D . ........................................... 4592 % A P roxy................ ........................................... 3088 .6 9 6 (9 .1 ) .4 8 6 (4 .4 ) .329 ( 4 .3 ) .288 (3 .0 ) - .1 0 0 ( -0 .8 ) - .2 8 6 ( -1 .6 ) .071 (0 .6 ) .2 2 0 (1 .4 ) - .1 8 4 (-1 .4 ) - .1 5 2 ( -0 .8 ) .0 8 2 ( 0 .6 ) .1 0 8 (0 .7 ) .2 2 8 (1 .5 ) .351 (1 .6 ) .0 4 7 (0 .3 ) .148 (0 .8 ) .1 8 9 (1 .2 ) .2 5 6 ( 1 .2 ) .0 3 6 ( 0 .2 ) - .0 2 3 ( - 0 .1 ) C. Based on quarterly data seasonally adjusted, 1960-67 % AT R * ................... .................................. - .4 0 6 7 % A D D ..................... .................................. - .7 1 2 4 % AT D & D D ........................................... 2781 % A P roxy................ ............................................2319 .8 1 6 (9 .9 ) .6 5 5 ( 4 .4 ) .7 3 9 ( 4 .9 ) .7 3 4 ( 4 .2 ) - .1 5 7 ( -1 .4 ) - .1 5 6 ( -0 .8 ) - .3 2 9 ( -1 .6 ) - .2 1 2 ( -0 .9 ) - .0 7 3 (-0 .9 ) - .0 9 4 ( “ 0 .7 ) - .2 8 2 ( -2 .0 ) - .2 5 9 ( -1 .6 ) .4 2 4 (2 .8 ) .3 9 6 (1 .4 ) .7 8 9 (2 .8 ) .5 6 5 ( 1 .7 ) .363 ( 3 .1 ) .5 3 0 ( 2 .5 ) .5 9 8 ( 2 .8 ) .6 1 2 ( 2 .5 ) N o t e . — * sh ow s where reserve requirem ents against U .S . G o v t, dem and d ep osits have been subtracted from the reserve m easures. Standard errors o f estim ates (SEE) are a t annu al rates, “ t ” values are in parentheses. SHORT-RUN TARGETS FIGURE 2 Residuals from equations for percentage changes in DEMAND DEPOSITS D a ta not s e a s o n a l ly ad j u s te d Per c e nt ' 15 i is M o n th ly d a ta show n at annual rates. E q u a tio n s em ploy seasonal dum m ies. FIGURE 3 Residuals from equations for percentage changes in DEMAND PLUS TIME DEPOSITS Pe r ce n t .....'■'] 15 ...i M o n th ly d ata show n at an n u a l rates. E q u a tio n s em ploy seasonal dum m ies. 20 FIGURE 4 Residuals from equations fo r percentage changes in the PROXY 52 Dat a not s e a s o n a lly a d ju s te d Per ce nt ; 15 FIG U R E 5 Residuals from equations for percentage changes in DEMAND DEPOSITS Data s e a s o n a l ly ad ju ste d Per c e n t j 15 J 15 SHORT-RUN TARGETS TABLE 5: Current-Period “ Exogenous” Variable: T ota l Reserves Regressions of Monetary Aggregates on Total Reserves D ep en d en t variable C on stant % A TR %A TR -1 %A T R -2 /?2 D .W . SE E Sim ple R 2 of % ATR A. B ased on monthly data without seasonal adjustm ent, 1965-68 W ithou t season al dum m ies: % A D D ............................................... .2919 % A T D & D D .................................. . 5895 % A P roxy........................................... .4321 W ith season al dum m ies: % A D D ................................................ 2 .8 1 0 2 % A T D & D D .................................. 1.3 7 0 5 % A P roxy........................................... 1.4 1 2 8 %A D D ................................................. .2339 .811 (4 .3 ) .4 1 2 (4 .9 ) .479 (6 .8 ) -.3 2 3 ( ~ 1 .7 ) - .0 6 8 ( —0 .8 ) .0 3 6 (0 .5 ) “ .1 7 8 ( -1 .0 ) - .0 2 7 ( —0 .3 ) .0 1 6 ( 0 .2 ) .4 3 9 5 .4 0 1 3 .4381 .3 9 9 8 .5 5 1 8 .5 2 1 2 2 .3 3 1 8 .1 8 .4 0 2 2 .2 6 8 .1 3 .4 3 0 2 .2 6 6 .8 0 .5 4 9 .1 3 3 ( 1 .3 ) .158 (2 .1 ) .281 (3 .5 ) - .0 7 0 (-0 .7 ) .0 5 4 (0 .7 ) .205 (2 .5 ) .0 3 6 ( 0 .4 ) .1 1 2 ( 1 .5 ) .1 4 2 (1 .8 ) .9 4 5 2 .9 2 2 0 .8549 .7933 .8089 .7 2 7 8 1 .8 0 6 .5 6 1 .1 9 4 .7 7 1 .3 6 5 .1 2 B. Based on monthly data seasonally adjusted, 1965-68 % A T D & D D .................................. .4 8 3 4 % A P roxy........................................... .2 5 4 8 .325 (2 .6 ) .3 7 5 ( 4 .5 ) .6 6 4 (9 .4 ) - .0 7 5 ( -0 .6 ) .0 4 4 ( 0 .5 ) .1 5 8 (2 .2 ) .083 ( 0 .7 ) .0 9 5 ( 1 .2 ) .0 2 6 ( 0 .4 ) .1 4 9 4 .0 9 1 4 .3 9 1 0 .3 4 9 5 .7311 .7 1 2 8 1 .7 8 5 .0 7 1 .3 7 1 .5 0 3 .3 9 .3 6 2 1 .8 5 2 .9 0 .6 9 7 N o t e . — Standard errors o f estim ate (SEE) are a t annual rates, “ t" values in parentheses. TABLE 6: Current-Period “Exogenous” Variable: T otal R eserves A d ju ste d f o r R eserves R equ ired A gain st U.St G o vern m en t D eposits Regressions of Monetary Aggregates on Total Reserves Adjusted for Reserves Required Against U.S. Government Deposits C on stant D ep en d en t variable % A TR* %A T R -1 * % A T R -2 * R* Ri D .W . SE E Sim ple R 2 of % ATR A. Based on monthly data without seasonal adjustm ent, 19 6 5 -6 8 W ith ou t season al dum m ies: % A D D .................................................. .02 4 3 % A T D & D D .................................... .4 8 8 9 % A P roxy ............................................. .4 6 6 5 W ith season al dum m ies: % A D D .................................................. 2 .5 2 4 5 % A T D & D D ..................................... 1 .1 8 2 4 1 .3 5 9 2 .9 9 8 (7 .4 ) .4 8 0 ( 8 .2 ) .3 7 3 (5 .3 ) - .0 9 3 (-0 .7 ) .0 1 6 ( 0 .3 ) .0 7 0 (1 .0 ) - .0 4 3 (-0 .3 ) .0 2 0 ( 0 .3 ) - .0 1 2 ( -0 .2 ) .6513 .6 2 7 6 .6 6 4 0 .6411 .4 4 0 7 .4 0 2 6 .2 9 5 (3 .1 ) .2 4 4 (3 .5 ) .233 (2 .7 ) .0 6 8 (0 .7 ) .1 3 5 ( 1 .8 ) .2 3 7 ( 2 .6 ) .079 (0 .8 ) .1 3 7 ( 2 .0 ) .1 7 0 ( 2 .0 ) .9 5 4 2 .9 3 4 8 .8 7 8 0 .82 6 3 .7841 .6925 2 .3 9 1 4 .3 4 .648 2 .2 7 6 .2 9 .6 6 3 2 .1 9 7 .6 0 .4 2 3 2 .0 4 6 .0 0 1 .3 8 4 .3 7 1 .3 8 5 .4 5 B. Based on m onthly data seasonally adjusted, 1965-68 % A D D ...................................... ...........................0428 % A T D & D D ......................... ...........................3564 ...........................1912 .5 9 7 ( 5 .2 ) .5 3 6 (7 .3 ) .521 ( 5 .0 ) - .0 1 8 ( -0 .1 ) .0 9 6 ( 1 .3 ) .3 5 4 ( 3 .3 ) .1 1 3 ( 1 .0 ) .1 1 0 ( 1 .5 ) .0 7 2 ( 0 .7 ) .42 0 7 .3 8 1 2 .6 2 8 7 .6 0 3 4 .5 4 7 6 .5168 1 .9 6 4 .1 9 .4 0 8 1 .5 8 2 .6 5 .588 1 .6 4 3 .7 7 .4 1 4 * W here reserve requirem ents against U .S . G overn m en t d em and d ep osits have been subtracted from the reserve m easures. o t e . — Standard errors o f estim ates (SEE) are a t annu al rates, "t" values in parentheses. N what is often considered the range between prudent extremes of “tight” and “easy” mone tary policy. IMPLICATIONS OF THE REGRESSION RESULTS. First of all, the results provide no support whatever for the view expressed, for example, by Allan Meltzer that “month-tomonth changes in money .. . can be kept within a very narrow range” .3 His conclusion is ap parently based on an examination of equations relating monthly dollar changes in M 1 and M 2 to current and one-period lagged changes in the monetary base and Treasury deposits at com mercial banks. The R 2’s of Meltzer’s equations range from 0.70 to 0.86, thereby leading to his optimistic view of the prospects for control. As the results for similar equations presented above show, however, R 2's may indeed be high, especially when unadjusted data are used and seasonal dummies are included. Nevertheless, the standard errors remain quite large. Even if the standard errors had been substantially smaller, moreover, Meltzer’s conclusion would still remain subject to the reservations regard ing this type of regression equation noted on page 48. While the equations do suggest that the Desk would be likely to miss rather badly any given deposit or credit target in any given month if it simply used one of these equations to guide its actions, the results do not neces sarily mean that there is no way of hitting such targets with tolerable accuracy on a monthly basis. First, better equations could probably be devised with further experimentation. More variables exogenous to the banking sector could be included. More complex lag struc tures could be investigated. Systematic allow ance for autocorrelation could be made, and so forth. Second, informed judgment by the Desk might yield better ways of determining appropriate growth rates for nonborrowed re serves than any equation could provide. Third, even if such equations were relied on fairly 3 “Controlling Money,” Federal Reserve Bank of St. Louis Review (May 1969), pp. 18 and 19. mechanically, there can be little doubt that re sults would be improved by making midmonth adjustments in the targeted behavior of non borrowed reserves for the balance of the month in light of any misses in the deposit or credit target occurring earlier in the month. Finally, even if there proved to be no way of hitting such targets as the monetary growth rate with tolerable accuracy on a monthly basis, this does not mean that such targets could not be hit with acceptable accuracy on the average over a span of months. The opera tion of a strategy with a multimonth horizon might well require the FOMC to review de viations of actual results over the preceding period from targeted values presented to the Account Manager at the previous meeting. While the operation of a multimonth strategy is beyond the scope of this paper, some equa tions using quarterly data but otherwise similar to the ones discussed above are presented in Part C of Tables 3 and 4 (pp. 49 and 50). Here the equations were estimated on percent age changes in seasonally adjusted quarterlyaverage values of the variables for the period 1960-67. The standard errors in the deposit and credit proxy equations are all about 2 per cent (annual rate), much lower than for the monthly equations, as expected.4 For the seven quarters beyond the sample period (that is, all of 1968 and the first three quarters of 1969), the average absolute prediction error for de mand deposits was a 2 . 1 per cent annual rate if no allowance were made for Treasury de posits. Adjusting for behavior of reserves re quired behind Treasury deposits actually wors 4 If quarterly percentage changes in total reserves are used, there is only a negligible reduction in the standard error for the two deposit measures. If total reserves are used with adjustments for reserves re quired behind Treasury deposits, the standard errors for the two deposit totals drop moderately to about a 1.6 per cent annual rate. Again warning is made of the implicit assumption that total reserves less re serves required behind Treasury deposits can be per fectly controlled. Indeed, as Part C of Table 3 shows, nonborrowed reserves can only be used to control total reserves on a quarterly basis up to a standard error of 1.2 per cent (annual rate). SHORT-RUN TARGETS F IG U R E 6 Percentage changes in DEMAND DEPOSITS Per c e n t Da ta sea s o n a lly a d j u s t e d 15 * B ased on d ata in T a b le 3, P a rt C. Q u arterly d ata show n at an n u a l rates. ened the results slightly, giving an average prediction error of 2.3 per cent (annual rate) in the seven-quarter period. Results beyond the sample period for equa tions for time and private demand deposits combined were less satisfactory, with the equa tion consistently and substantially overstating the growth rate of the deposit total. The aver age absolute prediction error for these equa tions amounted to a 4.5 per cent annual rate. Predicted and actual figures for the sample pe riod and for the seven-quarter extrapolation period for demand deposits and for demand deposits plus time deposits are shown in Fig ures 6 and 7 It is possible to take either an optimistic or a pessimistic view of the results using quarterly data. One could say that the standard errors of around 2 per cent per annum for the sample period and of about that size in the case of de mand deposits in the seven-quarter extrapola tion period are not large and could be substan 5 Prediction errors beyond the sam ple period for equations using current changes in total reserves were about the same as for the equations using cu r rent changes in nonborrow ed reserves. tially reduced in practice by means of midcourse corrections. The poor results for de mand plus time deposits in the 1968-69 pe riod could be dismissed as simply failing to allow for the profound effects of Regulation Q. Such a deficiency, it might be argued, could easily have been overcome in practice. The pessimistic view, however, would be that all the equations show a number of quarters both within and without the sample period where errors amount to a 2 per cent annual rate or more and that, for percentage changes in quar terly average levels, this is simply not a very good performance in a world where an annual rate of 2 to 6 per cent is accepted by many as defining the limits of prudent policy. Probably a sensible conclusion would be somewhat as follows: (1) There is no existing evidence to demonstrate the possibility of tight control over monetary and credit growth rates — even over quarterly-average periods and even if such control is sought relentlessly to the exclusion of other possible considerations. (2) Nevertheless, existing evidence does give reasonable grounds for hope that such control would in fact be possible over quarterly peri ods if midquarter corrections and the use of FIGURE 7 Percentage changes in TIME plus DEMAND DEPOSITS 1 B ased on d a ta in T a b le 3, P art C. Q uarterly d a ta show n at an n u al rates. judgment can be brought in to substantial ad vantage. The proposition that the Federal Reserve could control monetary growth rates or the credit proxy with tolerable accuracy on a quarterly-average basis if it sought to do so without regard to any other possible constraints on its behavior should perhaps not seem terri bly controversial. The problem of just what sacrifices might in fact have to be made as regards the money market effects of such a pur suit is discussed in the next section. THE EFFECTS OF QUANTITY TARGETS ON MONEY MARKET STABILITY THE NATURE OF THE PROBLEM. Every sort of evidence suggests that the FOMC has relied largely on money market targets in the period from the TreasuryFederal Reserve accord through the late 1960’s. These targets included free reserves, borrowed reserves, the Federal funds rate, the rate on 3-month Treasury bills, and other rates. The emphasis on these various measures has no doubt fluctuated over time. In most cases, the actual target has probably represented essentially a weighted average of these varia bles, with the weights never quantified or spelled out— though defined to some extent in the discussion at the Committee’s meetings. These variables tend to be reasonably collinear in any given period. Thus it is possible, for example, to spell out levels of free reserves that would be compatible with certain levels of borrowed reserves, the Federal funds rate, and the bill rate. Stabilizing one or more of these measures in effect fixes the “tone” of the money market— since that elusive concept ap pears to be essentially coterminous with a weighted average of these variables. The single major exception to the complete dominance of money market targets has apparently been the use of the proviso clause.'1 In obedience to this clause, money market targets have been al tered between FOMC meetings on several occasions when the bank credit proxy has 6 See pages 64 and 65. SHORT-RUN TARGETS deviated substantially from its projected growth rate. A basic feature of money market targets is that their use requires accommodation of fluc tuations in levels of required reserves during the period within which the target variable is to be stabilized at a given value. Since fluctua tions in required reserves ultimately reflect fluc tuations in the demand for bank credit and de posits at the level of money market rates associated with the given money market target, a money market target basically means accom modating fluctuations in the demand for credit and deposits. This is true whether the fluctua tions are seasonal or trivial random move ments or whether they are related to changes in business conditions or to shifts in the underlying structural demand equations for bank credit and deposits. To argue that the use of money market tar gets necessarily involves accommodation of de mand shifts is not to argue that the System is wholly passive with respect to quantities when ever it employs money market targets. The level at which the target is set, ceteris paribus, does influence the rate of growth of the aggre gates for reasons developed by many writers both before, but especially after, James Meigs’ well-known treatment of the problem. 7 Despite this influence, however, the essential points are ( 1 ) that once a money market target has been set, the System reacts essentially passively to shifts in demands as long as the target is un changed, and ( 2 ) that the ex-post statistical relationship between levels of money market targets and rates of growth of reserves, money, bank credit, and so forth have tended to be extremely weak. This point has also been rather well established by many writers from Meigs on. As a result of these considerations, many have argued that the use of money mar ket targets has in practice deprived the System 7 See references to the literature cited in Thomas Mayer, "Monetary Policy in the United States** (New York: Random House, 1968) Chapter 3, pp. 79-109. of any effective means of controlling aggre gates. Defenders of money market targets have argued that the System should accommodate, at least in the short run, most shifts in the demand for bank credit and deposits since such shifts are predominantly seasonal and temporary in nature. Failure to accommodate them, according to this view, would simply produce economically undesirable fluctuations in money market conditions. Whatever the merits of these arguments, it remains true that just as stabilizing money market conditions involves the accommodation of fluctuations in demand, so would the use of quantity targets involve fluctuations in money market conditions. If you wish to stabilize the price of any good, the amount you supply will reflect fluctuations in the demand schedule for the good; conversely, if you wish to stabilize the amount you supply, you must allow fluctu ations in demand to be reflected in price fluc tuations. A major question therefore is how much money market instability would be pro duced by attempting to follow quantity targets and how serious a problem would such in stability turn out to be. Unfortunately, there appears to be no way of providing confident answers to these ques tions in advance. Experience would have to tell the story. In principle, a correctly specified money market model could be used to simulate the interest rate effects of any given rate of growth of nonborrowed reserves. Unfortunately, no suitable model exists and the construction of such a model would undoubtedly be a major task. Even if a suitable model did exist, more over, the answers it would grind out might have relevance only for a short time. It seems reasonably likely that money market institu tions themselves would evolve under the pres sure of changed conditions, and that the ulti mate impact on money market rates of ceas ing to accommodate demand shifts would be different from the initial impact. Suppose, for example, that the System re placed money market targets with a target stated in terms of week-by-week stability in the rate of growth of nonborrowed reserves, not seasonally adjusted. All our experience suggests that sharp week-to-week fluctuations in demand for bank credit and deposits as well as marked seasonal patterns in these de mands would lead to erratic and large move ments in the Federal funds rate and related rates and would substantially magnify any sea sonal patterns that may exist in money market rates. How far out on the maturity spectrum the rate fluctuations introduced by such a change in the System’s modus operandi would extend is a question. At the very short end of the market, there is every reason to believe that conditions would be substantially different from what they are today. After banks and other financial institutions began to acquire some experience with the new environment, however, they might well discover ways of adapting to it that would themselves tend to dampen rate instability in the market. For ex ample, by borrowing ahead, borrowers could avoid having to pay very high rates at periods of seasonal tension. This would tend to diffuse rate pressures over time. Similarly, lenders could take advantage of seasonal rate pres sures by building up adequate loanable funds in advance, and this too would tend to diffuse rate pressures. In general, institutions could be expected to learn to respond more flexibly to take advan tage of rate fluctuations—thus increasing the supply elasticity of funds and thereby dampen ing the fluctuations themselves. Banks would probably want to keep stronger average basic reserve positions, with an increased willingness to buy or sell Federal funds depending upon rate conditions. While the precise nature of these institutional adaptations cannot be fore seen, it seems clear that some developments along these general lines would occur, tending to dampen random and seasonal fluctuations in money market rates. AN EXPERIMENT. While, as noted above, only the simulation of a complete money market model could, even in principle, give an accurate indication of the kind of money mar ket instability that would be created in the short run by quantity targets, the following cruder procedure may give some rough insight into the dimensions of the problem. In general, the method used here consists of computing the weekly levels of free reserves as they would have been in a particular historical pe riod if the System had provided a constant week-by-week growth in nonborrowed reserves in that period, given the historical pattern of actual changes in required reserves. An equa tion relating the Federal funds rate to free re serves and the discount rate is then used to es timate what the funds rate would have been had the System followed the quantity target. The computed rate is then compared with the actual pattern of the rate for the period. The period for the initial test covered July, August, and September 1967, a period in which policy was unchanged and in which the summary money market measures remained quite stable. For example, the average Federal funds rate in those 3 months was 3.79, 3.90, and 4.00 per cent, respectively; borrowings av eraged $87 million, $89 million, and $90 mil lion, respectively; and free reserves averaged $272 million, $298 million, and $268 million, respectively. Between the week of June 28 and the week of September 27, nonborrowed re serves rose by a total of $732 million, not sea sonally adjusted. This increase occurred in an irregular fashion, however, since week-to-week fluctuations in such reserves roughly matched week-to-week fluctuations in required reserves as the Desk went about the business of stabilizing money market conditions. Let us suppose, contrary to fact, that the System had produced the $732 million in crease in nonborrowed reserves that occurred in this period through steady, equal weekly in crements of about $56 million. Let us also suppose, however, that week-to-week changes in required reserves under this hypothetical sit uation would have been the same as they in fact were during the period. In Table 7 the re sulting hypothetical weekly levels of free re serves are compared with the levels that ac- SHORT-RUN TARGETS TABLE 7: Actual and Hypothetical Reserve Measures N on b orrow ed reserves W eek A ctual 1967— June 28 H ypo thetical 1 Free reserves R equired ■ reserves H y p o (actual) thetical (2 -3 ) A ctual (1) (2) (3) (4) (5) 2 3 ,4 0 6 2 3 ,4 0 6 2 3 ,1 1 6 290 290 July 5 1 2 . .. . . 19 26 23,531 2 3 ,9 9 7 2 3 ,8 3 8 2 3 ,9 6 7 2 3 ,4 6 2 2 3 ,5 1 9 2 3 ,5 7 5 2 3 ,631 2 3 ,4 2 2 2 3 ,4 2 3 2 3 ,6 5 3 2 3 ,5 8 9 40 96 -7 8 42 109 574 185 378 A u g. 2 ... 9 16 23 30 2 3 ,8 5 8 2 3 ,8 6 9 2 3 ,6 3 4 2 3 ,7 2 6 2 3 ,4 2 9 2 3 ,6 8 7 2 3 ,7 4 3 2 3 .8 0 0 2 3 ,8 5 5 2 3 ,9 1 2 2 3 ,6 7 9 2 3 ,5 8 9 2 3 ,3 8 1 2 3 ,3 0 0 2 3 ,2 1 5 8 154 418 555 697 179 280 253 426 214 Sept. 6 13 20 2 7 , 2 3 .8 4 6 2 3 ,9 6 9 2 4 ,2 1 0 2 4 ,1 3 8 2 3 ,9 6 9 2 4 ,0 2 5 24,081 2 4 ,1 3 8 2 3 ,5 9 3 2 3 ,6 5 3 2 3 ,9 0 8 24 ,0 0 1 376 372 173 137 253 316 302 137 1 C om puted by dividing actu al change in nonborrow ed reserves between w eeks o f June 28 and Sept. 27 into 13 equal weekly increm ents. tually prevailed. The actual level of free reserves fluctuated, for the most part, reasona bly narrowly around the average for the pe riod. Actual week-to-week fluctuations presuma bly reflected not only misses by the Desk due to misallowance for operating factors, but also deliberate changes reflecting allowances by the Desk for shifts in the distribution of reserves within the banking system and other familiar considerations. Compared with the actual course of free re serves, the hypothetical level shows a distinct time path (Figure 8 ). Thus the hypothetical level (based on constant increments in unad justed nonborrowed reserves) rises strongly during August, reflecting the seasonal weak ness in required reserves, and thereafter de clines as the September tax date puts upward pressure on required reserves, and, in this hy pothetical world where such seasonal pressures are not accommodated, on money market con ditions as well. It can be validly argued, of course, that the assumption on which this exercise rests— namely, that weekly movements in required re serves would be the same in the hypothetical situation as in the actual situation—is false, at least to some degree. Presumably the growing money market ease through late August pic tured in the hypothetical situation would bring forth greater credit and deposit demands and hence larger required reserves than actually occurred, with the reverse process occurring as stringencies developed in September. Since such developments would no doubt have oc curred to some extent, the seasonal movement in free reserves generated by the hypothetical example has to be regarded as defining the outer limits of the possible effects on the money market of a policy of rigid weekly in crements in nonborrowed reserves during the period. The true pattern of free reserves under such a policy would no doubt show a some what milder seasonal pattern. Furthermore, the experiment was conducted assuming constant weekly increments in nonborrowed reserves FIGURE 8 FREE RESERVES, JUNE 28-S EP T. 27, 1 9 6 7 Millions of dollars *........................................*....................... H yp oth etical, based on constant increm ents in: 28 JUNE 5 12 19 JULY 26 2 9 16 AUGUST 23 30 6 13 20 -i iooo 27 SEPTEMBER * C o m p u ta tio n s are the sam e as th o se d escrib ed in th e tex t, e x c e p t that the D e s k is a ssu m ed to su p p ly a c o n sta n t in cre m ent in se a so n a lly a d ju s te d n o n b o rr o w ed reserv es o v er the p erio d (a s d eterm in ed fro m w eek ly s e a so n a l f a c t o r s ) . D a ta are w e e k ly a v e r a g es o f d a ily figures. without seasonal adjustment. Simple transla tion to a rule of constant increments in season ally adjusted nonborrowed reserves would tend to smooth the fluctuations in free reserves generated by the use of a strict quantity target. 8 In any case, the effects on money market rates of the hypothetical policy of nonaccom modation can be estimated with the aid of an equation relating the Federal funds rate to the level of free reserves and the discount rate. One such equation, estimated on biweekly re 8 Some experiments along these lines suggest that the smoothing effects would be quite substantial. See Figures 8-11. serve averaging periods from mid-1966 to mid-1968 is shown below (“t” values in par entheses). rff = 2.30 *- m 2 R f + M 9rd (12.0) (5.2) TABLE 8: Derivation of Hypothetical Federal Funds Rate 1967— June 2 8 . . . H ypoth etical com p uted Federal fu n d s rate A ctual Federal fu n d s rate 0 4 .0 7 4 .0 7 July 5 ... 1 2 . .. 1 9 . .. 2 6 ... + + + + .1 3 8 .9 5 6 . 526 .6 7 2 3 .7 3 3 .9 8 3 .5 4 3 .9 3 3 .8 7 4 .9 4 4 .0 7 4 .6 0 A u g. 2 ... 9 ... 1 6 . .. 2 3 ... 3 0 . .. + .3 4 2 + .2 5 2 - .3 3 0 - .2 5 8 -.9 6 6 3 .7 5 4 .0 2 4 .0 5 3 .9 8 3 .5 9 4 .0 9 4 .2 7 3 .7 2 3 .7 2 2 .6 2 Sept. 6 ___ 1 3 . .. 2 0 ... 2 7 ..., - .2 4 2 - .1 1 2 + .2 5 8 0 4 .0 2 3 .9 8 4 .0 0 4 .0 0 3 .7 8 3 .8 7 4 .2 6 4 .0 0 R 2 = .915, R 2 = .912, SEE = .22 percentage point This equation can be regarded as a reduced form, derivable from the simple model pre sented on page 42 if the discount rate is in cluded, as it should be, in the demand equa tions for excess and borrowed reserves. The free reserves variable is of course nonborrowed reserves minus the predetermined level of re quired reserves and may itself be determined as an exogenous policy variable, deliberately fixed (subject to random errors) by the Desk. If this equation were used directly to com pute the hypothetical Federal funds rate, the results would differ from the actual rate not only because of differences between hypotheti cal and actual levels of free reserves, but also because of the error term in the equation. To avoid this muddying of the waters, the hypo thetical funds rate was computed, instead, by obtaining the difference between hypothetical and actual free reserves in each week from Table 7, multiplying this difference by the free reserves coefficient in the equation ( — .0 0 2 ), and adding the result to the actual level of the funds rate. The resulting hypothetical time path of the Federal funds rate is shown in Table 8 . Both the actual and hypothetical funds rates are shown in Figure 10. The range of the funds rate under the hypo thetical program of steady increments in non borrowed reserves is, of course, much larger than the range under a regime of accommodat ing fluctuations in required reserves. Thus in the hypothetical case, the funds rate ranges from a high of 4.94 per cent in the week of July 1 2 to a low of 2.62 per cent in the week of August 30. In fact, the rate ranged from a high of only 4.07 per cent in the week of June 28 to a low of only 3.54 per cent in the week H yp oth etical free reserves less actu al free reserves tim es ( —.002) W eek of July 19. Thus the hypothetical spread was 232 basis points, as compared with an actual spread of only 53 basis points. Computed week-to-week fluctuations in the funds rate were also substantially larger under a policy of constant weekly increments in non borrowed reserves than they were in fact dur ing the period. Thus the average absolute weekly change in the level of the funds rate was 0 . 2 2 percentage point. The computed av erage weekly change was almost 2 Vi times as large, or 0.53 percentage point. Roughly similar results, as to effects on both the range of the funds rate and the average size of its week-to-week fluctuations were ob tained for each of the three subsequent 3 FIGURE 9 FREE RESERVES, SEPT. 27-D EC . 27, 1 9 6 7 ^ Millions of dollars I H yp oth etical, based on co n sta n t increm ents in: •• ^ Seasonally adjusted nonborrowed reserves* Unadjusted nottborrowed reserves ' r- SV i a \y \ / Actual ; 800 ; BOO i 400 200 + 0 27 SEPT. 4 11 18 OCTOBER 25 ' 200 8 15 22 NOVEMBER 29 6 13 20 27 DECEMBER C o m p u ta tio n s are th e sa m e a s th o se d e sc rib e d in th e te x t. e x c e p t th a t th e D e s k is a ssu m e d to su p p ly a c o n s ta n t in c re m en t in s e a s o n a lly a d ju s te d n o n b o rr o w ed r eserv es o v e r the p e r io d ( a s d e ter m in ed fr o m w e e k ly s e a so n a l f a c to r s ) D a ta a re w e e k ly a v e r a g es o f d a ily figures. SHORT-RUN TARGETS month periods. The technique for computing the funds rate was exactly the same as was used for the period June-September just de scribed. In each period, the assumed week-toweek increase in nonborrowed reserves was the average actual weekly increase from the first week of the period to the last week of the period. The results are summarized below in Table 9. EVALUATION. The experiment reported above is hardly a sufficient basis for judging the amount of money market instability that might be associated with rigid adherence to a quantity target. In two respects at least, it seems to overstate the likely degree of instabil ity. First, as noted earlier, it assumes that re quired reserves would not respond at all to the effects on free reserves and the funds rate of pumping in a constant increment of nonbor rowed reserves week by week. Actually, there would certainly be at least some response, and it would be in a stabilizing direction. Required reserves would tend to weaken under the pres sure of tight money market conditions and to strengthen under the encouragement of easy money market conditions, thereby themselves tending to modify the extremes of tightness and ease in the money market. Secondly, and also as noted earlier, the pro vision of a constant increment of seasonally adjusted nonborrowed reserves would certainly produce substantially milder seasonal move ments in money market conditions than would the provision of constant increments of zmseasonally adjusted reserves. A comparison of hypothetical paths for free reserves and for the Federal funds rate using equal seasonally adjusted increments with hypothetical paths using equal unadjusted increments for two FIGURE 10 FEDERAL FUNDS RATE, JUNE 28-SEPT. 27, 1967 Per cent I . : :■ H yp oth etical, based on co nstant increm ents m: ; S e a s o n a l ly a d j u t U d n o n b o r r o w e d 28 5 JUNE 12 19 26 JULY 2 9 16 AUGUST 23 re»«rv«s * 30 6 5.0 13 20 27 SEPTEMBER * C o m p u ta tio n s are the sa m e a s th o se d escrib ed in th e text, e x c e p t that the D e s k is a ssu m ed to su p p ly a c o n sta n t in cre m ent in s e a so n a lly a d ju s te d n o n b o rr o w ed reserves ov er the p erio d ( a s determ in ed fro m w e e k ly se a so n a l f a c t o r s ). D a ta are w eek ly a v e r a g es o f d a ily figures. different 3-month periods is shown in Figures 8-11. Of course, the implementation of a sea sonally adjusted nonborrowed reserve target would raise the thorny technical problem of developing satisfactory weekly seasonal adjust ment factors. Having said that certain features of the ex periment tend to overstate the degree of poten tial money market instability, however, the writer is inclined to the view that the degree of instability indicated is nevertheless rather sur prisingly mild. As Table 9 shows, the com puted average absolute weekly change in the Federal funds rate tends to be only around 50 basis points, certainly substantially larger than the average changes that actually occurred (around 17 basis points), but not more than the market would seem able to handle without undue stress. Similarly, the computed ranges of TABLE 9: Federal Funds Rate, Selected Periods A ctual values Period June 28 to Sept. Sept. 27 to D ec. D ec. 27 to M ar. M ar. 27 to June 2 7 ............................... 2 7 ............................... 2 7 ............................... 2 6 ............................... H igh L ow R ange 4 .0 7 4 .6 3 5 .4 0 6 .3 4 3 .5 4 3 .5 0 4 .5 5 5 .4 0 .5 3 1 .1 3 .8 5 .9 4 C om puted values Average absolute w eekly change 0 .2 2 0 .1 5 0 .1 3 0 .1 6 H igh Low R ange 4 .9 4 4 .8 3 5 .9 2 6 .5 7 2 .6 2 3 .2 5 4 .4 9 4 .5 6 2 .3 2 1 .5 8 1 .4 3 2 .0 1 Average a bsolute weekly change 0 .5 3 0 .5 1 0 .6 9 0 .4 3 61 FIGURE 11 FEDERAL FUNDS RATE, SEPT. 27-DEC. 27, 1967 Per c e n t " ..... '■ " " .......... H y p o th e tic a l, b a s e d on c o n s ta n t increments in : as one goes further back in time, because of the lesser importance of the funds market. 5.5 S e a s o n a lf y a d j u s t e d n o n b o r r o w e d r e s e r v e * * SOME M IXED STRATEGIES— BLENDING MONEY M ARKET AND QUANTITY CONSIDERATIONS IN FRAM ING TARGETS . . . 27 SEPT. 4 II . . , 18 OCTOBER 25 1 8 15 22 NOVEMBER 29 0 6 13 20 27 DECEMBER * C om p u tatio n s are the sam e as those described in the text, except that the D esk is assum ed to supply a co n stan t in c re m ent in seasonally adjusted nonborrow ed reserves over the period (as determ ined fro m weekly seasonal fa c to rs). D ata are weekly averages of daily figures. the funds rate over the periods tested, about 150 to 225 basis points, were substantially larger than the ranges that actually occurred, about 50 to 115 basis points, but again, seem ingly not beyond the limits of manageability considering that the funds rate is a 1-day rate. Moreover, as Figures 8-11 suggest, these ranges could well be narrowed considerably by even a crude allowance for seasonal fluctua tions in reserve demands associated with sea sonal fluctuations in required reserves. Perhaps the apparent mildness of the money market's reaction to a quantity target as indi cated in this experiment ought to be regarded with some degree of skepticism. One factor upon which the reasonableness of the calcula tions depends is, of course, the estimate of the coefficient of free reserves in the equation for the funds rate, —.002 in this case. How stable is this coefficient? How indicative is the —.002 estimate of what might be expected in the future? A similar equation covering a pe riod 18 months earlier, the beginning of 1965 to the end of 1966, gives a very similar result (the coefficient of free reserves is — .0018). For pre-1965 periods, the computed coefficient tends to be much smaller, but the relevance of results before 1965 is questionable because of the market convention that the funds rate would never go above the discount rate and, Having examined some features of monetary aggregates as FOMC targets, it now seems use ful to sketch some procedures through which improved control over these aggregates might be reconciled with the desire to moderate fluc tuations in the tone of the money market. These procedures involve “mixed strategies” in which both the monetary aggregates and meas ures of money market conditions have a spe cific role to play. Before looking at these mixed strategies, however, some salient features of pure money market and pure quantity strate gies are reviewed. PURE MONEY MARKET STRATEGIES. In a pure money market strategy, the Manager can be instructed to maintain marginal reserve measures at a certain level or within a certain range, or he can be instructed to hold money market rates, in recent years especially the Federal funds rate, at a certain level or range. The normal practice, as noted earlier, has been to use a somewhat vaguely defined blend of these two approaches. If the banking system’s aggregate demand schedules for excess and borrowed reserves remain stable,9 rates such as the funds rate and the marginal reserve measures will move fairly closely in step with each other. Hence it will make little difference whether the Committee’s instructions em phasize the marginal reserve measures or short term interest rates. Stabilizing free or borrowed reserves at some target level will effectively stabilize ihe funds rate, and vice versa. In fact, the aggregate demand schedules of the banking system for excess and borrowed 9 T hat is, the dem and schedu le defined w ith respect to the level o f m oney m arket interest rates. SHORT-RUN TARGETS reserves evidently show a fair degree of shiftability. This is due in large measure to shifts in the distribution of reserves between groups of banks with very different individual demand schedules. The country-city bank shift is the one most often cited. However, not all causes of shifts in the banking system’s aggregate demand schedules need be as shortlived as those related to shifts in the distribution of reserves generally are. For example, shifts in bankers’ expectations about future reserve needs may shift their demand schedules for excess and borrowed reserves at crucial junctures. Thus it has been argued that the demand for free reserves shifted to the right in the sum mer of 1966 as the risk of a huge September run-off in CD’s began to seem more real to bankers. According to this argument, banks sought both to build up excess reserves and to reduce borrowings so as to strengthen their claim to future discount-window accommo dation when the period of peak strain actually arrived. Whatever the causes of shifts in the demand schedules for excess and borrowed re serves, such shifts drive a wedge between short-term rates, such as the funds rate, and aggregate levels of free and borrowed reserves. As a result, it does make some difference whether a pure money market strategy fo cuses primarily on stabilizing marginal reserve measures or primarily on the Federal funds and related rates. If attention is focused on interest rates, the Desk will find itself accommodating not only shifts in the demand for deposits and bank credit, a feature of all pure money market strategies, but also shifts in the banking sys tem’s aggregate demand for free reserves. If the Desk is instructed to hold the Federal funds rate at around x per cent, it must re spond equally to both types of shifts. A surge in the demand for deposits (bank credit) at current interest rates will expand the amount of required reserves, and in order to keep the funds rate from rising the Desk will have to supply nonborrowed reserves. Similarly, an at tempt on the part of banks to build up excess reserves or to repay borrowings and thus clear the books for subsequent borrowings will also tend to push up the funds rate. Again the Desk will have to supply enough nonborrowed reserves to keep the funds rate from rising. On the other hand, if the money market strategy is framed exclusively in terms of the marginal reserve measures, only shifts in the public’s demand for deposits and credit will be accommodated. A rise in the level of free re serves desired by the banking system at given interest rates will not be met by the Desk with a corresponding increase in nonborrowed re serves. As a result, actual free reserves remain unchanged, consistent with their targeted be havior, while interest rates rise, and, ceteris paribus, rates of growth of the deposit and bank credit aggregates tend to decline. Since it is difficult to see what policy purpose is served by the interest rate and deposit/credit effects of unexpected shifts in the aggregate demand schedule for free reserves, the interest rate variant of the money market strategy seems to have advantages over versions relying on mar ginal reserve measures. Clearly, the use of such reserve measures can produce wholly unin tended tightening or easing both of money mar ket rates and of the aggregates when demand schedules for these reserves shift. 10 PURE QUANTITY STRATEGIES. At the opposite extreme of the pure money market strategy, whether in its free reserves or Federal 10 A formal elaboration of the implications of in terest rate targets for the behavior of marginal re serve measures and the growth of aggregates and of the implications of marginal reserve targets for the behavior of interest rates and the growth of aggre gates is given in Richard G. Davis, “Open Market Operations, Interest Rates, and Deposit Growth,” Quarterly Journal of Economics, (Aug. 1965), pp. 433-42. Occasions in which shifts out in the demand schedule for free reserves may have resulted in inad vertent tightening are examined in two articles by Jack M. Guttentag. “The Strategy of Open Market Operations,’* Quarterly Journal of Economics, (Feb. 1966), pp. 1-30, discusses such shifts the first half of 1960, while “Defensive and Dynamic Open Market Operations, Discounting, and the Federal Reserve System’s Crisis-Prevention Responsibilities,” Journal of Finance, (May 1969), pp. 249-63, dis cusses the summer of 1966. funds rate variant, is the pure quantity strat egy. This strategy involves exclusive use of rates of growth in some monetary aggregates as targets to be pursued without any regard for the resulting effects on money market condi tions. The main features of this approach have been discussed in earlier sections and need only be recapitulated very briefly here. A de sired growth rate in bank credit, M u M s, or some other aggregate would be picked by the FOMC for the month ahead. Conceivably the Committee itself, or more likely the staff, would translate this target into a monthly rate of growth in nonborrowed reserves and, ulti mately, into week-by-week targets for such re serves. The pure quantity target need not, of course, involve anything so crude as a constant week-by-week increment in nonseasonally ad justed nonborrowed reserves during the period between FOMC meetings. Indeed it almost certainly would not. Seasonally adjusted data might be used, for example, and, given the in evitable “misses” and the existence of some kind of “error response” mechanism as de scribed in an earlier section, the pure quantity target would in fact probably involve week-toweek changes in the nonborrowed reserve in crements sought by the Manager. By defini tion, however, these changes would never be chosen in light of their impact on money mar ket conditions, but solely in terms of their ap propriateness for hitting a monthly target for M u bank credit, or whatever variable the Committee has in mind. Clearly the result would be larger fluctuations of a short-term and perhaps medium-term nature than pres ently exist in free reserves, the Federal funds rate, and other measures of money market conditions. While something very akin to the pure money market target was used over a period of many years, it came under increasingly heavy criticism. The pure quantity target, on the other hand, has never been tried. In view of the possible risks posed by the pure quantity target to the money and capital markets—most of them risks of essentially unknown and per haps unknowable magnitude—many would no doubt argue that a pure quantity target should not be tried. In these circumstances, the mid dle ground between pure money market and pure quantity targets is of considerable inter est. Such a middle ground would hopefully contain approaches that would retain some so licitude for money market conditions while providing a real measure of control over the monetary aggregates. THE CREDIT PROXY PROVISO CLAUSE. The first operational result, insofar as open market strategies are concerned, of the increased concern within the System over the behavior of monetary aggregates was apparently the “proviso” clause. The inclusion of such a clause in the Committee’s regular directive to the Account Manager represents, however, only the most cautious of steps outside the familiar world of the pure money market target. As it has been used, the proviso clause requires the Manager to shift the money market targets in the appropriate offsetting direction if growth in the bank credit proxy is deviating significantly from the figures projected at the time of the FOMC meeting. There have, of course, been many doubts and criticisms raised in connection with the proviso clause. Some would prefer to substitute other variables for the credit proxy as being more economically meaningful. Others feel that the proviso clause has in practice proved too vague to give the Manager sufficient guid ance. Thus there are always uncertainties as to just when deviations in the proxy from projec tions become substantial enough to require modification of the money market targets and uncertainties as to how large any such modifi cations should be. From the point of view of the present discussion, however, the chief problem with the proviso is that it does rela tively little to augment the System’s control over quantities. Indeed, in the minds of some within the System, it has not even been in tended or expected to have such an effect. In the first place, the wording of the proviso in terms of deviations from “currently pro SHORT-RUN TARGETS jected” growth rates stops far short of indicaing a desired growth rate, that is, a genuine target growth rate. Secondly, the proviso clause falls far short of providing a program for hitting such a target. The proviso clause does provide for a shift in the money market target in a somewhat easier or tighter direction if the proxy falls substantially short of, or rises substantially higher than, its projected growth rate. All the existing literature, as noted earlier, indicates that, ceteris paribus, the ef fects of these changes in the money market target should, in fact, tend to move the proxy in the desired direction. But there is no way of knowing how great the influence will be or what rate of change in the proxy will, in fact, be associated with the revised money market target. Moreover, the proxy and other quantity targets will continue to fluctuate in response to shifting demands under the new money market target, just as they did under the old one— although presumably the fluctuations will be around a higher (or lower) average than would have obtained under the old target. Without denying the very real usefulness of the proviso clause in protecting the System against large, unforeseen, and undesired move ments in the rate of growth of the proxy, it is evident that the proviso clause moved the Sys tem only a little closer to real control over quantities than it had been in the days of the pure money market target. In view of this situa tion, it seems worthwhile to consider some other ways of trying to blend quantity targets with a reasonable degree of money market orderliness. USING MONEY MARKET TARGETS AS A TACTICAL DEVICE IN A QUANTITIES-ORIENTED STRATEGY. One way of using money market targets as a tactical device in a quantities-oriented strategy really involves no departure at all from the pure money mar ket target. It does, however, require a more flexible use of such targets. Over much of the period since the accord, and even at present, the FOMC has apparently tended to identify its money market targets with the “tightness” or “ease” of policy. In the Brunner-Meltzer terminology, it has tended to treat its “target” as, simultaneously, its “indicator.” A change in the money market targets that the Manager is instructed to maintain is identified as a change in “policy.” In one sense, the identifica tion of changes in the money market target with changes in “policy” is a merely semantic matter. Nevertheless, the consequences of this identification have been far from trivial. Thus it becomes a major act for the Committee to change its money market target since, by defi nition, this is a change in “policy.” As a result, the target may go essentially unchanged or may be modified only slightly and gradually over fairly long periods. Often, events may have to become rather radically out of joint with the Committee’s intentions before enough momentum is generated to produce a clear and decisive change in the money market target. This sort of “inertia” can lead to long periods in which fluctuations in the rates of change of monetary aggregates remain almost wholly at the mercy of fluctuations in demand condi tions. Periods in which there have been only minor, if any, modifications in the money mar ket objectives but in which rather major, and often unwanted, accelerations or decelerations in the monetary aggregates have nevertheless developed have not been rare. If the Committee were to drop its tendency to identify changes in money market targets with changes in policy, a very different sort of situation could well develop. In the first place, a change in the money market target instruc tions given the Manager would very likely come to be thought of as involving only a rou tine technical adjustment—a change in tactics, rather than a fairly weighty decision to be made only after substantial evidence of unac ceptable developments has accumulated. Con sider, for example, a situation in which the Committee identifies “policy” with the rate of growth of the bank credit proxy. In that case the first paragraph of the directive might de scribe current economic conditions and the current objectives of policy, as it does now. The second paragraph might then go on to say that in these circumstances, a growth rate in the proxy of approximately x per cent per annum seems appropriate. In a final sentence, it might then state that for the period ahead, such a growth rate could best be fostered by such and such money market conditions. These conditions could be stated either as ranges for specific money market measures or simply described by some qualitative phrase, with numerical values understood from the dis cussion at the Committee meeting. The actual content of this final sentence with its reference to the money market target would have to be determined on the basis of staff projections. As long as policy were to re main “unchanged,” the rate of monetary growth referred to in the second paragraph as appropriate in view of the objectives stated in the first paragraph would remain unchanged. The money market conditions target, however, might be expected to change routinely at every meeting, even with “unchanged policy.” This being the case, monthly-average levels of free reserves and the Federal funds rate might be expected to fluctuate more frequently and more widely than they do at present. The gen eral procedure of sticking to money market targets, but modifying them more or less rou tinely in the service of some more basic quan tity objective has the advantage of requiring only a fairly modest departure, operationally, from a pure money market target. While money market targets would be changed sub stantially more often and perhaps by substan tially larger amounts than under a pure money market target, the state of the money market would still be the Desk’s primary week-byweek concern. The money market would not be left to its own devices; there would be no more risk of the daily-average rate for Federal funds jumping from 2 per cent one week to 1 2 per cent the next than there is at present. It is very difficult to see how the health of the finan cial markets could in any way be risked by following such a procedure. The main objection to the proposal is that while it certainly promises closer control over aggregates than exists under the regime of rel atively inflexible money market targets, it may not go far enough. One reason is that, as long as the money market targets remain rigid in the period between meetings, the System’s re sponse to shifts in bank credit and deposit de mands within that period remains essentially accommodative and passive. Second, there is still the problem of the very loose relationship between money market variables and monetary and credit growth rates. It may be very hard to find the right money market targets given the desired monetary growth rate. Once the money market target is fixed, moreover, the resulting behavior of the growth rate may show unacceptably wide deviations from its expected response. QUANTITY TARGETS WITH MONEY MARKET MODIFIERS. The next step along the road that leads from pure money market targets to pure quantity targets would be a procedure in which the FOMC instructs the Desk to hit a quantity target over the month, but to hit this target in a way that takes ac count of the impact on money market condi tions. Presumably the general format of the operational part of the directive under such a regime would be something like this: “Open market operations shall be conducted in such a way as to encourage the bank credit proxy to grow at an annual rate of about x per cent, while smoothing fluctuations in money market conditions to the extent possible consistent with this objective.” There are any number of ways by which such a directive might be carried out in prac tice. It may be useful to give one rather con crete but also rather mechanical procedure as an illustrative example. Suppose the Commit tee wants to see the proxy grow at an 8 per cent annual rate over the month ahead. As discussed in earlier sections, this desired 8 per cent rate of growth must then, by one tech nique or another, be converted into an appro priate monthly percentage change in nonbor rowed reserves. Given the average level of SHORT-RUN TARGETS such reserves in the previous month, this change can, in turn, be translated into an aver age level for the month ahead. Now there are any number of possible weekly patterns of changes (or levels) of nonborrowed reserves compatible with the desired monthly average level. Thus there is considerable leeway in making decisions about individual weekly changes (levels) in nonborrowed reserves for the 4 or 5 weeks covered by the period in question. The aim of the Desk in carrying out the Committee’s directive should then be to choose levels (or changes in) nonborrowed re serves week by week that ( 1 ) average out to the desired level over the month, and at the same time (2 ) minimize week-to-week fluctua tions in the tone of the money market. One very reasonable way of interpreting ( 2 ), the money market “modifier,” would be to pick weekly levels of nonborrowed reserves (consistent with the desired monthly average) that seem likely to minimize week-to-week fluc tuations in free reserves. If it is assumed that the bulk of week-to-week fluctuations in re quired reserves are seasonal, and if weekly seasonal factors are computed, a rough pattern for week-to-week fluctuations in required re serves for the month can be projected. Given these projected week-to-week fluctuations in required reserves, the familiar reserve identity associates with every possible weekly change in nonborrowed reserves a corresponding weekly change in free reserves. Thus, given the required reserve projections and the reserve identity, we can solve for that set of week-toweek changes in nonborrowed reserves consist ent with the targeted monthly-average level that (a) minimizes the average absolute weekly change in free reserves, or (b) mini mizes the sum of the squares of the weekly changes, or (c) equalizes weekly changes, or (d) satisfies some other criterion that seems to capture the idea of smoothing out changes in the tone of the money market. The possibilities outlined above may seem to suggest that the problem could be solved with mathematical rigor. While this is true in prin ciple for any well-defined notion of “smooth ing” free reserves, there would obviously have to be much fudging in practice. First, not all weekly changes in required reserves would be precisely seasonal. Indeed, if they were, there would hardly be any point to the exer cise. Second, weekly seasonal factors always involve heavy doses of judgment. Third, there would be misses in hitting nonborrowed re serves. Fourth, midmonth corrections would probably have to be made on the monthly nonborrowed reserves objective whenever it became apparent that the primary objective, the monthly growth rate in the bank credit proxy, was not turning out as targeted. Yet de spite these problems, and others that could be mentioned, it still seems reasonable to hope that any given desired monthly change in non borrowed reserves could be distributed over the month in a way that takes advantage of prior knowledge about seasonal changes in re serve needs and thereby minimizes money market instability. This is really all the pro posal amounts to. A MONEY MARKET PROVISO. A final possible version of the “mixed strategy” idea would be to use a quantity target with a money market conditions proviso, exactly the reverse of the procedure currently in use. In spirit, this suggestion is very similar to the one just discussed. That proposal involves aiming directly for some monthly-average value of a quantity variable, but adjusting the week-byweek path of developments in a way most likely to minimize money market instability. In the present proposal, a monthly value of some quantity variable would again be the objective, but there would be no specific attempt to make week-by-week changes in nonborrowed reserves such as to minimize fluctuations in free reserves. Instead, the concern for reasona ble money market stability would be imple mented by absolute constraints on the permit ted range in the level (or weekly change) in some money market variable such as free re serves or the funds rate. Thus the operational part of the directive might read something like can be better served by seeking direct control over the tone of the money market, and thereby exerting an important conscious influ ence on related financial markets. It would be grossly simple-minded to interpret this issue as a question of “Keynesianism” versus “mone tarism.” Nevertheless, it is true that some major questions currently agitating monetary economics are involved. While the relative values to be attached to control over aggregates versus control over the money market are beyond the scope of this paper, considerable attention has been given to the trade-offs between these two objectives. Ex perience seems to indicate that the System can exert a very high degree of control over money market conditions if it chooses to disregard quantity considerations. By contrast, there is no experience to show what degree of control over monetary aggregates might be possible if such control were to be pursued exclusively and without regard to the effects on the money market. Similarly, there is no experience to show what the cost in terms of money market instability might be. The evidence adduced in this paper has not been able to provide firm answers to these questions. In the nature of the case, a high degree of uncertainty is bound to remain, unless and until the FOMC actually experiments with procedures that depart from current and past practices. Despite the lack of adequate evidence on the controllability of quantities and on the costs of such control in terms of money mar ket instability, some tentative judgments on these matters can be made. Thus, the pros SOME GENERAL CONCLUSIONS pects for close control of aggregates over In the last analysis, a decision about targets monthly periods do not look terribly bright. for open market operations has to be made on The slippage between current monthly changes the basis of considerations that go beyond the in nonborrowed, or even total, reserves and relatively narrow focus of this paper. Conse " current changes in the major monetary aggre quently any attempt to make recommendations gates appears to be rather large. An advance here would be misplaced. A basic question, for allowance—even one that is perfectly correct— example, is whether it is better for the Federal for the reserves that will be needed to back Reserve to attempt to exercise reasonably close movements in Treasury deposits helps, but ap control over monetary and bank credit growth parently not enough. Further allowance for rates, or alternatively whether its basic aims other types of deposit movements that are simi this: “Open market operations shall be con ducted in such a way as to encourage the bank credit proxy to grow at an annual rate of about x per cent, except that operations shall be modified when needed to prevent undue stringency or ease in the money market.” Again, the last clause could be quantified in the directive itself, or the acceptable limits of fluctuations in free reserves or the funds rate could be more informally communicated. As in previous examples of quantity targets, more over, the targeted x per cent growth in the credit proxy would of course have to be trans lated into an appropriate rate of growth for the month in nonborrowed reserves. In one sense, the money market proviso ap proach is somewhat more conservative than the approach presented in the previous section since it puts absolute limits on the amount of money market instability that would be permit ted in pursuit of the basic quantity target. Thus, for example, if the permissible limits of fluctuations in net borrowed reserves were placed at $800 million to $ 1 , 2 0 0 million, this pursuit would simply have to be abandoned in any week when the quantity objective ap peared to call for a change in nonborrowed re serves that would, in turn, imply a level of free reserves outside the permitted range. Ob viously the significance of the money market proviso would depend, in practice, on how wide a range in free reserves (or the funds rate) were to be allowed. SHORT-RUN TARGETS larly insensitive to System operations within a given month and that can therefore be projected more or less independently of the assumed sup ply of reserves would provide additional help. Time deposits other than large certificates of deposit appear relevant in this connection. Mid month adjustments in the planned supply of reserves to compensate for unsatisfactory per formance in the first part of the month would probably also help significantly in improving control over the month-by-month movements in the aggregates. Nevertheless, despite the im provements obtainable from these various de vices, it is still likely to turn out that the sys tematic response of the banking system to given changes in the rate at which reserves are supplied within a given month would be only moderate relative to other, largely unpredicta ble determinants of deposit and bank credit behavior within that same month. On balance, it appears likely that even a policy designed to zero in on the growth rate of some aggregate would still leave the month-to-month behavior of that aggregate im portantly conditioned by these random, hardto-predict developments. Hence the short-term behavior of the aggregates would continue to display a substantial amount of statistical “noise.” It does not at all follow from this, however, that the influence of the System might not be dominant in the longer run. A policy of aiming at the growth rate of, say, the money supply might be able to fix the actual growth rate of that target averaged over a mul timonth period with a satisfactory degree of precision. The results obtained for quarterlyaverage figures can be interpreted as reasona bly encouraging—again assuming the regres sion results can be materially improved upon by midcourse corrections, the use of judg ment, and so forth. No doubt the results could be improved further if still longer periods were used. Unfortunately, of course, the need to av erage the behavior of a quantity target over relatively long periods to obtain an acceptable degree of control means that a change in the setting of the target might have a reliable and clearly visible effect on the actual behavior of the target variable only after a similarly long period of time. The results of this study suggest that the Committee could adopt the use of explicit quantity targets without producing an unac ceptable degree of short-term instability in the money market. In part, this conclusion rests on the evidence presented in the section begin ning on page 56. That section suggests that movements in the Federal funds rate induced by supplying nonborrowed reserves (not sea sonally adjusted) at a constant rate would not be intolerably large. Everyone will recognize the insufficiency of this evidence taken by itself, however. More fundamentally, therefore, the conclusion rests on the belief that allowance for seasonal changes in required reserves—or, better yet, adoption of one of the “mixed strategies” presented in the previous section— would permit fluctuations in money market conditions to be held within tolerable bounds. Probably the worst that could result from the adoption of one of these mixed strategies would be that neither the aggregate nor the money market would turn out to be regulated with much precision. It could be that given the difficulty of precise control of the aggregates in the short run under even the best of circum stances, and given the compromises that might be needed to hold money market fluctuations within acceptable limits, the behavior of the aggregate target might continue to be domi nated by random, or at least uncontrolled, fac tors. At the same time, both the marginal re serve measures and the Federal funds rate would surely show a less steady, “rational” pattern than is presently the case. As long as laboratory experiments on these matters are impossible, however, such risks are inevitable. Whether they should be taken depends heavily on how much importance is attached to achieving meaningful control over monetary aggregates—as opposed merely to exerting a rather loose “influence” over these magnitudes. by Leonall C. Andersen SELECTION OF A MONETARY AGGREGATE FOR USE IN THE FOMC DIRECTIVE CONTENTS 73 INTRODUCTION 73 A GENERAL FRAMEWORK OF THE INFLUENCE OF MONETARY ACTIONS ON THE ECONOMY 74 74 75 MONETARY AGGREGATES AND MOVEMENTS IN GNP Response of GNP to each monetary aggregate Basis for selecting a monetary aggregate 79 RECOMMENDATION 80 APPENDIX: Estimation Aspects of the “Simultaneous Equations Bias” Issue by H. Albert Margolis SELECTION OF A MONETARY AGGREGATE INTRODUCTION One suggestion for change in the Federal Open Market Committee directive is to place main emphasis on a stipulated movement in a monetary aggregate. Presumably this would be the best aggregate available for assisting the FOMC in achieving its ultimate economic goals. There are two general ways in which a monetary aggregate could be incorporated into the directive. First, the desired rate of change in the aggregate could be specified directly in the second paragraph of the directive. This paragraph contains the specific instructions of the FOMC to the New York Federal Re serve Bank for the conduct of open market operations between Committee meetings. The Manager of the open market desk at the New York Federal Reserve Bank is assigned the re sponsibility for carrying out the directive. Sec ond, a desired rate of change in the chosen ag gregate could be specified in the first paragraph of the directive, with instructions given to the Manager in the second paragraph in terms of changes in some other variable that could be more readily observable by the Manager and might be subject to his more direct control. Achievement of the specified movement in this latter variable by the Manager would be ex pected to produce the FOMC’s desired rate of change in the monetary aggregate. In either case, the explicit goal of the FOMC is desired movements in one monetary aggregate. The two procedures just outlined differ only in the short-term operating instruction given to the Manager. In this paper six monetary aggregates are considered for inclusion in the directive—non borrowed reserves (Nb), total member bank reserves (TR), the monetary base (B ), the narrowly defined money stock (M J , the money stock plus time deposits at commercial banks (M2), and bank credit (BC). This paper is concerned primarily with properties of each aggregate as they relate to the ability of the Federal Reserve System to achieve its ulti mate goals of desired real product growth and price level stability. Although other papers dis cuss in detail the ability of the Manager to control various aggregates, this paper merely takes a brief look at this problem. A GENERAL FRAMEWORK OF THE INFLUENCE OF MONETARY ACTIONS ON THE ECONOMY The general framework used in this paper for relating the influence of monetary actions to movements in real output and the price level differs greatly from that incorporated in most large-scale econometric models. Mone tary actions, summarized by changes in some monetary aggregates, and changes in Federal Government expenditures are viewed as the main determinants of total spending measured in current-dollar gross national product (nomi nal GNP). A given change in nominal GNP is then divided between a change in real output and a change in the GNP deflator. An impor tant factor in explaining this division is the dif ference between potential real output in the quarter and actual real output in the preceding quarter. Another factor is past price move ments. The specific model relating a particular summary measure of monetary influence (Mx) to output and the price level is presented elsewhere. 1 In contrast, most econometric models use a building-block approach, which considers that the major influence of monetary actions on both output and the price level is primarily in direct—for example that it operates through interest rates. One building block consists of the major components of GNP and their deter minants, which include fiscal actions and other exogenous variables. A second building block, the financial sector, determines a market rate of interest. Finally, the price level is deter N o t e .— The author is Vice President, Federal Re serve Bank of St. Louis. 1 Leonall C. Andersen and Keith M. Carlson, “A Monetarist Model for Economic Stabilization.” Re view, Federal Reserve Bank of St. Louis (Apr. 1970). mined by a Phillips curve equation or a wage/ price mark-up equation. Joint simulations of the three blocks are used to allow interactions among the three blocks. Frequently, the com ponents of GNP in real terms are summed, and this sum is multiplied by the price level to pro duce an estimate of nominal GNP. MONETARY AGGREGATES AND MOVEMENTS IN GNP This section presents empirical evidence that is used to select the best monetary aggregate from those under consideration for inclusion in the FOMC directive. First, the relation be tween changes in GNP and changes in each aggregate is measured by regression analysis. Second, three criteria are presented for the selection of the “best” aggregate, and relevant data are developed for application of the crite ria to each aggregate. RESPONSE OF GNP TO EACH MONE TARY AGGREGATE. As mentioned earlier, monetary actions and changes in Government expenditures are viewed as the major determi nants of movements in GNP. Six individual re gression equations are run in which quarterly changes in GNP are regressed on current and lagged changes in each of the six monetary ag gregates along with, in each regression, current and lagged changes in Government expendi tures on goods and services plus transfer pay ments (AE). Ordinary least-squares estimates of parameters are made, by using Almon lags with a fourth degree polynomial and coefficients for t + 1 and t — n — 1 constrained to zero. The length of the lag period (n ) is determined by the minimum standard error of estimate. The regression results are presented in Table 1. The fits of the equations seem to be very good, considering that first differences are used. The smallest R 2 is 0.52 for nonbor rowed reserves, and the largest is 0.67 for bank credit. The Durbin-Watson statistic indicates small likelihood of serial correlation in any of the residuals. Most of the regression coeffi cients are statistically significant from zero at the 5 per cent level. Some may be surprised by the positive coef ficients for Government expenditures for a few quarters followed by negative coefficients. In every regression, the sum of the coefficients for AE is not statistically significant from zero at the 5 per cent level. Each regression may be viewed as measuring the response of GNP to changes in a monetary aggregate with Govern ment expenditures held constant and its re sponse to changes in Government spending with the monetary aggregate held constant. In the latter case, Government expenditures are financed by taxing or borrowing from the pub lic. In such an instance, Government expendi tures may, over time, crowd out an equivalent amount of private expenditures, thereby ac counting for the observed pattern of regression coefficients. In most econometric work, the question of simultaneous-equation bias is always present. The appendix discusses this question in some detail and highlights the unsettled nature of this problem. In summary, formal discussions of bias are based on the asymptotic properties of large-size samples, and little is known about bias in the limited, finite samples available in economic research, and even these discussions do not apply to the case in which lagged endo genous variables appear. Moreover, one of the papers cited in the appendix shows that, in small samples with no lagged endogenous vari ables in a regression, if ordinary least squares (OLS) are biased, then two-stage least squares (TSLS) estimates are also biased, although under certain circumstances the degree of bias is smaller. TSLS estimates are commonly used to handle the bias problem. Bias, however, is not the only undesirable property of an estimation procedure; a large variance of parameter estimates is also unde sirable. It is well known in statistics that par ameters estimated by OLS have smaller vari ances than those estimated by TSLS. Thus, in selecting estimation procedures, one may have SELECTION OF A MONETARY AGGREGATE sample period. Third, the aggregate should perform best with regard to the ability of the Federal Reserve to control its movements. In the sample period 1953-1 to 1969-III, there is virtually no difference in the fit of the regressions for the equations involving M lf A/2, B, and BC (Table 1 ). The R 2's range from 0.65 for the monetary base to 0.67 for bank credit and, similarly, the standard errors of esti mates (SEE) range from 3.79 to 3.93. The fit of the regressions involving TR and Nb are not so close—with R 2's of 0.52 and 0.59 and SEE’s of 4.26 and 4.63. For these last two re gressions the Durbin-Watson statistics are also lower. On the basis of sample-period statistics, M lt M2, B, and BC all seem to perform equally well, and they all perform better than Nb and TR. The ex ante forecasting ability of each equa tion was tested for successive eight-quarter pe riods beginning with 1965-1. For example, each regression equation was estimated for 1953-1 to 1964-1V; then quarterly forecasts of changes in GNP for 1965 and 1966 were made by using the parameters estimated for the sample period. This procedure assumes to trade off bias against larger variance of pa rameter estimates. The appendix cites both a demonstration and experimental results to show that the mean-squared error statistic— which combines bias and variance of parame ter estimates into one number—can, in certain circumstances, be considerably greater for TSLS than for OLS. In view of the unsettled nature of these issues in economic research, there is no clear-cut case for asserting that there is obviously a large bias in the OLS pa rameter estimates presented in Table 1, or that the bias is of such a magnitude as to more than offset the gain from a smaller variance of parameter estimates. BASIS FOR SELECTING A MONETARY AGGREGATE. Three criteria are used in this study for selection of a monetary aggregate for monetary management. First, it should per form best in terms of goodness of fit relative to the five other aggregates in the sample pe riod used to relate changes in GNP to changes in each monetary aggregate. Second, and more importantly, the aggregate selected should pro duce the smallest forecasting errors in ex ante forecasts of changes in GNP made beyond the TABLE 1: Minimum Standard Error Regressions (1953-I-1969-III) (1) Item Q uarters t t-1 t-2 t-3 t-4 t-5 t-6 t-7 t-8 t*9 t -io t-11 t-12 t-13 t-14 t-15 t-16 AMi 1 .2 3 1 * 1 .7 8 7 * 1 .6 0 0 * .8 7 5 * .091 (2) AE .5 8 9 * .4 4 2 * - .0 2 1 - .4 4 0 * - .5 1 3 * AA /i - .0 7 4 .5 4 8 * .9 3 5 * .7 3 2 * .1 5 6 (3) AE .2 7 7 .3 7 0 * .0 8 4 - .4 4 4 * - .7 4 0 * AB 3 .3 1 0 * 6 .2 4 8 * 7 .0 5 8 * 5 .1 5 9 * 1 .1 4 5 - 3 .2 1 3 * - 4 .9 6 9 * (4) AE .4 1 5 * .1 6 4 - .2 3 5 * - .4 5 9 * -.3 8 2 * - .0 7 1 .2 1 0 1 .2 6 4 .7 5 0 .057 2 .2 9 7 * - .4 5 3 A TR 2 .1 9 2 7 .3 4 9 * 1 1 .0 2 1 * 1 0 .8 5 0 * 6 .5 6 8 * (5) AE .3 8 6 .3 4 2 * .081 - .1 9 2 - .2 8 1 ANb (6) AE ABC 2 .3 1 5 * 3 .8 4 2 * 4 .7 4 1 * 5 .1 5 3 * 5 .2 0 5 * 5 .0 0 4 * 4 .6 4 3 * 4 .1 9 8 * 3 .7 2 9 * 3 .2 7 7 * 2 .8 7 1 * 2 .5 1 8 * 2 .2 1 3 1 .931 1 .6 3 4 .1 9 2 * .2 6 5 * .2 5 4 * .1 9 0 .0 9 8 .0 0 2 - .0 8 3 - .1 4 3 * - .1 7 2 * - .1 6 6 * - .1 2 6 - .0 5 8 .0 2 6 .1 1 2 .179 .2 0 3 .0 7 2 .6 3 1 * .8 9 3 * .6 0 6 * .0 4 2 .2 5 5 .3 3 5 * .0 7 4 - .3 9 9 * - .6 6 2 * 5 5 .2 8 7 * .9 2 7 2 .2 4 3 * - .3 9 6 .1 5 4 Sum 5 .5 8 3 * C on stant 2 .6 6 9 * .6 6 .6 6 .6 5 .5 9 .5 2 .6 7 D .W , 1 .7 5 1 .7 0 1 .7 3 1 .5 0 1 .3 0 1 .7 3 SE E 3 .8 8 3 .8 6 3 .9 3 4 .2 6 4 .6 3 3 .7 9 1 .5 3 2 * V alues are significant a t the 5 per cen t level. 1 4 .7 3 8 * 2 .5 0 6 * - .3 5 8 AE 3 7 .9 7 9 * .8 4 6 .3 3 6 - 2 .6 6 4 .9 7 2 that the values of the independent variables for the forecast period were known in 1964-IV. Next, the regressions were rerun to include an additional year, and quarterly forecasts were made for the next 2 years. This procedure was repeated until the sample period ending with 1968-IV was reached. Two statistics are developed for each suc cessive 2 -year forecast period to compare the forecasting abilities of the six equations. The average-squared residual between the actual and the forecasted quarterly changes in GNP are calculated for each year of a forecast period and for each whole 2-year period. The second statistic is the standard error of forecast for each quarter of a 2-year forecast period. The standard error of forecast takes into considera tion the stochastic element in each equation, the variance of the parameters estimated for the sample period, and the variability of the independent variables in the forecast period. Table 2 presents the average squared residu als for each equation. For every 2-year forecast period as a whole, the equation for M t has the lowest average squared residual. Although in a few cases some of the other five equations for an individual year have smaller average squared residuals than the M t equation, their average squared residuals vary considerably more from year to year than in the case of the M 1 equation. For example, forecasts of changes in GNP based on a 1953-65 regres sion for BC have average squared residuals of 2.7 for 1966 followed by 60.1 for 1967. The comparable averages for M x are 10.7 and 12.1. The average standard error of forecast and its variance over each 2 -year forecast period are presented in Table 3. In most forecast pe riods there is little difference between the aver age standard error of forecasts for each regres sion, except for Nb, which consistently has the largest average standard error of forecast. However, there is considerable difference in the variability of the standard errors of fore cast. In almost every case, its variance for forecasts of changes in GNP based on Mi re gressions is relatively small and for three fore cast periods is the smallest. This paper covers only two aspects of the System’s ability to control monetary aggre gates. These are the magnitude of the control problem and the present flow of information on which control would be based. The prob- TABLE 2: Average Squared Residuals of GNP Forecasts Based on Monetary Aggregates Sam ple period Forecast period AM i AMi AB AT R ANB ABC 1965 1966 2 0 .9 7*1 2 2 .4 1 3 .9 3 5 .2 6 3 .5 3 1 .2 1 9 .2 3 8 .5 1 0 .3 2 5 .7 6 .1 A verage 1 4 .0 1 8 .2 4 9 .4 2 5 .2 2 4 .4 1 5 .9 1966 1967 1 0 .7 1 2 .1 9 .3 4 7 .5 2 8 .3 6 9 .7 1 6 .2 5 3 .1 1 7 .1 1 0 5 .0 2 .7 6 0 .1 A verage 1 1 .4 2 8 .4 4 9 .0 3 4 .7 6 1 .1 3 1 .4 1967 1968 1 6 .1 1 1 .8 4 4 .8 1 3 .4 5 2 .0 1 .3 4 5 .8 1 .1 6 5 .5 1 6 .6 6 2 .0 3 .8 A verage 1 4 .0 2 9 .1 2 6 .7 2 3 .5 4 0 .6 3 2 .9 1968 1969* 1 1 .2 7 .0 8 .9 3 8 .2 2 .2 1 9 .4 1 .1 2 8 .7 3 6 .7 3 4 .2 0 .8 2 4 .6 A verage ~9A 2 3 .6 1 0 .8 1 4 .9 3575 1 2 .7 1969* _9_.1 2 8 .0 ---- 2 0 .6 _--- 3 0 .3 ---- 2 0 .2 ----- - --- 9 .1 2 8 .0 2 0 .6 3 0 .3 2 0 .2 2 6 .5 1953 -I-1 9 6 4 -IV 1 9 53 -I-1 9 6 5 -IV 1953-1-1966-IV 1 9 5 3 -I-1 9 6 7 -IV 1 953 -I-1 9 6 8 -IV A verage * F o reca st period co n sists o f o n ly three quarters. 2 6 .5 SELECTION OF A MONETARY AGGREGATE TABLE 3: Comparison of Standard Error of Forecast Forecast period 1 Sam ple period A Mi AM t 1 953-I-1964-IV M ean V ariance 5 .2 6 .2 5 5 .2 5 .6 2 1 953-I-1965-IV M ean V ariance 5 .6 7 .1 7 1953-I-1966-IV M ean V ariance ATR A NB ABC 5 .1 6 *27 5 .0 5 7 .0 9 .86 4 .6 7 .10 6 .2 6 .4 8 5 .7 1 .1 5 5 .6 1 .1 5 8 .3 9 1 .5 8 5 .2 8 .2 3 5 .4 1 .0 6 5 .7 1 .4 9 5 .4 2 .1 9 5 .2 4 8 .7 5 .5 3 S . 16 1953-I-1967-IV M ean Variance 4 .9 0 .0 3 5 .0 1 .3 2 5 .0 1 .1 7 4 .9 9 .0 4 8 .6 9 5 .1 3 .11 .12 1 953-I-1968-V M ean V ariance 5 .7 0 5 .2 5 .1 7 4 .6 5 .0 7 4 .7 5 .0 3 7 .7 4 .0 6 5 .0 6 (*> .12 .10 .12 .02 1 The forecast period fo r each o f the sam ple periods ending 1964, 1965, and 1966 is eight quarters; for the sam ple ending 1967, seven quarters; and fo r the sam ple ending 1968, three quarters. * L ess than .005. lem of control is investigated by partitioning the six monetary aggregates into two classes. The first class consists of those considered to be more closely related to GNP—M u M 2, and BC; the second set consists of those considered to be subject to closer Federal Reserve control —B, TR, and Nb. The Brunner-Meltzer framework provides an approach for investigating the ability of the Federal Reserve to control M u M2, and BC. This approach views each of these aggregates as the product of the appropriate multiplier (mi) and the monetary base. The values of the mi’s reflect actions of the public, commercial banks, and the Government as they influence movements in each of the three aggregates. The monetary base reflects actions of the Federal Reserve. To reach a desired level of one of these aggregates, the monetary base would be changed to compensate for movements in the appropriate multiplier. Within the multiplier-base framework, there is a smaller problem of controlling M t due to actions of the public, commercial banks, and the Government (variations in the M t multi- 2 Monthly averages for BC were approximated by averaging end-of-month data for the current and pre vious month. This procedure may tend to overstate the variability in BC. plier) than for M 2 and BC. Multipliers were developed for M u M2, and BC (monthly aver ages of unadjusted data 2 for January 1960 to November 1969), and the standard deviations and ranges of monthly changes in each multi plier were calculated. The standard deviations are: 0.008 for AM u 0.017 for AM2, and 0.018 for A BC. Given the monetary base and the level of each aggregate for November 1969, these standard deviations in annual rate of change in each aggregate are: 3.6 per cent for Mj, 4.0 per cent for M 2, and 4.3 per cent for BC. Similar rates of change for the range of variation in these multipliers are: 23 per cent for M u 26 per cent for M 2, and 46 per cent for BC. Thus, there is smaller variability in due to variations in its multiplier, thereby creating a lesser control problem than in the cases of M 2 and BC. The question remains of the ability of the Federal Reserve to control the monetary base relative to its ability to control member bank reserves and nonborrowed reserves. Table 4 lists the factors determining each of these ag gregates. If the Federal Reserve were to meet a desired level of one of these aggregates, its holdings of U.S. Government securities would be adjusted so as to offset movements in the sum of all other factors. Changes in monthlyaverage levels (unadjusted data, January 1960 to November 1969) of factors other than Sys tem holdings of U.S. Government securities were calculated for B, TR, and Nb. The stand ard deviations of these changes are little dif ferent for each aggregate— 0.412 for total re serves, 0.437 for nonborrowed reserves, and 0.443 for the monetary base. These standard deviations in terms of annual rates of change from November 1969 levels are substantially different—7 per cent for the monetary base, 18 per cent for total reserves, and 2 0 per cent for nonborrowed reserves. It appears that the control problem, as measured above, is less for the monetary base than for the two reserve ag gregates. The preceding analyses of controlling each of the six monetary aggregates considered only the comparative magnitudes of variations in each aggregate from sources not under direct System control. The ability to forecast such variations was not examined on a comparative basis. Finally, let us consider the data require ments for controlling B, TR, and N b. Table 4 demonstrates the difference among these three aggregates regarding the data required to con trol each. All of the data required to control the monetary base are required to control the other two. In addition, both TR and Nb re quire information on currency in circulation and its distribution beween member banks, nonmember banks, and the nonbank public. Currency movements can cause wide seasonal movements in TR and Nb, thereby adding to the control problems for these two aggregates beyond those caused by similar movements in factors common to all three of these aggre gates. From a data standpoint, the monetary base appears easier to control than total re serves or nonborrowed reserves. T A B L E 4 : F a c t o r s A f f e c t in g M o n e ta r y B a s e , T o t a l R e s e r v e s , a n d N o n b o r r o w e d R e s e r v e s o f t h e B a n k in g S y s t e m , J u ly 1 9 6 9 1 M o n th ly averages o f d a ily figures in m illio n s o f d ollars Item Federal R eserve c r e d it..................................................................................................... H o ld in g s o f se c u r itie s ............................................................................................. M em ber bank borrow ings fro m F .R ............................................................... O th er b o rro w in g s...................................................................................................... F .R . flo a t....................................................................................................................... O ther F .R . a s s e t s ....................................................................................................... G o ld s to c k .............................................................................................................................. T reasury currency o u tsta n d in g ..................................................................................... T reasury cash h o ld in g s..................................................................................................... D e p o s its a t the F .R . (other than m em ber bank d e p o sits).............................. T rea su ry ......................................................................................................................... F o r e ig n ........................................................................................................................... O th e r ............................................................................................................................... O ther F .R . lia b ilities an d c a p ita l................................................................................. C u rrency in c ir c u la tio n .................................................................................................... S ources o f b ase T otal reserves a n d n o n b o rro w ed reserves o f ban k in g system s 6 0 ,8 8 8 5 4 ,2 9 8 1 ,1 9 0 * 6 0 ,8 8 8 5 4 ,2 9 8 1 ,1 9 0 2 ,6 8 4 2 ,6 7 0 2 ,6 8 4 2 ,6 7 0 1 0 ,3 6 7 6 ,7 3 7 — 657 — 1 ,7 3 2 — 1 ,1 1 7 — 142 — 473 10 ,3 6 7 6 ,7 3 7 — 657 — 1 ,7 3 2 — 1 ,1 1 7 — 142 473 — 2 ,0 3 8 Source b a s e ............................................................................................................................ R eserve adjustm ents *....................................................................................................... 7 3 ,5 6 5 3 ,8 7 6 M on eta ry b a s e ...................................................................................................................... 7 7 ,4 4 1 — 2 ,0 3 8 — 5 1 ,2 5 6 M em b er bank reserves w ith the F .R .............................................................................................................................................................................................. C u rrency held by m em b er banks (a llo w ed a s required reserv es)..................................................................................................................................... 2 2 ,3 0 9 4 ,6 7 1 T o ta l reserves o f m em ber b a n k s...................................................................................................................................................................................................... C urrency held by no n m em b er b a n k s ............................................................................................................................................................................................. R eserve adjustm ents *............................................................................................................................................................................................................................ 2 6 ,9 8 0 «1, 432 3 ,8 7 6 A ggregate reserves o f the banking sy s te m ................................................................................................................................................................................... M em b er bank borrow ings from the F .R ...................................................................................................................................................................................... 3 2 ,2 8 8 — 1 ,1 9 0 N o n b o rro w ed reserv es.......................................................................................................................................................................................................................... 3 1 ,0 9 8 1 N o t adjusted fo r season al variation. * In clu d es $46 m illion a ccep tan ces n o t sh ow n separately. * A d justm ents for ch an ges in reserve requirem ents, sh ifts in d ep osits betw een tim e d ep o sits and d em and dep osits, an d sh ifts a m o n g classes o f bank s. * E stim ated . S o u r c e .— F ederal R eserv e Bulletin. SELECTION OF A MONETARY AGGREGATE RECOMMENDATION The results of this study, in my opinion, suggest that among the six monetary aggre gates investigated, M 1 would be the best to in clude in the FOMC directive. The regressions for the sample period of changes in GNP on the six aggregates indicate that total reserves and nonborrowed reserves would be inferior to the other four. The ex ante forecasting experi ment indicates that Afa performs the best. With regard to our ability to control movements in M u Mo, and (Bank Credit), evidence was pre sented that Mx would be subject to closer con trol. These results lead me to recommend that M t be incorporated in the Committee’s direc tive, preferably in the instructions to the Man ager. If it is believed that there exists a major problem of controlling M x and that it would be desirable to give operating instructions in terms of some other aggregate, I recommend the monetary base. Among the aggregates in fluenced more closely by the Federal Reserve (Nb, TR, and B ), control of the monetary base appears to have fewer problems. In such a case, however, I also recommend that de sired movements in Mx be specified in the first paragraph of the directive. Implicit in this study is the assumption that the Manager’s in structions be in terms of annual rates of change in quarterly averages of M x and/or B. APPENDIX: Estimation Aspects of the “Simultaneous Equations Bias” Issue This appendix surveys the general state of knowledge regarding the problem o f “simultane ous equations bias.” Such bias is said to arise when one applies ordinary least squares (OLS) estimation procedures to an equation in which a critic feels at least one o f the independent varia bles is not predetermined, that is, all of the inde pendent variable’s current and past values are not independent o f the current random disturbance term ([2], p. 353). In statistical terms, if all but one o f the varia bles in an estimated equation are exogenous, then OLS yields estimators with desirable properties. In particular, the bias is zero, that is, the differ ence between the expected value o f the estimator and the true parameter is zero.1 If there is more than one endogenous variable in the estimated equation, “classical least-squares applied to the individual structural equations yield biased esti mates o f their parameters” ([6], p. 385). This is the standard textbook assessment o f the situation. Oi’s ([11], p. 36) statement is that when “two or more variables in an equation are simultaneously determined by som e larger system o f equations,” then OLS “will produce asymptotically biased pa rameter estimates.” 2 N o t e .— The author is an Economist at the Federal Reserve Bank of St. Louis. 1 The definitions of technical terms used in this ap pendix can be found in most econometric texts (for example 2, 6, 7). For the convenience of the reader, a few basic definitions are given here in a relatively informal phrasing. An estimator is consistent if its probability distribution tends to become stacked com pletely above the true parameter as the sample size increases beyond a certain value. An estimator is asymptotically unbiased if the mean of the estimator approaches the true parameter as the sample size in creases. The former is a stronger property than the latter. They are both “large sample” properties. 2 Sawa ([13], p. 932) shows that under very special by H*Albert Margolis For purposes of specific discussion, let us take an equation that has been estimated by ordinary least squares in which changes in G N P are con sidered to be a function o f present and lagged values o f changes in indicators o f monetary influ ence— for example, the m oney supply— and pres ent and lagged values o f changes in Government expenditures. The question is raised as to how much reliance can be placed on the estimated coefficient o f the money supply in the current time period. The critics w ho raise this question assert that the m oney supply is an endogenous variable. In such a case, a statistically complete system would require at least one additional equation accounting for the behavioral m ode by which G N P reacts back on the m oney supply. M ost textbooks point out that, “In recursive models 3 o f the type advocated by W old . . . single-equations least-squares estimators are consist ent ([5], p. 14).” But this is simply begging the question in the equation under discussion. Critics o f the equation are asserting that the system is not causal— that is, that there is indeed reverse causation from G N P to m oney. It seems natural to ask these critics to indicate the form o f the re verse causation. The question might be stated informally, “When should one o f the explanatory variables be considered endogenous and an additional struc ture equation added to the model?” Christ ([2], p. 157) elaborates on this point as follows: circumstances the small sample bias of both OLS and TSLS (two stage least squares) disappears. 3 A system is recursive if the equations in it can be ordered so that in the first equation only one en dogenous variable appears and in each succeeding equation only one new endogenous variable appears in addition to previously included endogenous varia bles. In addition, the covariance matrix of the dis turbance terms is diagonal. SELECTION OF A MONETARY AGGREGATE For there is no point in the enlargement of most models at which a <?Onvincing stand can be made against such arguments for the addition of another equation—unless it is the point where all possible variables have already been in cluded, and of course the model would then be utterly unmanageable. What the economist should do in practice, therefore, in my opinion, is to stop adding equations and variables when he believes that the variables he chooses to call exogenous meet the definition closely enough so that the errors incurred through the discrepancy are small in comparison with the degree of ac curacy that he thinks is desirable for his pur pose (or is attainable). This is necessarily a somewhat arbitrary decision, for, .unlike the other variables, the random disturbances by their nature can never be observed either. These decisions, like other decisions about what the form of each equation is to be and what varia bles are to be excluded from each, must be made on the basis of whatever presumptions seem plausible in the light of economic theory and experience. The model itself can be defini tively tested only after it is confronted with new data. If it proves reasonably accurate all may be well, and if not, it is likely that at least one wrong assumption was made somewhere. This seems to say that there is no statistical way to test whether an assumption that a variable is exogenous is correct.4 A survey of basic econometric texts shows two discussions in which techniques are mentioned by which the single-equation format is retained.5 Christ ([2], pp. 4 5 7 -6 3 ) follow s Bronfenbrenner [1] and suggests that attempts be made to deter mine the range o f possible error in the estimate by making assumptions about the unobservable error term in the equation. This seems extremely arbitrary. Kane ([8], pp. 3 1 3 -1 8 ) gives an ex ample in which Ferber avoids an overestimate of the marginal propensity to consum e by estimating the marginal propensity to save. In other words, Hi as in favor o f a particular hypothesis is re versed (but not eliminated). This strategem is 4 To emphasize this, we should point out that a necessary condition for zero bias under OLS estima tion is that the independent variables not be corre lated with the error term ([4], p. 591). But the error term is unobservable. We have the residual as only an approximation, and the estimates are constructed so that the expected value of the product of the re sidual and the independent variable is zero. 5 Reference should be made to T. Haavelmo, who is usually credited with the discovery of the bias (cf.[10]). more easily adopted because in the example— as in most discussions o f simultaneous equation bias — the second equation is an identity. On the whole, these suggestions do not seem useful in the present circumstances. We proceed now to the question as to what the situation involves if we feel there should be a second endogenous variable in the estimated equation. In other words, let us examine the situ ation as envisioned by the critics. A s soon as the money supply is considered to be endogenous, conceptually we have a larger system in which one equation is the G N P equation; now lagged values o f an endogenous variable— m oney supply — occur in the G N P equation. W e are then faced with a choice among various estimation proce dures— OLS, TSLS, other forms o f limited infor mation methods, and full information methods. The latter two procedures have not been widely used in practice. The choice seems to narrow to one between OLS and TSLS. Fisher ([4], p. 602) points out that TSLS esti mators share with other limited information methods certain practical difficulties when used in economy-wide econometric models with lagged endogenous variables. The first stage o f estima tion (the reduced form) may be difficult. The in clusion o f lagged endogenous variables “raises considerable difficulties in the likely presence of serial correlation o f the disturbances.” A ccording to Walters ([9], p. 189), a choice between OLS and TSLS can be made on the basis o f the purpose o f the estimation. If our purpose is to predict an endogenous variable, OLS will yield unbiased and best estimators while those o f TSLS are biased and inefficient. On the other hand, if our purpose is to estimate struc tural parameters, then OLS gives biased and in consistent estimators while those o f TSLS are consistent “although biased in small samples.”6 This reference to the small sample properties o f TSLS estimators seems very relevant, and we pursue it by quoting first from several textbooks and finally by referring to a recent paper that ad dresses itself directly to this question. Fisk ([5], p. 6) comments: 6 The results indicated in this paragraph apply only if there are no lagged endogenous variables in cluded; otherwise there are no known finite sample results. It is not obvious that this lack of consistency (of single-equation least-squares) should always cause concern, particularly when dealing with small samples for which the alternative consist ent estimator may have grossly inflated variances compared with the biased estimator given by (least-squares). We must always balance the de sirability of consistent estimators against the other criteria by which we judge estimators— principally: degree of precision of the estimator and ease of calculation. Goldberger ([6], p. 360) points out: . . . for small samples the second moments of the classical least-squares estimators (about the true parameter values) may be less than those of the TSLS estimators— their variances may be sufficiently small to compensate for their bias. It should be emphasized, however, that as the sam ple size increases, the variance of both classical least-squares and TSLS tend to zero, but the bias of classical least-squares persists. Christ ([2], p. 466) gives a table o f properties o f various types o f estimators in a m odel that ad mits lagged endogenous variables. It shows that OLS yields inconsistent estimators in general, al though with a small variance. The various other estimation procedures are shown to yield consist ent estimators. It should be remembered, how ever, that consistency is an asymptotic property and is frequently used as a criterion only because there are very few results dealing with small sam ple properties. Am ong the first steps toward the latter goal— that is, working with the exact distribution func tions of OLS and TSLS estimators— is an important paper by Richardson and W u [12]. This paper shows that in a case similar to the G N P equation, with the important and vital dif ferences that the Richardson and Wu case does not admit lagged endogenous variables, TSLS es timators are unbiased if, and only if, OLS estima tors are unbiased. It is not possible to derive the size o f the bias from these results, but the rela tive bias o f TSLS and OLS as a function o f var ious parameters is tabulated. The values vary from almost one to almost zero as the sample size ranges from approximately 10 to approxi mately 100. In another paper Sawa [13] derives the exact sampling distributions o f OLS and TSLS estima tors of a structural parameter in a structural equation with two endogenous variables. The exact distributions are “essentially similar,” and in a numerical example he finds that the plotted distributions are “surprisingly similar,” with the bias o f each always in the same direction. These results do not apply in a rigorous sense to the case discussed in this study, because, thus far, they have not been extended to include lagged endogenous variables. On the other hand, in a very similar sense, the criterion o f consist ency is not relevant to models with a finite sam ple size but it is used for lack o f a better crite rion. The Richardson and Wu and the Sawa papers are among the few dealing with the exact distri bution o f the various estimators. M ost works dealing with small sample properties have used M onte Carlo experiments. Fisher ([4], pp. 6 0 4 05) points out that even these results do not apply to the case in which lagged endogenous variables appear. The quotes from Fisk and Goldberger suggest that the small variance o f OLS may compensate for any bias. The mean-square-error criterion dis cussed in Johnston ([7], pp. 2 7 6 -7 7 ) combines the effects o f these two pathologies, variance and bias. The author (p. 294) rejects OLS on the basis of the bias even though he cites some stud ies that show OLS performing well on the meansquare-error criterion. Goldberger ([6], pp. 3 6 2 63) also opts for TSLS even though in the principal study he indicates that “the variance o f OLS was sufficiently small to give it generally the smallest second m o m e n t. . Sawa ([13], p. 933) suggests that TSLS is pre ferable but with som e qualifications: Indeed the TSLS estimator seems to dominate the OLS estimator in every case, but, in certain cases, this dominance is not readily observable. Furthermore, the bias of the TSLS estimator is not negligible. Consequently, it may be said that the TSLS estimator is not such a good estimator as expected in finite samples. The Richardson and Wu paper ([12], pp. 1 1 13) gives more perspective on this issue by presenting a table that derives the exact ratio o f expected mean-square errors o f TSLS and OLS estimators as a function of several parameters. The values in the table range from 9.5 to 0.04 as the sample size ranges from approximately 10 to approximately 100. This means that it is possible for either TSLS or OLS to enjoy a considerable advantage over the other according to the mean- SELECTION OF A MONETARY AGGREGATE square-error criterion. The caveat is repeated that the formulas used in this study do not include the lagged endogenous case. But then, as we have seen, neither the theoretical large sample proper ties nor the Monte Carlo small sample results apply in a firm fashion to the present case. For a last note o f nihilism in the simultaneous-equation bias controversy we quote Fisher ([4], p. 590), who points out that “very few re sults are available on the relevant robustness— the relative degree to which they (various estimators) stand up to such things as multicollinearity, speci fication error, and serial correlation in the dis turbance of the m odel.” Evans ([3], p. 5) refers to results obtained by Klein, w hich suggest that, when multicollinearity is present, it is the more complex methods o f estimation other than OLS that are more susceptible to bias. Christ ([2], pp. 4 8 0 -8 1 ) summarizes by say ing “it is not yet clear that the least-squares method for structural estimation is dead and should be discarded. . . . For structural parame ters, least-squares sometimes are preferable to simultaneous-equations methods (probably espe cially where samples are small and specification error is present). . . Oi ([11], p. 45) quotes Theil approvingly: Therefore, after reviewing all arguments we should conclude that although the method of least-squares can no longer claim to have the brilliant properties which earlier econometricians thought it had, it can be regarded as one of the few one-eyed men who are eligible for king in the country of the blind— at least as far as ex perimental small-sample estimation unaided by significant a priori information is concerned. Estimates o f regression coefficients and the de termination of lag structures cannot be accom plished with certainty. The fact that we have lagged variables and that we have only small samples, while our statistical techniques refer chiefly to large sample properties with current values, only highlights the unsettled nature of measuring econom ic relationships. The contro versy over the choice o f estimation procedures is far from being settled when skeptics have a larger arsenal o f weapons to use in critizicing research results than constructive researchers have for their purposes. “These problems must be faced, however, if an attempt is to be made to estimate the structure o f the econom y by empirical meth ods. Rejecting all empirical results out o f hand because o f visible disagreement among different studies will not help bring about a solution to these problems.” ([3], p. 2) REFERENCES Books Periodicals and Other 1. Bronfenbrenner, Jean. “Sources and Size of Least-Squares Bias in a Two-Equation M odel.” H ood, William C., and Koopmans, Tjalling C. (eds.). Studies in E conom etric M eth o d. N ew York: John W iley & Sons, Inc., 1953. 2. Christ, Carl F. Econom ic M odels and M eth ods. N ew Y ork : John W iley & Sons, Inc., 1966. 3. Evans, M ichael K. M acroeconom ic A ctivity: Theory, Forecasting, and C on trol. N ew York: Harper & Row, 1969. 4. Fisher, Franklin M. “D ynam ic Structure and Estimation in Economy-W ide Econometric M odels.” Duesenberry, James S. et aL The Brookings Q uarterly E conom etric M odel o f the U nited States. Chicago: Rand M cN ally and Co., 1965. 5. Fisk, P. R. Stochastically D epen den t Equa tions: A n Introductory T ext fo r E conom e tricians. N ew York: Hafner Publishing Company, 1967. 6. Goldberger, Arthur S. E conom etric T h eory. N ew York: John W iley & Sons, Inc., 1964. 7. Johnston, J. E conom etric M o d els. N ew York: M cGraw-Hill Book Company, 1963. 8. Kane, Edward J. E conom ic Statistics and E conom etrics. N ew York: Harper & Row, 1968. 9. Walters, A . A. A n Introduction to E conom e trics. London: M acmillan and Co., 1968. 10. Haavelmo, T. “M ethods o f Measuring the Marginal Propensity to Consum e,” Jour nal o f The A m erican Statistical A ssocia tion, Vol. 42 (Mar. 1947), pp. 1 0 5-22. 11. Oi, Walter Y. “On the Relationship A m ong Different Members o f the K-Class,” Inter national E conom ic R eview , V ol. 10 (Feb. 1 9 6 9 ). 12. Richardson, D avid H ., and Wu, De-M in. “A N ote on the Comparison o f Ordinary and Two-State Least-Squares Estimators.” U n published research paper, University of Kansas, 1969. 13. Sawa, T. “The Exact Sampling Distribution o f Ordinary Least Squares and Two-Stage Least Squares Estimators,” Journal o f The A m erican Statistical A ssociation, V ol. 64 (Sept. 1969), pp. 9 2 3 -3 7 . by John Kareken, Thomas Muench, Thomas Supel, and Neil Wallace DETERMINING THE OPTIMUM MONETARY INSTRUMENT VARIABLE CONTENTS 87 INTRODUCTION 88 I. QUADRATIC UTILITY AND A SIMPLE ECONOMIC STRUCTURE 89 II. QUADRATIC UTILITY AND A COMPLEX ECONOMIC STRUCTURE Distributions of parameters, disturbances, and exogenous variables Results 90 91 92 III. THE REAL INCOME-VARIANCE OF PRICE UTILITY FUNCTION 93 IV. CONCLUSION 94 APPENDIX: Holbrook and Shapiro on the Optimum Monetary Instrument Variable OPTIMUM MONETARY INSTRUMENT VARIABLE INTRODUCTION For some time monetary economists and officials have been debating how central banks ought to operate. Should the Federal Reserve, for example, seek to control one or another of the monetary aggregates? And if so, which one? Or should it control some interest rate or rates? We do not know how the Federal Reserve, or for that matter any other central bank, ought to operate. We do, though, know what seems to us a not unreasonable way of decid ing; a way, that is, of determining the opti mum monetary instrument variable. And in this paper we explain or, better, illustrate our way. The central bank that is certain about the economic structure constraining it or does not care about the variance of policy outcomes can, with complete indifference, use any possi ble instrument variable. It is difficult, however, to imagine any central bank being certain or not caring about the variance of policy out comes. The presumption must therefore be that most if not all central banks have a true choice to make: namely, which of all possible instrument variables to use. 1 And what we would have central banks do is decide by max imizing their respective expected utilities; or in other words, by comparing the maximum ex pected utilities associated with all the various possible instrument variables. What in effect we would have the Federal Reserve do is cal culate alternative opportunity loci, there being one such for each possible instrument variable, and then, having specified values for its target variables, determine which of these loci or constraints allows it to achieve the greatest ex N o t e .— Messrs. Kareken, Muench, and Wallace, all of the Economics Department, University of Minne sota, are Consultants to the Federal Reserve Bank of Minneapolis; Mr. Supel is Senior Economist at that Bank. They would like to thank Arthur Rolnick for his very considerable help, and the Minneapolis Bank for its financial support. 1 See W. Poole, “Optimal Choice of Monetary Pol icy Instruments in a Simple Stochastic Macro Model,” Quarterly Journal of Economics, (May 1970), and J. Kareken, “The Optimal Monetary Instrument pected utility. We would have the Federal Re serve do this not once but, since the cost is not much, at the beginning of every policy period. Nor is it impractical to suggest this. How ever complex the underlying economic structure, opportunity loci can be calculated. In Section I, we use a very simple economic structure. But we do so only because our pur pose there is to explain our way of determin ing the optimum monetary instrument variable; and it is convenient in explaining to use a sim ple structure. In Sections II and III, wherein we derive actual Treasury bill rate and demand deposit loci, we use the Federal Reserve-MITUniversity of Pennsylvania economic structure, which is very complex. 2 We do then provide some numbers or ex perimental findings. We would caution, how ever, against paying much attention to them. They are not, we think, even suggestive of how the Federal Reserve ought to operate. We decided to include them in the paper only be cause they show that our way of determining the optimum monetary instrument variable is practical. But our way is practical or feasible only for myopic central banks, for those concerned only about current-period developments or, by way of approximation, willing to pretend that they are. It is no accident that in Sections I and II we take utility as depending simply on current-period nominal gross national product and in Section III as depending on currentperiod real GNP and the current-period change in the price level. Had we taken utility as de pending on future-period values as well, we would not have been able to go further; we would not have been able to show the practic ability or feasibility of calculating and compar ing the maximum expected utilities associated with the various possible instrument variables. Variable,” Journal of Money, Credit, and Banking, (Aug. 1970). 2 Hereinafter, we shall refer to the FR structure. There is, we understand, a new version. If so, we used an old version, the one described in part by F. de Leeuw and E. Gramlich in “The Federal ReserveMIT Econometric Model,” Federal Reserve Bulletin, (Jan. 1968). It is not known what policies are optimal for a central bank that is uncertain about the true values of structural parameters and whose con cern extends into the future. We might have proposed comparing the ex pected utilities of arbitrary rather than optimal policies. But which ones? Or we might have proposed that variances of structural parame ters be ignored. It seemed to us, however, that uncertainty about parameters is an important fact of life and that we ought therefore to take utility as depending only on current-period val ues of target variables. Some readers might want to object that the Treasury bill rate and the stock of demand deposits are not possible Federal Reserve in strument variables. We believe, though, that the Federal Reserve if it wanted to could de termine the bill rate exactly. It would only have to announce a price for bills. And is coming quite close to some preassigned value for, say, the 3-month average of demand de posits impossible? We think not. But it does not really matter if we have been inept in se lecting possible Federal Reserve instrument variables. Our way might be used for choosing between (or among) other possible instrument variables. Let the economic structure be (1) Y = so + + ex and (2) m — S2 + s$Y + Sat + e% Equation 1 describes nominal aggregate de mand as a function of the interest rate, r, and equation 2 the condition for equality between the actual stock of demand deposits, m, and the desired stock. The monetary authority is uncertain about the values of the parameters, s0, su . . . , s4 and about the values of the dis turbances ex and e2. If r is used as the instrument variable, the reduced-form equation for Y is equation 1. If m is used as the instrument variable, it is (3) Y = sh + s&n + e 3 where SiSo S2Si s& = ----- 1------S4 + ^ 3 *6 = ----:----£4 + S\Sz and I. QUADRATIC UTILITY AND A SIMPLE ECONOMIC STRUCTURE Let the monetary authority’s utility function be U = - C T - Y) 2 where Y is nominal current-quarter GNP and Y is the desired or target value of Y . Then EU = - V Y - (EY - Y) 2 where E stands for expected value and V for variance. Iso-expected utility contours are parabolas, symmetric about E Y = Y, in the positive quadrant of the (E Y , V Y ) plane. The relevant opportunity loci, or constraints subject to which EU is maximized, are therefore all at tainable combinations of E Y and V Y . £3 = s^ei — S\ € 2 -----------------------$4 + SiS3 From these reduced forms, the two loci can be obtained. To illustrate, from equation 1, (4) and EY(r) = Es 0 + rEsi + Ee 1 (5) VY(r) = V(So + *0 + r2Vsi + 2rC(s0 + ei, Si) where C stands for covariance. Solving equa tion 4 for r and substituting the result into equation 5 gives the r-locus (6 ) VY(r) = Co + cxEY{r) + c&EY{r)Y OPTIMUM MONETARY INSTRUMENT VARIABLE where VsJE(s0 + e1)Y Co = K(so + ei) + (ESly 2E(sq + ei)C(sQ+ ^1? $i) Esi Ci = 2C(s0 + eu ^i) Es i 2VsiE(s0 + ei) (^ i) 2 c2 = Vsi/(Esi) 2 Equation 6 gives all combinations of E Y and V Y attainable when r is used as the instrument variable. The opportunity locus for m, the m-locus, is obtained in the same way as the r-locus was, but from equation 3. As we show in Section II, traditional or classical estimation of equations 1 and 2 pro vides the basic information needed to determine numerical values for the coefficients of the rlocus (that is, for c0, cu and c2) and for the coefficients of the m-locus. And with numerical opportunity loci, the monetary authority can determine its optimum instrument variable. All it has to do is specify a target value for Y . It is worth pausing briefly here to consider what it means to determine numerical opportu nity loci by traditional estimation of the economic structure. Each variance of possible outcomes of Y , for example VY(r), combines true randomness in the economy and uncer tainty about the values of structural parameters. Indeed, VY(r), like VY(m), is a forecast vari ance; that is to say, a variance of forecast Y around “true” or actual Y. To be sure, the randomness of “true” Y is entirely attributable to ex and e2. But the monetary authority, in making its instrument variable choice, must also be influenced by how certain it is about parameter values. Suppose that when m is used as the instrument variable, Y is partly deter mined by some parameter the value of which is extremely uncertain; and when r is used as the instrument value, Y is not determined even in part by this parameter. If at all averse to risk, the monetary authority should then, ceteris paribus, use r as its instrument variable. II. QUADRATIC UTILITY AND A COMPLEX ECONOMIC STRUCTURE The FR economic structure is, as we have said, very complex. There are many behavioral equations, some of which are nonlinear. It can be written Fi(x, z, r, au et) = 0 (z = 1, 2, . . . , K ) where x is a vector of the current values of endogenous variables, K in number; z is a vec tor of contemporaneous, nonpolicy exogenous variables; r is the rate on 3-month Treasury bills; ai is a vector of parameters; and e » is a disturbance. If all nonlinear terms in x and r are approximated by first-order Taylor expan sions, then the structure can be written (7) Ax = Br + C where A is a KxK matrix with elements a»y; B is a Kx 1 matrix with elements bn\ and C is a Kx 1 matrix with elements Cf. Also, a<j = 2 , r°, a,-, et) bn = z, r°, at, et) and Ci = hi(xQ, z, r°, au e{) where jc° and r° are the values of x and r used in making the model linear. It follows that ( 8) Xi = Y — dur + dio and (9) x 2 = m = d2\r + d‘>[ where dn = A j 1 B, A'l being the/th row of A~l, and dj 0 — A f 1 C. Then Y(r) = dnr + rfio and = Dnm + Dio So what is required are estimates of the first two moments of the vectors (dlu d10) and (Dn , Z)10). But since the d's and D’s are com plicated functions of the underlying random variables—the parameters an, the disturbances eif and the contemporaneous values of noninstrument exogenous variables Zi—their distri butions cannot be derived analytically from the distributions of the underlying random vari ables. 3 It is possible, though, to sample from the distributions of the underlying random vari ables, insert the sampled values into equation 7, and solve for values of the reduced-form coefficients, the d’s and the D’s. By repeated sampling, a set of values of the d’s and the D’s is built up, from which moments can be estimated and numerical opportunity loci derived.4 It is also possible to proceed differently. Relevant opportunity loci can be determined point by point from a nonlinear structure. For each of a set of values of r and each of a set of values of m, a sample of values of Y is gen erated and estimates of the first two moments are calculated. We decided against proceeding this way in part because of the cost. A great many simulations would have been required: 2 (nxp) simulations, in fact, in order get p points on each locus, using n observations on Y for each point. 3 Even for very simple economic structures, such as that of Section II, it is difficult if not impossible to derive the distributions of the reduced-form coeffi cients as functions of the moments of the structural parameters. To determine the numerical loci implied by the structure of Section II, it would therefore also be necessary to sample from the joint distributions of the structural parameters that are consistent with statistical estimation. 4 Why derive numerical opportunity loci rather than calculate expected utilities? It is just that to cal culate expected utilities, Y, the desired or target value of Y, must be known or assumed. But having derived numerical loci, one may find dominance in a neighborhood of some reasonable value for Y — that one variance is smaller than the other at every value of E Y in the neighborhood. Clearly, deriving numeri cal loci is for outsiders. DISTRIBUTIONS OF PARAMETERS, DISTURBANCES, AND EXOGENOUS VARIABLES. We assumed that the mean of each parameter in Fi is equal to the corre sponding estimate, that the variance-covariance matrix of a set of parameters is equal to a con stant times the variance-covariance matrix of the corresponding estimators, and that the variance of the disturbance in Fi is equal to a constant times the corresponding residual variance. 5 Sample values of «i, the vector of parameters in the ith equation, not the original structure, were generated jointly according to the matrix equation <*i — + R{V where a* is the vector of point estimates of aif Ri is a matrix such that RiR'i equals the es timated variance-covariance matrix of at, and v is a vector of random variables chosen inde pendently of one another from a normal distri bution with mean zero and variance one trun cated at plus and minus two. The disturbance for the zth equation was generated according to ei = <iiV where ^ is the estimated residual standard error for the zth equation and v is a single in dependent drawing from the same truncated normal. It follows that the expected value of is that the variance-covariance matrix o f i s 0.77 times the variance-covariance ma trix for that the mean of ex is zero, and that its variance is 0.77 v f. (The constant is 0.77 because we inadvertently failed to recognize that the variance of the truncated normal is 0.77 and not unity.) We chose a truncated distribution for v be cause many of the equations of the FR struc ture are in a form inconsistent with an un5 Thus, the data requirements for each estimated equation are the point estimates of the coefficients, the point estimates of the residual variance, and the inverse of the relevant cross-product matrix of the independent variables. The coefficient and residual variance estimates were available, but the cross-prod uct matrices had to be re-estimated. OPTIMUM MONETARY INSTRUMENT VARIABLE limited range for the disturbance. For example, several of the estimated equations for interest rates are linear, so that disturbances from a distribution with unlimited range could produce negative interest rates. Also, we did a certain amount of linearization and thereby changed some estimated equations which originally had forms that constrained the dependent variables to proper ranges. There are quite a few noninstrument exoge nous variables in the FR structure that can be treated as random. These include population, Federal Government expenditures, and exports. We assumed that these variables are generated by second-order autoregressive schemes, Z t ,% — P o t + P u Z t-i,i + P u Z t-2 ,i + U t, i The /?’s were taken as fixed and equal to the estimated coefficients from an ordinary least squares regression of z% on two lagged values of itself over the period 1952-Q1 to 1968-Q4. (It was an oversight that we did not also take the fi's as random.) The disturbance, ut)i, was treated as random with mean zero and variance equal to 0.77 times the estimated residual variance from that regression. The distributions of the exogenous variables can play an important role in determining the better instrument variable. In a simple model, the less variance in the aggregate demand schedule the more likely is it that the interest rate is the better instrument variable. Inability to forecast exogenous variables like govern ment expenditures and exports contributes di rectly to variance of aggregate demand. Thus, if there are schemes that forecast those vari ables with smaller error variance than do our autoregressive schemes, our failure to use them would seem, on the whole, to favor demand deposits as the optimum monetary instrument variable. RESULTS. We derived opportunity loci for the first quarter of 1969 using 100 random drawings. 6 With r as the instrument variable 6 Deriving loci for 1969-Q1, we linearized the FR structure around values for 1968-Q4. E(Y) = 884.9 - .819r and V(Y) = 361.0 - 2(.671)r + ,088r2 where r is measured as a per cent per annum and Y is measured in billions of dollars at an annual rate. Therefore, the r-locus is V(Y) = 102,012 - 231.4 £(7) +.13166 [E(Y)¥ The highest value of E(Y) for which the locus has any meaning is E(Y) — 884.9, since there r = 0. At r = 10, E(Y) = 876.7. The locus is drawn in Figure 1 for approximately that range of values. We would expect our estimated locus to most closely approximate the locus obtained from the original nonlinear model in the vicinity of r = r°, the value around which we linear ized, or in the vicinity of E(Y) — 880.3. With m as instrument, E(Y) = 805.8 + .495m and V(Y) = 1067.0 - 2(5.178)w + .0365m2 where m is in billions of dollars. Therefore, the m-locus (also shown in Figure 1) is V(Y) = 114,713.7 - 26\2E(Y) + .14909[£(7)]2 Note that in Figure 1 m dominates r as an instrument variable. For any expected value of y, the variance of Y is smaller with m as the instrument variable than with r as the instru ment variable. But the difference between the FIGURE 1 MEAN-VARIANCE LOCI E|Y] V|Y) 92 variances at, say, E(Y) = 880 is 20, and 20 is not a significant difference. For a sample of 100 drawings, a 90 per cent confidence interval around the variance of the w-locus at E(Y) — 880 ranges from 269 to 432, whereas the cor responding interval for the r-locus ranges from 285 to 457. There is, therefore, considerable overlap of the confidence intervals. given value of the variance of the deflator, fair gambles on X are always rejected. The relevant opportunity loci consist of all attainable com binations of E log X and 104E[(P — P°)/P0]2. These were obtained for r and for m as follows. Whether r or m is used as the instrument variable, there are reduced-form equations for both real income and the deflator. Let X — b\r + &2 P — b ^ -f- b\ III. THE REAL INCOME-VARIANCE OF PRICE UTILITY FUNCTION We also derived the first-quarter 1969 opportu nity loci relevant for maximization of expected utility, where U = \ogX - 6[100(P - P°)/i >0] 2 X is real GNP in 1958 prices, P is the GNP deflator, and P° is the deflator for the fourth quarter of 1968. Iso-expected utility contours for this function are straight lines with slope b in the [E log X , 104E(P - P°/P°)2] plane. The log function implies risk aversion; at a FIGURE 2 THE MEAN INCOME-PRICE VARIANCE LOCI E log X be those for r. Thus, E(log X) = E log (ihr + b2) so E log X cannot be written as a function of r and of the moments of bx and b2. It is pos sible, however, to compute E log X for each value of r in a reasonable range. We let r range from 1 per cent to 10 per cent. For each value of r, we computed and averaged log ( b j + b2) over the sample of values of bx and b 2 and took the resulting average as our estimate of E log X . From the reduced form for P, we have OPTIMUM MONETARY INSTRUMENT VARIABLE ings, 90 per cent confidence intervals around those estimates are almost coincident. = V W b 'E Q x * ) + £(&,*) + (P0) 2 - 2P°(rEbs + Ebt) + 2rE(b,bi)] IV. CONCLUSION Selected values of E log X and 104E are given in Table 1. Some values for the m~ locus, which were obtained in the same way using the reduced-form equations for my are also given in Table 1. Both loci are shown in Figure 2. TABLE 1: Selected Values For Real Income—Price Variance Loci /71-loCUS r-locus r S lo g * 1 6 .5 6 9 6 .5 6 8 6 .5 6 7 6 .5 6 6 6 .5 6 5 6 .5 6 4 6 .5 6 3 6 .5 6 3 6 .5 6 2 6.561 2 3 4 5 6 7 8 9 10 io < £ | .2 5 2 .2 5 0 .247 .2 4 4 .241 .239 .2 3 6 .2 3 3 .231 .2 2 8 m E lo g X 140.1 142.1 144.1 146.1 148.1 150.1 152.1 154.1 1 5 6.1 158.1 6 .5 5 9 6 .5 6 0 6 .5 6 1 6 .5 6 2 6 .5 6 3 6 .5 6 4 6 .5 6 6 6 .5 6 7 6 .5 6 8 6 .5 6 9 10«E ( t )' .2 2 2 .2 2 6 .2 2 9 .2 3 2 .2 3 5 .2 3 9 .2 4 2 .2 4 5 .2 4 9 .2 5 2 Once again m dominates r; at each value of E log X the variance of the deflator is smaller for the /rz-locus than it is for the r-locus. The difference, however, is miniscule. At E log X = 6.5647, which corresponds to r = r° for the r-locus, the percentage variance of the deflator for the w-locus is 0.2393, while that for the r-locus is 0.2397. For a sample of 100 draw As indicated in the introduction, we think that little attention should be paid to our ex perimental findings. It is not only because our samples were too small, but also because, to calculate numerical loci, it is necessary to as sume a utility function and, what is more, an economic structure. And to accept calculated loci, or a comparison thereof, is to accept the assumed utility function and the assumed struc ture. Even if our samples had been larger, we would not then have cared to press our findings. Before doing that, we would want to average over time 7 and several economic structures. But more fundamentally, we feel that no monetary authority should decide once and for all, by statistical inference, which of its possible instrument variables to use. Unless faced with prohibitive costs, it should decide which vari able to use at the beginning of every policy period or possibly every quarter. This ulti mately is why we could in good conscience content ourselves only with offering a way of determining the optimum instrument variable (and with a sample of only 1 0 0 drawings). 7 See the appendix, wherein we appraise the at tempt of Holbrook and Shapiro to determine empiri cally the optimum monetary instrument variable. APPENDIX: Holbrook and Shapiro on the Optimum Monetary Instrument Variable There has been one attempt that we know of, by Holbrook and Shapiro (H&S), to determine empirically the Federal Reserve’s optimum in strument variable. 1 What H&S did was to cal culate and then to compare certain variances of real GNP, variances associated with three pos sible monetary instrument variables: the nar rowly defined stock of money, the monetary base, and, what would seem a rather surprising choice, the average rate on long-term Treasury bonds. 2 What they found was, for every calen dar quarter in a long stretch of years, a smaller variance for the narrowly defined money stock than for both the monetary base and, by a much wider margin apparently, the average Treasury bond rate. Thus, their tentative con clusion was that, in setting its policy, the Fed eral Reserve ought to use the narrowly defined money stock. But H&S calculated and so compared the wrong variances. They went astray, we suspect, because they forgot that there must be dis turbance terms in their structural equations. Whatever the explanation, though, they cannot be regarded as having made a case, even a highly provisional case, for the narrowly de fined money stock as the Federal Reserve’s optimum monetary instrument variable. H&S distinguished between actual GNP, de noted here by 7, and predicted GNP, denoted here by yp. Suppose that 1 See Robert Holbrook and Harold Shapiro, “The Choice of Optimal Intermediate Economic Targets,” American Economic Review, May 1970, pp. 40-46. 2 Holbrook and Shapiro referred to the narrowly defined stock of money, the money base, and the av erage rate on Treasury bonds as possible intermedi ate target variables. But they assumed that the Fed eral Reserve is able to determine exactly any one of these three variables, so it is quite proper for us to refer to them here as possible instrument variables. (1) C = axY + Ux (2 ) (3) / = v= #2 + 03 r + U% cisY H- Qtfti -f- C/3 #4 (4) Y = C+ I where C and / are, respectively, consumption and investment, r and m are the two possible monetary instrument variables, respectively, the rate of interest and the stock of money and Ulf U2, and U3 are random disturbances. 3 Then (5) Y(r) = 1 1 (i02 + a3r + Ui + U2) - ai and (6 ) Y(m) = 1 1 — 01 — 0305 (0 2 + 0304 + azdtfn + Ui + U2 + 0»t/*) where Y (r) is actual GNP with r as the instru ment variable and Y (m ) is actual GNP with m as the instrument variable. Also (5a) 1 y»(r) = 5 " (0 2 + - ai 1 0 *r) and (6 a) yp(m) = 1 - -x -s - (# 2 + 1 — d i — <3305 0304 + 030em) where yp(r) is predicted GNP with r as the instrument variable, yp(m) is predicted GNP with m as the instrument variable, and 0 * is the estimator of a 3 This economic structure is far simpler than the one specified by H&S. But since we want only to illus trate wherein they went wrong, we do not need even a faintly realistic structure or more than two possible monetary instrument variables. H&S failed to include disturbances in describing their model, but they must surely belong there, for otherwise the model must be rejected unless the data fit it exactly. OPTIMUM MONETARY INSTRUMENT VARIABLE The loss function explicitly assumed by H&S is (7) L(x) = [yp(x) - Y(x)Y (x = r, m) where Y p(r) and Y p{m) are the first-order Taylor expansions of, respectively, >>p(r) and yp(m) around the point a = (a*, a2> . . . , o6). Since (8 ) Yp(r) = y ~ ~ t« 2 + fl> + (fli - ai)^] I —Qi and (9) W = ! + ^ + a3ma6 + (Si — ai)7 + (dz — a3)r + ^ 3(^ 5 ~ 05 )^ it follows that 4 (10) EL{r) = E[Yp(r) - 7(r) ] 2 _ ™ + „(]* ± £ ) and (11) EL{m) = E[Yp(m) ~ i'M P PW \ 1 — fll — G3^5 / £L (r) is the expected loss with r as the instru ment variable and EL{m) is the expected loss with m as the instrument variable. The straightforward procedure would seem to be to minimize E L (r) by the choice of r and to minimize EL(m) by the choice of m and then to compare the respective minima. But doing so would amount to assuming that the monetary authority does not care about the expected value of Y. H&S therefore assumed that “the policy maker . . . select(s) the value of each intermediate target variable such that the expected value of income is equal to desired income, and then . . . choose(s) among (in strument) variables that one which minimizes the expected squared deviation of actual from desired income.” So H&S would have the 4 This formula for forecast error holds exactly only in the post-sample period, for in the sample pe riod there is also a covariance term. monetary authority minimize E L (x ), but sub ject to the constraint (12) EYp(x) = Y where Y is the target value of Y . But they themselves did not compute their constrained minima of E L (x ), that is, EL(x). They forgot to calculate the second terms on the right-hand sides of equations 1 0 and l l . 5 This is hardly a minor oversight. Those terms would remain even if the sample size were indefinitely large. And we suspect that for their estimated model the omitted terms are relatively large. A ranking of instruments by V Y p(x) in no way implies a ranking by EL(x). Even if H&S had not forgotten the second terms on the right-hand sides in equations 1 0 and 1 1 , they would have ended up calculating the wrong variances. For in calculating vari ances, they used actual values of both r and m (that is, ra and ma). And as is easily shown, E Y p(ma) is not in general equal to E Y p(ra). The expectation of Yp at r = ra is, by equation 8, (13) EYp(ra) = 7 1 ~ ~ [a2 + W a] — a\ assuming unbiased estimators of the a^s. From equation 3, it follows that (14) mtt = — [ra — at — a 5y(ra) — t/3] and from (5) that (15) ma = + (X\Q\ - [(1 - ai - aza5)ra — “b O 5U2) ~b U3 — O 1 C /3 ] But the expectation of Y p(m) at m = ma is, by equations 9 and 15, (16) EYp(ma) - - j [a2 + aarj Q z ia ^ U i (1 + CI5U2 — ai - aza5) (1 - ax) 5 In footnote 7, p. 45, they recognize but do not deal with this omission. So at any point in time E Y p(ma) ^ E Y p(ra) unless, by some chance, all Ui$ happen to be zero. Thus, if actual or observed values of both (all) possible instrument variables are used in calculating variances, the resulting variances will correspond to different mean values of Y p, and a comparison of variances corresponding to the same value of E Y P, which is what H&S proposed, is not achieved. by James L. Pierce THE TRADE-OFF BETWEEN SHORT- AND LONG-TERM POLICY GOALS CONTENTS 99 INTRODUCTION 99 SHORT-RUN VS. LONG-RUN GOALS 101 SOME SIMULATION EXPERIMENTS 103 CONCLUSIONS TRADE-OFF BETWEEN POLICY GOALS INTRODUCTION The existence of long lags in the response of the real sectors of the economy to changes in monetary policy is well documented. These lags may require an horizon for monetary pol icy strategies that spans many calendar quar ters. Even if long planning horizons are desira ble, specific operating strategies still must be adopted for the actual short-run conduct of monetary policy. These, however, should be consistent with the long-term goals. If shortrun considerations—such as stabilization of money market interest rate movements—cause modification of the operating strategy, the long-run goals in terms of income, employ ment, and the price level may suffer. This paper discusses some of the areas in which short- and long-term goals may conflict and at tempts to evaluate the costs to the long-term targets of imposing short-run side conditions on policy actions. SHORT-RUN VS. LONG-RUN GOALS Available econometric evidence indicates that variations in monetary policy instruments can exert little influence on the nonfinancial sectors of the economy in the short run. Ex periments with a recent version of the Federal Reserve-MIT model indicate that, other things equal, a $ 1 billion increase in the money stock in a given quarter will produce only a $0.3 billion increase in nominal gross national product in that quarter. Further, inspection of the coefficients for the relevant equations in the model suggests that even this small response is probably overstated. It is in teresting to note that the long-run multiplier relation between money and nominal GNP is substantial. Other things equal, a $ 1 billion permanent rise in the money stock leads to a permanent increase in nominal GNP of ap proximately $3.2 billion. N o t e . —The author, who is Associate Adviser, Di vision of Research and Statistics, would like to thank William Poole for his constructive comments on an earlier version of this paper. Given the short-run multiplier, attempts to establish short-run (quarter by quarter) con trol over the economy may require variations in policy instruments that are unacceptably large. An example may clarify the issue. As sume that during a generally inflationary pe riod, the decision is made to attempt to stop the inflation within a single quarter. To ac complish this end, a sharp rise in interest rates, and probably a substantial reduction in the levels of the monetary aggregates, would be required during the quarter. Even if this strategy were successful, a new problem would immediately develop. With the passage of time beyond the quarter, the economy would con tinue its deflationary adjustment—probably at an increased rate—in response to the mone tary restriction. If an overresponse of the economy to the original policy restriction is to be avoided, policy must reverse itself immedi ately by sharply reducing interest rates and ex panding the monetary aggregates. This easing of policy would require in turn a restrictive policy the next quarter. Thus, by never looking more than one quarter ahead, large short-term reversals of policy would be required to stabi lize the economy. Whether this myopic strategy of trying to hit targets in the real sector on a quarter-by-quarter basis can be successful over the long run depends, among other things, upon the existing parameters of the system. 1 It is quite possible that pursuit of such a strategy would have no long-run future because ever larger changes in monetary policy instruments would be required to achieve stability in the real sector. Even if the strategy produced permanent economic sta bility, it could create extreme fluctuations in financial markets. It is quite possible, however, that large fluc tuations in financial variables would alter in 1 For a simple treatment of this problem see E. Gramlich, “The Usefulness of Monetary and Fiscal Policy as Discretionary Stabilization Tools,” pre sented at the Conference of University Professors sponsored by the American Bankers Association, Sept. 1969. terest rate expectations enough to weaken greatly the efficacy of the myopic policy strat egy. Rapid reversals of monetary policy may encourage investors to expect wide fluctuations in short-term interest rates. In this situation, efforts to reduce long-term rates would be thwarted by investor expectations of a rise in rates in the near future. Thus, the pursuit of the myopic policy strategy could be self-defeat ing. There are two obvious ways to approach the problem posed by the small amount of short term control over the economy. First, mone tary policy could pursue the myopic rule of attempting to hit a target quarter by quarter but could subject the strategy to constraints imposed by financial conditions. Thus, a spe cific target value for employment or for the price level would be pursued provided the act of attempting to hit the target did not cause excessive fluctuations in interest rates. If inter est rates moved more than was deemed desira ble, policy instruments would be changed suffi ciently to bring interest rates within the allowable range. The imposition of such con straints could greatly reduce the ability of monetary policy to achieve short-term goals. The second approach would involve a lengthening of the policy-planning horizon. In this situation, policy would take a view longer than one quarter into the future. The aim would be to achieve the best path of, say, em ployment over some interval of time consistent with acceptable performance of financial mar kets. Extension of the horizon would allow problems of the real sector and of the financial sector to coexist on a more equal basis. No immutable constraints would be placed on the system by money market conditions if the planning horizon could be extended. However, by giving up some short-term control over var iables in the real sector, it should be possible to reduce fluctuations in financial variables to more manageable proportions. Conceptually, it should be possible to deter mine the trade-off between ( 1 ) short-term control over employment and prices and ( 2 ) stability of the financial sector. In general, a lengthening of the policy-planning horizon to promote short-run stability in financial markets will come at the cost of reduced control over nonfinancial variables. Alternatively, a shorten ing of the planning horizon will come at the cost of increased short-run fluctuations in fi nancial variables. Lengthening the horizon for major policy goals raises some obvious problems. Because the long-term goals of employment and prices are relatively far in the future, it is easy to give them a back seat to the short-run stabili zation problems often encountered in financial markets. The problem with this approach is that overattention to short-run problems may have important implications for the paths re quired to hit desired long-run targets. Further, if short-run constraints are continually im posed, it may be impossible to hit the long-run goals in the time specified. Under those cir cumstances it may be necessary to lengthen the horizon and to accept the ensuing costs of less desirable performance of the real sector. The previous paragraph suggests that over the longer run the goals of price and output stability may not conflict with the goal of money market stability. Overzealous attempts to stabilize the money market in the short run may distort output and prices to the point that large changes in interest rates are required in the longer run to bring the economy under control. By allowing wider short-run fluctua tions in money market conditions, it might be possible to avoid large swings in interest rates over the longer run. The discussion suggests that, given a set of initial conditions in the economy, there is an optimal policy strategy available. The strategy determines simultaneously the length of the planning horizon, the paths of target variables such as employment and prices over the pe riod, and the expected stability of financial markets. The determination of specific strate gies is a problem in optimal control theory and is beyond the scope of this paper. Instead, the paper attempts to assess the trade-offs involved and illustrates problems that may arise from pursuing particular policy strategies. TRADE-OFF BETWEEN POLICY GOALS SOME SIMULATION EXPERIMENTS This section describes some simulation ex periments that were conducted to illustrate the problems encountered when short-term and long-term goals conflict. The structure of a re cent version of the FR-MIT model was used for the simulation exercises. 2 The first experiment assumes a monetary policy that focuses on the rate of growth of the money stock provided the change in the Treasury bill rate over any quarter does not exceed some arbitrary value. An unconstrained growth in money is assumed to promote de sired long-run behavior of the real sector. However, if the policy-determined money stock for a quarter led to a projected change in the bill rate over that quarter that exceeded the constraint value, then the money supply was changed sufficiently to bring the change in the bill rate back to its allowable range. In those situations in which monetary policy is at tempting to offset either boom or recession, this constrained policy would lead to a per formance of the economy that is inferior to one which is unconstrained. If shifts in the demand for money are the source of wide interest rate fluctuations when policy is attempting to hit a money stock tar get, the situation is changed. Here, it would be appropriate to introduce interest rate con straints. Such constraints would automatically satisfy the demand for money after some point. Limiting interest rate movements in this case would promote long-run stability. 3 The re sults of the simulation experiments suggest, however, that one should have strong reasons for believing that shifts in money demand are 2 Some of the simulation results reported here are drawn from an earlier paper on a related topic. See J. Pierce, “Some Rules for the Conduct of Monetary Policy," in Controlling Monetary Aggregates (Fed eral Reserve Bank of Boston, 1969). 3 For a theoretical discussion of the desirability of interest rate versus money stock stabilization in a stochastic world, see W. Poole, “Optimal Choice of Monetary Policy Instruments in a Simple Stochastic Macro Model,” Quarterly Journal of Economics, Vol. 84 (May 1970), pp. 197-216. causing wide quarter-to-quarter fluctuations in interest rates. If unexpected shifts in aggregate demand are the cause, long-run goals may suf fer greatly. To illustrate the problems that arise during periods of excess aggregate demand, various simulations of the FR-M IT model were run for the 1963-68 period. First, a control simu lation was run that took all exogenous varia bles at their historical values but assumed that the money stock grew at a constant annual rate of 4.25 per cent. This was the constant rate at which the initial money stock in 1962IV had to grow to achieve its actual value in 1968-IV. Then additional simulation experi ments were conducted by applying the same exogenous variables and the same 4.25 per cent money growth rate to the model provided that the Treasury bill rate did not change dur ing the quarter by more than a specified abso lute amount. If the bill rate fell outside the al lowable range, bank reserves and the money supply were changed sufficiently to bring the bill rate back to the nearest boundary of the range. All other exogenous variables were as sumed to remain unchanged. Several absolute change values were attempted; results for ab solute changes of 30 basis points and 10 basis points are reported. The results indicate that the placement of sufficiently narrow bounds on the change in the bill rate can have a large impact on the simulated value of GNP. Figure 1 shows the differences between the simulated values of GNP for the steady rate of growth of money and those subject to maximum absolute changes in the bill rate of 30 and 10 basis points, respectively. In both cases, because in terest rates could not rise in the later periods, there was a tendency to add to the existing excess demand conditions. As indicated earlier, if interest rate fluctua tions are caused by erratic shifts in the de mand for money, then stabilization of interest rates may be a reasonable course of action. The simulation results suggest, however, that inter est rate stabilization can be costly during peri ods of strong excess demand. 1 0 2 FIG U RE 1 Effect on GNP of 4 .2 5 % GROWTH IN MONEY SUBJECT TO MAXIMUM ABSOLUTE CHANGE IN BILL RATE Deviations from STRAIGHT 4 .2 5 % MONEY-GROW TH SIM ULATION Bill ion s o f dol lars It is interesting to note that if stabilization of financial markets takes the form of con straining the rate of growth of the money stock, the problems encountered during periods of shifting aggregate demand are diminished. Assume that monetary policy attempts to hit an employment target by setting market inter est rates at appropriate levels. Introducing a constraint on the allowable range of growth rates of the money stock in this situation can under some circumstances lead to improved performance of the economy. If it happens that the interest rate selected is not the cor rect one because aggregate demand is either stronger or weaker than expected, variations in the rate of growth of the money stock can pro vide important evidence of this condition. For example, if aggregate demand is stronger than expected, given the interest rate and the de mand for money, the growth in the money stock will be greater than expected. If the ac celeration in the growth rate of money is taken as a signal to raise the interest rate, the growth rate of money will fall and the excessive growth in aggregate demand will be reduced. If the unexpected growth in the money stock is the result of a shift in the demand for money, then the monetary expansion should be accommodated. In this situation, interest rates should not rise. There is really no way to avoid making judgments concerning the causes of fluctuations in the money stock and in interest rates. If the source is unexpected strength or weakness in aggregate demand, one course of action is called for. If the source is erratic shifts in the demand for money, quite a different policy reaction is required. The pur pose of the simulation experiments was not to “prove” that aggregate demand is always the cause of money market fluctuations. Rather, the purpose of the exercises was to illustrate the potential costs of pursuing a policy strat egy that implicitly assumes that money market fluctuations are caused primarily by an erratic, unpredictable demand for money. Simulation experiments with the model were conducted to measure the impact of con straints on the growth rate of money. The con trol simulation was one in which the interest rate was made to rise at a constant annual rate from a base period of 1963-1 to achieve its ac tual value in 1968-1. In this simulation, the money stock is endogenous. Additional policy simulations were then conducted in which con straints on the growth rate of money were im posed on this interest rate policy. If the rate of growth of the endogenous money stock fell outside the allowable range, the interest rate was changed sufficiently to bring the growth in money back to the nearest boundary of its al lowable range. Figure 2 shows the difference between the values of GNP from the control simulations and those for maximum ranges of 3 to 5 per cent and of 3.5 to 4.5 per cent in the annual growth rate of money. The results indicate that this combination of interest rate and money supply policies would have been beneficial over the period of simulation. Further simulation experiments were con ducted taking the conditions of the 1960-61 recession as the starting point for the policy exercises. The results were similar to those de scribed above for periods of excess demand. Control simulations were conducted for the pe TRADE-OFF BETWEEN POLICY GOALS riod 1960-III to 1968-1 under the assumption of a constant rate of growth of the money stock. Given the actual history of the exoge nous variables in the system and given the ini tial conditions, the time required to get ini tially to full employment was a decreasing function of the money growth rate. Particularly rapid growth rates, however, lead to substan tial overshooting and can create chronic excess demand. Quite predictably, imposition of a constraint on policy in the form of maximum allowable quarterly changes in the Treasury bill rate made it more difficult to hit the full employment target. The interest rate constraint produced a slowing of the rate of expansion of output and employment from the recession base and lengthened the time necessary to hit a full employment target. The results also indi cate that the degree of the slowdown of eco nomic expansion resulting from the constraint depends upon how quickly the target level of employment is to be reached and how narrow is the allowable range of the quarterly change in interest rates. FIGURE 2 Effect on GNP of CONSTANT BILL-RATE GROWTH SUBJECT TO MAXIMUM MONEY-GROWTH RATES D eviations from STRAIGHT BILL-RATE-GROWTH SIMULATION B i ll i o n s o f d o l l a r s ; 3-5% |r—--------------- — '' ' "■' '' ' ' '- ' • I- + 0 -i 15 - ''63 '64 ’■ '65 ■L* *66 ‘67 * '68 It should be emphasized that a restriction on changes in interest rates is potentially less disruptive to the economy than is a restriction on the level of rates. Constraints on the maxi mum short-term change in interest rates can retard but not arrest desired adjustments of the economy. The existence of ceilings or floors on the level of interest rates may prevent the ad justments from ever occurring. Pegging the level of interest rates can lead to a total loss of control by policy over output, employment, and prices. The recession results for a money supply constraint are also similar to those obtained for the excess demand case. A monetary policy that attempts to achieve its objectives through influencing money market conditions— interest rates— can be enhanced in the recession case by imposing a constraint on the rate of growth of money. If the course of aggregate demand proves to be other than expected, variations in the interest rate promoted by the constraint imposed by an allowable range of growth in money rates will serve to push the rate of ex pansion in the desired direction. CONCLUSIONS The brief discussion in the preceding section suggests that high priority should be placed on coordinating short-run operating procedures with the longer-run goals of monetary policy. Failure to achieve such coordination can lead to a serious reduction in the ultimate effective ness of monetary policy. Stabilizing short-term interest rate fluctuations can lead to destabiliz ing shocks to the real sectors of the economy. Better information on the stability of the de mand functions in the economy is sorely needed. The focus of policy on money market conditions may be badly misplaced if the money demand function is relatively stable and predictable through time. Certainly the hypoth esis that the demand for money is erratic and unpredictable is not well documented. It is cu rious, therefore, that policy decisions should depend so strongly on money market condi tions. It might be argued that the central bank is obligated to stabilize the markets for debt in- 104 strumpnts. An unfortunate paradox can result here. An overly zealous attempt to stabilize in terest rates can so disturb the real sectors of the economy as to lead ultimately to extreme variations in market interest rates. The experi ence of the last few years appears to bear out this contention. It would appear that a mone tary policy based almost exclusively on stabiliz ing short-run money market conditions is a luxury we can ill afford. On a conceptual basis the appropriate course of action for policymaking appears to be clear. Given staff projections of the course of the economy over the coming year or so, the instruments of monetary policy should be set to promote the desired time paths of variables such as employment and prices over the period. In order to make such decisions meaningful, several policy alternatives should be presented showing alternative time paths for the target values in the real sector. The policy alternatives should be compared both in terms of the expected values of such variables as output, employment, and prices, and in terms of the dispersion of these projec tions around their expected values. In assess ing the variability of the projections, it is nec essary to provide evidence as to the possible impacts on the projections of various shocks to the system. How sensitive are the projections to shifts in the demand for money or in the demand for investment goods? An analysis of the impact on the projections of alternative as sumptions concerning the values of certain key exogenous variables such as Government spending is also crucial. Furthermore, it is quite likely that the sensitivity of the projec tions to shocks and alternative values of exogenous variables is not independent of the existing state of the economy. At times projections are quite insensitive to fairly large changes in the underlying specifications of the system, but at other times they are extremely sensitive to these specifications. It is essential, therefore, that evidence be provided concern ing the likely dispersion of relevant variables around their projected values. The fluctuations in interest rates and mone tary aggregates implied by the various policy alternatives should also be projected. On the basis of all of this information, trade-offs be tween expected money market stability and the behavior of variables in the real sector can be assessed. The need for reliable econometric models and for seasoned judgment in these ex ercises is obvious. At this point, our ability to generate the required set of projections is quite limited. These limitations suggest that policy strategies should be fairly simple and straightforward. Elaborate policy strategies do not seem consistent with our ability to assess and trace through time the impact of policy acts on the economy. Given a policy strategy over the coming year or so, how can the strategy be reduced to day-by-day operating procedures? Here, there is need for a document that presents projec tions of financial conditions to be expected over the near term. A blending of projections obtained from quarterly and monthly econo metric models is sorely needed. Conceptually, such blends are difficult but possible. On the basis of these short-term projections and the basic policy strategy mentioned above, specific operating instructions can be formulated. Here, limitations on the ability to make short-term projections suggest that the operating proce dures adopted should be fairly simple. We now come to the central problem. How can we continue to link the basic policy strategy with operating procedures as the economic forecasts are modified and as monetary policy strays off course? As policy is currently con ducted, there is no effective means of varying the basic strategy as new information comes in, and there is no way to relate changing con ditions to actual operating procedures. Ideally, we would like to generate new long-term forecasts each quarter and to map out new alternative policy strategies each quar ter. Often, however, the new information that comes in leads to conflicting conclusions about changes in the future course of the economy. Further, econometric models and other proce- TRADE-OFF BETWEEN POLICY GOALS dures often do not predict with sufficient ac curacy to allow useful quarter-by-quarter changes in implied operating strategy. The dis cussion of the original projections also suggests that the initial strategies may at times be very much in doubt. A possible strategy under these conditions is to set quarterly operating instructions in terms of some combination of interest rates and money stock. A policy that sets an interest rate subject to constraints on the rate of growth of money is a very appealing candi date. By setting a range to the allowable growth of money, shifts in the money demand function are automatically accommodated up to the extreme points of the range. The width of the range should depend in part on esti mates of likely quarterly fluctuations in the de mand for money. In setting the range, how ever, it must be recalled that the wider the allowable range, the greater the potential loss in output and employment when variations in aggregate demand are the cause of money growth fluctuations. For this reason, a rela tively narrow band, for example, 4 to 6 per cent, seems desirable as a working principle. Certainly, if there are persuasive arguments explaining why an unusual shift in money de mand occurred in a particular quarter, then a growth rate of the money stock outside the range should be allowed. The point is, how ever, that relaxation of the constraints should be a rare event. In every case when such an action is being considered, the burden of proof should rest squarely on those who believe that an unexpected movement of money outside the range is caused by money demand and not by aggregate demand. Further, the longer the con dition of unusually high or low money growth persists at existing interest rates the greater should be the presumption that the interest rate is inappropriate and should be changed. These recommendations do not call for a drastic departure from current procedures; they call primarily for greater attention to be paid to the long-run objectives of economic stabilization policy. Such objectives are de signed to put short-run stabilization of money market conditions in the context of possible costs to the economy in terms of income, em ployment, and prices. Truly effective implementation of policy re quires that operating strategies intended to achieve desired long-term goals be set forth explicitly. Such strategies must be followed under conditions of great uncertainty about the course of the exogenous variables in the sys tem and about the performance of our models. In such a situation it would appear to be a mistake to focus attention primarily on the un certainties of the money market. Monetary policy decisions must come to grips with the uncertainties we face with respect to aggregate demand. A policy strategy that relies as much as possible on projections but that also com bines a setting of interest rates with allowable ranges on the money growth rate appears to be most appropriate for the near future. by Benjamin M. Friedman TACTICS AND STRATEGY IN MONETARY POLICY CONTENTS 109 INTRODUCTION 109 I. STRATEGY, TACTICS, AND THE DECISION PROCESS Frequency of Committee decisions Strategy and the sequential decision process Tactics 109 110 111 112 118 II. ECONOMIC SPECIFICATION AND APPROPRIATE TACTICAL RESPONSE Economic content of tactical decisions Role of economic specification: four different cases Conclusions 119 119 120 121 125 III. QUESTIONS OF STABILITY A two-equation stock-flow model Extensions of the model Stability conditions in context of the model Conclusions 125 IV. OPTIMAL FILTERING OF OPERATIONS DATA Available Federal Reserve data Conceptual use of probability distribution parameters Example of a data filtering scheme Alternative schemes Conclusion 112 114 125 128 129 131 132 132 V. SUMMARY OF CONCLUSIONS FOR MONETARY POLICY 134 REFERENCES TACTICS AND STRATEGY INTRODUCTION In this paper monetary policy is viewed from the standpoint of operational policy deci sion-making. Section I divides the monetary policy deci sion process into two separate phases—strat egy and tactics—applying heuristic arguments of largely intuitive appeal. The strategy phase involves quarterly decisions outlining a plan for monetary policy over the next several quarters. The tactics phase involves shorterrun technical decisions concerning implementa tion of the first quarter of the strategy and deals with the question of how best to adjust for apparent deviations of the monetary policy instruments from their planned targets. Sections II, III, and IV, applying more for mal analysis, take up several problems that are especially relevant at the tactics stage. Section II examines the influence of economic specifi cation in making tactical decisions, showing that different specifications may imply no re sponse at all to past operating misses of the monetary policy instruments or may imply compensating responses of a number of forms. Section III considers questions of stability in volved in choosing the appropriate speed of tactical action. Section IV, which includes ap plications based on Federal Reserve data, shows the implications for monetary policy tactics of using data that are imperfect and subject to subsequent revision. Section V briefly restates the major conclu sions of Sections I through IV for monetary policy. I. STRATEGY, TACTICS, AND THE DECISION PROCESS As now constituted, the decision process of the Federal Open Market Committee seems to involve an independent monetary policy deci sion at each meeting of the Committee, held once every 3 or 4 weeks. Although Committee members may apply to these decisions as short or as long a time horizon as they see fit, the decision in fact commits the Federal Reserve System only for the time interval until the Committee’s next meeting. Some three times per year, major staff reassessments of the eco nomic outlook occur in the form of audiovisual chart shows presented to the Committee, but chart show meetings do not necessarily involve a different form either of discussion or of de cision on the Committee’s part. Most currently available estimates suggest that monetary policy affects real spending only after substantial time lags. The Federal ReserveMassachusetts Institute of Technology econo metric model, for example, suggests that mone tary policy actions have little effect for the first two to four subsequent quarters and that twothirds of the effect of such action has occurred only after some 2 years. 1 The loosely anchored relationships between monetary policy and real spending decisions suggest that frequent and abrupt policy shifts will have little effect, or in any case an unpredictable effect, on real spending. FREQUENCY OF COMMITTEE DECI SIONS. The Committee in fact does not shift policy at every meeting or every other meeting. It has followed a more slowly moving proce dure of establishing a monetary policy stance and then maintaining it for some months. Various shadings of this stance may occur, but a fundamental revision is likely to happen only at larger intervals. It therefore seems un necessary to preserve an operating machinery under which the Committee may shift policy at each meeting, when in practice this potential flexibility remains virtually unused. Maintaining this unused flexibility would not necessarily be detrimental to efficient decision making, were it not for limited resources on the part of both staff and principals. Taking N o t e .— The author, who is lunior Fellow of the Society of Fellows, Harvard University, and Consult ant, Board of Governors, Federal Reserve System, is grateful to Mrs. Irene Welch of the research staff of the Reserve Bank of Boston for carrying out the sta tistical computations involved in preparing the tables presented in the text. 1 de Leeuw and Gramlich (1968). decisions hurriedly at frequent intervals may well be inefficient if it precludes less frequent but more intensive discussions and examina tions of the relevant financial and general eco nomic developments, both observed and antici pated. Hence, a primary reason for taking major monetary policy decisions less frequently is to permit more thorough exploration of the out look and the available alternatives,- accompa nied by more meaningful Committee discus sions in terms of the goals of monetary policy. Useful staff support to such Committee decisions may involve presentations along the lines of the present chart shows, expanded to include projections of the most likely conse quences of several different patterns of mone tary policy, as well as an analysis of the cur rently attainable trade-offs among different policy goals. A further, related reason for reducing the frequency of major Committee decisions con cerns the dangers and safeguards built into a decision-making process by the design of its machinery. Incremental decision-making, with each decision considered independently, may in any organization lead at times to faulty ac tions and missed opportunities that a more unified decision process can help to avoid. The problem is that past Committee decisions sim ply become part of the data, while future ones remain unconsidered; hence the true unity and interdependence of the series of decisions is not evident. Consider, for example, a Committee deci sion of whether or not to move to a tightmoney policy: A Committee member may well ask, under the current decision machinery, what is the cost of delaying the move until the next Committee meeting, if the move is in fact advisable at all. Given the currently available state of knowledge about economic relation ships, the answer must be that the cost of this several weeks’ delay is so small and uncertain as to be virtually unidentifiable. The problem arises in that the same question, asked at six successive Committee meetings, may elicit the same answer of nearly negligible cost to the in cremental delay each time; yet the true cost of delaying the move for 6 months may be not only identifiable but in fact quite substantial. Two conclusions emerge from this discus sion: First, decision-making opportunities should occur with quantum time intervals great enough to render individual decisions meaning ful; and, second, the relation among interde pendent decisions should be explicitly evident, even when those decisions follow one another in time. If the Committee’s major policy decisions should occur less frequently than every 3 or 4 weeks, what time interval is then appropriate? Although a number of possibilities are perhaps workable, for several reasons 3 months seems to be the best. Much of the relevant economic data to be used in considering monetary policy decisions are available only on a quarterly basis; real expenditures information in the na tional income accounts is perhaps the leading example. Further, our knowledge of economic relationships is based largely on a quarterly discrete time conception of the economic system; judgmental analysts seem to think primarily by quarters, and the available econometric work is mostly in quarterly form. In addition, referring again to the observable impact of monetary phenomena on real spend ing, the quarter is probably the smallest time unit for which meaningful information is now identifiable. STRATEGY AND THE SEQUENTIAL DECISION PROCESS. Collecting the several conclusions derived above gives a brief outline of the strategy stage of the decision-making process for monetary policy: The Open Mar ket Committee may meet quarterly to analyze recent and prospective economic develop ments; to study the outlook for the foreseeable future in the light of the possible patterns of monetary policy and the projected conse quences of each; to consider the attainable trade offs among different monetary policy goals; and to take a decision on the course of mone tary policy. In sum, the strategy decision con TACTICS AND STRATEGY siders the entire relevant economic picture and derives the best monetary policy decision to meet it. The committee may well continue to meet on its current more frequent schedule, but it would take major monetary policy decisions only quarterly. A further question is whether the Commit tee should decide the course of policy only until the time of the next major decision meet ing, or should instead attempt actively to in clude a longer time horizon in its plans. If five or six quarters into the future is the furthest ahead that analysts can reasonably look, using currently available methods, and if the lag required for monetary policy to take effect is approximately two quarters, then the furthest horizon for which the Committee can viably plan policy actions is three to four quarters into the future. Is there any advantage to the Committee’s formulating a hypothetical policy for this three- or four-quarter period, which would extend considerably past the time of the next Committee meeting? The decision systems elaborated by Holt, Theil, and Simon2 suggest that formulating long-range plans and revising them at shorter intervals may well be more efficient than merely planning for the shorter interval be tween decisions. Even though in practice one may choose to plan only from January to April, for example, and wait to examine the situation in April before finalizing further plans, tracing through the anticipated course of events after April and considering April’s deci sion in advance is a useful exercise for im proving January’s decision. This principle is especially valid in light of the lags associated with monetary policy, which imply that Janu ary’s decision may have little or no effect on the goals of the policy before July. In addition, this system of coordinated hypothetical planning for the future helps bring into focus the unity and interdependence of the entire series of Committee decisions. Hence at its quarterly meetings the Commit 2 Holt (1962), Theil (1964), and Simon (1956). tee may formulate a strategy for several quarters, of which the immediate quarter be comes actual policy to be followed until the Committee’s next major decision meeting. At each such meeting the Committee updates its strategy by one quarter; it revises the previous decision’s hypothetical second-quarter plan to become an immediate-quarter plan, which is then the Committee’s actual policy for opera tional purposes. In this way a sequential de cision process enables the Committee to take as long a view as is possible in formulating its strategy, while preserving as much flexibility as is probably necessary. TACTICS. Once the Committee has speci fied its strategy for the quarter immediately ahead, several operational problems are likely to arise which require decisions of a subor dinate nature. For example, suppose that the Committee has expressed its strategy as achiev ing over the quarter a movement of X in M, where X is an appropriately chosen number and M is some selected financial variable (or X may be a vector of numbers and M a list of variables). In effect the Committee has specified a target path for variable M. If M were directly within Federal Reserve control, or if the operational levers available to in fluence M had no uncertainties attached to their use, then following the Committee’s strategy would be a straightforward and un ambiguous process. In practice, however, market activity, as well as Federal Reserve operations, works to determine most of the interesting candidates for M; and a multitude of uncertainties and shifting structures characterize the entire finan cial system. Hence a series of subordinate de cisions— a set of tactics— are necessary to adjust operating procedures in the light of un foreseen situations as they arise. Such tactics govern Federal Reserve response to sudden surprise movements of financial variables and resulting deviations of M from the target path specified in the quarter’s strategy. Since further decisions are still necessary once the Committee has formulated its strat egy, one may perhaps question the efficacy of the two-stage decision process developed here. One rationale for it is that the Open Market Committee may reserve strategy decisions to itself, while delegating tactical matters to a subordinate group subject to its review, much as it currently delegates many operating de cisions to the Manager of the Open Market Ac count. Alternatively, if the Committee wishes to reserve both strategy and tactics for its own decisions, allocating some meetings to strategy and others to tactics is probably a more effi cient use of time than the system currently in practice. Finally, since strategy and tactics are differ ent decisions on two distinct sets of questions, separating the two— even if only in discussion —should enhance clarity and thereby help de cision-makers to operate more efficiently. A strategy is the result of an examination of the entire relevant economic outlook; it expresses the best monetary policy response to that situa tion. Tactics embrace the more technical oper ational difficulties involved in meeting the monetary policy targets specified in the strat egy. II. ECONOMIC SPECIFICATION AND APPROPRIATE TACTICAL RESPONSE Retaining the division of monetary policy decisions into the two levels of strategy and tactics, the discussion of this and the following two sections examines several issues especially relevant for the tactics stage. An assumption which therefore underlies this discussion is that the basic elements of the prevailing monetary policy strategy are fixed inputs in evaluating a tactical problem. More specifically, assume throughout these three chapters that the Open Market Committee has identified the financial target variable (or list of variables) M and has specified a desired movement of X (where X is a vector if M is more than one variable) for the immediate quarter. The value of M at the beginning of the quarter and the desired movement X together suffice to define a target path for M. A typical tactical problem arises if, after one month of the quarter, incoming reports indicate that M has strayed away from this target path. The basic decisions in the tactical stage of the monetary policy process involve confronting this problem and formulating appropriate Fed eral Reserve responses. The following discussion explores three sets of issues related to this central tactical prob lem of observed deviations of M from its tar get path: In this section, how does the specification of the economic transmission of the effects of monetary policy influence the appropriate tac tical response? In Section III, what precautions are neces sary to prevent these responses from destabil izing, rather than stabilizing, the economy? In Section IV, what are the implications of using data subject to revision, in evaluating the situations to which monetary policy tactics are to respond? ECONOMIC CONTENT OF TACTICAL DECISIONS. Although tactical decisions in the sense used here may be technical and opera tional in nature, they do have substantial eco nomic content. Perhaps the clearest example of this fact is the influence upon tactical decisions of one’s specification of the effects of monetary policy on the economy. Consider the following simplified illustration: The Open Market Committee has identified one financial variable M as its target variable and has set down its strategy for the quarter as a movement of X in M. For the purpose of this example, let M be some monetary aggregate and let X be equivalent to a given increase in M. (X < 0 implies a desired decrease in M.) Then, M x* = Mo + Ai *M M 2* = M x* + A2*M Mz* = M2* + A 3*M = Mo + X Ai* M + A2*M + A3*M = X TACTICS AND STRATEGY where M 0 = actual value of M at the begin ning of the quarter, Mi* = desired value of M at the end of the zth month, and A{*M = de sired change in M during the /th month. For further simplicity assume that the initial intention is to spread the total movement X evenly over the quarter. 3 Then Ai *M = Ai* M = A3*M = \^X The Manager of the Open Market Account will then conduct open market operations over the first month in the manner that he thinks is most likely to achieve the desired movement At*M. A large number of factors beyond the Manager’s immediate control, however, also in fluence movements in M .4 As a result, his abil ity to achieve an exact total of M x* at the end of the month is limited. In more general terms, the Manager may miss A |* M and M i * and in stead achieve A\M and Mi, where AiM = actual change in M during the /th month, and Mi = actual value of M at the end of the /th month. A number of responses to this situation are possible: For example, the Manager may at tempt to rectify the entire error during the next month, so as to return M to its target path by the end of that month. This plan leads to a revised desired change in M in month / + 1: A = Ai+x*M + Ai*M - A,-Af M i+i* - M { where Ai**M = revised desired change in M during the /th month. M i+1* remains un changed. Alternatively, the Manager may attempt to spread the correction process so as to return M to its target path only at the end of the quarter. This plan involves revising both the desired change in M in month / 4- 1 and the desired level of M at the end of the month i + l: Ai+1**M = Ai+l*M + (Ai*M - AtM) = M i+1* - ^ (Ai*M - AiM) where N = 3 — i — number of months remain ing in the quarter, and Mi** = revised desired value of M at the end of the /th month. In the special case of equal desired monthly incre ments, mentioned above, when, with 2 months remaining, Ai+x*M = Ai+i*M = ± X these relationships simplify to A<+1**M = \ (X - A,M) M i+1** = Mi + 1 (X - AiM) In the special case that i ~ 2, that is, when the Manager has kept M to its target path for the first month of the quarter but could not prevent a deviation in the second, this plan of achieving the full correction only by the end of the quarter coincides with the earlier plan of achieving the full correction in the first pos sible month. Neither of these alternatives includes any scheme for “compensation” for past devia tions. If, for example, M has strayed below its target path during the first month of the quar ter, one tactical response could be to force it above its target path by an equivalent amount and for an equivalent period of time. Assum ing that the first month’s deviation has been accumulating gradually, the revised program becomes M 2** = M 2*+ M i * - M x A2**M = A2*M + 2 (Ai*M - AiM) 3 Note that making each movement Ai*M the same differs from pursuing a constant rate of growth in each period. 4 Among these factors are float, currency in circu lation, Treasury deposits at Federal Reserve Banks, gold and foreign accounts, Federal Reserve foreign currency holdings, and so forth. See Maisel (1969). A3**M = A3*M - (Ai*M - AXM) M3* remains unchanged. These three limited programs by no means exhaust the possible tactical responses to a de- viation of M from its target path, nor is the Manager restricted to revising his operations monthly. Weekly data reports for some finan cial variables and daily reports for others should help him to recognize incipient devia tions quickly. Nevertheless, some deviations will almost certainly occur, and the simplified illustrations above suggest at least two ques tions relevant to deciding upon appropriate re sponses: Should the Manager correct devia tions in M as rapidly as possible, or should he spread the correction process so that M re turns to its target path only by the end of the quarter? Further, should he compensate for unintended deviations of M in one direction by deliberately inducing controlled deviations in the other direction, or should he simply restore M to its target path with no compensation for past errors? ROLE OF ECONOMIC SPECIFICA TION: FOUR DIFFERENT CASES. One factor determining the answers to these ques tions, particularly that of compensation, is the specification of the relationship between M and the economic variables that are the ulti mate goals of monetary policy. In the strategy stage of the policy-decision process, the Open Market Committee has determined variable(s) M and the desired movement(s) X by refer ring to some relationships, however vaguely conceived, of the form Y = y(M ) where Y — the ultimate policy goal variable (or vector of variables) and y is some func tional relation (or set of relations). In general, four separate possibilities are available for this relationship; only two of these have similar implications for purposes of tactical decisions. Although the reality of the economy is perhaps closer to a continuous time mathematical representation, preservation of the discrete system with a time interval of 1 month is preferable here for ease in ex position. The following discussion retains the frame work used above, in which the set strategy for the quarter calls for a movement X in M, and the Manager sets out to achieve X in three equal monthly movements, each equal to ^X. The object of working through the logic and algebra of these straightforward exercises is to illustrate the significant influence upon tactical decisions of one’s specification of the relation Y = y(M ). Case A: One unlikely specification is that Y depends on M with these properties: First, the proper argument in the relation is the movement in M, that is, AM, not the value of M itself. Second, there are no continuing lags in the system’s response to this movement in M. This second restriction means that a AM in one month influences Y in one month only (not necessarily the same m onth); similarly, it means that Y in any month is influenced by AM in only one month (again not necessarily the same m onth). Hence, while the restriction per* mits Y to depend upon one lagged value of AM, it precludes the dependence of Y on lagged values of itself, since such a lagged re sponse relation would enable AM in any given month to influence Y in a series of months through a Koyck-type distributed lag.5 In this case, there need be no compensation for past errors of any form. Assume, for ex ample, that AM in any month influences Y as described above, with a 6-month lag. Then AiM influences Yi+6. If a fM deviates from its specification in the Committee’s strategy deci sion, it is of no benefit to adjust At+1M to com pensate. Doing so merely causes Y i+J to de viate from its desired value. Under the specifi cations of Case A, when the /th month ends, all Y through Ki+6 are beyond the control of monetary policy. Monetary policy can still in fluence Y beginning with Y Ul\ but these 7 ’s are in no way affected by the error in A{M, and so compensation for that error is pointless and even harmful. Hence the Manager should con tinue to try to achieve AM equal to j X 5 Koyck (1954). in TACTICS AND STRATEGY each subsequent month, and the target path for M over the quarter shifts vertically to reflect Mi as its new starting point. The restrictive specification of Case A ren ders it highly unrealistic, and only methodo logical completeness justifies its inclusion here. Case B: A slightly less restrictive specifica tion is that Y depends on M with these prop erties: First, the proper argument in this relation is the value of M itself. Second, as in Case A, there are no continuing lags in the system’s re sponse to the value of M. Again, this second restriction means that a value M in one month influences Y in one concurrent or succeeding month only; similarly, it means that Y in any month is influenced by the value M in only one concurrent or preceding month. Again, as in Case A, this restriction precludes the depend ence of Y on lagged values of itself. In this case, however, one form of compensa tion is in order. If M deviates from its target path in the zth month, retaining the 6-month lag assumption used above, Yi+6 takes on an un desired movement that is beyond the influence of monetary policy once the ith month has ended. Nevertheless, permitting M to persist in this deviation past the ith month perpetuates the undesired movements in Y past Yi+6. Hence it is essential to restore M to the appropriate initial target path. The Case B answer to the compensation question, then, is as follows: When AiM has deviated from At*M, so that M i deviates from Mi*, Aj+/**M replaces Ai+;*M according to some scheme to compensate for the original error, where } extends to as many months as are necessary to return M to its original target path. If this correction process is to take more than 1 month, M i+/** replaces M i+j*, where j extends to one less than the number of months necessary to return M to its initial target path. Hence in Case B there is compensation in the sense of responding to A*M different from Ai*M by letting Ai+;**M differ from Ai+/*M in the opposite direction. There is no compensa tion in the sense of responding to a level M i different from M i* by letting M*+/** differ from M i+;* in the opposite direction. The speed of adjustment conclusion for Case B, arguing narrowly from the causal effects of M on Y, is to return M to its target path as rapidly as possible, since continuing the devia tion through Mi+j** perpetuates the undesired effects on Y through Yi+J+e. The more interesting Cases C and D repeat the argument specifications for Cases A and B, only without the extremely restrictive assump tion of no continuing lags in the response of Y to the movement AM (Case C) or the value M (Case D ). Relaxing this restriction admits more realistic and believable specifications in which Y depends on lagged values of itself. The tactical conclusions for Case C emerge to be similar to those for Case B, while Case D introduces a new form of compensation. The formal analysis is as follows: Case C: Here the proper argument in the relation between Y and M is the movement in M, that is, AM. Since Y i may depend on a series of lagged Aj-yM (or, equivalently, on some one Ai_/M and also on Y*_i), a AM in one month influences Y in a number of suc ceeding months. In this case, as in Case B, it is necessary to provide compensating movements in AM, in order to return M to the original target path (or near it, as in Example 2 below). If A{M has differed from Ai*M, recasting the 6-month lag assumed above into a 6-month no-response period, Yi+6 takes on an undesired movement that is beyond the influence of monetary policy once the zth month has ended. The specification in Case C, however, indicates that the error in A*M itself, if not offset by compen sating revisions of Ai+y*M to Ai+j**M, will lead to undesired movements in Y i+j+6. Example 1 (rectangular lag): Suppose that Y| depends equally on A,_6M through A^nM, that is, the lag is rectangular and persists for two quarters after an initial no-response period of two quarters. Then, if AiM has differed from Ai*M, setting 116 At-+i**M = Ai+i* M + Ai*M - A{M = M i+ i* — Af,- returns M to its target path after only 1 month’s deviation. Y i +6 will have an undesired com ponent which no further monetary policy can correct; assuming no further errors in M , how ever, y i+7 through Yuii will be on target. A problem emerges only for Yi+12, which still de pends on A i+1M but not on A iM. If no further corrective adjustment occurs, Yi+12 will have an undesired element equal in size and opposite in direction to that which occurred in Y i+e. Here arises one of the few differences in compensation action conclusions between Case B and Case C: Under the former specification, the compensation in A which returns M to its target path, limits the undesired move ments in Y to Yj+6 only. With the perpetuating lags of Case C, however, the error compensa tion adjustment in A i+1**M , which also returns M to its target path, causes Yi+as to differ from its desired value. If the speed of adjustment is such that the compensation occurs not just in Aj+i**M but in A w h e r e j extends to as many months as are necessary to return M to its target path, then the situation is somewhat more complex. Y takes on diminishing deviations through all Yi+j+5, instead of just in Yi+6; similarly, Y takes on diminishing deviations through all Yi+/+u, instead of just in Y i+12. As in Case B, however, it seems desirable, in light of the arguments considered here, for the speed of adjustment to be as rapid as possible, thus minimizing the deviations of Y in the two periods beginning with Y i+G and Y*+12. The existence of the second deviation period for Y, beginning with Y i+12, leads to a further complexity. While this second deviation period is due to the compensating tactics of monetary policy, further monetary policy tactics can off set at least part of it. Whether or not to do so depends on the specific loss attached to devia tions of Y from its desired values through time. If, for example, the relevant loss function is a sum of absolute values of deviations of Y from its desired values, then, within the context of the rectangular lag on AM, it is a matter of indifference whether or not to offset the second deviation period. Alternatively, if the loss func tion is a sum of squared deviations of Y , then it is preferable to replace one large deviation with several small ones, and so this second off setting action can potentially reduce the total loss sustained. More formally, suppose, in line with the above discussion, that the y function includes this component: Yi = . . . + fi £ j =6 A,WM + . . . Suppose further that AiM has contained an error of e, that is, e = AiM - A?M and that A has fully offset this error, re turning M to its original target path by the end of month i + 1 : Af+1M = Ai+1**M = A{+i*M - e Then the only difference between Y and its desired value in the first deviation period is Yu g Y i+d* = /3e where Yi* — the desired value of Y in period i. Hence the loss associated with the first devia tion period, by using a simple quadratic penalty function, is Li - (Yi+ 6 - Yi+s * y = (0e)2 = /32e2 This loss is unavoidable once AiM has differed from A f*M and occurs whether or not monetary policy attempts to offset the second deviation period in Y. If monetary policy fore goes such secondary offsetting action, the loss from the secondary deviation is U = (Yi+12 - Y £ l2)* = ( - 0 e) 2 = 0 V One possibility for offsetting action is simply to set Ai+7**M = A,+v*M + ~ ( and Ai+s**M = A<+g*M - i e TACTICS AND STRATEGY The resulting secondary loss is u = (Yi+i2 - r i+12* y + (Yi+n - - W + f r y Yi+li* y - l ' " Hence the result of this secondary offsetting action is to reduce the loss associated with the secondary deviation by 50 per cent and the loss associated with the entire episode by 25 per cent, with the losses measured by a simple quadratic function. Example 2 (Koyck lag): Suppose that Y { depends on A*.5M and all preceding AM, that is, Ai-jsM for all j > 0 , with geometrically declining weights. Then Yi = . . . j8 £ (1 - X )'A ,_ 6M + . . . j= l 0 < X< 1 This geometric lag pattern is probably more realistic than the rectangular pattern assumed above in Example 1. An error in the zth month of AiM - Ai*M = e then leads to a primary deviation in Yi+6 and an associated Lx = (Y{+ 6 - 1^6 * ) 2 = 08(1 - X)e) 2 = (1 - \ ) 2j3V A compensatory tactic setting At*+i**M = Ai+1*M - (1 - X)e permits Y to sustain no deviation from its de sired value through time, beginning with Yi+7. Furthermore, revising Ai+1*M to A i n this manner eliminates the secondary deviation problems inherent in the rectangular lag of Ex ample 1. Hence the only deviation of Y from Y* for the entire episode is that contained in Yi*. This result highlights a second difference be tween the Case B and Case C compensation conclusions. Under the former specification, revised level replaces M i+j* only for one less month than necessary to return M to its target path, after allowing for speed of ad justment considerations. Under the Case C specification with a geometric lag pattern, level differs from Mi+;* by Mi+j** = M i+j* + Xe for all j > 0. This change implies a vertical shift in the entire target path for M, similar to that required in Case A, making the new starting point and proceeding with move ments of Ai+;-+1*M as previously planned. Hence, while Case B always involves returning M to its original target path, Case C may or may not involve such a return, depending upon the specific lag pattern involved. The clear implication for monetary policy is that tactics become much simpler if a geo metrically declining lag pattern maintains in the Y — y ( M ) relation. Even if different lag pat terns maintain, however, a corresponding sim plicity in tactics may result from the introduc tion of uncertainty discounting into the loss function. In other words, the economy may be sufficiently stochastic that random events render planning for the more distant future increas ingly futile after some point. In terms of the discussion of primary and secondary deviations of Y from Y*, unforeseen circumstances un associated with monetary policy may arise, causing movements in Y and at times calling for revisions in A* M and M*, thereby elimi nating the rationale behind intricate plans cal culated to minimize secondary deviations. Case D: The final category of specification requires that the proper argument in the Y = y ( M ) relation is the value M itself, as in Case B; while the Yi may depend on a series of lagged M i- jt Hence M in one month influ ences Y in a number of succeeding months. A relation of this sort introduces a new dimension of compensation for previous opera tional errors in keeping M to its designated target path. If A\M has differed from Aj*M, causing Mi to differ from Mj*, an adjustment is necessary not merely to return M to its target path but also to cause an offsetting deviation of equivalent magnitude and duration in the opposite direction. Analogously to Case C, only substituting levels of M for movements AM, a deviation of even one from the correspond ing Mi* will cause undesired movements in a series of subsequent Yi if not compensated. Consider again the rectangular and geometric lag examples: Example 1 {rectangular lag): The direct analog to Case C is to compensate for a devia tion of Mi from Mi* by setting M i+1 ** = Mi+i* + Mi* - Mi and retaining M i+j+1* unchanged for all j > 0 , assuming that speed of response considera tions call for correcting the error entirely in the first month. This step to minimize the primary deviation in Y implies revisions in monthly movements A*i**M = At+i*M + 2(Ai*M - A{M) and Af+a**M = Ai+2 *M - (Ai*M - A{M) and retention of Ai+;-+2*M unchanged for all j > 0. Hence M returns to its target path only in the second month following the initial error in M,. This response accomplishes full compensa tion on M in the first month after the initial error, returning M to its original target path at the end of that month. Preserving the assumption of a rectangular lag of 6 months, this scheme restricts the primary deviation in Y to Y i+e only. The argument concerning sub sequent tactics to prevent secondary deviations of ^i+i+12 from Yi+j+i 2 * is directly analogous to that presented for Case C, Example 1 . One assumption underlying Case D is that the value M\ represents a mean value over the /th month. 6 If Mi instead represents the level at the close of the /th month, it is important to estimate during how much of the month that level maintained, before formulating tactics. The duration and magnitude of the compensa tory portion of M's actual path both must match the corresponding features of the error 6 This usage represents a change from that intro duced on p. 113 in which is the actual value of M at the end of the /th month. portion of the path. The argument works in the other direction as well; an end-of-month Mi equal to Mi* does not indicate success of mon etary policy operations if the actual level pre vailing through most of the month differs sub stantially from that implied by the M * path. Example 2 (Koyck lag): Again in analogy to the Case C treatment of the geometric lag, suppose that the Y = y(M) relation contains the following term: Yi = . .. + fi £ (1 - + ... j=l If M i has deviated from M i* , adjusting M i+1* by = Mi+i* - (1 - X) (Mi - Mi*) leaving M i+j+1* unchanged for all / > 0, con fines the undesired movement in Y to Y i+Q* This step implies revisions in the monthly movements of Ai+i**M = Ai+i*M - (2 - X) (Mi - Mi*) Ai+2**M = A,-+a*Af — (1 — X) (Mi - Mi*) and leaves Auj+z*M unchanged for all / > 0 . Like Case B, Case D requires ultimately return ing M to its originally specified target path. CONCLUSIONS. To summarize the conclu sions of this section, optimal monetary policy tactics depend fundamentally on the specifica tion of the relationship between the financial variables chosen as monetary policy targets and the variables that represent the goals of the policy. Tactics may require no response at all to operating errors or very complex re sponses; may call for redefining a new target path for M or for retaining the original target path; or may imply the adequacy of a onceand-for-all response to operating errors or the necessity of acting to prevent future policyinduced instabilities. The one seemingly consistent conclusion in all four possible cases is that compensatory responses, if warranted at all, should take place as rapidly as possible; Section III examines this tentative conclusion more closely and exposes it to additional con siderations. Before proceeding, however, one further TACTICS AND STRATEGY issue deserves explicit treatment. The above discussion, for all cases, has used a time pe riod of 1 month and has often considered up to 1 2 periods into the future; a fixed strategy decision, under the program formulated in Sec tion I, covers only 3 months. There are two justifications for this apparent contradiction. First, these illustrations are more general in their application than the dis cussion—couched in more specific terms for expository purposes—may indicate. Second, considering immediate actions in terms of their implied effects on future actions—even when those future plans will be revised—is an aid to proper policy formulation. This principle holds for sketching tactical decisions beyond the point at which a new strategy will supplant the current one, as well as for formulating stra tegic decisions for quarters beyond the imme diate one. III. QUESTIONS OF STABILITY The discussion in Section II reviews different possible specifications of the relationship be tween the instruments and the goals of monetary policy, drawing the implications of each specification for appropriate policy tactics in response to an unintended deviation of an instrument from its designated path. In those cases of specification that warrant compensa tory movements in monetary policy instru ments, the analysis tentatively suggests that the best tactics would always be to accomplish the full compensation as rapidly as possible. Doing so would avoid further undesired movements in the policy goal variables beyond that caused by the initial error in the instruments. The cor responding analysis for strategic shifts in policy, to offset exogenous shifts in demands and sup plies in the economy, appears to be similar, also calling for as rapid a policy response as possible. This analysis, while perhaps adequate to in dicate the importance of specification of the Y = y(M ) relation, is too restrictive to deal effectively with the speed-of-response question. In particular, it omits considerations of stabil ity which arise from feedback effects of Y upon M in the private economy, independent of Federal Reserve action. A TWO-EQUATION STOCK-FLOW MODEL. To be more specific, consider a two-variable illustration of monetary policy control: Let Y be some measure of output (equal to income) of the economy, and let M represent the existing stock of some set of fi nancial assets. Analysis that has previously treated growth models with capital stock accu mulation has a monetary analog. The follow ing model uses these familiar tools to deal with a growth model that has an accumulating stock of financial assets and to draw implica tions for desired speeds of response in mone tary policy tactics. 7 Suppose that the Y = y(M ) relationship is homogeneous of degree one; the linear forms of the four specifications of Section II may all meet this condition under certain assumptions. It then follows that growth in either of Y or M must imply an equiproportional growth in the other. For any period of time, the growth rate of Y is Yt - Yp Yq while that for M is M t - Mo Mo Suppressing the time unit and dealing in a con tinuous time framework, the corresponding rates of growth are (3.1) = D log Y and ~ D M = D \ o g M respectively, where D is the differential opera tor with respect to time. The condition of equiproportional growth, implied by the homo7 This discussion follows closely the format pre sented in Chapter 10 of Allen (1967). 120 geneous degree one condition on Y — y(M ) relation, is that over time the Yt - Y 0 Mt - M Q Yo ~~ Mo In particular, if the economy satisfies this con dition at every point in time, D log Y — D log M Let g indicate this common growth rate for Y and M. Then integration over time of the growth-rate equation where log Y = constant + gt or (3.6) v 7 By the same steps, (3.3) M = Moe“ where M — M 0 at time t — 0. These results depend only on the homoge neous degree one properties of the Y — y(M ) relation. Since the model theoretically repre sents a real economy with a stock of finan cial assets, however, it is possible to impose stock and flow conditions upon the system. In equilibrium, the economy must satisfy both sets of conditions. For a stock condition, posit a simple fixed velocity relation for Y — y ( M ) : 0 for all t . For a flow condition, assume that income earners desire to add a fixed fraction of cur rent income to their stock of asset M; further assume that the Federal Reserve System con ducts monetary policy so as to accommodate this desire. Then . sY=-DY V I D Y = D log Y = sv (3.7) D M = s(vM) Hence this model gives an attainable steadystate growth only for (3.8) Y = Yoeot where v > 1 j-j D M = D log M — sv where A is an arbitrary constant of integra tion. Replacing arbitrary A with an initial con dition Y 0 at time t = 0 gives Y = vM, or M = ^ Y , < s < These two conditions together yield a solu tion for the common growth rate attainable si multaneously for Y and M f from elimination of M from equations 3.4 and 3.5: Y = Ae** (3.4) 0 and so g = sv. The same solution results from elimination of Y from equations 3.4 and 3.5: D log Y = g yields (3.2) D M = s Y for all t (3.5) D log Y = D log M = g = sv EXTENSIONS OF THE MODEL. So far the model does not allow for exogenous shifts in the asset-creating propensities of the private sector. Suppose that the Federal Reserve has provided sufficient reserves to enable the stock of asset M to grow at rate g = sv, while in come Y has also grown at g = sv. Further suppose that at time t = 0 a new asset be comes available, leading income earners to al locate to this new asset some fixed amount of the income that previously they would have al located to accumulation of asset M. Mean while, suppose that the initial velocity relation Y — y ( M ) continues to hold. This change means a shift in the flow condition from equa tion 3.5 to (3.9) sY = D M + A for alU > 0 where A is the fixed accumulation of the new asset. The stock condition, equation 3.4, re mains unchanged. The solution of the system now changes, to reflect the change in the flow condition. The differential equation 3.6, formed by eliminating TACTICS AND STRATEGY M from both stock and flow conditions, be comes (3.10) sY - A = -D Y V Y now follows the path through time which is consistent with the solution of differential equation 3.10. That solution is (3.11) + ( r ° - j ) egt where Y = Y 0 at time t — 0 (initial condition) and g — sv as before. Equation 3.11 is a generalization of the simpler path of equation 3.2, which is consist ent with the model with no alternative asset, as solved in equation 3.6. g — sv remains in equation 3.11 the only possible rate of steadystate growth consistent with the homogeneous degree one properties of the stock condition, but here there are more possible results. A — 0 leads, as expected, to the same path as before. A — sY0 indicates, using the new flow condi tion 3.10, that the economy directs all asset creation to asset A , leaving no growth in asset M; hence, by stock condition 3.4, 7 = -s for all / ~ > 0 indicating the establishment of a stationary state. 0 < A =£ sY0 leads to a steady growth (if A < sYo) or decline (if A > sY 0) in Y and in M . The path of M, analogous to equa tion 3.11, is (3.12) time. If A varies over time in a pattern indi cated by some function A ( t ), then differential equation 3.10 becomes M =^ + where M 0^ ~ — Y0, and g = sv. Equation 3.12 v is itself a generalization of the simpler path for M of equation 3.3, which is consistent with the model with no alternative asset, as solved in equation 3.7. A further extension of the model is to re duce the restrictions on the exogenous finan cial asset accumulation A . Specifically, it is un realistic to assume that A is constant through (3.13) sY - ^D Y - A(i) = 0 The solution of this equation is (3.14) Y = f(t) + B e ° * where the particular integral /(/) depends on the function A (t), and B is an arbitrary form determined by the initial condition Y0. STABILITY CONDITIONS IN CON TEXT OF THE MODEL. This machinery, developed at some length, is now available to explore problems of stability of the economy and its susceptibility to monetary policy con trol. Here stability bears the traditional sense of Harrod’s usage, especially with reference to the “knife-edge” problem. 3 So far the analysis has assumed fulfillment of the stock condition 3.4 both initially, giving Yo = vM 0 and through all time, giving Yt - vMt for all t. Following Allen’s treatment, 9 the basic stability questions arise if output Y and stock of the asset M are out of line, that is, deviate from this stock condition. Such devia tions, or disequilibria, may stem from errors on the part of policy-makers or from unforeseen exogenous disturbances in the economy. The analysis here may treat the discrepancy either as existing initially, making M 0 out of line with Y0f or as arising in the paths of M and Y through time. Any resulting response on the part of the Federal Reserve to try to adjust M introduces a servo-mechanism control device into the system. 10 The disequilibrium situation invali8 Harrod (1948). s»Allen (1967). 10 A servo-mechanism is a controller which sets the magnitude and/or direction of the controlling im pulse as a function of the state of the system to be controlled. 122 dates the stock condition 3.4; if the policy re sponses follow a set pattern, dictated by prede termined strategic or tactical principles, they then replace stock condition 3.4 by a new stock condition. The new condition represents both the reliance on and the accumulation of asset M in the private sector, as well as the discretionary influence of the Federal Reserve. If the Federal Reserve responses follow a sufficiently regular pattern to permit their being expressed as a new stock condition, it is possible to analyze the resulting system to de termine just what effects such Federal Reserve actions will produce, in terms of their stabiliz ing or destabilizing influence on the economy in its disequilibrium state. Conversely, examin ing different response systems in the abstract yields conditions that any actual policy scheme must meet to ensure that it stabilizes, rather than destabilizes, the economy. This approach to stability is relevant both to the longer-term strategy decisions of the Open Market Committee and to nearer-term exer cises in monetary policy tactics. As in the dis cussion of Section II about different forms of compensating movements, the specification of the particular relationships in the system is crucial to the conclusions derived. Further, for at least one important conclusion, the actual values of the coefficients in the relations as sume a new significance. Before illustrating these results with an ex ample, it is best to determine the definition of stability that best fits the intent of monetary policy. Suppose that the Open Market Com mittee has specified a desired path of M through time, denoted M*, and that associated with M* is a path for Y, denoted Y*. Suppose, further, that M* and Y* call for M and Y to meet the stock and flow conditions 3.4 and 3.5 at all times. Then these paths call for ex ponential growth of both M and Y at rate g = sv,7 and M = — y Y at all times. deviate from M*, for example, M > M*, which implies M > — Y*. Call this point in time t — 0 , and consider the problem of gen erating growth paths of Y and M from the initial condition M 0 > — Y 0, v Two alternative definitions of stability of the system are possible: The first requires these generated growth paths to converge back onto the original desired paths Y * and M*. This requirement is very severe and is not well suited to the current stock-flow model in which Y represents income and M the stock of a set of financial assets. It is more appropriate to a model in which Y represents the goals of policy in a more stationary form, for example, the unemployment rate and the rate of price in crease, and in which M has similar character istics and stands itself for rates of asset growth or for interest rates. A second definition of stability, identified by Jorgenson as “stability in the sense of Harrod, ” 12 requires only that the growth rates of Y and M converge over time to the rate g = sv. Y and M find new paths, then, which do not necessarily regress to the initially de sired paths Y* and A/*, but which do meet the conditions of the original system in that both Y and M come to grow at rate g. If the Federal Reserve can stabilize the economy in this sense, it will be offsetting the initial exogenous disturbance in the stock condition just suffi ciently to allow the economy to proceed as before, only consistently with the new, shifted stock condition. In the context of the stockflow model as developed here, with the cur rent definitions of Y and M, this definition of stability seems appropriate. Adopting the second, or Jorgenson-Harrod, definition of stability, consider again the model with a unique steady-state growth solution given by O Now suppose that some exogenous shift in the financial determinants11 of M causes M to 11 See footnote 4, p. 113. (3.2, 3.3) Y = Y 0eot and M = M 0eot where g = sv, provided that the initial values 12 Jorgenson (1960). TACTICS AND STRATEGY Y0 and M 0 satisfy Y 0 = vM0. Assume now, without loss of generality, that Y 0 < vM0. Re taining M * as the desired path of M, the initial condition is M 0 > M 0*. For M to have achieved a level greater than its target, it must have grown at greater than the desired rate. Let Combining the two, the new planned rate of growth for M is 8 1 M — Af** ~ T ' M ~ 8 1/ ~ T\ 1 Y**\ ~ v M ) by using 7** = VM** * = \-.DM M Hence the Federal Reserve, in this disequilib rium situation, imposes in place of the unsatis fied stock condition 3.4 a growth rate for M over time of Then Now assume that the Federal Reserve adopts a policy to reduce M to a new path M** which will satisfy the stock condition Y = vM (equa tion 3.4). Further suppose that the intended tactics of policy are to have the stock condition satisfied by the end of T time periods. The specific mechanics of the tactics may be as fol lows: A new desired path Y** replaces the original path Y*. This new path, together with the stock condition 3.4, implies some new path M** for M. Since M 0 > — Y0, clearly M ** < M 0, since v 0 the paths Y** and M** proceed from Y0 and — Y0, respectively. Then the tactics call for v moving M onto M** by the end of T time periods. During these T time periods, there fore, there is yet a different planned target path M*** that the tactics indicate for M to follow; originates at the point M 0 and converges to M** at the end of the Tth time period. Such a policy 13 involves reducing M in each period up to the Tth by an amount equal to — (M — M /* ). The corresponding proporT f tional rate of reduction, analogous to the pro portional rate of growth g, is ( i m \ *** dm ) k r _ u D + k [_8 T\ _ i M where is the speed of response of the tactical response process, and the differential form is standard for simple exponential lags. 14 Equation 3.15 and flow condition 3.5 to gether form a system that contains enough in formation to solve for x***, defined now as the rate of growth of M that monetary policy tac tics should achieve. Use / i **** = \*** = D log M*** to substitute in condition 3.5 to obtain Y = - Mx*** s Hence 1 Y _ x*** vM g (3.16) Substitute 3.16 into equation 3.15 to obtain 1 M - M** T ' M ‘• • - s T i t ' - f O - r ) ] which is a first-order differential equation in 13 The policy described returns M to M** only asymptotically; it does not accomplish the full correction in T periods. jc* * * * 14 For reference, see Allen (1967), pp. 88 ff. 124 (3.17) Dx*** + k ( l x*** The solution of this differential equation is (3.18) x*** = g + (x 0 - g)e~kll~(,Fr)]1 where x*** = x 0 at i = 0 is the initial condi tion of the system. Hence if x 0 = £, there is no problem and x*** = g for all f, thus giving the familiar steady-state growth solution of equation 3.3. In the initial disturbance, or disequilibrium, case, however, x 0 =^=g> and so the full gen eralized equation 3.18 is necessary to determine x***. In particular, the coefficient on t in the exponential term of the equation determines the system’s behavior over time. Let represent this coefficient. The speed of response factor k, constrained by k > 0 , merely deter mines how fast or slow x*** is to follow any given pattern; it does not itself influence whether that pattern will be stable or unstable. The part of c within the brackets, in particular the relation between parameters g and T, de termines the stability for x*** over time. Three cases arise: Case 1: If T = — , then c vanishes, and so g the exponential term in equation 3.18 for x*** remains constant at unity for all time. Hence x*** = x0 > g for all t. While x*** does not diverge further from g, it does not converge to g either, and so this case does not meet the Harrod definition of stability. Case 2; If T < ~ then k > 0 implies that c > 0. Since c is the coefficient on time itself, the exponential term in equation 3.18 increases through time, and the growth rate x*** steadily diverges from g. In policy terms, this means that the Federal Reserve response to x 0 > g would have to make M grow at an ever-faster rate for all time, just to aim at returning M to M ** by the end of the ever-receding Tth pe riod. This case clearly is unstable. Case 3: If T > —, then k > 0 implies c < 0. Hence the exponential term in equation 3.18 decreases through time, vanishing in the limit to yield x*** = g. In policy terms, the Federal Reserve response to x 0 > g would be to make M grow at x*** > g for some time, eventually returning to the original steady-state growth rate g. This case is clearly stable, and so the policy-oriented definition of stability be comes clear: A policy-response system is stable if a finite initial disturbance leads to a finite amount of compensation to be effected in some period of time, thereby permitting the system to return to equilibrium and to the rates of growth which maintained before the disturbance. Hence T > — is necessary for the stability of the sys tem. Proceeding to derive the actual pattern of monetary policy, it is a straightforward exercise to use equations 3.15, 3.5, and 3.18 to obtain the target paths M *** and Y **, and the cor responding path A/**. Using x*** = D log M*** and integrating 3.18 will yield (3.19) log Mo = gt + I ( x o - g) [1 - '] Using 5.5 yields (3.20) Y ** Af* -k[l ~(rr) It The target path M**, representing not the path which monetary policy forces M to follow but rather that which it always seeks to make M approach by — of the discrepancy per period, is simply M** = -i- y**. v TACTICS AND STRATEGY In the limit of the stable case, M*** returns to M **, and x*** returns to g. M never re turns to the original target path M*; in the limit (3.21) Af*** = M * * = zM oeot where (3.22) log z = ^ (x 0 - g) J ? T _ t Similarly, Y** never returns to its original path Y* but rather follows the path of equa tion 3.20. Since M *** = M ** in the limit, for the stability case, the system returns to equilibrium with the original stock condition satisfied. CONCLUSIONS. The object of this some what tortuous exercise has been to illustrate the danger that the Federal Reserve System may itself destabilize the economy by causing movements in monetary policy instruments that attempt to do too much too soon. This point is relevant for decisions at both the strategic and the tactical levels. The specifica tions and parametric values of the economic relationships involved determine the conditions for actual stabilization. The important result in the model used here is that T > y is necessary for stability of the system. This conclusion contradicts the pre sumption of Section II that, in planning move ments of monetary policy instruments to cor rect for past exogenous disturbances or errors, proper tactics call for effecting the entire com pensation as quickly as possible. The analysis of this section, which includes not only the Y = y ( M) relation but also the feedback effects of Y on the accumulation of M t shows that this presumption is incorrect. Considera tions of stability force the compensation speed of response to be less than a certain rate, as determined by both the specification of the re lations involved and the values of the system’s individual parameters. IV. OPTIM AL FILTER IN G OF OPERATIONS DATA The discussion of Sections II and III, while explicitly acknowledging random events in the form of disturbances to the process of mone tary operations, has assumed complete cer tainty in the knowledge of past events. In the notation used in these sections, the actual value of the financial variable Mi? as well as its movement AiM, is an available datum as of the close of the /th time period. In reality, however, the available data are merely estimates that are based on sampling and reporting machineries and are subject to revision. This information-generating process is familiar in Bayesian analysis of decisions and in control theory. 15 Using certain advance in formation, the estimator formulates a subjec tive prior probability distribution for the varia ble in question. He then uses the newly available sampling and reporting data to up date this prior distribution into a posterior probability distribution on the same variable. When he receives yet another set of sample in formation, he treats this posterior distribution as a prior distribution (prior in the sense of its being prior to the second sample) and repeats the updating process to produce a new poster ior probability distribution. He may repeat this process as often as new information continues to arrive, producing as many posterior distri butions as there are distinct samples of data. AVAILABLE FEDERAL RESERVE DATA* Actual practice differs from this ideal ized conception only by being less complete. Each Friday the Federal Reserve Board staff produces a “Perspective on Bank Reserve Utilization.” This report presents, among other information, point estimates for the monthly movements in a number of monetary aggre gates. These point estimates are the means of the probability distributions considered in the Bayesian formulation of the information proc 15 Standard references are Pratt, Raiffa, Schlaiffer (1965) and Bryson and Ho (1969). and ess. The last Perspective of any month gives, for that month’s movements in a given mone tary variable M, a projection which for pur poses of this discussion is equivalent to the mean of a prior probability distribution. As the month ends and more complete sample data arrive, the staff uses the newly available information to update this estimate, producing a new estimate which is then the mean of a posterior probability distribution. This new esti mate appears in the first Perspective following the month’s end. As another week passes and still more sam ple data arrive, the staff treats the previous week’s posterior distribution as the current week’s prior distribution and updates it to pro duce a new posterior distribution for its expec tation of the movement in M in the month re cently ended. The mean of this second posterior distribution enters the second Perspective after the end of this month as a new, revised point estimate for the movement in M for this month. Current practice is to repeat this process not less than eight nor more than 15 times, and so the staff estimate of a month’s movement in M appears in the first eight and perhaps up to the next seven weekly Perspectives. Although the staff may not directly concep tualize its data-revision activities within a Bayesian framework, the updating process nev ertheless follows this general pattern. Similarly, while the staff may not explicitly view its point estimates as means of posterior probability dis tributions, the numbers generated are in each case the expected posterior values of M (pos terior in the sense of following upon all infor mation received through that time). Since the available monetary data represent the means of probability distributions, it is likely that other parameters of these distribu tions also yield information of potential useful ness for monetary policy and particularly for monetary policy tactics. The problem again is that the staff does not explicitly derive these probability distributions as such and hence cannot directly quote their various parameters. One approach—not a very good one for this application—would be to have the staff indi cate its confidence in each reported estimate, perhaps by bracketing its point estimate within an interval wide enough to reduce the subjec tive probability of the true number’s lying out side this band to one-third. Then assuming, for example, normal properties for the distribution itself, this band would be two standard devia tions wide. While formulating confidence intervals in this subjective manner may be a useful procedure in developing judgmental projections of future events, it is not well suited to the problem of developing levels of confidence in reported data subject to revision. A more direct approach is to analyze, ex post, the record of the staff’s data estimation machinery, testing the relation between data estimates for successive weeks and the num bers later accepted as the true numbers. Tables 1 through 4 show the results of such an exam ination applied to Perspective reports during 1968 and the first half of 1969.16 In Table 1, for example, the lines corre spond to different financial variables, all mone tary aggregates. The first column gives, for each variable, the variance of the first reported estimate of a month’s movement about the “true” value. 17 The next seven columns give, for each variable, the analogous variances for the second through the eighth estimates. For monthly changes in total reserves, for example, Table 1 shows that the first reported estimate has variance of 6,630 about the true value. By the fourth reported estimate, that is, allowing a reporting lag of 4 weeks, this variance falls to 814; by the eighth reported estimate, the vari ance falls to 328. Tables 2 and 3 repeat the same format, giv ing, respectively, the standard deviations and average absolute errors corresponding to Table 16 Specifically, the Perspective reports used span the period from January 1968 through July 1969. 17 For purposes of this analysis the “true” value is the last one reported in a Perspective table, before the month is dropped from the listings. TACTICS AND STRATEGY TABLE 1: Variances of Reporting Errors Weekly Estimates of Monthly Changes in Monetary Aggregates W eeks after m onth -end V ariable 1 2 3 5 4 7 8 740 1 ,1 4 0 886 641 980 872 328 668 628 .0 3 6 .065 .0 0 0 .0 6 6 .0 0 4 .0 1 3 .049 .0 3 5 .0 5 4 .001 .061 .0 0 3 .0 0 7 .0 4 7 .001 .0 1 7 .001 .0 1 6 .0 0 2 .007 .0 1 6 6 7 8 27 34 30 25 31 30 18 26 25 .1 9 0 .2 5 5 .2 5 7 .0 6 3 .1 1 4 .221 .1 8 7 .2 3 2 .0 3 2 .2 4 7 .0 5 5 .0 8 4 .2 1 7 .0 3 2 .1 3 0 .0 3 2 .1 2 6 .0 4 5 .0 8 4 .1 2 6 6 7 8 6 M illio n s o f d ollars, squared T otal reserves................................................................................ N on borrow ed reserves.............................................................. T otal required reserves.............................................................. 6 ,6 3 0 3 ,7 8 0 2 ,8 6 0 2 ,5 7 0 2 ,0 3 0 2 ,3 0 0 1 ,5 1 0 1 ,7 2 0 1 ,5 5 0 T otal m em ber bank d e p o sits.................................................. T otal m on ey su p p ly .................................................................... C u rrency.......................................................................................... D em an d d ep o sits......................................................................... U .S . G overnm ent dem and deposits (m em ber b a n k s). T im e deposits (com m ercial b an k s)...................................... M on ey supply an d tim e d e p o sits.......................................... .0 5 9 .0 9 8 .0 0 3 .0 8 8 .0 2 4 .0 1 6 .1 3 5 .057 .0 5 0 .0 0 6 .0 3 9 .0 0 9 .0 1 8 .0 6 8 .031 .0 5 2 .0 0 2 .0 4 5 .003 .0 1 3 .0 4 6 814 1 ,3 9 0 885 810 1 ,2 4 0 884 B illions o f dollars, squared .0 3 7 .0 5 5 .0 0 2 .053 .0 0 3 .0 1 3 .051 .028 .055 .001 .0 5 5 .0 0 2 .011 .0 4 6 TABLE 2: Standard Deviations of Reporting Errors Weekly Estimates of Monthly Changes in Monetary Aggregates W eeks after m onth -end V ariable 1 2 3 4 5 M illio n s o f dollars T otal reserves................................................................................. N on b orrow ed reserves.............................................................. T otal required reserves.............................................................. 81 62 53 51 45 48 39 42 39 T otal m em ber bank d e p o sits.................................................. T otal m on ey su p p ly.................................................................... .2 4 3 .3 1 3 .0 5 5 .2 9 7 .1 5 5 .1 2 6 .3 6 7 .239 .2 2 4 .0 7 7 .1 9 7 .0 9 5 .1 3 4 .261 .1 7 6 .2 2 8 .0 4 5 .2 1 2 .0 5 5 .1 1 4 .2 1 4 29 37 30 28 35 30 B illion s o f dollars D em an d d e p o sits......................................................................... U .S . G overnm ent dem and d ep osits (m em ber b a n k s). T im e d ep osits (com m ercial b a n k s)...................................... M on ey supply and tim e d e p o sits......................................... .1 9 2 .2 3 5 .0 4 5 .2 3 0 .0 5 5 .1 1 4 .2 2 6 .1 6 7 .2 3 5 .0 3 2 .235 .0 4 5 .1 0 5 .2 1 4 —_ TABLE 3: Average Absolute Reporting Errors Weekly Estimates of Monthly Changes in Monetary Aggregates W eeks after m onth -end Variable 1 2 3 4 5 M illio n s o f dollars T otal reserves................................................................................. T otal required reserves.............................................................. 57 46 35 32 30 27 25 25 21 .1 8 4 .2 7 9 .0 2 6 .2 5 3 .111 .111 .321 .1 4 7 .1 8 4 .0 4 7 .1 4 7 .0 5 3 .0 8 4 .2 2 6 .111 .1 8 4 .021 .1 3 7 .0 2 6 .0 6 3 .1 7 4 16 22 13 15 19 13 13 17 12 9 11 11 5 7 9 .0 6 3 .1 5 3 .0 5 3 .1 1 6 .0 0 5 .111 .0 1 6 .0 3 2 .1 1 6 .0 0 5 .0 5 8 .0 0 5 .0 4 7 .011 .0 3 2 .0 6 8 B illion s o f dollars T otal m em ber bank d e p o sits.................................................. T otal m on ey su p p ly .................................................................... U .S . G overnm ent dem and dep o sits (m em ber b a n k s). T im e d ep osits (com m ercial b a n k s)...................................... M on ey supply and tim e d e p o s its ......................................... .1 0 5 .1 7 4 .0 1 6 .1 5 3 .0 3 2 .0 6 8 .1 6 8 .0 6 3 .1 4 2 .0 0 5 .1 3 7 .021 .0 5 3 .1 4 2 --- .1 4 7 .0 2 6 .0 4 7 .1 3 7 TABLE 4: Average Reporting Errors Weekly Estimates of Monthly Changes in Monetary Aggregates W eeks a fter m onth -end Variable 1 2 3 4 5 8 6 7 -1 2 -1 6 -7 -9 -1 1 -8 -5 -7 -6 - .0 2 1 .0 5 8 .0 0 0 .053 — .0 0 5 - .0 2 6 .0 3 2 - .0 3 2 .021 - .0 0 5 .0 1 6 - .0 1 6 - .0 1 1 .011 .0 0 5 .0 3 7 - .0 0 5 .037 - .0 0 1 - .0 1 1 .0 3 7 M illio n s o f d ollars T otal reserv es................................................................................ N on borrow ed reserves.............................................................. T otal required reserves............................................................. -4 3 -2 8 -2 1 -2 6 -2 5 -2 1 -1 9 -2 0 -1 6 T otal m em ber bank d e p o sits................................................. T otal m on ey su p p ly .................................................................... C urrency.......................................................................................... D em and d e p o sits......................................................................... U .S . G overnm ent dem and deposits (m em ber b an k s). T im e dep osits (com m ercial b a n k s)...................................... T im e supply and time d ep o sits....... ..................................... .0 0 5 .0 5 8 - .0 0 5 .0 3 2 .047 .0 0 5 .111 - .0 1 1 .068 .005 .0 4 2 .011 - .0 3 2 .0 3 7 .0 1 6 .0 5 8 .0 0 0 .0 5 3 - .0 0 5 - .0 1 1 .0 4 7 -1 3 -1 5 -8 -1 3 -1 8 -8 B illions o f dollars 1. Table 4 gives the corresponding mean er rors; nonzero mean errors indicate bias in the reporting process and are probably due to the effect on the computations of data series revisions. 18 As an illustration of the significance of these data, compare the standard deviations for two monetary aggregates that the Open Market Committee may wish to control, for example, the money supply and total reserves of mem ber banks. In the first report after the end of the month, the money supply estimate is ac curate to within a standard deviation of $313 million, or approximately yG of a per cent on a base of some $ 2 0 0 billion; the total reserves estimate is accurate to within a standard devia tion of $81 million, or approximately 16 of a per cent on a base of some $27 billion. 19 These errors correspond to a 2 and a 4 per cent annual rate of change, respectively. By the fourth report after the end of the month, the money supply estimate is accurate to within a standard deviation of $235 million, or approximately y1 0 of a per cent, and the total reserves estimate is accurate to within a standard deviation of $29 million, or approxi 18 The calculations reported in the tables have al ready made some adjustments for series revisions. 19 The magnitude of reporting error in member bank total reserves may seem surprisingly large. These errors in fact reflect the “as-of adjustments” made within the Federal Reserve Banks to correct for accounting errors in debiting and crediting mem ber bank accounts. .011 .0 5 8 - .0 0 5 .0 5 8 - .0 1 1 - .0 0 5 .0 5 3 - .0 1 1 .047 .005 .053 .0 0 0 - .0 1 1 .0 3 7 mately y1 0 of a per cent. These errors both correspond to a VA per cent annual rate of change. By the eighth report after the end of the month, the money supply estimate is ac curate to within a standard deviation of $130 million, or approximately 7/100 of a per cent; the total reserves estimate is accurate to within a standard deviation of $18 million, or ap proximately 7/100 of a per cent. Both of these errors correspond to a 1 per cent annual rate of change. Hence, under current reporting and estima tion systems, money supply information is in the first instance more accurate than total re serves information; but this difference effec tively vanishes with an allowed reporting lag of one month or longer. CONCEPTUAL USE OF PROBABILITY DISTRIBUTION PARAMETERS. The object of the above exercise has been to obtain pa rameters of the successive probability distribu tions corresponding to successive data reports, while circumventing the tedious and somewhat inapplicable procedure of having the staff esti mate these parameters subjectively. The re ported data in the Perspectives are the means of these distributions. By assuming that the sampling, reporting, and updating machineries have not changed radically in the past 2 years, it is possible to accept the data in Table 1 as approximations to the variances of these prob ability distributions. That such variances, or any corresponding parameters, are useful inputs in formulating TACTICS AND STRATEGY monetary policy tactics remains to be shown. Heuristically, the argument is as follows: As elaborated earlier, there is usually an ad vantage to discovering operational errors in the monetary policy instrument variables and to undertaking the proper responses as quickly as possible. Even in circumstances in which considerations of stability lead to spreading compensatory reactions over some substantial period, it is usually advantageous to begin these reactions at the earliest possible time. Simultaneously, however, basing tactical de cisions on incorrect data leads to incorrect de cisions. It is possible to undertake compensa tory responses to errors that have not occurred, as well as to over- or underestimate the amount of correction necessary. It is possi ble, though less probable, to undertake a com pensatory response in the wrong direction, if the data report is such as to change the signs of the true differences that have arisen between the actual and the desired values of the mone tary policy instruments. Hence, from the stand point of avoiding mistakes due to faulty data, it is best to postpone taking action until the data are more secure. Two influences therefore oppose each other —one tending to accelerate and one to delay tactical action. The solution to the dilemma must in most cases be some compromise. Just what this answer means in operational terms, however, is not obvious. The following more formal analysis should clarify the role of the data probability distribution parameters in de termining the best tactics in any given situa tion. EXAMPLE OF A DATA FILTERING SCHEME. Consider, for example, a simple Y = y(M) model of the form (4.1) Yi = . . . + p E i —K M i- i + • • • where none of the omitted terms is a lagged value of Y. This model is the rectangular lag model of Example 1, Case D, in Section II. To recall, Case D is interesting in that operational errors in this specification call for subsequent compensatory responses in both the levels M and the changes AM. Here K is the number of time periods in the no-response section of the lag in the effect of M on Y , while N time pe riods is the total length of the lag. More specifically, consider the case in which K — 0 and N = 2. While this lag pattern is unrealistic, it has two advantages: First, the mathematical manipulations of the problem in crease in number as (N — K + 1 ) 2 increases, and a simple example should suffice to illustrate the main points of the analysis without intro ducing unnecessary complexities. Second, if p = Vs, then this expression is simply an averaging term; replace Yi on the left-hand side by some term in M, and the equation expresses the average value of M in a quarter as onethird times the sum of M in each month of the quarter. Hence the K = 0, N = 2 model is ap plicable both to tactical decisions as discussed above and to the somewhat narrower, more specific problem of achieving a predetermined average M over any given quarter. Suppose, then, that month i — 2 has ended, and that the tactical problem is to select the proper plan for month / — 3. Suppose also that the circumstances are the following: 1. Y 3* is fixed, from the currently operative strategy. 2. Mt*, M2*, and M3* are also fixed from the currently operative strategy. 3. M x and M 2 are known; further, M x and M 2 ^=M2*. 4. The loss associated with Y3 is the simple quadratic L 3 = (Y3* — Y3) a. In the deterministic cases considered in Sec tions II and III, achieving (4.2) M 3** = M3* - (Mt - M2*) - (Mi - Mi*) will determine Y3 = Y3* and L 3 = 0. Suppose, however, that assumption 3 above does not hold. M x and M 2 are not known with certainty. Instead, define M*J = value of M at the end of the zth month, as reported at the end of the jth month. Then the most recent avail able estimates of M x and M 2 are M x2 and M 22> namely, the current report on both values as of the close of month j = 2 . The tactical decision of equation 4.2 may now become timates available k months after the fact, and 0 < yk < 1, all k > 0. Using the rule of equation 4.5 yields the relation between M3** and the actual previous experience of (4.3) M 3** = M 3* - (M 22 - M 2*) (4.6) M3** = M3* - - (M i 2 - Mi*) but acting in this fashion ignores the possibility of reporting errors, that is, the possibility that M 22 ^ M2 and M x2 ^ M2. One recourse is to apply some uncertainty discounting factor to the second and third terms in equation 4.3, in line with the classical formulations of optimal filtering. Define the variables a s Mi - Mi * €ik — Mij Mi*, Rik ss Mij - Mi, k = j i (4.4) €» = 6 ik Rik Equation 4.2, the rule with certain knowledge, is then M 3** = M 3* - 6 2 - 6 1 Equation 4.3, the equivalent using reported data estimates, is M 3** = M 3* “ €20 — €11 By using equation 4.4 and substituting, M 3** = M 3* - (62 + * 20) - (ei + Ru) The idea of applying a filtering process to the available data is to make the operating rule (4.5) M 3** = M z* — 70*20 — yien where y* = discount factor applied to data es + * 20) — 7i(*i + * 11) Assuming that the Manager in fact achieves this goal, so that M3 = M3**, a series of ma nipulations yields (4.7) Lz = 02[e2(l - 70 ) + «i(l - 7i) + 70*20 + 7i*n]2 The optimal filtering problem is to choose y0 and yx so as to minimize the expected value of L3, (4.8) E (L z) = PWV ~ 270 + 7o2) + €i2( 1 — 2 7 1 — 7i3) k =j - i Here = the true operating miss on M in the /th month; eu- = the estimated operating miss on M in the /th month, as reported k months after the end of the /th month; and Rin = the reporting error in the estimate of M in the /th month, reported k months after the end of the /th month. Then 7 0 (6 2 + 26162 (l — 70 — 71 + 7071) + 2e2/?2o(7o — 7 o 2) + + 261 /^1 1 (7 1 — 7 12) + 261 262 * 2 0 ( 7 0 — 7o7i) + * 1 1 (7 1 — 7o7i) 2 * 20* 11(7 0 7 1 ) + 7 o2* 202 + 7 i 2* i i 2] At this stage, several specific assumptions are necessary that are crucial to the specific results achieved, though not to the development of the argument. In other words, changing this set of assumptions would change the results of the computation without invalidating the process of deriving these results: First, assume, as suggested above, that the probability distribution of R a reflects the re cent history of A-month lagged observations on the variable M. Assume that the reporting record in the recent past for such estimates Rik is unbiased (mean zero) and has variance 0% -“. Hence (4.9) E(Rik) = 0, E(Rik*) = all / Second, assume that there is no relation be tween an error in the reported estimate of one month’s value of M and any errors in simul TACTICS AND STRATEGY taneously reported estimates of values of M for other months. Hence (4.18) (en2 + 2<7-i2) ti — *n2 + cr\ 2 (4.10) E(RikRjh) = 0, all / ^ j, These equations are analogous to 4.13 and 4.14, but are operational in the sense of con taining only the unknowns to be found ( 7 0 and yi), the reported operating misses, and vari ances drawn from analysis of the recent his tory of data-reporting errors. Again, this is a simultaneous system; the operational solution analogous to equations 4.15 and 4.16 is (/ - j ) = - ( k - h) Third, assume that there is no relation be tween errors in reported data estimates and the difference between actual M and target M* for any month or for any report. Hence (4.11) E(uR*) = 0, all /, k Applying these three assumptions to equa tion 4.8 yields (4.12) E{U) = /32[e22(l - 27o - To2) + Ci2(l — 2yi — 7 12) + 2e2ei(l — To — 7i + ToTi) + ToVo2 + T i 2^ i 2] The first-order minimum conditions are Applying these conditions yields (4.13) (e22 + <ro 2) t o = e22 + *2ei(l - 7 i) (4.14) (ei2 + <ti2) t i = *i2 + €26i ( 1 — To) Solving this pair of simultaneous equations for the discount parameters yields x *2Vi2 + €ie2<ri2 (4.15) 70 - £ iW + ^ 2 + ^2 ^2 (4.16) 7i - e iW + tif2<ro 2 + ej2<ro2 + The principal remaining difficulty with this pair of equations is that the true operating misses, the e* are unknown. The solution, therefore, is to return to equations 4 .13 and 4 .1 4 and substitute the appropriate eik and R ik expressions from 4.4 , and then to take ex pected values as before, using assumptions 4.9, 4 .10, and 4.11. The result of these operations is the pair of equations (4.17) (e2o2 + 2(7o2) to = *202 + cro2 + €20*11(1 — Ti) + €20*11(1 “ To) (4.19) _ eii2q~o2 + 2e20 V i 2 + e20en<7i2 + 2<70 V i 2 To_ 2e„W + 2a20Vi5 + 4<r0W (4.20) _ ^1 6202<rl 2 + 2 €liV 0 2 + €20*110'Q 2 + 2 q-qVi2 e o ci 2 + 2 €h20 -o2 + 4<ro2(ri2 2 2 2 These two equations express the optimal weights to be attached to the reported errors e20 and en in making a tactical decision about Af3** according to equation 4.5. Based on the relative magnitudes of e20, *u, oo2, anc* °i 2 discount parameters may vary substantially; in particular, their values are very sensitive to whether the reported errors e20 and etl are in the same or opposite directions. In some cases one of the two discount parameters may be greater than 1.0. The sum of the two, however, is always less than 2 . 0 when the reporting error variances are nonzero. (If these variances do go to zero, there is no uncertainty factor in the reported data, and there exists no true solution for y0 and y1? as seen from equations 4.13 and 4.14.) ALTERNATIVE SCHEMES. Equations 4.19 and 4.20 express the optimal uncertainty discount factors, consistent with the specifica tion of the model itself in 4.1, a quadratic loss function, and an operating rule as in 4.3 and 4.5. This solution is probably the best and most workable method, but others are possible. Al ternative solutions for choosing these uncer tainty discount factors could take two different paths from here, one simpler and one more complex. The simpler solution is to return to equa tions 4.13 and 4.14 and to make the same re cent historical observations about the operating misses e* as the analysis above has done for the data-reporting variances a*2. Specifically, data study may show that over some recent period the root mean-square operating miss has been some value e. Further study may show that the operating miss in one month is not necessarily correlated with a miss in the next. Then apply ing expected values of the form E(ei) = e and E(uej) = 0, / ^ j reduces equations 4.13 and 4.14 to (4.21) e2 70 ' * (4.22) ^ e2 €2 + <T\ 2 This nonsimultaneous form may in fact be preferable because of its simplicity. It implies setting the yfc on the basis of the relevant datareporting variances in relation to the operatingmiss variances of the recent past. It requires no specific calculation of new y a t each tactical decision, but only a periodic updating of the crft and e values to keep them consistent with current experience. The more complex solution than equations 4.19 and 4.20 involves reconsidering assump tions 4.10 and 4.11 in the light of further data study. First, it is possible that data-reporting errors in successive months are correlated. Sec ond, if the staff personnel evaluating the in coming reports are aware of the targets that monetary policy operations are trying to achieve, it is possible that a given month’s operating miss and that month’s data-reporting error are correlated. Relaxing these assump tions would involve substituting for some terms of equation 4.8 their estimated recent values, rather than dropping them altogether as is the case in equation 4.12. Hence the resulting rules analogous to equations 4.19 and 4.20 would be more complex, though perhaps more realistic, determinants of y0 and yt. Although the analysis above has presented only a simple example, for one particular specification of the Y = y( M) relation and for one loss function, the methods developed are applicable to more general circumstances, and the changes necessary to incorporate modifica tions follow directly from this procedure. Similarly, as in previous sections, the monthly time period used here is merely an expositional device. The Perspective tables and their implied probability distributions are available weekly, and Tables 1 to 4 summarize some parameters of these distributions. If, for example, tactics use rules of the form of equations 4.21 and 4.22 to set the uncertainty discount factors yic, then the relevant y&to use may change with each week; this procedure would reflect the greater confidence in data reports after the extra week’s time lag. CONCLUSION, The main point of this sec tion is that monetary policy tactics should take account of the possibility of errors in the avail able current data by applying some filtering process to these data. In actual practice the filter should take the form of a set of uncer tainty discount parameters to apply to data reports and estimates of particular vintages. A number of schemes, some simple and some complex, are available to compute these param eters, and there is room for choice among them; but failing to discount for data errors at all and ignoring the possibility of data revisions may lead to undesired results in monetary policy operations. V. SUMMARY OF CONCLUSIONS FOR MONETARY POLICY The major conclusions of Sections I through IV for monetary policy can be briefly restated as follows: 1. The monetary policy decision process should contain two phases— strategy and tac tics. Strategy involves quarterly decisions out lining a plan for monetary policy over the sev eral following quarters. Tactics involve TACTICS AND STRATEGY shorter-run technical decisions concerning im plementing the first quarter of the strategy and deal with the question of how best to adjust for apparent deviations of the monetary policy instruments from their planned targets. 2. The formulation of monetary policy strategy should follow a sequential decision making procedure, revising multiquarter strate gies once per quarter. At each decision, the immediate quarter of such a strategy becomes the currently effective operating policy. 3. Tactical decisions are not purely techni cal. An illustration of their real economic sub stance is the dependence of proper tactics upon the specification of the relation between the instruments and the ultimate goals of mone tary policy. Different specifications may imply no response at all to past operating misses of the monetary policy instruments or may imply compensating responses of a number of differ ent forms. 4. In planning monetary policy tactics, as well as strategy, it is important to take account of the implications for stability of the simulta neous structure of the financial and nonfinancial system. Because of feedback relationships between instrument variables and the rest of the economy, movements in the instruments from one level to another, should they be too rapid, may in fact destabilize the economy rather than stabilize it. 5. Because the data available to monetary policy decisions contain reporting errors and are subject to revision, it is necessary to apply a set of uncertainty discount factors to the data when devising appropriate responses to apparent operating misses in the monetary pol icy instruments. These uncertainty discount factors should effect a compromise between the desire to react quickly so as to prevent un wanted situations from persisting and the de sire to delay so as to have better data. REFERENCES 134 Books Allen, R. G. D . M acro-E conom ic T heory. Lon don: Macmillan, 1967. Bryson, Arthur E., and H o, Yu-Chi. A p p lied O p tim al Control: O ptim ization , Estim ation and C ontrol. Waltham: Blaisdell Publishing Com pany, 1969. Harrod, R. F. Tow ards a D yn am ic E conom ics. London: Macmillan, 1948. Koyck, L. M. D istribu ted Lags and Investm ent A nalysis. Amsterdam: N orth-H olland Publish ing Company, 1954. Maisel, Sherman J. “Controlling Monetary Aggre gates,” Controlling M onetary A ggregates. Bos ton: Federal Reserve Bank o f Boston, 1969. Pratt, John, Raiffa, Howard, and Schlailfer, R ob ert. Introduction to Statistical D ecision T h eory. N ew York: M cGraw-Hill, 1965. Theil, Henri. O ptim al D ecision R ules fo r G overn m ent and Industry. Amsterdam: North-Holland Publishing Company, 1964. Periodicals and Other de Leeuw, Frank, and Gramlich, Edward. “The Federal R eserve-M IT Econometric M odel,” Federal Reserve Bulletin, V ol. 54 (Jan. 1 9 6 8 ), pp. 11 -4 0 . Holt, Charles C. “Linear D ecision Rules for E co nomic Stabilization and Growth,” Q uarterly Journal o f Econom ics, V ol. 76 (Feb. 1962), pp. 2 0 -4 5 . Jorgenson, Dale W. “On Stability in the Sense of Harrod,” E conom ica, V ol. 27 (Aug. 1960), pp. 2 4 3 -4 8 . Simon, Herbert A . “D ynam ic Programming under Uncertainty with a Quadratic Criterion Function,” E conom etrica, V ol. 24 (Jan. 1956), pp. 7 4 -8 1 . CONTENTS 137 INTRODUCTION 138 I. THE THEORY OF MONETARY POLICY UNDER UNCERTAINTY Basic concepts Monetary policy under uncertainty in a Keynesian model 138 139 145 145 145 148 149 151 151 153 158 161 167 167 167 172 II. EVIDENCE ON THE RELATIVE MAGNITUDES OF REAL AND MONETARY DISTURBANCES Nature of available evidence Impact of an expenditure disturbance Evidence from reduced-form equations Evidence on stability of demand for money function III. A MONETARY RULE FOR GUIDING POLICY Rationale for a rule-of-thumb Post-accord monetary policy A monetary rule Tests of the proposed rule IV. SELECTION AND CONTROL OF A MONETARY AGGREGATE Basic issues Selection of a monetary aggregate Technical problems of controlling money stock 182 184 186 V. SUMMARY Purposes of the study The theory of monetary policy under uncertainty Evidence on relative magnitudes of real and monetary disturbances A monetary rule for guiding policy Selection and control of a monetary aggregate Concluding remarks 187- REFERENCES 179 179 179 181 by William Poole RULES-OF-THUMB FOR GUIDING MONETARY POLICY RULES-OF-THUMB FOR POLICY INTRODUCTION This study has been motivated by the recogni tion that the key to understanding policy prob lems is the analysis of uncertainty. Indeed, in the absence of uncertainty it might be said that there can be no policy problems, only adminis trative problems. It is surprising, therefore, that there has been so little systematic atten tion paid to uncertainty in the policy literature in spite of the fact that policy-makers have re peatedly emphasized the importance of the un known. In the past, the formal models used in the analysis of monetary policy problems have al most invariably assumed complete knowledge of the economic relationships in the model. Uncertainty is introduced into the analysis, if at all, only through informal consideration of how much difference it makes if the true rela tionships differ from those assumed by the pol icy-makers. In this study, on the other hand, uncertainty plays a key role in the formal model. Since this study is so long, a few comments at the outset may assist the reader in finding his way through it. The remainder of this in troductory section outlines the structure of the study so that the reader can see how the var ious parts fit together. The reader interested only in a summary of the analysis and empiri cal findings should read this introductory sec tion and then turn directly to the summary in Section V. This summary concentrates on the theoretical analysis while only briefly stating the most important empirical findings. It omits completely the technical details of both the theoretical and empirical work. The reader in terested in the technical details should, of N o t e .—The author is Senior Economist* Special Studies Section, Division of Research and Statistics, Board of Governors of the Federal Reserve System. Special thanks are due Miss Joan Walton for her as sistance in programming and other matters, Mrs. Lillian Humphrey for assistance with the figures, and Miss Debra Bellows for typing a long and messy manuscript through several drafts. The author, of course, is wholly responsible for any remaining er rors. course, turn to the appropriate parts of Sec tions I through IV. Insofar as possible these sections have been written so that the reader can understand any one section without having to wade through all of the other sections. Section I contains the theoretical argument comparing interest rates and the money stock as policy-control variables under conditions of uncertainty. The analysis is verbal and graphi cal, using the simple Hicksian IS-LM model with random terms added. This model is gen eral enough to include both Keynesian and monetarist outlooks, depending on the specific assumptions as to the shapes of the functions. Since the theoretical analysis emphasizes the importance of the relative stability of the ex penditures and money demand functions, an examination of the evidence on relative stabil ity appears in Section II. Given the conclusion of Section II on the superiority of a policy operating through ad justments in the money stock, the next ques tion is how the money stock should be adjusted to achieve the best results. While pol icy-makers generally look askance at sugges tions for policy rules, the only way that econo mists can give long-run advice is in terms of rules* That is to say, the economist is not being helpful at all if he in effect says, “Look at the rate of inflation, at the rate of unemployment, at the forecasts of the Government budget deficit, and at other relevant factors, and then act ap propriately.” Advice requires the specification of exactly how policy should be adjusted, and for this advice to be more than an ad hoc rec ommendation for the current situation, it must involve specification of how the money stock or some other control variable should be ad justed under hypothetical future conditions of inflation, unemployment, and so forth. The purpose of Section III is to develop such a rule-of-thumb, or policy guideline, based on the theoretical and empirical analyses of Sec tions I and II. A number of technical problems of mone tary control are examined in Section IV. After a short introduction to the issues, the first part of this section discusses the relative merits of a number of monetary aggregates including var ious reserve measures, the narrowly and broadly defined money stocks, and bank credit. The second part examines whether pol icy should specify desired rates of change of an aggregate in terms of weekly, monthly, or quarterly averages, or in some other manner. The third part examines in a very incomplete fashion a few of the problems of adjusting open market operations so as to reach the de sired level of an aggregate. Finally, Section V consists of a summary of Sections I through IV. To avoid undue repeti tion, woven into this summary section are a number of general observations not examined in the other sections. I. THE THEORY OF M ONETARY POLICY UNDER UNCERTAINTY BASIC CONCEPTS, The theory of optimal policy under uncertainty has provided many in sights into actual policy problems [8, 12, 21, 25]. While much of this theory is not accessible to the nonmathematical economist, it is pos sible to explain the basic ideas without resort to mathematics. The obvious starting point is the observation that with our incomplete understanding of the economy and our inability to predict accu rately the occurrence of disturbing factors such as strikes, wars, and foreign exchange crises, we cannot expect to hit policy goals exactly. Some periods of inflation or unemployment are unavoidable. The inevitable lack of precision in reaching policy goals is sometimes recog nized by saying that the goals are “reasonably” stable prices and “reasonably” full employ ment. While the observation above is trite, its im plications are not. Two points are especially important. First, policy should aim at minimiz ing the average size of errors. Second, policy can be judged only by the average size of er rors over a period of time and not by individ ual episodes. Because this second point is par ticularly subject to misunderstanding, it needs further amplification. Since policy-makers operate in a world that is inherently uncertain, they must be judged by criteria appropriate to such a world. Consider the analogy of betting on the draw of a ball from an urn with nine black balls and one red ball. Anyone offered a $2 payoff for a $1 bet would surely bet on a black ball being drawn. If the draw produced the red ball, no one would accuse the bettor of a stupid bet. Simi larly, the policy-maker must play the eco nomic odds. The policy-maker should not be accused of failure if an inflation occurs as the result of an improbable and unforeseeable event. Now consider the reverse situation from that considered in the previous paragraph. Suppose the bettor with the same odds as above bets on the red ball and wins. Some would claim that the bet was brilliant, but as suming that the draw was not rigged in any way, the bet, even though a winning one, must be judged foolish. It is foolish because, on the average, such a betting strategy will lead to sub stantially worse results than the opposite strat egy. Betting on red will prove brilliant only one time out of 10, on the average. Similarly, a particular policy action may be a bad bet even though it works in a particular episode. There is a well-known tendency for gam blers to try systems that according to the laws of probability cannot be successful over any length of time. Frequently, a gambler will adopt a foolish system as the result of an ini tial chance success such as betting on red in the above example. The same danger exists in economic policy. In fact, the danger is more acute because there appears to be a greater chance to “beat the system” by applying eco nomic knowledge and intuition. There can be no doubt that it will become increasingly pos sible to improve on simple, naive policies through sophisticated analysis and forecasting and so in a sense “beat the system.” But even with improved knowledge some uncertainty RULES-OF-THUMB FOR POLICY will always exist, and therefore so will the tendency to attempt to perform better than the state of knowledge really permits. Whatever the state of knowledge, there must be a clear understanding of how to cope with uncertainty, even though the degree of uncer tainty may have been drastically reduced through the use of modern methods of analy sis. The principal purpose of this section is to improve understanding of the importance of uncertainty for policy by examining a simple model in which the policy problem is treated as one of minimizing errors on the average. Particular emphasis is placed on whether con trolling policy by adjusting the interest rate or by adjusting the money stock will lead to smaller errors on the average. The basic argu ment is designed to show that the answer to which policy variable—the interest rate or the money stock—minimizes average errors de pends on the relative stability of the expendi tures and money demand functions and not on the values of parameters that determine whether monetary policy is in some sense more or less “powerful” than fiscal policy. MONETARY POLICY UNDER UNCER TAINTY IN A KEYNESIAN MODEL.1 The basic issues concerning the importance of uncertainty for monetary policy may be ex amined within the Hicksian IS-LM version of the Keynesian system. This elementary model has two sectors, an expenditure sector and a monetary sector, and it assumes that the price level is fixed in the short run.2 Consumption, investment, and government expenditures func tions are combined to produce the IS function in Figure 1, while the demand and supply of money functions are combined to produce the LM function. If monetary policy fixes the stock of money, then the resulting LM function is LAfi, while if policy fixes the interest rate at r0 1 For the most part this section represents a verbal and graphical version of the mathematical argument in [25]. 2 Simple presentations of this model may be found in [6, pp. 275-82] and [7, pp. 327-32]. FIGURE 1 the resulting LM function is LM 2. It is assumed that incomes above “full employment income” are undesirable due to inflationary pressures while incomes below full employment income are undesirable due to unemployment. If the positions of all the functions could be predicted with no errors, then to reach full em ployment income, Yh it would make no differ ence whether policy fixed the money stock or the interest rate. All that is necessary in either case is to set the money stock or the interest rate so that the resulting LM function will cut the IS function at the full employment level of income. Significance of disturbances. The positions of the functions are, unfortunately, never pre cisely known. Consider first uncertainty over the position of the IS function—which, of course, results from instability in the under lying consumption and investment functions— while retaining the unrealistic assumption that the position of the LM function is known. What is known about the IS function is that it will lie between the extremes of /Si and IS 2 in Fig ure 2. If the money stock is set at some fixed level, then it is known that the LM function will be L Mlf and accordingly income will be somewhere between the extremes of and Y*>. On the other hand, suppose policy-makers fol low an interest rate policy and set the interest rate at r0. In this case income will be some where between Y /, and Y 2', a wider range than 140 FIGURE 2 FIGURE 3 Y x to r*, and so the money stock policy is superior to the interest rate policy.3 The money stock policy is superior because an unpredicta ble disturbance in the IS function will affect the interest rate, which in turn will produce spend ing changes that partly offset the initial disturb ance. The opposite polar case is illustrated in Fig ure 3. Here it is assumed that the position of the IS function is known with certainty, while unpredictable shifts in the demand for money cause unpredictable shifts in the LM function if a money stock policy is followed. With a money stock policy, income may end up any where between Y x and Y But an interest rate policy can fix the LM function at LM 3 so that it cuts the IS function at the full employment level of income, Y ft With an interest rate policy, unpredictable shifts in the demand for money are not permitted to affect the interest rate; instead, in the process of fixing the inter est rate the policy-makers adjust the stock of money in response to the unpredictable shifts in the demand for money. In practice, of course, it is necessary to cope with uncertainty in both the expenditure and monetary sectors. This situation is depicted in Figure 4, where the unpredictable disturbances are larger in the expenditure sector, and in Figure 5 where the unpredictable disturbances are larger in the monetary sector. The situation is even more complicated than shown in Figures 4 and 5 by virtue of the fact that the disturbances in the two sectors may not be independent. To illustrate this case, consider Figure 5 in which the interest rate policy is superior to the money stock policy if the dis turbances are independent. Suppose that the disturbances were connected in such a way that disturbances on the LM2 side of the average LM function were always accompanied by dis turbances on the IS 2 side of the average IS function. This would mean that income would never go as low as Y u but rather only as low as the intersection of LM t and IS2, an income not as low as Y x under the interest rate policy. Similarly, the highest income would be given by the intersection of LM Z and ISU an income not so high as Y2'.4 4 The diagram could obviously have been drawn so that an interest rate policy would be superior to a money stock policy even though there were an inverse relationship between the shifts in the IS and LM func tions. However, inverse shifts always reduce the mar 3 In Figure 2 and the following diagrams, the out gin of superiority of an interest rate policy, possibly comes from a money stock policy will be represented to the point of making a money stock policy supe by unprimed Y*s, while the outcomes from an inter rior. Conversely, positively related shifts favor an in terest rate policy. est rate policy will be represented by primed Y's. RULES-OF-THUMB FOR POLICY FIGURE 4 FIGURE 5 Importance of interest elasticities and other parameters. So far the argument has concen trated entirely on the importance of the relative sizes of expenditure and monetary disturbances. But is it also important to consider the slopes of the functions as determined by the interest elasticities of investment and of the demand for money, and by other parameters? Consider the pair of IS functions, ISXand IS2, as opposed to the pair, IS 3 and IS4, in Figure 6. Each pair FIGURE 6 represents the maximum and minimum posi tions of the IS function as a result of disturb ances, but the pairs have different slopes. Each pair assumes the same maximum and minimum disturbances, as shown by the fact that the horizontal distance between and IS 2 is the same as between IS 3 and IS4. For convenience, but without loss of generality, the functions have been drawn so that under an interest rate policy represented by LAf2 both pairs of IS functions produce the same range of incomes. To keep the diagram from becoming too messy, only one LM function, LM U under a money stock policy has been drawn. Now consider disturbances that would shift LM X back and forth. From Figure 6 it is easy to see that if shifts in LM Xwould lead to income fluctuations greater than from 5V to 5V—which fluctua tions would occur under an interest rate policy — then an interest policy would be preferred regardless of whether we have the pair ISX and IS2>or the pair IS 3 and /S4. The importance of the slope of the LM function is investigated in Figure 7 for the two LM pairs, LM Y and LM 2, and LM S and LM 4. The functions have been drawn so that each pair represents different slopes but an identical range of disturbances. It is clear that if shifts in ISX are small enough, then an interest rate policy will be preferred regardless of which pair of LM functions prevails. Conversely, if a money stock policy is preferred under one pair of LM functions because of the shifts in the IS function, then a money stock policy will also be preferred under the other pair of LM functions. The upshot of this analysis is that the cru cial issue for deciding upon whether an inter est rate or a money stock policy should be followed is the relative size of the disturbances in the expenditure and monetary sectors. Con trary to much recent discussion, the issue is not whether the interest elasticity of the de mand for money is relatively low or whether fiscal policy is more or less “powerful” than monetary policy. To avoid possible confusion, it should be emphasized that the above conclusion is in terms of the choice between a money stock policy and an interest rate policy. However, if a money stock policy is superior, then the steeper the LM function is, the lower the range of income fluctuation, as can be seen from Figure 7. It is also clear from Figure 6 that under an interest rate policy an error in setting the interest rate will lead to a larger error in hitting the income target if the IS func tion is relatively flat than if it is relatively steep. But these facts do not affect the choice between interest rate and money stock policies. The “combination” monetary policy. Up to this point the analysis has concentrated on the choice of either the interest rate or the money stock as the policy variable. But it is also pos sible to consider a “combination” policy that works through the money stock and the inter est rate simultaneously. An understanding of the combination policy may be obtained by further consideration of the cases depicted in Figures 2 and 7. In Figure 8 the disturbances, as in Figure 2, are entirely in the expenditure sector. As was seen in Figure 2, the result obtained by fixing the money stock so that LM Xprevailed was su perior to that obtained by fixing the interest rate so that LM 2 prevailed. But now suppose that instead of fixing the money stock, the money stock were reduced every time the in terest rate went up and increased every time the interest rate went down. This procedure would, of course, increase the amplitude of interest rate fluctuations.5 But if the proper re lationship between the money stock and the interest rate could be discovered, then the LM function could be made to look like LM 0 in Figure 8. The result would be that income would be pegged at Y ft Disturbances in the IS function would produce changes in the interest rate, which in turn would produce spending changes sufficient to completely offset the effect on income of the initial disturbance. The most complicated case of all to explain graphically is that in which it is desirable to increase the money stock as the interest rate rises and decrease it as the interest rate falls. 5 The increased fluctuations in interest rates must be carefully interpreted. In this model the IS func tion is assumed to fluctuate around a fixed-average position. However, in more complicated models in volving changes in the average position of the IS function, perhaps through the operation of the in vestment accelerator, interest rate fluctuations may not be increased by the policy being discussed in the text. By increasing the stability of income over a pe riod of time, the policy would increase the stability of the IS function in Figure 8 and thereby reduce in terest rate fluctuations. RULES-OF-THUMB FOR POLICY FIGURE 8 143 FIGURE 9 In Figure 9 the leftmost position of the LM function as a result of disturbances is LM t when the money stock is fixed and is LM 2 when the combination policy of introducing a positive money-interest relationship is followed. The rightmost positions of the LM functions under these conditions are not shown in the diagram. When the interest rate is pegged, the LM func tion is LM 3. If either LM r or LM 2 prevails, the intersection with IS 1 produces the lowest in come, which is below the Y / level obtained with LM3. But in the case of LM 2, income at Y x is only a little lower than at Y /, whereas when IS 2 prevails, LM 2 is better than LM 3 by the difference between Y2 and Y 2 . Since the gap between Y 2 and Y 2 is larger than that be tween Ya and Y /, it is on the average better to adopt LM 2 than LM Z even though the ex tremes under LM 2 are a bit larger than under LM3. Extensions of model. At this point a natural question is that of the extent to which the above analysis would hold in more complex models. Until more complicated models are constructed and analyzed mathematically, there is no way of being certain. But it is possible to make educated guesses on the effects of adding more goals and more policy instruments, and of relaxing the rigid price assumption. Additional goals may be added to the model if they are specified in terms of “closer is bet ter” rather than in terms of a fixed target that must be met. For example, it would not be mathematically difficult to add an interest rate goal to the model analyzed above, if deviations from a target interest rate were permitted but were treated as being increasingly harmful. On the other hand, it is clear that if there were a fixed-interest target, then the only possible pol icy would be to peg the interest rate, and in come stabilization would not be possible with monetary policy alone. The addition of fiscal policy instruments af fects the results in two major ways. First, the existence of income taxes and of government expenditures inversely related to income (for example, unemployment benefits) provides au tomatic stabilization. In terms of the model, automatic stabilizers make the IS function steeper than it otherwise would be, thus reduc ing the impact of monetary disturbances, and reduce the variance of expenditures disturb ances in the reduced-form equation for income. This effect would be shown in Figure 6 by drawing ISt so that it cuts LM 2 to the right of y / and drawing IS 2 so that it cuts LM %to the left of y 2'. The second major impact of adding fiscal policy instruments occurs if both income and the interest rate are goals. Horizontal shifts in the IS function that are induced by fiscal pol icy adjustments, when accompanied by a coor dinated monetary policy, make it possible to come closer to a desired interest rate without any sacrifice in income stability. An obvious illustration is provided by the case in which the optimal monetary policy from the point of view of stabilizing income is to set the interest rate as in Figure 5. Fiscal policy can then shift the pair of IS functions, ISt and IS2, to the right or left so that the expected value of in come is at the full employment level. If the interest rate is not a goal variable, then fiscal policy actions that shift the IS func tion without changing its slope do not improve income stabilization over what can be accom plished with monetary policy alone, provided the lags in the effects of monetary policy are no longer than those in the effects of fiscal policy. An exception would be a situation in which reaching full employment with monetary policy alone would require an unattainable in terest rate, such as a negative one. These comments on fiscal policy have been presented in order to clarify the relationship between fiscal and monetary policy. While monetary policy-makers may urge fiscal action, for the most part monetary policy must take the fiscal setting as given and adapt monetary policy to this setting. It must then be recog nized that an interest rate goal can be pursued only at the cost of sacrificing somewhat the in come goal.6 6 An interest rate goal must be sharply distin guished from the use of the interest rate as a mone tary policy instrument. By a goal variable is meant a variable that enters the policy utility function. Income and interest rate goals might be simultane ously pursued by setting the money stock as the pol icy instrument or by setting the interest rate as the policy instrument. All of the analysis so far has taken place within a model in which the price level is fixed in the short run. This assumption may be re laxed by recognizing that increases in money income above the full employment level in volve a mixture of real income gains and price inflation. Similarly, reductions in money in come below the full employment level involve real income reductions and price deflation (or a slower rate of price inflation). The model used above can be reinterpreted entirely in terms of money income so that departures from what was called above the “full employ ment” level of income involve a mixture of real income and price changes. Stabilizing money income, then, involves a mixture of the two goals of stabilizing real output and of sta bilizing the price level. However, interpreted in this way the struc ture of the model is deficient because it fails to distinguish between real and nominal interest rates. Price level increases generate inflation ary expectations, which in turn generate an outward shift in the IS function. The model may be patched up to some extent by assum ing that price changes make up a constant fraction of the deviation of income from its full employment level and assuming further that the expected rate of inflation is a con stant multiplied by the actual rate of inflation. Expenditures are then made to depend on the real rate of interest, the difference between the nominal rate of interest and the expected rate of inflation. The result is to make the IS func tion, when drawn against the nominal interest rate, flatter and to increase the variance of dis turbances to the IS function. These effects are more pronounced: (a) the larger is the inter est sensitivity of expenditures; (b) the larger is the fraction of price changes in money income changes; and (c) the larger is the effect of price changes on price expectations. The con clusion is that since price flexibility in effect increases the variance of disturbances in the IS function, a money stock policy tends to be favored over an interest rate policy. RULES-OF-THUMB FOR POLICY II. EVIDENCE ON THE RELATIVE MAGNITUDES OF REAL AND MONETARY DISTURBANCES NATURE OF AVAILABLE EVIDENCE. Little evidence is available that directly tests the relative stability of the expenditure and money demand functions. It is necessary, therefore, to proceed somewhat indirectly. First, simulation of the FR-MIT m odel7 is used to show the probable size of the effect on gross national product (GNP), the GNP deflator, and the unemployment rate of an assumed expenditure disturbance. This evidence provides some indication of the ex tent to which the impact of an expenditure dis turbance depends on the choice between the money stock and the Treasury bill rate as monetary policy control variables. This evi dence bears only on the question of what hap pens if an expenditure disturbance occurs, not on the relative stability of the expenditure and money demand functions. However, this ap proach is useful when combined with intuitive feelings about relative stability. The second type of evidence, derived from reduced-form studies, is more direcdy related to the question of relative stability; neverthe less, it is not entirely satisfactory because the studies examined were not designed to answer the question at hand. To supplement these studies by other investigators, there follows a simple test of the stability of the demand for money function. IMPACT OF AN EXPENDITURE DIS TURBANCE. Simulation of the FR-M IT model provides some insight as to how the size of the impact of an expenditure disturbance depends on the choice of the monetary policy instrument. The simulation technique is neces sary because the FR-M IT model is nonlinear, making it impossible to obtain an explicit ex pression for the reduced form.8 However, com 7 For a general description of the model, see [14]. 8 See opposite column. parison of two sets of simulations provides some interesting results. Except as indicated below, the simulations all used the actual his torical values of the model’s exogenous varia bles and all simulations started with 1962-1, a starting date selected arbitrarily. The first set of five simulations assumes an exogenous money stock that grows by 1 per cent per quarter, starting with the actual money stock in 1961-IV as the base. To inves tigate the impact of a disturbance in an exog enous expenditures variable, the exogenous variable “Federal expenditures on defense goods” was set in one simulation at its actual level minus $10 billion; in another at actual minus $5 billion; and in three further simula tions at actual, actual plus $5 billion, and ac tual plus $10 billion. This procedure produces four hypothetical observations on “disturb ances” in defense expenditures, of —10, “ 5, + 5, and + 10, and the simulation provides four corresponding observations for the change in income (and other endogenous variables). By using income as an example, the change in an endogenous variable in response to a dis turbance in defense expenditures is the differ ence between income simulated by the model when defense expenditures were set at actual historical values and when set at actual plus 10, plus 5, and so forth. The income obtained in the simulations, even when defense expendi tures are set at actual levels, is not the same as the actual historical level of income both be cause the assumed monetary policy differs from the policy actually followed and because of errors in the model itself. By calculating the ratio of the change in an endogenous variable to the disturbance in de fense expenditures for the four observations, four estimates of the linear approximation to the reduced-form parameter, or multiplier, of 8 In a reduced-form equation, an endogenous (that is, simultaneously determined) variable is expressed as depending only on exogenous and predetermined variables (variables taken as given for the current period). defense expenditures are obtained, and these four estimates have been averaged to produce a single estimate. Since the effects of a disturb ance accumulate over time, the reduced-form parameter estimate has been calculated for the 12 quarters from 1962-1 through 1964-IV. Exactly the same procedure has been used for the simulations with a fixed rate for 3-month Treasury bills. Finally, the ratio of the param eter estimates for the reduced forms under the money stock and interest rate policies has been calculated with the parameter estimates from the simulations with the exogenous money stock in the numerator of the ratio. The reduced-form parameter estimates under the two monetary policies, and the ratios of these estimates, have been plotted in Figure 10 for 12 quarters for the reduced forms for nom inal GNP, for the unemployment rate, and for the GNP deflator. The results are striking. A substantial difference appears in the parame ters of reduced forms for the fourth quarter following the initial disturbance, and the dif ferences in the parameters become steady thereafter. By the 12th quarter the reducedform parameters for the money stock policy are only about 40 per cent of those for the interest rate policy. The interpretation of these results is that employment, output, and the price level are far more sensitive to disturbances in defense ex penditures under an interest rate policy than under a money stock policy. This conclusion presumably generalizes to expenditures varia bles other than defense expenditures, but the results would differ in detail because each ex penditures variable enters the F R -M IT model in a somewhat different way. It might be argued that these results suggest that there is no significant difference between interest rate and money stock policies because the reduced-form parameters are essentially identical up to about four quarters. Surely, so this argument goes, mistakes could be discov ered and offset within four quarters. There are two difficulties with this argument. The first is that the F R -M IT model may overstate the FIGURE 10 REDUCED-FORM PARAMETER ESTIMATES FOR FEDERAL DEFENSE EXPENDITURES FROM FR-MIT MODEL Per c e n t P er c e n t Q U ARTE RS FO LLO W IN G 1961 Q4 length of the lags and therefore understate the differences in reduced-form parameters for the two policies for the quarters immediately fol lowing a disturbance. But the second and more important reason is that it may not be easy to reverse the effects of the disturbance after the disturbance has been discovered. With an in terest rate policy, a very large change in the rate might be required to offset the effects ap pearing after the fourth quarter, and such a change might not be feasible, or at least not desirable in terms of its effects on security markets and on income in the more distant fu ture. The numerical results reported above de pend, of course, on the F R -M IT model, and this model is deficient in a number of respects. But any model in which, other things being equal, investment and other interest-sensitive expenditures decline when interest rates rise will show results in the same direction. RULES-OF-THUMB FOR POLICY These results may be extended to analyze the significance of errors in forecasting exogenous variables. Consider an explicit expression for the reduced form for income. Let the exoge nous variables such as government expendi tures, perhaps certain categories of investment, strikes, weather, population growth, and so forth, be X l9 X 2, . . . , X n, and let the coeffi cients of these variables be ora, when the interest rate is the policy instrument, and Ai, A2, . . . , An when the money stock is the instrument. Then the reduced form for income when the interest rate is the instrument is (1) y = <*0 + OL\Xi + <22^2 + . . . + anXn + ocrr + u where <xr is the coefficient of the interest rate and u is the random disturbance. On the other hand, when the money stock is the instrument, the reduced form is (2) Y = Xo + XiA’i + \ 2X 2 + . . . + \ nXn + AjvfM + v As discussed in Section II, the disturbance vt may have either a larger or a smaller vari ance than the disturbance ut. One factor tend ing to make v* smaller than ut is that a money stock policy reduces the impact of expenditures disturbances, but another factor, the introduc tion into the reduced form of money demand disturbances, tends to make vf larger. The net result of these two factors cannot be deter mined a priori. But in formulating policy it is not possible to reason directly from equations 1 and 2 because many of the X { cannot be predicted in advance with perfect accuracy. For scientific purposes ex post it may be possible to say that a change in income was caused by a change in some X u for policy purposes ex ante this scientific knowledge is useless unless the change in X { can be predicted. It is necessary to think of each X i as being composed of a predictable part, X if and an unpredictable part, E Xi = Xi + Ei For policy purposes the error term in the reduced form includes both the disturbances to the equation and the errors in forecasting ex ogenous variables. The two types of errors ought to be treated exactly alike in formulating policy. Equations 1 and 2 can then be rewrit ten as follows: (3) Y = Oto + Ctjtl + CX2X 2 + . . . + QtnXn + aTr + aiEi + a 2E 2 + . . . + anEn + u (4) Y — Xo + \\X i + X2Z 2 + . . . + \nXN + \m M + XlEl + \ 2E 2 + . . . + Xn£n + V For policy purposes the error term in the reduced-form equation 3 is the sum of the terms from atE lt through ut and in the reduced-form equation 4 the sum of the term E lt through v(. A systematic study of the importance of the Ei terms cannot be made because no formal record of errors in forecasting exogenous vari ables exists insofar as the author knows. How ever, some insight into the problem may be obtained by listing the variables that must be forecast. Which variables have to be forecast depends, of course, on the model being used. The larger econometric models generally have relatively few exogenous variables that raise forecasting problems because so many varia bles are explained endogeneously by the model itself. The FR-M IT model has 63 exogenous variables; some of these are relatively easy to forecast, but others are subject to considerable forecasting error. The latter include such vari ables as exports, number of mandays idle due to strikes, Armed Forces, and Federal expend itures. Furthermore, this model involves lagged endogenous variables in many equations; hence an inaccurate forecast of GNP next quarter will increase the error in forecasting GNP two quarters into the future, which in turn will lead to errors in forecasting GNP three quar ters into the future, and so forth. Errors in forecasting exogenous variables, therefore, pro duce cumulative errors in forecasting GNP in future quarters. In simpler models the forecasting problem is 148 more severe. Consider, for example, the oppo site extreme from the large econometric model, the single-equation model. Convenient repre sentatives of such models are those spawned in the controversy over the Friedman-Meiselman paper [2] on the stability of the money/income relationship. The various definitions of exoge nous, or “autonomous,” spending utilized by the various authors in this controversy are as follows: a) Friedman-Meiselman definition: Au tonomous expenditures consist of the “net private domestic investment plus the government deficit on income and product account plus the net foreign balance” [2, p. 184]. b) Ando-Modigliani definition: Autono mous expenditures consist of two var iables which enter the reduced form with different coefficients. One varia ble is “property tax portion of indi rect business taxes” plus “net interest paid by government” plus “govern ment transfer payment” minus “un employment insurance benefits” plus “subsidies less current surplus of gov ernment enterprises” minus “statistical discrepancy” minus “excess of wage accruals over disbursement.” The sec ond variable is “net investment in plant and equipment, and in residen tial houses” plus “exports” [10, pp. 695, 696, and 702]. c) DePrano-Mayer definition: The basic definition is “investment in producers’ durable equipment, nonresidential construction, residential construction, federal government expenditures on income and product account, and ex ports. One variant of this hypothesis subtracts capital consumption esti mates, and the other does not” [15, p. 739]. DePrano and Mayer also tested 18 other definitions of autono mous expenditures [15, pp. 739 and 740]. d) Hester definition: Autonomous ex penditures consist of the “sum of gov ernment expenditure, net private do mestic investment, and the trade balance” [19, p. 366]. Hester also ex perimented with three other defini tions involving alternative treatments of imports, capital consumption al lowances, and inventory investment [19, pp. 366, 367]. To a considerable extent the diversity in these definitions is misleading because except for the Friedman-Meiselman definition all the definitions are in fact rather similar. But whichever definition is used, it is impossible to escape the feeling that inaccurate forecasting of exogenous variables is likely to be a major source of uncertainty. And while this discus sion has taken place within the context of for mal models, exactly the same problem plagues judgmental forecasting. Every forecasting method can be viewed as starting from fore casts of “input,” or exogenous, variables and then proceeding to judge the implications of these inputs for GNP and other dependent, or endogenous, variables. Regardless of what type of model is used, it appears that for the foreseeable future it will be necessary to forecast exogenous variables that simply cannot be forecast accurately by using present methods. As a result, it seems very likely that the error term including fore cast errors has a far smaller variance in equa tion 4 than in equation 3. Indeed, it might be argued that as a source of uncertainty the E { terms are far more important than the u or v terms, and therefore that the smaller size of the Xi parameters as compared to the «i pa rameters is of great importance. If the parame ter estimates from the FR-M IT model are ac cepted, the standard deviation of the total random term relevant for policy (that is, in cluding errors in forecasting exogenous varia bles) would be over twice as large under an interest rate policy as under a money stock policy. If this argument is correct, shifting from the current policy of emphasizing interest rates to one of controlling the money stock might cut average errors in half, where errors are measured in terms of the deviations of em ployment, output, and price level from target levels for these variables. EVIDENCE FROM REDUCED-FORM EQUATIONS. Additional insight into the RULES-OF-THUMB FOR POLICY relative sizes of disturbances under interest rate and money stock policies may be obtained by examining the controversy generated by the Friedman-Meiselman paper on the stability of the money/income relationship [2]. In this paper equations almost the same as equations 1 and 2 above were estimated. The equation corresponding to equation 1 differs in that the exogenous variables were assumed to consist only of a single autonomous spending variable, as defined above. The equation corresponding to equation 2 has the same disability for our purposes, but it also did not include an interest rate as a variable. Before examining the implications of the Friedman-Meiselman findings for this study, it should be noted that their approach was sharply criticized in papers by Donald D. Hes ter [19], Albert Ando and Franco Modigliani [10], and Michael DePrano and Thomas Mayer [15], These critics particularly attacked the Friedman-Meiselman definition of autono mous expenditures, and proposed and tested the alternative definitions listed above. How ever, they also attacked the single-equation ap proach and recommended the use of large models instead. The tests of alternative equations must be regarded as inconclusive in terms of which variable—the money stock or autonomous spending—is more closely related to the level of income*9 Both approaches achieve values for R 2 of 0.98 or 0.99 so that the unexplained variance is very small in both cases. It seems very unlikely that the addition of an interest rate variable to the equations by using autono mous expenditures as the explanatory variable, which addition would make the equations cor respond to equation 1 above, would make any substantial difference. 9 For reasons that need not be explained here, most of this controversy was conducted in terms of equations with consumption rather than GNP as the dependent variable. In the Friedman-Meiselman study* however, results are reported for equations with GNP [2, p. 227]. Such results are also reported in [9, p. 17]. From this evidence it appears that ex post explanations of the level of income are about as accurate by using autonomous expenditures alone as are those by using money stock alone. But given the inaccuracies in forecasting au tonomous expenditures, it must be concluded that ex ante explanations by using the money stock are substantially more accurate than those with forecasts of autonomous expendi-' tures. From this evidence, the total random term in equation 4 appears to have a substan tially smaller variance than the total random term in equation 3. For the reasons mentioned by the Fried man-Meiselman critics, evidence from single equation studies cannot be considered definitive. But neither can the evidence be ignored, espe cially in light of the difficulties encountered in the construction and the use of large econometric models such as the FR-M IT model. EVIDENCE ON STABILITY OF DE MAND FOR MONEY FUNCTION. One of the shortcomings of the single-equation studies discussed above is that their authors paid too little attention to the stability of regression coefficients over time. Consider the following statement by Friedman and Meiselman: The income velocity of circulation of money is consistently and decidedly sta bler than the investment multiplier except only during the early years of the Great Depression after 1929. There is through out, including those years, a close and consistent relation between the stock of money and consumption or income, and between year-to-year changes in the stock of money and in consumption or income [2, p. 186]. This conclusion is based on correlation coeffi cients between money and income (or con sumption), but what is relevant for policy is the regression coefficient, which determines how much income will change for a given change in the money stock. In the FriedmanMeiselman study, a table [2, p. 227] reports the regression coefficient for income on money as being 1.469 for annual data 1897-1958. 150 However, the same table reports regression coefficients for 12 subperiods, some of which are overlapping, ranging from 1.092 to 2.399. With a few exceptions, most economists agree that velocity changes can be explained in part by interest rate changes.10 Thus, variabil ity in the regression coefficients when income is regressed on money is not evidence of the instability of the demand for money function. To obtain some evidence on the stability of this function, the following simple procedure was used. Quarterly data were collected on the money stock, GNP, and Aaa corporate bond yields for 1947 through 1968. A demand for money function was fitted by regressing the log of the interest rate on the log of velocity, and vice versa. The regressions were run for the four periods, 1947 through 1960, 1947 through 1962, 1947 through 1964, and 1947 through 1966. The results inside each estima tion period were then compared with the re sults outside the estimation period. The results of this process for the 1947-60 estimation period are shown in Figure 11. The observations for 1947 through 1960 are repre sented by dots, and the observations for 1961 through 1968 by X ’s. The two least-squares regressions— log interest rate on log velocity and vice versa— fitted for the 1947-60 period have been drawn. From Figure 11 it appears that the relationship since 1960 has been quite similar to the one prior to 1960. Table 1 presents the results of applying a standard statistical test to the regression and postregression periods to determine whether the demand for money function was stable. To understand this table, refer first to section A of the table, and to the 1947-60 estimation pe riod. Section A reports results from regressing the log of velocity on the log of the Aaa cor porate bond rate, and the first row refers to the regression for 1947 through 1960. The square of the regression’s standard error of es timate is 0.00517 with 54 degrees of freedom. There were 32 quarters in the postregression 10 F o r a convenient review of evidence on this sub ject, see [4], FIG U RE 11 VELOCITY AND INTEREST RATE REGRESSIONS Regressions Fitted to Quarterly Data, 1 9 4 7 -6 0 Log R R 6.69 1.9 ! A. 1 6.05 1.8 B- 1 * =i 5.47 1.7 4.95 1.6 . 4.48 1.5 * 4.06 1.4 i 3.67 1.3 r 3.32 1.2 5 3.03 1.1 2.72 1.0 t Observation in 1961 -6 8 period / / , . * *v/ / , / * 9* 2.46 .7 2.01 2.23 .9 1.0 1.1 1.2 Log V 1.3 1.4 1.5 2.46 2.72 3.03 3.67 4.06 4.48 4.95 3.32 1.6 V period 1961 through 1968, and for this period the mean-square error of velocity from the ve locity predicted by the regression is 0.00836. The ratio of the mean-square errors from re gression outside to those inside the estimation period is given in the column labeled “F.” Since the ratio of two mean squares has the F distri bution under the hypothesis that both mean squares were produced by the same process, an F test may be used to test whether the deTABLE 1: Tests of the Stability of the Demand for M oney Function by Using Quarterly Data A. Log velocity regressed on log Aaa corporate bond yield Estim ation period Regression (SEE)* 1 9 4 7 - 6 0 ..........00517 1 9 4 7 - 6 2 .......... 0 0 48 4 1 9 4 7 -6 4 ..........00509 1 9 4 7 -6 6 ..........00 50 2 Postregression d.f. M SE d.f. 54 62 70 78 .0 0 8 3 6 .0 0 7 4 6 .00 587 .0 0 9 8 6 32 24 16 8 Significance level 1 .6 2 1 .5 4 1 .1 5 1 .9 6 .10 . 10 > .2 5 . 10 B. Log Aaa corporate bond yield regressed on log velocity Estim ation period Regression (SEE)2 1 9 4 7 -6 0 .......... 00 6 8 4 1 9 4 7 - 6 2 .......... 0 0 61 4 1 9 4 7 - 6 4 .......... 0 0 57 0 1 9 4 7 - 6 6 ..........0053 7 ‘ M SE < (SEE)1. Postregression d.f. M SE 54 62 70 78 .0 0 5 S9 .00 723 .0 1 162 .0 2 1 9 2 32 24 16 Significance level 1 .1 6 * 1 .1 8 2.0 4 4.0 8 > .2 5 > .2 5 .02 5 .00 5 RULES-OF-THUMB FOR POLICY mand for money function has been stable. If the function has been stable, then errors from regression outside the period of estimation should be, on the average, the same size as the errors inside the period of estimation. For the 1947-60 regression being discussed, F = 1.62 and is significant at the 10 per cent level but not at the 5 per cent level. Looking at Table 1 as a whole it can be seen that, for three of the regressions, the errors out side the period of estimation are not statistically significantly larger than those inside the period of estimation. Indeed, for the bond rate regres sion for the 1947-60 period, the errors outside the period of estimation were actually smaller, on the average, than those inside the period of estimation. Over-all, however, these results taken at face value cast some doubt on the stability of the demand for money function. However, there is reason to believe that there are problems in applying the F test in this situation. The reason is that the residuals from regression exhibit a very high positive se rial correlation as indicated by Durbin-Watson test statistics of around 0.15 for all of the re gressions. What this means is that the effective number of degrees of freedom is actually less than indicated in the table, and with fewer de grees of freedom the F ratios computed have less statistical significance than the significance levels reported in the table. The only way around this problem is to run a more complex regression that removes the serial correlation of the residuals, but there is no general agree ment among economists as to exactly what variables belong in such a regression. The vir tue of the simple regressions of velocity on an interest rate and vice versa is that this form has been used successfully by many investiga tors starting in 1954 [22]. The appropriate conclusion to be drawn from this evidence would seem to be that the relationship between velocity and the Aaa cor porate bond rate is too close and too stable to be ignored, but not close enough and stable enough to eliminate all doubts. However, the question is not whether an ironclad case for a money stock policy exists but rather whether the evidence taken as a whole argues for the adoption of such a policy. While there is cer tainly room for differing interpretations of Fig ure 11 and Table 1, and of the other evidence examined above, on the whole all of these re sults seem to point in the same direction. It appears that the money stock rather than in terest rates should be used as the monetary policy control variable. III. A M ONETARY RU LE FO R GUIDING POLICY RATIONALE FOR A RULE-OF-THUMB. The purpose of this section is to develop a rule-of-thumb to guide policy. Such a rule— not meant to be followed slavishly—would incorporate advice in as systematic a way as possible. The rule proposed here is based upon the theory and evidence in Sections II and III and upon a close examination of post-accord experience. Individual policy-makers inevitably use in formal rules-of-thumb in making decisions. Like everyone else, policy-makers develop cer tain standard ways of reacting to standard situ ations. These standard reactions are not, of course, unchanging over time, but are adjusted and developed according to experience and new theoretical ideas. If there were no stand ard reactions to standard situations, behavior would have to be regarded as completely ran dom and unpredictable. The word “capricious” is often, and not unfairly, used to describe such unpredictable behavior. There are several difficulties with relying on unspecified rules-of-thumb. For one thing, the rules may simply be wrong. But an even more important factor, because formally specified rules may also be wrong, is that the use of un specified rules allows little opportunity for cumulative improvements over time. A policy maker may have an extremely good operating rule in his head and excellent intuition as to the application of the rule but unless this rule can be written down there is little chance that it can be passed on to subsequent generations of policy-makers. An explicit operating rule provides a way of incorporating the lessons of the past into cur rent policy. For example, it is generally felt that monetary policy was too expansive follow ing the imposition of the tax surcharge in 1968. Unless the lesson of this experience is incorporated into an operating rule, it may not be remembered in 1975 or 1980. How many people now remember the overly tight policy in late 1959 and early 1960 that was a result of miscalculating the effects of the long steel strike in 1959? Since the FOMC membership changes over time, many of the current mem bers will not have learned firsthand the lesson from a policy mistake or a policy success 10 years ago. If the FOMC member is not an economist, he may not even be aware of the 10-year-old lesson. It is for these reasons that an attempt is made in this section to develop a practical pol icy rule that incorporates the lessons from past experience. The rule is not offered as one to be followed to the last decimal place or as one that is good for all time. Rather, it is offered as a guide—or as a benchmark— against which current policy may be judged. A rule may take the form of a formal model that specifies what actions should be taken to achieve the goals decided upon by the policy makers. Such a model would provide forecasts of goal variables, such as GNP, conditional on the policy actions taken. The structure of the model and the estimates of its parameters would, of course, be derived from past data and in that sense the model would incorporate the lessons of the past. But in spite of advances in modelbuilding and forecasting, it is clear that forecasts are still quite inaccurate on the average. In a study of the accuracy of forecasts by several hundred forecasters between 1953 and 1963, Zarnowitz concluded that the mean absolute forecast error was about 40 per cent of the average year-to-year change in GNP [26, p. 4]. He also reported, “there is no evidence that fore casters’ performance improved steadily over the period covered by the data” [26, p. 5], Not only are forecasts several quarters ahead inaccurate but also there is considerable uncertainty at, and after, the occurrence of business-cycle turning points as to whether a turning point has actually occurred. In a study of FOMC recognition of turning points for the period 1947-60, Hinshaw concluded that [1, p. 122]: The beginning data of the Committee’s recognition pattern varied from one to nine months before the cyclical turn. . . . On the other hand, the ending of the rec ognition pattern varied from one to seven months after the turn. . . . With the ex ception of the 1948 peak, the Committee was certain of a turning point within six months after the NBER date of the turn. At the date of the turn, the estimated probability was generally below 50; it reached the vicinity of 50 about two months after the turn. This recognition record, which is as good as that in 10 widely circulated publications whose forecasts were also studied in [2], casts further doubt on the value of placing great reliance on the forecasts.11 Given the accuracy of forecasts at the cur rent state of knowledge,12 it seems likely that for some time to come forecasts will be used primarily to supplement a policy-decision-making process that consists largely of reactions to current developments. Only gradually will poli cy-makers place greater reliance on formal ^ 0r further analysis of forecasting accuracy, see 12 The accuracy of forecasts may now be better than in the periods examined in the studies cited above. But without a number of years of data there would be no way of knowing whether forecasts have improved, and so forecasts must in any case be as sumed to be subject to a wide margin of error at the present time. RULES-OF-THUMB FOR POLICY forecasting models.13 While a considerable amount of work is being done on such models, essentially no attention is being paid to careful specification of how policy should react to cur rent developments. While sophisticated models will no doubt in time be developed into highly useful policy tools, it appears that in the meantime relatively simple approaches may yield substantial improvements in policy. Given that knowledge accumulates rather slowly, it can be expected that carefully speci fied but simple methods will be successful be fore large-scale models will be. Careful specifi cation of policy responses to current develop ments is but a small step beyond intuitive policy responses to current developments. This step surely represents a logical evolution of the policy-formation process. POST-ACCORD MONETARY POLICY. That an operating guideline is needed can be seen from the experience since the Treasury-Federal Reserve accord. In order that this experience may be understood better, subperiods were defined in terms of “stable,” “easing,” or “firming” policy as determined from the minutes of the Federal Open Market Committee. The minutes used are those pub lished in the Annual Reports of the Board of Governors of the Federal Reserve System for 1950 to 1968. The definitions of “stable,” “easing,” and “firming” periods are necessarily subjective as are the determinations of dates when policy changed.14 The dating of policy 13 It may be objected that great reliance is already placed on forecasts, at least on judgmental forecasts. However, these forecasts typically involve a large ele ment of extrapolation of current developments. It seems fair to say that in most cases in which condi tions forecast a number of quarters ahead differ markedly from current conditions, policy has fol lowed the dictates of current conditions rather than of the forecasts. 14 The author was greatly assisted in these judg ments by Joan Walton of the Special Studies Section of the Board’s Division of Research and Statistics. Miss Walton, who is not an economist, carefully read the minutes of the entire period and in a large table recorded the principal items that seemed impor tant at each FOMC meeting. Having a noneconomist changes was based primarily on the FOMC minutes, although the dates of changes in the discount rate and in reserve requirements were used to supplement the minutes. “Stable” peri ods are those in which the policy directive was unchanged except for relatively minor wording changes. In some cases the directive was es sentially unchanged although the minutes re flected the belief that policy might have to be changed in the near future. While the Manager of the System Open Market Account might change policy somewhat as a result of such discussions, the unchanged directive was taken at face value in defining policy turning points. More difficult problems of interpretation were raised by such directives as “unchanged policy, but err on the side of ease,” or “resolve doubts on the side of ease.” Such statements were used to help in defining several periods during which policy was progressively eased (or tightened). For example, in one meeting the directive might call for easier policy, the next meeting might call for unchanged policy but with doubts to be resolved on the side of ease, and a third meeting might call for further ease. These three meetings would then be taken together as defining an “easing” period. However, unless accompanied by other FOMC meetings clearly calling for a policy change, statements such as those calling for an “un changed policy with doubts resolved on the side of ease” were interpreted as not calling for a policy change. Some important monthly economic time se ries for the post-accord period are plotted in Figure 12. The heavy vertical lines represent periods of “stable,” “easing,” and “firming” policy as indicated by “S,” “E,” and “F” at the bottom of the figure. Except for the unemploy ment rate, the average of each series for each policy period has been plotted as a horizontal line. read the minutes tempered the inevitable tendency for an economist to read either too much or too lit tle into the minutes. However, the final interpretation of the minutes rested with the author. 154 Two features of the post-accord experience are especially noteworthy. First, decisions to change policy have been taken about as close to the time when, in retrospect, policy changes were needed as could be expected in the light of existing knowledge.15 There have been mis15 F or additional views on the tim ing of Federal Reserve decisions, see [13] and [1]. takes in timing, but the over-all record is im pressive. The second major feature of this pe riod is that policy actions, as opposed to policy decisions, have been in the correct direction if policy actions are defined by either free reserves or interest rates, but not if policy actions are defined in terms of either the money stock or bank credit. FIGURE 12 POST-ACCORD MONETARY POLICY B illion s o f d ollars RULES-OF-THUMB FOR POLICY To examine the timing question in more de tail, a useful comparison is that between busi ness cycle turning points (as defined by the National Bureau of Economic Research) and decisions to change policy. The post-accord period begins at a time when the U.S. econ omy was beset by inflation stemming from the war in Korea. The dates of the principal changes in policy and of the business cycle peaks and troughs are listed in Table 2. The policy dates are those that define the beginning of the “stable,” “easing,” and “firming” peri ods indicated in Figure 12. The decision to ease policy was made prior to the business cycle peaks of July 1953 and May 1960. The decision in 1957 was made in 155 FIGURE 12 (CONCLUDED) POST-ACCORD MONETARY POLICY B illio n s o f d o llars 10 0 TREASURY BILL RATE 3 -m o n th ; 2 1 . . o Per cent ‘ 6 4 0 156 the fourth month following the cycle peak in July, but as can be seen from Figure 12, the unemployment rate had not risen very much through October. Given the amount of uncer tainty always present in interpreting business conditions, this lag must be considered to be well within the margin of error to be expected for stabilization policy. However, the easing policy decision in 1968 was clearly a mistake in retrospect but not in prospect given the ex pectations held by the majority of economists that the tax increase would significantly temper the economic boom. TABLE 2: Dates of Principal Monetary Policy Decisions and of Business Cycle Peaks and Troughs B u siness cycle Turning p oin t D ate Peak 1953, July Trough 1954, A ugust Peak 1957, July Trough 1958, April Peak 1960, M ay Trough 1961* February F O M C p olicy decisions P olicy Starting date A ccord Firm ing Stable E asing Stable Firm ing Stable Easing Stable F irm ing S table Easing Stable Firm ing Stable Firm ing Stable Firm ing Stable Firm ing Stable Firm ing Stable Firm ing Stable Easing Stable Firm ing Stable Easing Stable Firm ing Stable 1951— M ar. 1 -2 1952— Sept. 25 D ec. 8 1953— June II D ec. 15 1954— D e c . 11 1955— O ct. 4 1957— N o v . 12 1958— A pr. 15 July 29 1959— June 16 1960— M ar. 1 A ug. 16 1961— O ct. 24 N o v . 14 1962— June 19 July 10 D ec. 18 1963— Jan. 8 M ay 7 A u g. 20 1964— A u g. 18 1965— M ar. 2 D e c . 14 1966— Sept. 13 N ov. 1 1967— M ay 2 N o v . 27 1968— Apr. 30 July 16 A ug. 13 D ec. 17 1969— Apr. 29 Firming policy decisions were also generally well timed. Following the 1953-54 recession, decisions to firm policy in small steps were taken from December 1954 to September 1955, as unemployment declined to about 4 per cent of the labor force. During the recov ery period after the 1957-58 recession, firming decisions were taken from July 1958 to May 1959. There was also a series of firming deci sions taken from the end of 1961 to 1966. Es pecially noteworthy are those taken from De cember 1965 to August 1966, in response to the beginning of inflation associated with the escalation of military activity in Vietnam. The easing policy decisions taken in late 1966 and early 1967 were fully appropriate in light of the economic slack that developed in 1967. Even from the point of view of those who doubt the importance of fiscal policy, this rec ord of the timing of policy decisions in the post-accord period is remarkably good. The timing record does not suggest that much at tention was paid to forecasts, but this lack of attention was perhaps not unfortunate given the accuracy of forecasts during the period. From this point of view, the only real mistake was the easing decision taken in 1968. Of course, those who believe that a steady rate of growth of the money stock is better than any discretionary policy likely to be achieved in practice may read this record as supporting their thesis. But the post-accord record of the timing of policy decisions is certainly encour aging to those who believe that the lags in the effects of policy are short enough, and the ef fects predictable enough, to make discretionary monetary policy a powerful stabilization tool if only decisions can be made promptly. While the System’s performance in the tim ing of policy decisions has been commendable, the same cannot be said for the actions taken in response to the decisions. In the earlier dis cussion the purposely vague terms “easing,” “firming,” and “stable” were used to describe policy decisions. These terms were meant to convey the notions that policy-makers wanted, respectively, to accelerate, decelerate, or main tain the pace of economic advance. The ques tion that must now be examined is whether policy actions did in fact tend to accelerate, decelerate, or maintain the level of economic activity. Policy actions were in accord with policy decisions if these actions are measured by ei ther the 3-month Treasury bill rate or free re serves. The bill rate rose in “firming” periods, fell in “easing” periods, and tended to remain unchanged in “stable” periods. However, there RULES-OF-THUMB FOR POLICY was some tendency for the bill rate to rise in “stable” periods following “firming” periods, and to fall in “stable” periods following “eas ing” periods, a pattern not inconsistent with the interpretation of policy being offered in this study. Similar comments apply to free re serves. But the picture is quite different if policy actions are measured by the rate of growth of the money stock. Careful study of Figure 12 will make this point clear. The growth rate de clined in response to the “firming” policy deci sion in late 1952, and again in the “stable” pe riod in early 1953. This behavior was, of course, consistent with the “firming” decision. But the rate of growth declined further follow ing the “easing” decision in June 1953 and re mained low until the middle of 1954. The un employment rate rose rapidly from its low of 2.6 per cent at the cycle peak in July 1953 to 6.0 per cent in August 1954, the cycle trough; the money stock was at the same level in April 1954, 9 months following the cycle peak and 10 months following the decision to adopt an “easing” policy, as it had been at the peak. The same pattern that had appeared during the 1953-54 recession appeared again at the time of the 1957-58 recession. The rate of growth of the money stock declined in 1957 prior to the cycle peak. (The Treasury bill rate also rose substantially.) But after the de cision to adopt an “easing” policy in Novem ber 1957, the growth rate of the money stock declined further. From October 1957 to Janu ary 1958, the money stock fell at a 2.9 per cent annual rate; from the cycle peak in July to October it had fallen at a 1.5 per cent an nual rate. The rate of growth of the money stock in creased substantially in February 1958, and it remained at the higher level during the “sta ble” policy period April to July. There fol lowed a period of “firming” policy decisions from the end of July 1958 to May 1959; how ever, the average growth rate of the money stock during this period was virtually identical to the average in the preceding “stable” pe riod. But in the “stable” period from June 1959 to February 1960, the rate of growth of money, at —2.2 per cent, was much lower than in the preceding “firming” period. This rate of growth of money can hardly be consid ered appropriate in the light of the fact that except for one month the unemployment rate was continuously above 5 per cent. However, the picture was confused by a long steel strike. The decision to ease policy was taken on March 1, 1960, but the rate of growth of the money stock remained negative until July. The rate of growth of money fell following the “firming” policy decisions of October 1961 and June 1962. In spite of another firming de cision in December 1962 the rate of growth then increased, and it continued to rise during the “firming” period in 1963, maintaining the same rate in the following “stable” period. In August 1964, another “firming” decision was taken, and the growth rate trended down dur ing the “firming” period from August 1964 to February 1965. During the “stable” period from March to November 1965, the Vietnam war heated up. In the second half of 1965 the growth rate of money was 6.1 per cent compared with 3.0 per cent during the first half. The “firming” policy decision came in December, but the rate of growth of money averaged over 6 per cent for the months December through April 1966. At this point monetary growth ceased. In Jan uary 1967 the money stock was actually less than in May 1966—there having been no in crease in the growth rate in the months imme diately following the “easing” decision of November 1, 1966. The growth rate of money then accelerated during the “stable” period from May through October 1967; for the period as a whole growth averaged 8.7 per cent. In the following “firming” period November 1967 through April 1968, the rate of growth of the money stock was lower but it was still relatively high at 5.1 per cent. The growth rate then rose to 9.6 per cent in the “stable” period May through July 1968 and thereafter fell to a little less than 6 per cent in the July-November 1968 period following the “easing” decision of July 16, 1968. There ensued a “firming” period from De cember 1968 through April 1969. Although original figures indicated that monetary growth was relatively little during this period, a revi sion in the money stock series showed that the rate averaged 5.5 per cent for the period as a whole. The rate following April was lower, es pecially in the June-December 1969 period, which saw no net growth in the money stock. A broadly similar view of the timing of pol icy actions is obtained from a careful examina tion of the rate of growth of total bank credit. However, as shown in Figure 12, this series is quite erratic and much more difficult to inter pret than the series on the rate of growth of the money stock. The proper way to interpret these results would seem to be as follows. When interest rates fell in a recession, policy was easier than it would have been if interest rates had not been permitted to fall. But if the money stock was also falling, or growing at a below-average rate, policy was tighter than it would have been had money been growing at its long-run average rate. Similar statements apply to rising interest rates and above-average monetary growth in a boom. A MONETARY RULE. Given the argu ments of Sections I and II on the advan tages of controlling the money stock as op posed to interest rates, a logical first step in developing a policy guideline is to examine cases clearly calling for ease or restraint. Con sider first a recession. To insure that monetary policy is expansionary, the rule might be that interest rates should fall and the money stock should rise at an above-average rate. This pol icy avoids two possible errors. The first is illustrated in Figure 13. If the IS function shifts down from ISX to IS 2 while the LM function shifts from LM X to LM2, the in terest rate will fall from rx to r2. The shift from LMi to LMn could be caused by a shift in the demand for money with the stock of money unchanged. But this shift could also be caused by a decline in the stock of money, perhaps because of an attempt by policy-makers to keep the interest rate from falling too rapidly. However, in terms of income it is clearly bet ter to permit the interest rate to fall to r3 by maintaining the stock of money fixed, and bet ter yet to shift the LM function to the right of LM! by increasing the stock of money. The point is the simple one that monetary policy should not rely simply on a declining interest rate in recession but should also insure that the money stock is growing at an ade quate rate. The LM function may still shift to LMo in spite of monetaiy growth because of an increased demand for money; without the monetary growth, however, this shift in the de mand for money would push the LM function to the left of LM 2 and income would be even lower. The second type of error avoided by the proposed policy rule is illustrated in Figure 14. Again, it is assumed that the situation is one of recession. With a fixed money stock, an in crease in the demand for money will shift the LM function from LM t to LM2, tending to re duce income. However, if the interest rate is prevented from rising above rl9 the increased demand for money is met by an increased sup ply of money. Maintaining monetary growth and a declin ing interest rate in recession insures that the contribution of monetary policy is expansive. Increases in the demand for money, unless ac companied by a falling IS function, are fully offset by preventing increases in the interest rate. The greater the fall in the IS function the smaller the offset to an increased demand for money. However, in no case should a fall in the IS function be permitted to cause a fall in the money stock. The policy proposed does not, of course, guarantee an expansion of income. No such guarantee is possible because downward shifts in the IS function may exceed any specified RULES-OF-THUMB FOR POLICY shift in the LM function. But more important than theoretical possibilities are empirical probabilities. For all practical purposes the problem is not how to insure expansion in a recession but how to trade off the risks of too much expansion against too little. The discus sion of Figures 13 and 14 was entirely in terms of encouraging income expansion, or limiting further declines, in the face of de pressing disturbances. But disturbances may be expansionary in a recession, and such disturb ances may combine with expansionary policy to create overly rapid recovery from the reces sion. Consider again Figure 13, but suppose the initial position is as shown by IS 2 and LM2. If the interest rate is not permitted to rise, a shift to /Si will lead to a large increase in income to the level given by the intersection of ISX with a horizontal LM function drawn at r2. This situ ation can be avoided only if the interest rate is permitted to rise. The natural question is how the interest rate can be permitted to rise within a recession policy of pushing the interest rate down and maintaining above-average monetary growth. The answer is that the recession policy FIGURE 13 should be followed only if the interest rate can be kept from rising with a monetary growth rate below some upper bound. Exactly the same analysis running in reverse applies to a policy for checking an inflationary boom. In a boom interest rates should rise and monetary growth should be below average. However, there must be a lower limit on mon etary growth to avoid an unduly contraction ary policy. Having presented the basic ideas behind the formulation of a monetary rule, it is now necessary to become more specific about the rule. After specifying the rule in de tail, it will be possible to discuss the considera tions behind the specific numbers chosen. The proposed monetary policy rule-ofthumb is given in Table 3. The rule assumes that full employment exists when unemploy ment is in the 4.0 to 4.4 per cent range and that monetary growth in the 3 to 5 per cent range is consistent with price stability. At full employment the Treasury bill rate may rise or fall, either because of market pressures or be cause of small adjustments in monetary policy; however, monetary growth should remain in the 3 to 5 per cent range. 159 TABLE 3: Proposed Monetary Policy Rule-of-Thumb In per cent R u le fo r m o n th 1 U n em p loym en t rate previous m onth 0 - 3 . 4 ....................................... 3 . 5 - 3 . 9 ....................................... 4 . 0 - 4 . 4 ........................................ 4 . 5 - 4 . 9 ........................................ 5 . 0 - 5 . 4 ........................................ 5 . 5 - 5 . 9 ........................................ 6 . 0 - 1 0 0 . 0 ................................... D irection o f Treasury bill rate (3-m onth) G row th o f m oney stock (annual rate) R isin g R isin g R isin g or falling F alling Falling F alling Falling 2 1 -3 1 2 -4 3 -5 * 4 -6 * 5 -7 afr-8 6 -8 1 T he 3-m onth bill rate is to be adjusted in the indicated direction provided that m onetary grow th is in the indicated range. I f the bill rate change ca n n o t be achieved within the m onetary grow th rate guideline, then the bill rate guideline should be aband oned. s I f the bill rate the previous m on th w as b elow the bill rate 3 m on th s prior to that, then the upper and low er lim its on m onetary grow th are b oth increased by I per cent. * I f the bill rate the previous m onth was a b o v e the bill rate 3 m o n th s prior to that, then the upper and low er lim its on m onetary grow th are b oth reduced by 1 per cent. When unemployment drops below 4 per cent, the rule calls for a restrictive monetary policy. The bill rate should rise and monetary growth should be reduced. If the bill rate and monetary growth guidelines are not compati ble, then the monetary guideline should be binding. For example, suppose that unemploy ment is in the 3.5 to 3.9 per cent range. If mon etary growth below 2 per cent would be re quired to obtain a rising bill rate, then monetary growth should be 2 per cent and the bill rate be permitted to fall. If this situation persists so that the bill rate falls for several months in spite of the low monetary growth, then the limits on monetary growth should be increased as indicated in footnote 2 to Table 3. The reason for this prescription is that the bill rate on the average turns down 1 month before the peak of the business cycle [21, p. 111]. Unemployment, on the other hand, may increase relatively little in the early months following a cycle peak. Tying monetary growth to the bill rate in the way indicated in footnote 2 of Table 3 produces a more timely adjust ment of policy than relying on the unemploy ment rate alone. The proposed rule calls for a falling bill rate and a relatively higher rate of monetary growth as unemployment rises above the 4.0 to 4.4 per cent range. The rule for high un employment situations calls for adjusting the monetary growth rate downward when the bill rate is consistently rising as indicated by foot note 3 to Table 3. The reasoning behind this adjustment is exactly parallel to the reasoning above for low unemployment situations. The proposed monetary rule has the virtues of simplicity and dependence on relatively well-established economic doctrine. Because of its simplicity, the basic ideas behind the rule can be explained to the noneconomist. The simplicity of the rule also will make possible relatively easy evaluations of the rule’s per formance in the future if the rule is followed. With more complicated rules it would be much more difficult to know how to improve the rule in the future because it would be difficult to judge what part of the rule was unsatisfactory. Since, as has been repeatedly emphasized above, the rule is not proposed as being good for all time, it is best to start with a simple rule and then gradually to introduce more var iables into the rule as experience accumulates. In designing the rule, the attempt was made to base the rule on fairly well-established eco nomic knowledge. There is, of course, a great deal of debate as to just what is and what is not well established. What can be done, and must be done, is to explain as carefully as pos sible the assumptions upon which the rule is based, with full recognition that other econo mists may not accept these assumptions. First, the evidence for the importance of money is impressive. It seems fair to say that very few economists believe today that changes in the stock of money have nothing to do with business fluctuations. Rather, the argument is over the extent to which monetary factors are important. Some no doubt will feel that the 2percentage-point ranges on monetary growth specified by the rule are excessively narrow; however, it should be noted that a 4 per cent growth rate is double a 2 per cent growth rate. Also important is the fact that the rule is meant to serve as a guideline rather than be absolutely binding. Since policy should deviate RULES-OF-THUMB FOR POLICY from the rule if there is good and sufficient reason—such as wartime panic buying—a fur ther element of flexibility exists within the framework of the rule. The rule is specified in terms of changes in the bill rate and the monetary growth rate, with the monetary growth rate being tied to the unemployment rate and to changes in the bill rate in the recent past. This formulation has been designed to avoid what seem to be the most obvious errors of the past. Over the years the monetary growth rate has been low est at business cycle peaks and in the early stages of business contractions, and highest at cycle troughs and in the middle stages of busi ness expansions. The highest rate of monetary growth since the Treasury-Federal Reserve accord has been during the inflation associated with escalation of military operations in Viet nam. For purposes of smoothing the business cycle, so far as this author knows, there is no theory propounded by any economist that would call for high monetary growth during inflationary booms and low monetary growth during recessions. Such behavior of the money stock could only be optimal within a theory in which money had little or no effect on business fluctuations and in which other goals such as interest rate stability were important. Being based on the unemployment rate and bill rate changes in the recent past, the pro posed monetary rule does not rely on forecast ing. Nor does the rule depend on the current and projected stance of fiscal policy. Both of these factors ought to be included in applying the rule by adjusting the rate of growth of the money stock within the rule limits, or even by going outside the limits. But given the accu racy of economic forecasts under present methods, and given the current uncertainty over the size of the impact of fiscal policy (not to mention the hazards in forecasting Federal receipts and expenditures), it does not appear that these variables can be systematically in corporated into a rule at the current state of knowledge. TESTS OF THE PROPOSED RULE. Three types of evidence on the value of the rule are examined below. The first approach involves a simple comparison of the rule with the historical record to show that the rule would generally have been more expansionary (contractionary) than actual policy when ac tual policy—in the light of subsequent eco nomic developments—might be judged to have been too contractionary (expansionary). The second approach examines the cyclical behav ior of the estimated residuals from a simple demand for money function to show that it is unlikely that the proposed rule would interact with the disturbances to produce an exces sively inflationary or deflationary impact. Both these approaches are deficient because they rely heavily on the historical record, a record that would have been quite different had the rule been followed in the past. To avoid this difficulty, a third approach uses simulation of the FR-M IT model, but the results do not ap pear very useful because of shortcomings in this model. An impressionistic examination of the rule. Broadly speaking, the results of comparing the rule with the historical record since the Treas ury-Federal Reserve accord in March 1951 are these. The rule would have provided a substantially tighter monetary policy than the actual during the inflationary period from the accord until about September 1952. At that point, actual policy as measured both by the rate of growth of the money stock and by the 3-month bill rate became considerably tighter. In the last quarter of 1952, actual policy was in accord with the rule, but thereafter it tight ened even further. In the 9 months following the cyclical peak in July 1953, the money stock had a zero rate of growth while the un employment rate rose from 2.6 per cent to 5.9 per cent. Under the rule the rate of growth of the money stock would never have gone below 1 per cent and would have steadily increased as unemployment rose. Actual policy became more expansive in the second quarter of 1954, and the cycle trough was reached in August. However, the rule would have been considerably more expansive, and it would have remained more expansive than the actual all through the 1955-56 boom. Inasmuch as the unemployment rate remained near 4.0 per cent from May 1955 through Au gust 1957, the rule would have been too infla tionary during this period. However, it can be argued that monetary policy was overly restric tive before the cycle peak in July 1957, since in the year prior to the peak the money stock grew only by 0.7 per cent. Less subject to dis pute is the fact that policy was far too restric tive after the peak; in the 6 months following the peak the money stock fell at an annual rate of 2.2 per cent, and at the same time the unemployment rate rose from 4.2 per cent to 5.8 per cent. The rule would have been considerably more expansive all during the high unemploy ment period of 1958-59, and it would have prevented the declines in the money stock in late 1959 and early 1960. At the peak in May 1960 the unemployment rate was 5.1 per cent, and the money stock had fallen by 2.1 per cent in the previous 12 months. Unlike the pe riods following peaks in 1954 and 1957, pol icy became more expansive immediately after the May 1960 peak, although not so expansive as called for by the proposed rule. From the trough in February 1961 through June 1964, the unemployment rate never de clined below 5 per cent. Under the rule, policy would have been more expansive than the ac tual policy followed throughout this period, especially as compared with the MarchSeptember 1962 period, during which the money stock fell slightly. Unemployment fell rapidly in 1965 with the Vietnam build-up; the rule would have been more expansive than ac tual through July 1965 and then less expansive than actual through April 1966. Indeed, in the 9-month period prior to April 1966, with the unemployment rate falling from 4.4 per cent to 3.8 per cent, monetary growth accelerated to a 6.6 per cent annual rate; the proposed rule would have first called for monetary growth in the 3 to 5 per cent range, and then in the 2 to 4 per cent range starting in February 1966, fol lowing the drop in the unemployment rate below 4.0 per cent in January. Finally, the negative growth rates of money in the 1966 credit crunch would have been avoided under the rule, as would the high rates of growth in 1967 and 1968. This impressionistic look at the proposed rule may be supplemented by a simple scoring system for judging when the rule would have been in error. For each month during the sam ple period it was determined whether the rule would have been more or less expansive than the actual policy, or about the same as the ac tual policy. The unemployment rate 12 months from the month in question was used to indi cate whether or not the policy was correct, with a desired range of unemployment of 4.0 to 4.4 per cent. The rule was deemed to have made an error if: (1) the actual policy was in accord with the rule, but unemployment 12 months later was not in the desired range; (2) the rule called for a more expansive policy than the actual, and unemployment 12 months later was below the desired range; and (3) the rule called for a less expansive policy than the actual, and unemployment 12 months later was above the desired range. Since the latest data used in this analysis were for July 1969, comparison of the rule with actual policy ends July 1968. Starting the sam ple with 1952, the first full year after the accord, provides a total of 199 months. Based on the criterion described above, the rule would have been in error in 63 months. If the criterion is changed by substituting the unemployment rate 9 months ahead instead of 12 months ahead, the rule has 62 errors; using the unemployment rate 6 months ahead yields 59 errors. Some of these errors are of negligible im port. For example, in March 1953 the rule calls for a money growth rate of 2 to 4 per cent, but the actual was 1.9 per cent. Thus, RULES-OF-THUMB FOR POLICY the rule would have been more expansive than the actual this particular month, a mistake since unemployment was too low and inflation too high during this period. However, the rule would have been less expansive than actual in every one of the preceding 6 months and in all but one of the 6 months following this “mistake.” Except for scattered errors such as the one just discussed, most of the rule errors occurred in two separate periods. The first is the 2-year period following the cycle trough in August 1954, during which time the rule would have been too expansive. The second is the last half of 1964 and the first half of 1965, when the rule would have been too expansive in light of the subsequent sharp decline in un employment. Unless one has completed a careful exami nation of the data, there is a tendency to un derestimate how rapidly the economy can change. For example, from the cycle peak in July 1953 to the cycle trough 13 months later, the unemployment rate rose by 3.4 percentage points; and from the peak in July 1957 to the trough 9 months later in April 1958, it rose by 3.2 percentage points. Changes in the other direction have tended to be somewhat less rapid, but significant nonetheless. In the year following the trough in August 1954, the un employment rate declined 2.0 percentage points, and it declined 2.2 percentage points in the year following the trough in April 1958. In January 1965 unemployment was 4.8 per cent and the problem was still one of how to reach full employment. A year later the rate was 3.9 per cent and the problem was inflation. Thus, it appears that for the most part the rule would have been superior to policy actu ally followed. Of course, the rule is not in fallible and would have erred on a number of occasions. But in spite of these errors— and it should be recognized that some errors are inevitable no matter what rule or which discre tionary policy-makers are in charge—the pro posed rule has the great virtue of turning policy around promptly as imbalances develop. Relationship of the rule to monetary dis turbances. Since the rule was developed on the basis of the theoretical and empirical anal ysis of Sections I and II, which emphasized the relative stability of the demand for money, it is appropriate to conduct a systematic exam ination of the disturbances in the demand for money. It will be recalled that the rule was formulated in such a way as to insure expan sionary policy action in a recession and con tractionary policy action in a boom. However, it was recognized that disturbances in the ex penditure sector and/or in the monetary sector might reinforce policy actions leading to an ex cessively expansionary or contractionary effect on income. If there were a significant chance of these excessive effects occurring, then the rule proposed would be overly “aggressive” and a rule involving a smaller range of mone tary growth rates would be in order. To provide some evidence of the effect of disturbances in the money demand function, the residuals from the simple velocity function tested in Section II were examined carefully. The technique involved regressing velocity on the Aaa corporate bond rate, and vice versa, for the 1947-68 period and then comparing the residuals with turning points in the busi ness cycle. The reader may make these com parisons visually from Figure 15. At the bot tom of this figure cycle peaks and troughs are identified bv “P” and “T,” respectively. The residuals from the estimated equations suggest that the demand for money has con tractionary disturbances near business cycle peaks and expansionary disturbances near cycle troughs. The residuals have the same turning points for the regression of velocity on the interest rate as for the regression of the interest rate on velocity. The residual peaks occur at or before the cycle peaks, while the residual troughs occur at or after the cycle troughs. To assess the significance of these findings, consider the following simple view as to the dynamics of monetary effects. In the short run, be littie danger that the rule would be overly expansionary because after the cycle trough, while policy is still expansionary, contractive shifts in the demand for money occur. Simulations of the FR -M IT model. The final technique used to test the proposed mon etary rule was to simulate the FR -M IT model under the rule. As explained below, the results are of questionable value but are presented anyway for the sake of completeness and in order not to suppress results unfavorable to the proposed rule. To simplify the computer programming, the rule used in the simulations is not exactly the same as the one proposed in Table 3 above. The proposed rule, it will be recalled, involved a bill rate guideline and a money stock guide line. If, for example, the bill rate cannot be pushed up without pushing monetary growth income is a predetermined variable in the de mand for money function. An increase in the money stock makes the interest rate lower than it would be otherwise, and this eventually leads to expansion in investment and income. A downward disturbance in the demand for money function has the same effect. Given this view of monetary dynamics, Fig ure 15 suggests the following conclusions. Shifts in the demand for money tend to be contractive in their effect on income in the late stages of a business cycle expansion, implying that a restrictive monetary policy must not be pushed too hard. Then, shortly before the cycle peak, the shifts apparently tend to be come expansive. This effect is fortunate since it is only after the cycle peak that rising unem ployment would trigger a policy change under the proposed rule. However, there appears to RESIDUALS FROM VELOCITY REGRESSION COMPARED WITH BUSINESS-CYCLE TURNING POINTS Per cent rj LOG ■ VELO CITY \ RESIDUAL < Period r Regression: log V=.0 5 4 + .8 5 4 log R of e stim a tio n : 1 9 47 -6 8 ■ 2 :i R =.901 ; J RULES-OF-THUMB FOR POLICY FIG U R E 16 SIMULATIONS OF UNEMPLOYMENT IN FR-MIT MODEL Per c e n t ' ’58 '60 '62 ’64 ’66 4 ’68 1958 ’60 ’64 ’62 Per c e n t ’66 ’68 Per c e n t Pe r c e n t 4 Is 0 1960 A ’62 ’64 ’66 ’68 1962 '64 ’66 ’68 1964 ’66 ’68 A c tu a l S Rule sim ulation • Control sim ulation below the lower limit in the money guideline, the proposed rule calls for setting monetary growth at its lower limit. The simulation rule, on the other hand, ignores the bill rate guide line and simply sets the monetary growth rate at the midpoint of the range specified by the proposed rule. Another difference, and no doubt a more important one, between the proposed rule and the simulation rule is that the simulation rule had to be specified in terms of quarterly data since the F R -M IT model uses quarterly data. In the simulation rule, the growth of the money stock depends on the level of unem ployment determined by the model in the pre vious quarter. The growth rate of the money stock was modified by past changes in the bill rate, as in footnotes 2 and 3 to Table 3, ex cept that the relevant bill rate change was in terms of the previous quarter compared with the quarter before that. The simulation rule, then, reacts somewhat more slowly to unem ployment trends than does the proposed rule. In order to investigate the importance of the starting point, simulations were run with start ing dates in the first quarters of 1956, 1958, 1960, 1962, and 1964. The simulated unem ployment rate for the five simulations is shown in the five panels of Figure 16 by the curves marked “S.” The actual unemployment rate is shown by the curves marked “A,” and control simulations, to be explained below, by the un connected points. It is clear from Figure 16 that the simula tion rule for money growth produces an unsta ble unemployment rate. However, because of deficiencies in the model this result is probably not very meaningful. That the model is defec tive can be seen by comparing unemployment in the control simulations with the actual un employment. In the control simulations all of the model’s exogenous variables, including the money stock, were set at their actual levels.16 Even with the exogenous variables set at their actual levels, the simulated level of unemploy ment at times differs from the actual level. Because of the role of the stochastic disturb ances in the model, especially as they feed through lagged endogenous variables, it cannot be expected that control simulations will ex actly duplicate the actual results. But the fact that the control simulations differ from the ac tual by considerable margins over long periods of time strongly suggests that the money rule simulations do not provide much useful infor mation on the properties of the proposed rule. The simulations are valuable in one respect, however. An examination of Figure 16 strongly suggests that the money rule is inter acting with the rest of the model to produce a 16 The FR-M IT model was estimated with the money stock as an endogenous variable. There are separate equations for currency and demand deposits, both of which are endogenous, while unborrowed re serves are exogenous. In the simulations the money stock was made exogenous by suppressing the equa tion that makes demand deposits depend on unborrowed reserves. To simulate the effects of a particu lar rate of growth of money, the currency equation was retained, but demand deposits were set at what ever level was required to obtain the desired rate of growth of demand deposits plus currency. In the control simulations demand deposits were set at their actual levels, but currency remained an endogenous variable and differed somewhat from actual since simulated GNP differed somewhat from actual GNP. cycle of 5 to 6 years. Such a cycle is particu larly evident in the simulations starting in 1956 and 1958. That the monetary rule has very powerful effects in the model is shown by the simulations beginning in 1960 and 1962. In both simulations unemployment reaches a trough in 1964 and then rises in spite of the 1964-65 tax cuts and the stimulus of spending for military operations in Vietnam starting at the end of 1965. There is no doubt that the monetary rule is too aggressive within the context of the FR-M IT model. A simulation of a perfectly steady rate of growth of money is shown in Figure 17. The rate of growth in this simula tion is 2.76 per cent per year, the same as the actual rate of growth over the period 1955-IV through 1969-L In Figure 17, the curve la beled S2 is the simulated unemployment rate with the steady rate of growth of money. The simulated unemployment rate under the mone tary rule is shown by Sly which is the same as S in panel A of Figure 16. The unconnected points show the same control simulation as shown in panel A of Figure 16. It appears impossible to draw any firm con clusions from the simulations. However, the simulations clearly raise the possibility that the proposed monetary rule may produce ecoFIGURE 17 SIMULATIONS OF FR-MIT MODEL Per cent • Control simulation S2 Constant growth rate simulation Sj Rule simulation RULES-OF-THUMB FOR POLICY nomic instability. If anything, the proposed rule is too aggressive, and so policy should probably err on the side of producing growth rates in money closer to a steady 3 to 5 per cent rather than farther from the extremes in the proposed rule. IV. SELECTION AND CONTROL OF A MONETARY AGGREGATE BASIC ISSUES. Up to this point, the analysis has been entirely in terms of optimal control of the money stock. The theoretical analysis has been general enough that no pre cise definition of the money stock has been re quired. The empirical work, however, has used the narrow definition of demand deposits ad justed plus currency, for the simple reason that this definition seems to be the most appropriate one. In principle there is no reason not to look simultaneously at all of the aggregates and, of course, at all other information as well. But in practice, at the present state of knowledge, there simply is no way of knowing how all of these various measures ought to be combined.17 Furthermore, the selection of a single aggre gate for operating purposes would permit the FOMC to be far more precise in its policy de liberations and in its instructions to the Man ager of the Open Market Account. Thus, the best procedure would seem to be to select one aggregate as the policy control variable, and insofar as the state of knowledge permits, to in corporate other information into policy by making appropriate adjustments in the rate of growth of the aggregate selected. 17 This point is an especially important one since those favoring simple approaches are frequently cas tigated for ignoring relevant information, and for applying “simplistic solutions to inherently complex problems.” For this charge to be upheld, it must be shown explicitly and in detail how this other infor mation is to be used, and evidence must be produced to support the proposed complex approach. As far as this author knows, there is essentially no evidence sorting out the separate effects of various compo nents of monetary aggregates. In principle the aggregate singled out as the control variable should be subject to exact de termination by the Federal Reserve. The rea son is that errors in reaching an aggregate that cannot be precisely controlled may interact with disturbances in the relationships between the aggregate and goal variables such as GNP to produce a suboptimal policy. However, as argued later in this section, this consideration is likely to be quite unimportant in practice for any of the aggregates commonly considered. Therefore, the analysis of which aggregate should be singled out will be conducted under the assumption that all of the various aggre gates can be precisely controlled by the Fed eral Reserve. SELECTION OF A MONETARY AG GREGATE. At the outset it must be em phasized that the various aggregates frequently discussed are all highly correlated with one another in the postwar period. This is true for total bank credit, the narrow money stock, the broad money stock (narrow money stock plus time deposits), the bank credit proxy (total member bank deposits), the monetary base (member bank reserves plus currency held by the public and nonmember banks), and several other figures that can be computed. While these various aggregates are highly correlated over substantial periods of time, they showr significantly different trends for short periods. In selecting an aggregate, the most important considerations are the theoreti cal relevance of the aggregate and the extent to which the theoretical notions have been given empirical support. Both of these consid erations point to the selection of the narrowly defined money stock. The most important theoretical dispute is between those who emphasize the importance of bank deposit liabilities—the “monetary” view—and those who emphasize the impor tance of banks’ earning assets—the “credit” view. This controversy, which dates back well into the 19 th century, is difficult to resolve be cause historically banks have operated on a fractional reserve basis and so have had both earning assets and deposit liabilities. Since bal ance sheets must balance, bank credit and bank deposits are perfectly correlated except insofar as there are changes in nonearning as sets—such as reserves—or nondeposit liabil ities—such as borrowing from the Federal Reserve System. If these factors never changed, the perfect correlation between bank deposits and bank credit would make it impossible ever to obtain evidence to distinguish between the monetary and the credit views. Since the corre lation, while not perfect, has historically been very high, it has been very difficult to obtain evidence. Hence, it is still necessary to place major reliance on theoretical reasoning. There would be little reason to examine the issue closely if we could be confident that the very high correlation between deposits and bank credit would continue into the indefinite future. But there are already substantial differ ences in the short-run movements of bank credit and bank deposits, and these differences are likely to become greater and of a longerterm character in the future. Banks are raising increasingly large amounts of funds through nondeposit sources such as sales of commercial paper and of capital certificates and through borrowing from the Euro-dollar market and the Federal Reserve System. (Borrowings from the System would probably expand signifi cantly if proposed changes in discount-window administration were implemented.) The easiest way to examine the theoretical issues is to consider some hypothetical experi ments. Consider first the experiment in which the Federal Reserve raises reserve require ments by $10 billion at the initial level of deposits but simultaneously buys $10 billion in U.S. Government securities in the open mar ket. Deposits need not change, but banks must hold more reserves and fewer earning assets. Under the monetary view the effects would be nil (except for very minor effects examined below) because deposits would be unchanged, but under the credit view the effect would be a tendency for income to contract because bank credit would be lower. The monetary view is easily explained. Sup pose first that the banks initially hold U.S. Government securities in excess of $10 billion. When reserve requirements are raised, the banks simply sell $10 billion of these securities, and this is exactly the amount being purchased by the Federal Reserve. Thus, since deposits are unchanged and bank loans to the nonbank pri vate sector—hereinafter called simply the “private sector”—are also unchanged, there should be no effects on that sector. Now suppose that the banks do not have $10 billion in Government securities. In this case they must sell private securities, say cor porate bonds, to the private sector. The pri vate sector obtains the funds to buy these bonds from the sale of $10 billion of Government se curities to the Federal Reserve. The amount of credit in the private sector is again unchanged. The banks own fewer private securities, while the public owns more private securities and fewer Government securities. Thus, the amount of credit extended to the private sector need not change at all even though bank credit falls. However, two minor effects are possible: First, the Federal Reserve purchase of Government securities changes the composition of portfolios. Thus, even if banks have over $10 billion of Government securi ties, they may be expected to adjust their port folios by selling some Government securities and some private securities. For ease of expo sition, run-offs of loans may be included in the sale of private securities. The net result, then, is that the banks have more reserves, fewer Government securities, and fewer private se curities; the private sector has fewer Govern ment securities and fewer liabilities to the banks. The private sector may have—but it will not necessarily have—fewer claims within the sector. It is quite possible that private units may substitute claims on other private units for the Government securities sold to the Fed eral Reserve. Looked at from the liability side, those units initially with liabilities outstanding to banks may have those liabilities shifted to other pri RULES-OF-THUMB FOR POLICY vate sector units. This occurs, of course, when banks sell securities to the private sector or allow loans to run off that are then replaced by firms selling commercial paper to other firms, drawing on sources of trade credit, and/or borrowing from nonbank financial insti tutions. A net effect can occur only when the combined portfolios of banks and the private sector contain fewer Government securities, though more reserves, than before; such a change may be looked upon as a reduction in liquidity and thereby lead to a greater demand for money and a reduced willingness to under take additional expenditures on goods and services. The second effect of the hypothetical experi ment being discussed is that bank earnings will be reduced by the increase in reserve require ments. Banks will eventually adjust by raising service charges on demand deposits and/or re ducing interest paid on time deposits. For sim plicity, assume that the change in reserve re quirements applies only to demand deposits so that there is no reason for banks to change the interest paid on time deposits. With higher service charges on demand deposits, lower in terest rates on securities are required if people are to hold the same stock of money as before. Since the hypothetical experiment assumed that deposits did not change, interest rates must fall by the same amount as the increase in service charges, an effect that will tend to expand investment and national income. The portfolio effect tends to contract income while the service charge effect tends to expand income. These effects individually seem likely to be small, and the net effect may well be nil. In this regard, it is interesting to note that the relationship of velocity to the Aaa corporate bond rate is about the same for observations in the 1950’s as in the 1920’s [22, 23] in spite of the enormous changes in financial structure and in Government bonds outstand ing. Consider another hypothetical experiment —one that is in fact not so hypothetical at the current time. Suppose that banks suddenly start issuing large amounts of commercial paper and investing the proceeds in business loans. It is possible that the loans simply go to corporations that have stopped issuing their own commercial paper. In this case the bank would be purely a middleman with no effect on the aggregate amount of commercial paper outstanding. The increase in bank credit would not represent an increase in total credit. But, of course, banks issuing commercial paper must perform some function. This func tion is clearly that of increasing the efficiency of the financial sector in transferring funds from the ultimate savers to the ultimate bor rowers. The efficiencies arise in several ways. First, under fractional reserve banking, banks have naturally developed expertise in lending. It is efficient to make use of this expertise by permitting banks to have more lendable funds than they would have if restricted to demand deposits alone. The efficiency takes the form of fewer administrative resources being re quired to transfer funds from savers to bor rowers. The second form of efficiency results from the fact that financial markets function best when there is a large amount of trading in a standardized instrument. For example, the shares of large corporations are much more easily marketed than those of small corpora tions. Many investors want, and require, read ily marketable securities, and they can be per suaded to buy securities in small firms only if the yields are high. As a result funds may go to large corporations to finance relatively lowyielding investment projects while high-yielding projects available to small firms cannot be financed. Commercial banks, and other finan cial intermediaries, improve the allocation of capital by issuing relatively standardized secu rities with good markets and lending the pro ceeds to small firms. The question is whether there is any effect on economic activity from an increase in bank credit financed by commercial paper—assum ing that the money stock is not affected. To begin with, it must be emphasized that an in crease in the efficiency of investment does not necessarily affect the total of investment. The same resources may be absorbed either in building a factory that will produce a product that cannot be sold or in building a factory to produce a highly profitable product in great demand. Banks, and financial intermediaries in gen eral, have the effect of reducing somewhat the cost of capital for small firms. Because inter mediaries bid funds away from large corpora tions, the cost of capital for large corporations tends to be somewhat higher than it would be if there were no intermediaries. At this stage in the analysis the net effect on investment is impossible to predict since it depends on whether the reduction in investment by large corportions is larger or smaller than the in crease in investment by small corporations. In examining the effects of intermediation, however, another factor must be considered. Suppose it is assumed that the interest rates relevant for the demand for money are rates on high-quality securities. It was argued above that intermediation tends unambiguously to raise the yields on high-quality securities above what they otherwise would be. Since the as sumption throughout has been that the stock of money is unchanged, the level of income must increase if the quantity of money de manded is to be unchanged with the higher in terest rate of high-quality securities. The con clusion, therefore, is that the increase in bank credit is expansionary in the hypothetical ex periment being discussed. This conclusion, however, does not warrant the further conclusion that bank credit is the appropriate monetary aggregate for policy pur poses. The effect examined above occurs when any financial intermediary expands. Not only is there the problem that data for all intermedi aries are simply not available on a current basis but also there are serious problems in even defining an intermediary. A particularly good example of this difficulty is afforded by trade credit. A large nonfinancial corporation may advance trade credit to customers, many of whom may be small, and may also advance funds to suppliers through prepayments. The large corporation finances these forms of credit through the sale of securities, or through re tained earnings diverted from its own invest ment opportunities and/or from dividends. In this case the large corporation is serving ex actly the same function as the financial inter mediaries are. But tracing these credit flows is obviously impossible at the present time. Another problem with bank credit as a guide to policy is that changes in bank credit depend both on changes in bank deposits and on changes in nondeposit sources of funds. As demonstrated by the hypothetical experiments examined above, the effect of a change in bank credit depends heavily on whether or not deposits change. One final hypothetical experiment will be considered. Suppose the U.S. Treasury sells additional Government securities to the public to finance an increase in cash balances at com mercial banks. Since banks have received no additional reserves, total deposits cannot change. Deposits owned by the public are transferred to the Treasury. Bank credit is un changed, but the impact on the private sector is clearly contractionary. The private sector holds more Government bonds and fewer de posits. Equilibrium can be restored only through some combination of a rise in interest rates and a decline in income. The conclusion is that it appears to be fun damentally wrong for policy-makers to place primary reliance on bank credit. This is not to say that there is no information to be gained from analysis of bank and other credit flows. However, selection of bank credit as the mon etary aggregate would be a mistake. Instead, in formation on credit flows may be used to ad just the desired rate of growth of the money stock, however it is defined, although it is not clear that the knowledge presently exists as to how to interpret credit flows. From this analysis it appears that neither bank credit nor any deposit total that includes Treasury deposits is an appropriate monetary RULES-OF-THUMB FOR POLICY aggregate for monetary policy purposes. Before considering the narrow and broad definitions of the money stock, let us examine the mone tary base, total reserves, and unborrowed re serves. It is clear that different levels of the money stock may be supported by the same level of the monetary base. Given the monetary base, different levels of the money stock result from changes in reserve requirement ratios; from shifts of deposits between demand and time, which of course are subject to different reserve requirement ratios; from shifts of deposits among classes of banks with different reserve ratios; and from shifts between currency and deposits. These effects are widely understood, and they have led to the construction of mone tary base figures adjusted for changes in re serve requirements. Similar adjustments are applied to total and nonborrowed reserves. If enough adjustments are made, the adjusted monetary base is simply some constant frac tion of the money stock, while adjusted re serves are some constant fraction of deposits. It is obviously much less confusing to adopt some definition of the money stock as the ap propriate aggregate rather than to use the ad justed monetary base or an adjusted reserve figure. There can be no doubt that FOMC instruc tions to the Manager in terms of nonborrowed reserves would be more precise and more eas ily followed than instructions in terms of the money stock. But the simplicity of reserve in structions would disappear if adjusted reserves were used, for then the Manager would have to predict such factors as shifts between de mand and time deposits, the same factors that must be predicted in controlling the money stock. No one would argue that such factors —and others such as changes in bank borrow ings and shifts in Treasury deposits—should be ignored. If the FOMC met daily, instruc tions could go out in unadjusted form with the FOMC making the adjustments. But surely this technical matter should be handled not by the FOMC but by the Manager and his staff in order to permit the FOMC to concentrate on basic policy issues. The only aggregates left to consider are the narrowly and broadly defined money stocks. There is a weak theoretical case favoring the narrow definition because time deposits must be transferred into demand deposits or cur rency before they can be spent. The case is weak because the cost of this transfer is rela tively low. If the cost were zero, then there would be no effective distinction between de mand and time deposits. Indeed, since time de posits earn interest, all funds would presuma bly be transferred to time deposits. No strong empirical case exists favoring one definition over the other. The broad and narrow money stocks are so highly correlated over time that it is impossible to distinguish separate effects. It appears, however, that there is a practical case favoring the adoption of the narrow money stock. Time deposits include both passbook accounts, which can be readily transferred into demand deposits, and certifi cates of deposit, which cannot. Since CD’s ap pear to be economically much more like com mercial paper than like passbook time accounts, they ought to be excluded from the broadly defined money stock. There is, of course, no reason why CD’s cannot be excluded from the definition of money. The problem is that banks may in the future invent new instruments that will be classified as time deposits for regulatory pur poses but that are not really like passbook ac counts. In retrospect it may be clear how the new instrument should be treated, but the situa tion may be confused for a time. The same sort of problem exists with demand deposits—con sider the compensating balance requirements imposed by many banks—but it seems likely that the problem will remain more serious for time deposits. In summary, there is a strong case favoring the selection of some definition of the money stock as the monetary aggregate, and there ap pears to be a marginal case for preferring the narrowly defined money stock. 172 TECHNICAL PROBLEMS OF CON TROLLING MONEY STOCK. In the pre ceding sections it has been argued that the monetary policy control instrument should be the money stock. The purpose of this section is to investigate some of the technical prob lems in controlling the money stock. The first topic examined is that of the form of instruc tions to the Manager of the System Open Mar ket Account. Following this discussion is an examination of the feedback method of control. Finally, there is an examination of the signifi cance of data revisions. All of this discussion is in terms of the narrowly defined money stock, but much of it also applies to other ag gregates. Specification of the desired money stock. There are two major issues connected with the form of FOMC instructions to the Manager. The first is whether the desired money stock should be expressed in seasonally adjusted or unadjusted form, while the second is whether the desired money stock should be expressed in terms of a complete path week by week over time or of an average over some period of time. The first issue turns out to be closely re lated to the question of data revisions, and so its discussion will be deferred for the moment. It is to the second issue that we now turn. Since required reserves are specified in terms of a statement-week average, the state ment week is the natural basic time unit for which to measure the money stock, and the measure takes the form of the average of daily money stock figures over the statement week. The fact that daily data may not be available on all components of the money stock does not affect the argument; however estimated, the weekly-average figure is the most appropri ate starting point in the analysis. The weekly money stock is clearly not sub ject to precise control because of data lags and uncontrollable random fluctuations. Further more, no one believes that these weekly fluctuations have any significant impact. The natural conclusion to be drawn is that there is no point in specifying instructions in terms of weekly data but rather that some average level over a period of weeks should be used. Upon closer examination, however, this conclusion can be shown to be unjustified. The difficulty in expressing the instructions in terms of averages can be explained very simply by two examples. To keep the exam ples from becoming too complicated, it will be assumed that instructions take the form of simple rates of growth on a base money stock of $200 billion. The neglect of compounding makes no essential difference to the argument. For the first example, assume that the policy instruction is for a growth rate of 4 per cent per annum, which is $8 billion per year or about $154 million per week. If the money stock grew by $154 million per week for 8 weeks, then the figure for the eighth week would be above the base week figure by an amount representing a 4 per cent annual growth rate. The average of weeks 5 through 8 would be above the average of weeks 1 through 4 by $616 million, an amount also representing a 4 per cent annual growth rate. So far, there is no reason to favor the path specification over a specification in terms of 4-week averages. Now suppose that the increase in weeks 1 through 4 was on schedule, but that a large uncontrollable increase of $500 million oc curred in the fifth week. Starting from a baseweek figure of $200 billion, the average money stock for weeks 1 through 4 would be $200,385 billion, and if the instruction were in terms of 4-week averages it would specify an average money stock of $201,001 billion for weeks 5 through 8. Since by hypothesis the money stock grew by $154 million in each of the first 4 weeks, in the fourth week the level was $200,616 billion. The jump of $500 million in the fifth week would take the level to $201,116 billion, a figure already above the desired average of $201,001 billion for weeks 5 through 8. To reach this desired average given the jump in week 5, the money stock in weeks 6 through 8 would have to average less than $201,001 bil lion, and so the money stock would have to be RULES-OF-THUMB FOR POLICY forced below the level of the fifth week for weeks 6 through 8. Furthermore, as the reader may calculate, it would be necessary to have higher than normal weekly growth in weeks 9 through 12 if the average of these weeks were to be above the average of weeks 5 through 8 by $616 million. On the other hand, if the in struction were in terms of the desired weekly path, the instruction would read that the de sired money stock in the eighth week was $201,232 billion, and therefore the Manager would not have to force the money stock down in weeks 6 through 8. Instead, he could aim for a growth of about $39 million in each of the weeks 6 through 8 to bring the level in week 8 to the desired figure of $201,232 bil lion. From this example it can be seen that speci fication in terms of averages of levels of the money stock forces the Manager to respond to random fluctuations in a whipsawing fashion. Since week-by-week fluctuations have essen tially no significance, there is no point in wrenching the financial markets in order to undo a random fluctuation. If averaging is to be used, the average should be specified in terms of the desired average weekly change over, say, the next 4 weeks rather than in terms of the average level of the next 4 weeks. Specification in terms of the average weekly change is equivalent to a specification stating that the Manager should aim for a particular target level in the fourth week. The second example illustrating the hazards of specification in terms of the average level will show what happens when policy changes. As before, assume that the money stock in the base week is $200 billion and that the desired growth is at a 4 per cent rate in weeks 1 through 4. In this example it is assumed that there are no errors in hitting the desired money stock. Thus, the money stock is as sumed to grow by $154 million per week, reaching a level of $200,616 billion in the fourth week and an average level of $200,385 billion for weeks 1 through 4. Now suppose that in week 4 the FOMC de cides on a policy change and specifies a 1 per cent growth rate for the money stock for weeks 5 through 8. If the specification were in terms of the average level, then it would re quire an increase in the average level of $154 million, which would bring the average level to $200,539 billion for weeks 5 through 8. But the figure for week 4 is already $200,616 bil lion, and so the money stock in weeks 5 through 8 would have to average less than the figure already achieved in week 4. Thus, after a steady 4 per cent growth week by week, an average-level policy specification would actually require a negative week-byweek growth before the new 1 per cent growth rate could be achieved. On the other hand, a policy specification in terms of the weekly path would require a weekly growth of $38.5 mil lion each week for weeks 5 through 8. To make the point clear, this example was constructed so that the policy shift from a 4 to a 1 per cent growth rate would actually re quire a negative growth rate for a time on a week-by-week basis when the instructions are in terms of average levels. In general, when average levels are used, a policy shift to a lower growth rate will require in the short term a growth rate lower than the new policy rate set, and a policy shift to a higher growth rate will require a short-term growth rate above the new policy rate. Since policy-makers will typically want to shift policy gradually, the levels specification is especially damaging be cause it in fact instructs the Manager to shift policy more rapidly than the policy-makers had desired. It should be noted that the larger the number of weeks included in the averagelevel specification, the more severe this prob lem becomes. Because the money stock cannot be con trolled exactly, there is a natural tendency to feel that instructions stated in terms of aver ages are more attainable. In actuality, of course, this effect is illusory; averaging pro duces a smaller number to measure the errors, but does not improve control. Nevertheless, if averages are to be used in the instructions, the above examples demonstrate that the averages should be calculated in terms of weekly (or perhaps monthly) changes but not in terms of averages of levels. Use of average changes does have one ad vantage, however. An instruction in this form permits the Manager to correct an error in week 1 over the next few weeks rather than instructing him to correct the error entirely in week 2. As explained above, an instruction in terms of the average weekly change over the next 4 weeks is equivalent to an instruction in terms of the desired level in week 4, leaving unspecified the desired levels in weeks 1 through 3. Control through the feedback principle. It is useful to begin by comparing the problems of controlling the money stock with the problems of controlling interest rates. In controlling in terest rates, the availability of continuous read ings on rates makes it possible for the Man ager to exercise very accurate control without understanding the causes of rate changes. Being in continuous contact with the market, the Manager can intervene with open market purchases or sales as soon as the Federal funds rate, the Treasury bill rate, or any other rate starts to change in an undesirable fashion. This feedback control is not exact since inter est rate information arrives with some lag, and there are other lags such as the time required to decide upon and execute an open market transaction and the time it takes for the mar ket to react to the transaction. More precise control over interest rates could be achieved if the Manager were willing to announce Federal Reserve buying and sell ing prices for, say, 3-month Treasury bills available to all comers. This is essentially the way in which Government securities were pegged during World War II. In principle, there is no reason why such a peg could not be operated in peacetime, although it would cer tainly be desirable to change the peg fre quently, perhaps as often as every day or even every hour. However, in terms of actual be havior of interest rates there is no significant difference between a frequently adjusted peg and continuous intervention by the Manager as described in the previous paragraph. The main point of this discussion of interest rate control is to emphasize that with frequent interest rate readings it is not necessary to know exactly what causes interest rate changes. In time the Manager develops a feel for the market that enables him to guess ac curately which interest rate changes are tempo rary and which are likely to be “permanent” and so require offsetting open market opera tions. Furthermore, his feel for the market will enable him to know how large the operations should be. Finally, when he guesses wrongly on these matters, his continuous contact with the market enables him to correct mistakes rapidly. The same arguments apply to controlling the money stock. The difference between inter est rate control and money stock control is a matter of degree rather than kind. Data on the money stock become available with a greater lag, and the data are more subject to revision. But since it is not necessary to control the money stock down to the last dollar, the ques tion is whether it is technically possible to have control that is accurate enough for policy purposes. The answer to this question would certainly appear to be in the affirmative. The weekly-average figure for the money stock is released to the public 8 days following the end of the week to which the average re fers. Of course, data are available internally with a shorter lag. Since the policy rule in the previous section is based on controlling the monthly-average money stock, it would appear that the data are at the present time available with a short enough lag that feedback methods of control are feasible. To see how feedback control would work, suppose that the Manager were instructed to come as close as possible to a target money stock of M 4* in week 4 of a 4-week operating horizon. The Manager knows that the weekly change in the money stock depends on open market purchases, P, which he controls, and RULES-OF-THUMB FOR POLICY many other factors as well, which for simplic ity of exposition will be denoted by one factor, z. These factors cannot be predicted exactly, and so the Manager will think of z as consist ing of a predictable part, z, and an unpredicta ble part, u. These relationships may be ex pressed as At the end of the first week the Manager has the estimate, M ly for the money stock for that week, and again it is assumed that he wants to spread the desired change M* — Mt equally over the next 3 weeks. Thus, the Man ager sets P2 according to (5) (7) AM = aP + z = aP + z + u where a is the coefficient giving the change in money per dollar of open market purchases. If there were no errors in measuring the money stock, the analysis could be completed on the basis of equation 5. But of course there are errors in measuring the money stock. To analyze the significance of measurement errors, let Mi be the money stock for week / as meas ured at the end of week Z.18 Also, let M tf be the final “true” money stock figure for week i, and let ei = Mi — M i. The Manager starts out the 4-week period with an estimated money stock of M0 for week zero. Of course, the figure for M0 is a pre liminary one, but revisions in this figure as more data accumulate will affect the estimates for the money stock in later weeks and so affect the Manager’s actions in later weeks. It will be assumed that he wants to increase the money stock by equal amounts in each week to reach the desired figure of M4* in week 4. In week 1, therefore, he wants to produce a change in the money stock Va (M4* — M0). Substituting this figure into equation 5 we obtain 1/4 (M4* - M 0) - aPi + Thus, the Manager sets (6) Pi = I Px + m according to [ J (M 4* - Mo) - z j 18 If a money stock estimate is not directly avail able at the end of week i, one can be constructed by taking the estimate from actual deposit data for week i —l and adding to it a projection for the effects of open market operations and other factors for week /. This projection would, of course, come from equation 5. Pi = ^ M ») - * * ] Similarly, he sets P3 and P4 according to equa tions 8 and 9. (8) (9) {Mi*~ Mi) ~ *3] Mt) - z 4] Pt = - a From equations 9 and 5 it can be seen that the actual money stock in week 4 is (10) M 4 = W + AT* - M z - z4 + z = M4*+ ez + Ui This expression for the fourth week of a plan ning period generalizes to the /ith week of a planning period of any length merely by re placing the subscript 4 by the subscript n. We can, therefore, express the annual rate of growth, g, over an n week period by (U> 52 / M nf — M<>f\ ) _ 52 / « . » - M J n \ M o f + 52 , ~ (e « -l + Wn) From equation 11 it can be seen that the ac tual growth rate, g, equals the desired growth rate plus an error term that becomes smaller as n becomes larger. This analysis shows that a feedback control system that continuously adjusts open market operations as data on the money stock in the recent past become available can achieve a target rate of growth with a margin of error that is smaller the longer the period over which the rate of growth is calculated. It also provides a framework in which to examine the relative importance of operating errors, the uit and data errors, the e%. To obtain an accurate estimate of the sizes of these errors is beyond the scope of this study. However, a very crude method may be used to obtain an estimate of the maximum size of the total error. Monthly money stock changes at annual rates were computed for the period January 1951 through September 1969 on the basis of seasonally adjusted data. This time period yields a total of 225 monthly changes. Then each monthly change was ex pressed in terms of its deviation from the aver age of the changes for the previous 3 months. For example, the September deviation was cal culated by subtracting from the September monthly change the average of the changes for August, July, and June. The use of deviations allows in part for longer-run trends in the money stock, which trends are assumed to be readily controllable. Since the deviations were calculated over a period during which little or no attention was paid to controlling the money stock, they surely represent an upper limit to the degree of volatility in the money stock to be expected under a policy directed at control of the money stock. These monthly deviations have a standard deviation of 3.12 per cent per annum. Apply ing equation 11, except for replacing 52 by 1 to reflect the fact that the rates of change were expressed at annual rates in the first place, it is found that the standard deviation over a 3month period would be 1.04 per cent per annum. If it is assumed that these deviations are normally distributed, the conclusion is that over 3-month periods the actual growth rate would be within plus or minus 1.04 per cent of the desired growth rate about 68 per cent of the time, and would be within plus or minus 2.08 per cent about 95 per cent of the time. Inasmuch as these limits would be cut in half over 6-month periods, the actual growth rate 95 per cent of the time would be in the range of plus or minus 1.04 per cent of the desired growth rate.19 When it is recalled that 19 If the calculations are based on the variability of the monthly changes themselves rather than on the deviations of the monthly changes, the results are not these calculations are based on an estimate of variability over a period in which very little at tention was paid to stabilizing money stock growth rates, it is clear that fears as to the ability of the Federal Reserve to control the money stock accurately are completely unfounded.20 This conclusion justifies the approach used at the beginning of this section on the selection of a monetary aggregate, at least for the nar rowly defined money stock and most probably for other aggregates as well. That approach, it will be recalled, analyzed the selection issue on the assumption that every one of the aggre gates considered could be precisely controlled for all practical purposes. There can be no doubt that errors in reaching targets for goal variables such as GNP, at the present state of knowledge, are due almost entirely to incom plete knowledge of the relationships between instrument variables (such as various aggre gates and interest rates) and the goal varia bles, and hardly at all to errors in setting in strument variables at desired levels. Problems of data revisions and changing seasonality. Another topic that needs examina tion is the effect of data revisions. While week ly-average data are released with an 8-day lag, these figures are subject to revision. Not much weight can be given to early availability of data that are later revised substantially. To in vestigate this problem, two money stock series were compared, one “preliminary” and one greatly changed. The standard deviation of the monthly changes over the same period used before is 3.53 per cent per annum, which yields a 95 per cent chance of the growth rate being in a range around the desired rate of plus or minus 2.36 (1.18) per cent per annum for 3-month (6-month) periods. 20 Compare “First, however, it may be worthwhile to touch on the extensively debated subject whether the Federal Reserve, if it wanted to, could control the rate of money supply growth. In my view, this lies well within the power of the Federal Reserve to accomplish provided one does not require hair-split ting precision and is thinking in terms of a time span long enough to avoid the erratic, and largely mean ingless, movements of money supply over short peri ods.” [3, p. 75] RULES-OF-THUMB FOR POLICY “final.” Since the analysis below is based on published monthly data, it obviously provides little insight into the accuracy of weekly data. However, since policy instructions may be based on monthly data, the analysis is of some value in assessing data accuracy. Furthermore, the conclusions on the importance of revisions in seasonal factors can be expected to hold for the weekly data. A “preliminary” series of monthly growth rates of the money stock was constructed by calculating the growth rate for each month from data reported in the Federal Reserve Bulletin for the following month. For example, the Bulletin dated September reports money stock data for 13 months through August; it is the annual rate of change of August over July that is called the “preliminary” August rate-ofchange observation. The “final” series is the annual rate of growth calculated from the monthly money stock series covering 1947 through September 1969, reported in the Fed eral Reserve Bulletin for October 1969, pp. 790-93. Data were gathered on both a season ally adjusted basis and an unadjusted basis for January 1961 through August 1969. The correlation between the preliminary and final seasonally adjusted series is 0.767, while for the unadjusted series the correlation is 0.997. Another way to compare the prelimi nary and final series is to examine the differ ences in the two series.21 For the seasonally adjusted data, the differences have a mean of 0.122 and a standard deviation of 3.704, and the mean absolute difference is 2.891. On the other hand, for the seasonally unadjusted data the differences have a mean of 0.150 and a standard deviation of 1.366, and the mean ab solute difference is 0.955,22 21 The analysis of the differences inadvertently runs from February 1961 through August 1969 while the correlation analysis runs from January 1961 through August 1969. 22 To take account of the fact that the “final” money stock series may be further revised for months near the October 1969 publication date of this series, the analysis of differences between the preliminary and final series was also run on the pe- These results make it abundantly clear that the major reason why the preliminary and final figures on the money stock differ is revision of seasonal adjustment factors. While such revi sions may produce substantial differences be tween preliminary and final monthly growth rates, the differences must be lower for the av erage of several months’ growth rates. The reason, of course, is that revision of seasonal factors must make the figures for some months higher and those for other months lower, leav ing the annual average about unchanged. The significance of revisions in seasonal fac tors can be understood only after a discussion of the significance of seasonality for a money stock rule. If the monetary rule were framed in terms of the seasonally unadjusted money stock, the result would be to introduce sub stantially more seasonality into short-term in terest rates than now exists. It can be argued not only that greater seasonality in interest rates would not be harmful but also that it would be positively beneficial. Greater season ality in interest rates would presumably tend to push production from busy, high-interest sea sons into slack, low-interest seasons. Although the argument for seasonality in in terest rates could be pushed further, there is an important practical reason for not initially adopting a money rule stated in terms of the seasonally unadjusted money stock. The rea son is that the rule ties the growth rate of the money stock to the seasonally adjusted unem ployment rate and to the interest rate. The rule has been developed through an examina tion of past experience. If the seasonal were taken out of the money stock, a different sea sonal would be put into interest rates, and posriod February 1961 through December 1968. The mean difference, the standard deviation of the differ ences, and the mean absolute difference, are, respec tively, for the seasonally adjusted data 0.026, 3.779, and 2.922, while the figures for the seasonally unad justed data are 0.038, 1.280, and 0.890. In spite of the fact that the “final” series is not really final for 1969 data, the average differences are generally larger for the longer period due to the relatively large data revisions in the middle of 1969. 178 sibly into the unemployment rate as well. Sea sonal factors for these variables, especially for the unemployment rate, determined from past data would no longer be correct if the money stock seasonal were removed. Seasonally ad justing the unemployment index by the old factors could produce considerable uncertainty over the application of the monetary rule. Thus, application of the rule through the sea sonally unadjusted money stock, if desirable at all, should only come about through gradual reduction rather than immediate elimination of seasonality. A further reason for a gradual ap proach would be to permit the financial mar kets to adjust more easily to changed seasonal ity. The point of this discussion is not to urge acceptance of a rule framed in terms of the unadjusted money stock, since this step would not be initially desirable in any case. Rather, the point is to emphasize that seasonality is in the money stock only in order to reduce the seasonality of other variables, primarily inter est rates. The seasonality of the money stock, unlike variables such as agricultural produc tion, is not inherent in the workings of the economy but rather exists because the Federal Reserve wants it to exist. The money stock can be made to assume any seasonal pattern the Federal Reserve wants it to assume. The monetary rule should be framed, at least initially, in terms of the seasonally ad justed money stock—using the latest estimated seasonal factors. In subsequent years changes in these seasonal factors should not result from mechanical application of seasonal adjustment techniques to the money stock data but rather should be the result of a deliberate policy choice. The policy choice would be based on the desire to change seasonality of other varia bles. For example, if it were thought desirable to take the seasonality out of short-term inter est rates, the seasonal factors for the money stock would then be changed to take account of changes in tax dates and other factors. Under a money stock policy, whether or not guided by a monetary rule, revised seasonal factors cannot properly be applied to past data. If the changes are applied to past data with the result that some monthly growth rates of adjusted data become relatively high while others become relatively low, the conclusion to be drawn is not that policy was mistaken as a result of using faulty seasonal factors. Instead, the conclusion is merely that seasonal policy differed in the past from current policy or from the seasonal pattern assumed by the in vestigator who computed the seasonal factors. Seasonal policy can be shown to be “wrong” only by showing that undesirable seasonals exist in other variables. One final problem deserves discussion. While it appears from the analysis of season ally unadjusted money stock data that revi sions of the data are relatively unimportant, at least from the evidence for 1961-69, how should the policy rule be adjusted when there are major data revisions—as in the middle of 1969? For example, suppose that revisions in dicate that monetary growth has been much higher than had been expected, and higher than was desirable. On the one hand, pol icy could ignore the past high rate of growth and simply maintain the current rate of growth of the revised series in the desired range. On the other hand, the policy could be to return the money stock to the level implied by apply ing the desired growth rate to the money stock in some past base period. The first alternative involves ratifying an undesirably high past rate of growth, while the second may involve a wrenching change in the money stock to return it to the desired growth path. The proper pol icy would no doubt have to be decided on a case-by-case basis. However, a useful pre sumption might be to adopt the second alter native, but to set as the base the money stock 6 months in the past and to return to the de sired growth path over a period of several months. Improving control over the money stock. The analysis above has shown that under pres ent conditions the money stock can be con trolled quite accurately. However, it should be RULES-OF-THUMB FOR POLICY emphasized that there are numerous possibili ties for improving control. Although detailed treatment of this subject is beyond the scope of this study, a few very brief comments ap pear appropriate. There are three basic methods for improving control. The first method is that of improving the data. The more quickly the deposit data are available, the more quickly undesirable movements in the money stock can be recog nized and corrected. And the more accurate the deposit data, the fewer the mistakes caused by acting on erroneous information. It is clear that expenditures of money on expanding the number and coverage of deposit surveys and on more rapid processing of the raw survey data can improve deposit data. The second method of improving control is through research, which increases our under standing of the forces making for changes in the money stock. For example, transfers be tween demand and time deposits might be more accurately predicted through research into the causes of such transfers. The third method of improving control is through institutional changes. To reduce fluc tuations in excess reserves and thereby achieve a more dependable relationship between total reserves and deposits, the Federal funds mar ket might be improved by making possible transfers between the East and West Coasts after east coast banks are closed. Also helpful would be a change from lagged to contempor aneous reserve requirements. More radical re forms such as equalization of reserve require ments for city, country, and nonmember banks and elimination of reserve requirements on time deposits should also be considered. V. SUMMARY PURPOSES OF THE STUDY. The primary purpose of this study has been to argue that a major improvement in monetary policy would result through a systematic policy approach based on adjustments in the money stock. Equal emphasis has been placed on the “systematic” part and the “money stock” part of this approach. The analysis has proceeded first by showing why policy adjustments should be made through money stock adjustments, and second by showing how these policy ad justments might be systematically linked to the current business situation through a policy guideline or rule-of-thumb. A third, and sub sidiary, part of this study is an analysis of the reasons for preferring the money stock over other monetary aggregates, and of some of the problems in reaching desired levels of the money stock. It has been emphasized throughout that this policy approach is one that is justified for the intermediate-term future on the basis of knowledge now available. The specific recom mendations are not intended to be good for all time. Indeed, the approach has been designed to encourage evaluation of the results so that the information obtained thereby can be incor porated into policy decisions in the future. THE THEORY OF MONETARY POLICY UNDER UNCERTAINTY. Since policy makers have repeatedly emphasized the im portance of uncertainty, it is necessary to analyze policy problems within a model that explicitly takes uncertainty into account. In particular, only within such a model is it possi ble to examine the important current issue of whether policy adjustments should proceed through interest rate or money stock changes. A monetary policy operating through inter est rate changes sets interest rates either through explicit pegging as was used in World War II or through open market operations di rected toward the maintenance of rates in some desired range. Under such a policy the money stock is permitted to fluctuate to what ever extent is necessary to keep interest rates at the desired levels. On the other hand, a pol icy operating through money stock changes uses open market operations to set the money stock at its desired level while permitting inter est rates to fluctuate freely. If there were perfect knowledge of the rela- 180 tionships between the money stock and interest rates, the issue of money stock versus interest rates would be nonexistent. With perfect knowledge, changes in interest rates would be perfectly predictable on the basis of policyinduced changes in the money stock, and vice versa. It would, therefore, be a matter of pref erence or prejudice, but not of substance, whether policy operated through interest rates or the money stock. To analyze the interest versus money issue, then, it is necessary to assume that there is a stochastic link between the two variables. And, of course, this is in fact the case. There are two fundamental reasons for the stochastic link. First, the demand for money depends not only on interest rates and the level of income but also on other factors, which are not well understood. As a result, the demand for money fluctuates in a random fashion even if income and interest are unchanged. If the stock of money is fixed by policy, these ran dom demand fluctuations will force changes in interest and/or income in order to equate the amount demanded with the fixed supply. The second source of disturbances between money and interest stems from disturbances in the relationship between expenditures—espe cially investment-type expenditures—and inter est rates. Given an interest rate fixed by pol icy, these disturbances produce changes in income through the multiplier process, and these income changes in turn change the quantity of money demanded. With interest fixed by policy, the stock of money must change when the demand for money changes. On the other hand, if the money stock were fixed by policy, since the expenditure disturb ance changes the relationship between income and interest, some change in the levels of in come and/or interest would be necessary for the quantity of money demanded to equal the fixed stock. Money stock and interest rate policies are clearly not equivalent in their effects, given that disturbances in money demand and in ex penditures do occur. Since the effects of these policies are different, which policy to prefer depends on how the effects differ and on pol icy goals. At this level of abstraction, it is clearly appropriate to concentrate on the goals of full employment and price stability. Unfor tunately, the formal model that has been worked out, which is examined carefully in Section I above, applies only to the goal of stabilizing income. If “income” is interpreted to mean “money income,” then the goals of employment and price level stability are in cluded but are combined in a crude fashion. The basic differences in the effects of money stock and interest rate policies can be seen quite easily by examining extreme cases. Sup pose first that there are no expenditure dis turbances, so there is a perfecdy predictable relationship between the interest rate and the level of income. In that case, a policy that sets the interest rate sets income, and policy-mak ers can choose the level of the interest rate to obtain the level of income desired. When the interest rate is set by policy, disturbances in the demand for money change the stock of money but not the level of income. On the other hand, if policy sets the money stock, then the money demand disturbances would affect interest and income leading to less satis factory stabilization of income than would occur under an interest rate policy. The other extreme case is that in which there are disturbances in expenditures but not in money demand. If policy sets the interest rate, expenditure disturbances will produce fluctuations in income. But if the money stock is fixed, these income fluctuations will be smaller. This point can be seen by considering a specific example such as a reduction in in vestment demand. This disturbance reduces in come. But given an unchanged money demand function, with the fall in income, interest rates must fall so that the amount of money de manded will equal the fixed stock of money. The decline in the interest rate will stimulate investment expenditures, thus offsetting in part the impact on income of the initial decline in the investment demand function. With expend- RULES-OF-THUMB FOR POLICY itures disturbances, then, to stabilize income, it is clearly better to follow a money stock policy than an interest rate policy. The conclusion is that the money versus in terest issue depends crucially on the relative importance of money demand and expendi tures disturbances. It is especially important to note that nothing has been said about the size of the interest elasticity of the demand for money, or of the interest elasticity of invest ment demand. These coefficients, and others, determine the relative impacts of changes in money demand and in investment and govern ment expenditures when the changes occur. The interest versus money issue does not depend on these matters, however, but only on the relative size and frequency of dis turbances in the money demand and expendi tures functions.23 The analysis above is modified in detail by considering possible interconnections between money demand and expenditures disturbances. It is also true that in general the optimal pol icy is not a pure interest or pure money stock policy, but a combination of the two. These matters, and a number of others, are discussed in Section I. EVIDENCE ON RELATIVE MAGNI TUDES OF REAL AND MONETARY DIS TURBANCES. Resolution of the money ver sus interest issue depends on the relative size of real and monetary disturbances. Unfortun ately, there is no completely satisfactory body of evidence on this matter. Indeed, because of the conceptual difficulties of designing empiri cal studies to investigate the issue, the evi dence is unlikely to be fully satisfactory for some time to come. Nevertheless, by examin ing a number of different types of evidence, a substantial case can be built favoring the use of the money stock as the policy control varia ble. Before discussing the evidence, it is neces sary to define in more detail what is meant by “disturbance.” Consider first a money demand 23 For a full understanding of this important point, the reader should refer to the analysis of Section I. disturbance. The demand for money depends on the levels of income and of interest rates, and on other variables. The simplest form of such a function uses GNP as the income varia ble, and one interest rate—say the Aaa corpo rate bond rate—and all other factors affecting the demand for money are treated as disturb ances. To the extent possible, of course, these other factors should be allowed for, but for policy purposes these factors must be either continuously observable or predictable in ad vance so that policy may be adjusted to offset any undesirable effects on income of these other factors. Factors not predictable in ad vance must be treated as random disturbances. Similarly, expenditures disturbances are de fined as the deviations from a function linking income to the interest rate and other factors. These other factors would include items such as tax rates, government expenditures, strikes, and population changes. Again, for policy pur poses these factors must be forecast, and so errors in the forecasts of these items must be included in the disturbance term. It is impor tant to realize that the disturbances will be de fined differently for scientific purposes ex post because the true values of government spend ing and so forth can be used in the functions once data on these items are available. In the discussion of the theoretical issues above it was noted that an expenditure dis turbance would have a larger impact on in come under an interest rate policy than under a money stock policy. Simulation of the FR-M IT model provides the estimate that the impact on income of an expenditures disturb ance, say in government spending, is over twice as large under an interest rate policy as under a money stock policy. An error in fore casting government spending, then, would lead to twice as large an error in income under an interest rate policy. Since there is no system atic record of forecasting errors for variables such as government spending and strikes, there is no way of producing evidence on the size of such forecasting errors. However, after listing the variables that must be forecast, as is done in Section II, it is difficult to avoid feeling that errors in forecasting are likely to be quite significant. These real disturbances, including forecast errors in government expenditures, strikes, and so forth, must be compared with the disturb ances in money demand. The reduced-form studies conducted by a number of investigators provide some evidence on this issue. These studies compare the relative predictive power of monetarist and Keynesian approaches in ex plaining fluctuations in income. From these studies the predictive power of both ap proaches appears about equal. However, the predictive power of the Keynesian approach relies on ex post observation of “autonomous” expenditures, and it is clear that these expendi tures are subject to forecasting errors ex ante whereas the money stock can be controlled by policy. The evidence from the reduced-form studies suggests that when forecast errors of autono mous expenditures are included in the disturb ance term, the disturbances are larger on the real side than on the monetary side. There are many difficulties with the reduced-form ap proach and so these results must be interpreted cautiously. Nevertheless, the results cannot be ignored. The final piece of evidence offered in Sec tion II is a study by the author of the stability of the demand for money function over time. Using a very simple function relating the in come velocity of money to the Aaa corporate bond rate, he found that a function fitted to quarterly data for 1947-60 also fits data for 1961-68 rather well. The reader interested in the precise meaning of “rather well” should turn to the technical discussion in Section II. Evidence on relative stability is difficult to obtain and subject to varying interpretations. No single piece of evidence is decisive, but all the various scraps point in the same direction. The evidence is not such that a reasonable man can say that he has no doubts whatsoever. But since policy decisions cannot be avoided, the reasonable decision based on the available evidence is to adopt the money stock as the monetary policy control variable. A MONETARY RULE FOR GUIDING POLICY. The conclusion from the theoretical and empirical analysis is that the money stock ought to be the policy control variable. For this conclusion to be very useful, it must be shown in detail how the money stock ought to be used. It is not enough simply to urge policy-makers to make the “appropriate” ad justments in the money stock in the light of all “relevant” information. There is no general agreement on exactly what types of adjustments are appropriate. However, it would probably be possible to ob tain agreement among most economists that ordinarily the money stock should not grow faster than its long-run average rate during a period of inflation and should not grow slower than its long-run average rate during recession. But many economists would want to qualify even this weak statement by saying that there may at times be special circumstances requir ing departures from the implied guideline. Others would say that there is no hope at present of gauging correctly the impact of spe cial circumstances (or even of “standard” cir cumstances) so that policy should maintain an absolutely steady rate of growth of the money stock. The basic issues are, first, whether policy makers can forecast disturbances well enough to adjust policy to offset them, and second, the extent to which money stock adjustments to offset short-run disturbances will cause unde sirable longer-run changes in income and other variables. The theoretical possibilities are many, but the empirical knowledge does not exist to determine which theoretical cases are important in practice. It is for this reason that a systematic policy approach is needed so that policy can be easily evaluated and improved with experience. Policy could be linked in a systematic way to a large-scale model of the economy. Target values of GNP and other goal variables could be selected by policy-makers, and then the RULES-OF-THUMB FOR POLICY model solved for the values of the money stock and other control variables (for exam ple, discount rate) needed to achieve policy goals. While this approach may be feasible in the future, it is not feasible now because a suf ficiently accurate model does not exist. In stead, policy decisions are now made largely on the basis of intuitive reactions to current business developments. Given this situation, the obvious approach is to specify precisely how policy decisions ought to depend on current developments, and this is the approach taken in Section III. The specifi cation there takes the form of a policy guide line, or rule-of-thumb. The proposed rule is purposely simple so that evaluation of its mer its would be relatively easy. Routine evaluation of an operating guideline would over time pro duce a body of evidence that could be used to modify and complicate the rule. But it is nec essary to begin with a simple rule because the knowledge that would be necessary to con struct a sophisticated rule does not exist. The proposed rule assumes that full employ ment exists when the unemployment rate is in the 4.0 to 4.4 per cent range. The rule also as sumes that at full employment, a growth rate of the money stock of 3 to 5 per cent per annum is consistent with price stability. There fore, when unemployment is in the full em ployment range, the rule calls for monetary growth at the 3 to 5 per cent rate. The rule calls for higher monetary growth when unemployment is higher, and lower mon etary growth when unemployment is lower. Furthermore, when unemployment is relatively high the rule calls for a policy of pushing the Treasury bill rate down provided monetary growth is maintained in the specified range; similarly, when unemployment is relatively low the rule calls for a policy of pushing the bill rate up provided monetary growth is in the specified range. Finally, the rule provides for adjusting the rate of growth of money accord ing to movements in the Treasury bill rate in the recent past. The exact rule proposed is in Table 3 (p. 160) and the detailed rationale for the various components of the rule is ex plained in the discussion accompanying that table. The rule is specified throughout in terms of 2 per cent ranges for the rate of growth of the money stock on a month-by-month basis. By expressing the rule in terms of a range, leeway is provided for smoothing undesirable interest rate fluctuations and for minor policy adjust ments in response to other information. Fur thermore, it is not proposed that this rule-ofthumb or guideline be followed if there is good reason for a departure. But departures should be justified by evidence and not be based on vague intuitive feelings of what is needed since the rule was carefully designed from the theo retical and empirical analysis of Sections I and II, and from a careful review of post-ac cord policy. There is no way of really testing the pro posed rule short of actually using it. However, it is useful to compare the rule with post accord policy. A detailed comparison may be found in Section III, pp. 153-57. A summary comparison suggests, however, that for the period January 1952 through July 1968 the rule would have provided a less appropriate policy than the actual policy in only 63 of the 199 months in the period. The rule was judged to be less appropriate if it called for a higher— lower—rate of monetary growth than actually occurred and unemployment 12 months hence was below—above—the desired range of 4.0 to 4.4 per cent. The rule was also judged less appropriate than the actual policy if actual policy was not within the rule but unemploy ment nevertheless was in the desired range 12 months hence. The rule actually has slightly fewer errors if the criterion is unemployment either 6 or 9 months following the months in question. The rule has the great virtue of turning pol icy around promptly as imbalances develop and of avoiding cases such as the 2.2 per cent rate of decline in the money stock from July 1957 through January 1958, during which time the unemployment rate rose from 4.2 per cent to 5.8 per cent. Furthermore, it seems most unlikely that the rule would produce greater instability than the policy actually fol lowed. Actual policy has, as measured by the money stock, been most expansionary during the early and middle stages of business cycle expansions and most contractionary during the last stages of business expansions and early stages of business contractions. Unless a very improbable lag structure exists, the rule would surely be more stabilizing than the actual his torical pattern of monetary growth. SELECTION AND CONTROL OF A MONETARY AGGREGATE. The analysis in this study is almost entirely in terms of the narrowly defined money stock. The reasons for using the narrowly defined money stock as op posed to other monetary aggregates may be stated fairly simply. Some economists favor the use of bank credit as the monetary aggregate because they view polic]' as operating through changes in the cost and availability of credit. The major difficulty with this view is that there is no un ambiguous way of defining the amount of credit in the economy. And even if a satisfac tory definition could be worked out, there is no current possibility of obtaining timely data on the total amount of credit or of controlling the total amount. The definitional problem arises largely from the activities of financial intermediaries. Sup pose, for example, that an individual sells some corporate debentures and invests the proceeds in a fixed-income type of investment fund, which in turn uses the funds to buy the very same debentures sold by the individual. If both the debentures and the investment fund shares are counted as part of total credit, then in this example total credit has risen without any additional funds being made available to the corporation to finance new facilities and so forth. As another example, it is difficult to see that it would make any substantial difference to ag gregate economic activity whether a corpora tion financed inventories through sales of com mercial paper to the public or through borrowing from banks that raised funds through sales of CD’s to the public. Since there are numerous close substitutes for bank credit, the amount of bank credit is most un likely to be an appropriate figure to empha size. Furthermore, since bank credit is only a small part of total credit there is essentially no possibility of controlling total credit, however defined, through adjustments in bank credit. Ultimately the issue again becomes that of the stability of various functions. If the de mand and supply functions for all of the var ious credit instruments, including those of fin ancial intermediaries, were stable and were known, then it would be possible to focus on any aggregate that was convenient. For if all the functions were known, then there would be known relationships among various credit in struments, the money stock, and stocks and flows of goods. But the demand and sup ply functions for the various credit instru ments are not known, and it is unlikely that they ever will be known with any degree of precision. There are two basic reasons for this state of affairs. The first, and less important, is that given the great degree of substitutability among credit instruments, substitutions are constantly taking place as a result of changes in regulations, including tax regulations. But second and more important, individual credit instruments are greatly influenced by changes in tastes and technology, factors that econo mists do not understand well. As an example of the effects of regulations, consider the substitution in recent years of de bentures for preferred stock as a result of the tax laws permitting deduction of interest. As examples of the effects of changes in tastes and technology, consider the inventions of new instruments such as CD’s and the shares in dual-purpose investment funds. Furthermore, the relationships among credit instruments will change as attitudes toward risk change due to numerous factors including perhaps fading memories of the last recession or depression. Money viewed as the medium of exchange RULES-OF-THUMB FOR POLICY seems to be substantially less subject to changes in tastes and technology than do other financial assets. Of course, money is not im mune to these problems, as shown by the un certainty presently existing over the impact of credit cards. But a great deal of empirical work on money has been completed and the major findings have been substantiated by a number of different investigators. And the in terpretation of the empirical findings is usually clear because the empirical work has been conducted within the framework of a welldeveloped theory of money. There is, on the other hand, no satisfactory theory of bank credit to guide empirical work and to permit interpretation of the significance of empirical findings. For these reasons, and others, bank credit does not appear to be an appropriate monetary aggregate for policy to control. However, be cause bank credit and the money stock were so highly correlated in the past, it must be ad mitted that it probably would not have made much difference which one was used. From re cent experience, however, it appears that changes in banks’ nondeposit sources of funds are likely to become more, rather than less, important, and so in the future the correlation between money and bank credit is likely to be lower than in the past. If this prediction is cor rect, then the issue is a significant one. As a monetary aggregate, to be used for policy adjustments, the money stock has clear advantages over the monetary base and var ious reserve measures. These aggregates are al most always examined in adjusted form, where the adjustments allow for such factors as changes in the currency/deposit ratio, in re serve requirements, and in shifts between time and demand deposits. The adjustments are made because the effects of these various fac tors are understood and are thought to be worth offsetting. The adjustments have the ef fect of making the base an almost constant fraction of the money stock, or making total reserves an almost constant fraction of demand deposits. It obviously makes more sense to look directly at the money stock, especially since given the nature of the adjustments it is no easier to control the adjusted base or ad justed total reserves than to control the money stock. The final aggregate to be considered is the broadly defined money stock—the narrow stock plus time deposits. No strong case can be made against the broad money stock. From existing empirical work both definitions of money appear to work equally well. The theo retical distinction between demand deposits and passbook savings deposits depends on the costs of transferring between the two types of deposits, and these costs appear to be quite low. However, CD’s do appear to be theoreti cally different and probably should be ex cluded from the definition of money. The major reason for excluding all time deposits from the definition is that in the future banks may invent new instruments that will be classi fied as time deposits for regulatory purposes but for which the matter of definition as money may not be at all clear. The issue of controllability is a technical one and need not be discussed carefully in this summary. However, two conclusions may be stated. First, instructions from the FOMC to the Manager of the Open Market Account should take the form of a specified average weekly change in the money stock over the pe riod between FOMC meetings. Such an in struction must be distinguished from one in terms of the average level of the money stock over the period between FOMC meetings. The average-level specification has several technical difficulties and should be avoided. The second conclusion is that it is possible to control the rate of growth of the money stock over a 3-month period in a range of 1 per cent on either side of a desired rate of growth. This conclusion is based on an analy sis of monthly changes in the money stock over the 1951-68 period, a period during which little or no attention was paid to stabi lizing monetary growth, and it takes the histor ical record at face value. Assuming that efforts to control the money stock would in fact suc ceed in part rather than make money growth less stable than in the past, the estimate of plus or minus 1 per cent is an upper limit to the errors in controlling the growth rate of money over 6-month periods. CONCLUDING REMARKS. The orienta tion throughout this study has been the re direction of monetary policy on the basis of currently available theory and evidence. The recommendations are not utopian; in the au thor’s view they are supported by current knowledge and are operationally feasible. The approach has been in terms of what ought to be done in the near future, rather than in terms of what might be done eventually if enough in formation accumulates. No effort has been made to slide over gaps in our knowledge; rather, the emphasis has been on how policy should be formed given the huge gaps in our knowledge. Indeed, it is precisely these gaps in our knowledge that lead to the conclusion favoring policy adjustments through the money stock. It is the contention of this study that policy can be improved if there is explicit recognition of the importance of uncertainty. As much at tention should be given to the consequences of errors in projections as to the projections themselves. Policy may be improved more by “don’t know” answers to questions than by projections believed by no one. This is the static view. If policy can be im proved now through greater attention to uncer tainty, in the long run it can be improved fur ther only through a reduction in uncertainty. This longer view underlies the proposal for a policy rule-of-thumb. Policy successes and failures ought to be incorporated into a policy design in a form that will repeat the successes and prevent the recurrence of the failures. Policy-making will always require judgment, but the judgment will be applied to changing problems at a moving frontier of knowledge. A systematic formulation of policy will speed the accumulation of knowledge so that the policy problems of today will become the technical staff problems of tomorrow. RULES-OF-THUMB FOR POLICY REFERENCES Books 1. Fels, Rendigs, and Hinshaw, C. Elton, Fore casting and Recognizing Business C ycle Turning Points . New York: National Bureau of Economic Research, 1968. 2. Friedman, Milton, and Meiselman, David. “The Relative Stability of Monetary Ve locity and the Investment Multiplier in the United States, 1897-1958.” Commission on Money and Credit, Stabilization Poli cies. Englewood Cliffs, N.J.: PrenticeHall, Inc., 1963. 3. Holmes, Alan R. “Operational Constraints on the Stabilization of Money Supply Growth,” Controlling M onetary A ggre gates. Boston: Federal Reserve Bank of Boston, 1969. 4. Laidler, David E. W. The D em and fo r M oney: Theories and E vidence. Scranton, Pa.: International Textbook Company, 1969. 5. Mincer, Jacob (ed.). Econom ic Forecasting and E xpectations . New York: National Bureau of Economic Research, 1969. 6. Reynolds, Lloyd G. E conom ics . 3rd ed. Homewood, 111.: Richard D. Irwin, Inc., 1969. 7. Samuelson, Paul A. E conom ics , 7th ed. New York: McGraw-Hill, 1967. 8. Theil, Henri. O ptim al D ecision R ules for G overnm ent and Industry . Amsterdam: North-Holland Publishing Company, 1964. Periodicals and Other 9. Andersen, Leonall C., and Jordan, Jerry L. “Monetary and Fiscal Actions: A Test of Their Relative Importance in Economic Stabilization.” R eview , Federal Reserve Bank of St. Louis (Nov. 1968), pp. 11-24. 10. Ando, Albert, and Modigiliani, Franco. “The Relative Stability of Monetary Velocity and the Investment Multiplier,” A m erican E conom ic R eview , Vol. 55 (Sept. 1965), pp. 693-728. 11. . “Rejoinder,” A m erican E conom ic R e view, Vol. 55 (Sept. 1965), pp. 786-90. 188 12. Brainard, William. “Uncertainty and the Effectiveness o f Policy,” A m erican E co nom ic R eview : Papers and Proceedings o f the 79th Annual M eeting o f the A m erican E conom ic A ssociation, Vol. 57 (M ay 1967), pp. 4 1 1 -2 5 . 13. Brunner, Karl, and Meltzer, Allan H. “The Federal Reserve’s Attachment to the Free Reserve Concept.” Prepared for the Sub comm ittee on D om estic Finance, Banking and Currency Committee, H ouse of Rep resentatives, 88th Cong., 2d sess. Wash ington, D.C.: U .S. Government Printing Office, 1964. 14. de Leeuw, Frank, and Gramlich, Edward. “The Federal Reserve— M IT Econometric M odel,” Federal Reserve B ulletin, Vol. 54 (Jan. 1968), pp. 1 1-40. 15. DePrano, M ichael, and Mayer, Thomas. “Tests o f the Relative Importance o f A u tonom ous Expenditures and M oney,” A m erican E conom ic R e vie w , Vol. 55 (Sept. 1965), pp. 7 2 9 -5 2 . 16. . “Rejoinder,” A m erican E conom ic R e view , Vol. 55 (Sept. 1 9 6 5 ), pp. 7 9 1 -9 2 . 17. Friedman, Milton, and Meiselman, David. “Reply to A ndo and M odigliani and to DePrano and M ayer,” A m erican E co nom ic R eview , V ol. 55 (Sept. 1 9 6 5 ), pp. 7 5 3 -8 5 . 18. . “Reply to D onald Hester,” R eview o f E conom ics and Statistics, V ol. 46 (N ov. 19 6 4 ), pp. 3 6 9 -7 6 . 19. Hester, Donald D . “Keynes and the Quan tity Theory: A Com m ent on the Friedman-M eiselman CMC Paper,” R eview o f E conom ics and Statistics, Vol. 46 (N ov. 1964), pp. 3 6 4 -6 8 . 20. . “Rejoinder,” R eview o f E conom ics and Statistics, V ol. 46 (N ov. 19 6 4 ), pp. 3 7 6 -7 7 . 21. Holt, Charles C. “Linear D ecision Rules for Econom ic Stabilization and Growth,” Quarterly Journal o f E conom ics, Vol. 76 (Feb. 1962), pp. 2 0 -4 5 . 22. Latane, Henry A . “Cash Balances and the Interest Rate— A Pragmatic Approach,” R eview o f E conom ics and Statistics, Vol. 36 (N ov. 1954), pp. 4 5 6 -6 0 . Periodicals and other (Cont.) RULES-OF-THUMB FOR POLICY Periodicals and other (Cont«) 23. -------. “Income Velocity and Interest Rates— A Pragmatic Approach,” R eview o f E co nomics and Statistics, Vol. 42 (N ov. 1 960), pp. 4 4 5 -4 9 . 24. Moore, Geoffrey H ., and Shiskin, Julius. In dicators o f Business Expansions and Con tractions, Occasional Paper 103. N ew York: National Bureau of Econom ic Re* search, 1967. 25. Poole, William. “Optimal Choice of M one tary Policy Instruments in a Simple Sto chastic Macro M odel,” Quarterly Journal o f Economics, Vol. 84 (M ay 1 970), pp. 197-216. 26. Zarnowitz, Victor. A n A ppraisal o f ShortTerm E conom ic Forecasting, Occasional Paper 104. N ew York: National Bureau o f Econom ic Research, 1967. 189 By Stephen H. Axilrod APPENDIX: MONETARY AGGREGATES AND MONEY MARKET CONDITIONS IN OPEN MARKET POLICY CONTENTS 195 DIRECTIVES OF THE FOMC 198 POLICY FORMATION 200 ROLE OF MONETARY AGGREGATES 206 DAY-TO-DAY OPEN MARKET OPERATIONS 210 APPENDIX: Current Economic Policy Directives Issued by the FOMC This appendix is reprinted from the Federal Reserve B ulletin, February 1971, pages 79-104. Monetary Aggregates and Money Market Conditions in Open Market Policy THERE HAS BEEN WIDESPREAD discussion over the past year or so about the emphasis given to monetary and credit aggregates, as compared with traditional operating variables such as money market conditions, in the formulation and conduct of the Fed eral Reserve System’s open market policy. This article discusses the role— in the decision-making process of the Federal Open M arket Committee (F O M C )1 and in the day-to-day conduct of Federal Reserve open market operations— of aggregates such as the money supply and bank credit, in comparison with other financial variables. Such aggregates, of course, represent only a few of the many financial variables, including interest rates and credit flows through nonbank institutions and the market di rectly, that are evaluated in monetary policy decisions and their implementation. And financial conditions as a whole are eval uated against the underlying purpose of monetary policy— the encouragement of a healthily functioning economy, both do mestically and in relation to the rest of the world. The policy decisions of the FOMC are based on a full-scale evaluation by Committee members of likely tendencies in critical 1 The Federal Open Market Committee is the statutory body responsible for open market operations (purchase and sale of U.S. Government securities in the open market), the most flexible and frequently used instrument by which monetary policy affects bank reserves, bank credit, money supply, and ulti mately over-all credit conditions. The FOMC consists of the seven members of the Board of Governors of the Federal Reserve System, the President of the Federal Reserve Bank of New York, and four of the remaining 11 Reserve Bank presidents serving in rotation. The Chairman of the Board of Governors has traditionally been elected by the Committee to serve as Chairman of the Open Market Committee, and the President of the Federal Reserve Bank of New York has traditionally been elected Vice-Chairman. measures of economic performance such as output, employment, prices, and the balance of payments. In deciding on the stance of monetary policy, the Committee considers whether these ten dencies in domestic economic activity and the balance of pay ments appear desirable, and if not, how they might be influ enced by changes in financial conditions— including the pace of monetary expansion, credit availability, interest rates— and by expectational factors. Once a general policy stance is adopted, guidelines are set for the day-to-day conduct of operations in the open market. During 1970 somewhat more emphasis was placed on the behavior of monetary aggregates— such as the money supply and bank credit— in providing guidance for the day-to-day conduct of open market operations. Since it has always been recognized that the effect of mone tary policy stems from its influence on bank credit, money, in terest rates, and financial flows generally, the greater emphasis placed on monetary aggregates basically represented a modifica tion of operating procedures rather than a change in the funda mental objective of policy. Under conditions of uncertainty— for example, uncertainty about the impact on interest rates of expectational factors or about the strength of future demands for goods and services— some emphasis on the aggregates helps to guard against the risk that open market operations might in the end supply either too large or too small amounts of bank reserves, credit, and money as a result of unexpected and un desired shifts in demands for goods and services and for credit. At the same time, however, an approach that utilizes aggre gates as one operating guide must take account of shifts in the demand for money and liquidity at given levels of income. Such shifts would have to be accommodated through open m arket operations in order to help provide the money and liquidity de manded if interest rates and credit conditions generally were not to become unduly tight or easy. Thus, the longer-run path for monetary aggregates needs to be evaluated in relation to emerg ing credit conditions and tendencies in economic activity, to help determine if demands for liquidity have been properly assessed. And whatever longer-run path for the aggregates may be in cluded as guidance for open market operations, short-run, selfcorrecting variations in money and credit demands need to be ac commodated in order to avoid inducing unnecessary, and possibly destabilizing, fluctuations in money market conditions. In practice, allowance has to be made— in the formulation of MONETARY AGGREGATES AND MARKET CONDITIONS monetary policy and in the guides to the conduct of policy— for uncertainties with respect to both the demand for goods and the demand for money and liquidity. And trends in monetary aggregates, interest rates, and other financial variables have to be evaluated in relation to the continuing flow of evidence as to the likely course of economic activity. DIRECTIVES OF THE FOMC The monetary policy decisions of the FOM C— which in recent years has generally met about every 4 weeks— are embodied in the Committee’s current economic policy directive, voted on near the end of each meeting. This directive is issued to the Federal Reserve Bank of New York, which, because it is located in the Nation’s central money and credit market, under takes open market operations for the Federal Reserve System. The directive is carried out by a senior officer of the Bank, who is designated by the FOMC as Manager of the System Open Market Account. Both the form and the content of the FOMC directive have changed over the years. Since 1961 the directive has contained two paragraphs. The first paragraph has contained statements about recent key economic and financial developments, and also a general statement of current goals of the FOMC with respect to economic growth, price stability, and the balance of pay ments.2 The second paragraph contains the FOM C’s instructions to the Account Manager for guiding open market operations in the interval between FOMC meetings. The second paragraph is, - For illustrative purposes the first paragraph of the directive issued on Dec. 16, 1969, is quoted below: ‘The information reviewed at this meeting indicates that real economic activity has expanded only moderately in recent quarters and that a further slowing of growth appears to be in process. Prices and costs, however, are continuing to rise at a rapid pace. Most market interest rates have advanced further in recent weeks partly as a result of expectational factors, including concern about the outlook for fiscal policy. Bank credit rose rapidly in November after declining on average in October, while the money supply increased moderately over the 2-month period; in the third quarter, bank credit had declined on balance and the money supply was about unchanged. The net contraction of outstanding large-denomination CD’s has slowed markedly since late summer, apparently reflecting mainly an increase in foreign official time deposits. However, flows of consumer-type time and savings funds at banks and nonbank thrift institutions have remained weak, and there is con siderable market concern about the potential size of net outflows expected around the year-end. In November the balance of payments deficit on the liquidity basis diminished further and the official settlements balance reverted to surplus, mainly as a result of return flows out of the German mark and renewed borrowing by U.S. banks from their foreign branches. In light of the foregoing developments, it is the policy of the Federal Open Market Commit tee to foster financial conditions conducive to the reduction of inflationary pressures, with a view to encouraging sustainable economic growth and attain* ing reasonable equilibrium in the country’s balance of payments/’ 195 in essence, a highly condensed summary of the Committee s dis cussion and conclusions as to the sort of operations that will be required to reach its longer-run policy goals. These directives are made public after a 3-month lag in a “record of policy actions,” which also includes a resume of prevailing economic and financial conditions and of the Committee’s discussion of policy implications at the meeting. The nature of the operating instructions in the second para graph of the directive has changed from time to time. Money market conditions have remained as important guides in de termining day-to-day open market activity. Though emphasis on various money market indicators has varied over the years in light of changing economic and financial circumstances, money market conditions have generally been construed to include member bank borrowings at the Federal Reserve discount win dow, the net reserve position of member banks (excess reserves of banks less borrowings from the Federal Reserve), the interest rate on Federal funds (essentially reserve balances of banks that are made available to other banks, usually on an overnight basis), and at times the 3-month Treasury bill rate. At times when it was framing the operating instructions con tained in the second paragraph of its directive solely in terms of money market conditions, the FOMC was nevertheless con cerned with developments in monetary aggregates and financial conditions generally as they affect the broad objectives of policy. Beginning in 1966, the Committee supplemented the reference to money market conditions in the second paragraph with a reference to certain monetary aggregates, such as bank credit, and later the money supply.3 The desired behavior of aggregates has been given increased emphasis since early 1970. From mid-1966 through 1969 the reference to aggregates was generally to bank credit and was contained in a so-called proviso clause. The second paragraph of the directive issued on December 16,1969, is illustrative: “To implement this policy, System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining the prevailing firm conditions in the money market; provided, however, that operations shall be modified if bank credit appears to be deviating significantly from current projections or if unusual liquidity pressures should develop.” was also occasional reference to such aggregates in directives during the first half of the 1960’s. MONETARY AGGREGATES AND MARKET CONDITIONS In 1970 monetary aggregates came to play a more prominent role in the phrasing of the second paragraph, and references were made to the money supply as well as to bank credit. In the directive issued on March 10, 1970, the Committee stated more directly its desires with respect to the aggregates rather than referring to them in the form of a proviso clause. The second paragraph of the directive of that date read as follows: “To implement this policy, the Committee desires to see moderate growth in money and bank credit over the months ahead. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining money market conditions consistent with that objective.” The operating instructions in the second paragraphs of FOMC directives are not confined to money market conditions and a de sired pattern of behavior in the monetary aggregates. The Sys tem Account M anager has also been directed, when appropriate, to take account of Treasury financings, liquidity pressures, and the possible impacts of bank regulatory changes in the process of achieving satisfactory conditions in the money market and satisfactory performance of monetary aggregates. As the nature of economic and financial problems has altered, so has the phrasing of the second paragraph of the directive. For instance, the second paragraph of the directive issued on May 26, 1970, emphasized the need to moderate pressures on financial markets; it read as follows: “To implement this policy, in view of current m arket uncer tainties and liquidity strains, open market operations until the next meeting of the Committee shall be conducted with a view to moderating pressures on financial markets, while, to the ex tent compatible therewith, maintaining bank reserves and money market conditions consistent with the Committee’s longer-run objectives of moderate growth in money and bank credit.” The short-run bulge in bank credit expansion expected to re sult from the Board’s action around midyear in suspending ceil ings on maximum interest rates payable by banks on large cer tificates of deposit in the 30- to 89-day maturity range was taken into consideration in the phrasing of the second paragraph of the directive issued by the FOM C on July 21,1970: “To implement this policy, while taking account of persisting market uncertainties, liquidity strains, and the forthcoming Treasury financing, the Committee seeks to promote moderate growth in money and bank credit over the months ahead, allow ing for a possible continued shift of credit flows from m arket to banking channels. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consist ent with that objective; provided, however, that operations shall be modified as needed to counter excessive pressures in financial markets should they develop.” And in the directive issued on August 18, 1970, an easing of conditions in credit markets was taken as an objective of open market operations parallel with desires with respect to m onetary aggregates, as follows: “To implement this policy, the Committee seeks to promote some easing of conditions in credit markets and somewhat greater growth in money over the months ahead than occurred in the second quarter, while taking account of possible liquidity prob lems and allowing bank credit growth to reflect any continued shift of credit flows from market to banking channels. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consistent with that objective, taking account of the effects of other monetary policy actions.” The first and second paragraphs of all directives issued from December 16, 1969, through December 15, 1970, are shown in the appendix to indicate the variety of considerations that the FOMC takes into account in formulating its policy and framing its operating instructions. FORMATION The FOMC s basic concern is with the real economy— produc tion, employment, prices, and the balance of payments. But the Committee must translate its broader economic goals into the monetary and credit variables over which the Federal Reserve has a direct influence. Thus, whatever emphasis is given to the financial variables that influence day-to-day open m arket opera tions, it is recognized that the immediate targets of day-to-day operations are not the goals of monetary policy, but rather that those targets are set with a view to facilitating the achieve ment of the broader financial and economic objectives of the FOMC. In setting its immediate operating targets, the FO M C neces sarily reviews past and prospective relationships between finan cial conditions and economic objectives. A benchmark in this MONETARY AGGREGATES AND MARKET CONDITIONS review is provided several times a year in a presentation by the staff to the Committee of an interrelated set of longer-run eco nomic and financial projections. These exercises review in detail recent economic and financial developments, assess the outlook for and impact of fiscal policy, and trace the likely patterns of change in such measures as income, output, employment, prices, and the balance of payments for a period of about a year ahead. Provisional estimates are also presented of the flow of funds— including various monetary aggregates— and interest rates ex pected to be consistent with these patterns of economic develop ment. A reappraisal of current tendencies in and prospects for economic activity, financial flows and credit market conditions, and the balance of payments is presented to the FOM C by the staff on the occasion of each meeting. Included in the regular documentation is an analysis of relationships among money market variables, paths for monetary aggregates, and interest rates broadly considered for a period several months ahead. At each FOMC meeting, most of the time is given over to a free interchange of views by Committee members of their assessment of the current economic situation and outlook and of the related appropriate monetary policies. As the discussion proceeds, each Committee member indicates his assessment of the basic tendencies in economic activity, prices, employment, and so forth; his appraisal of recent financial developments in relation to desired economic goals; and what steps might be taken through open market operations (or other policy instruments that interact with open market operations) to help achieve financial conditions suitable to economic goals. It may develop, for instance, that most or all Committee mem bers believe that economic prospects are deviating from those that had previously been expected and desired. If so, the Com mittee may wish to modify its objectives concerning money market conditions and desired rates of expansion in monetary and credit aggregates, so as to promote over-all financial and credit conditions that are more conducive to desired economic conditions. O r it may turn out that economic activity is develop ing about in line with expectations but seems to be entailing a pattern of financial flows different from that originally expected. Still another possibility is that the relationship that is developing between the variables specified for the System Account M anager for purposes of guiding day-to-day open m arket operations and )0 broader financial flows and interest rates is not what was ex pected. Under any of such circumstances, the FOMC could react by changing its operating instructions. The operating instructions in the second paragraph of the directive are expressed qualitatively. But the specific variables involved— money market conditions and monetary and credit aggregates— are typically indicated in terms of ranges in the discussion. Over the past year the operating instructions embodying the Committee’s policy thrust have changed in two general ways. First, as has been noted, somewhat more emphasis has been placed on monetary aggregates as a target for open market opera tions rather than as an outgrowth of such operations. Second, the time horizon for a path of monetary and credit aggregates (in relation to money market conditions and other financial variables) has been viewed as encompassing several months or, expressed in calendar quarters, at least one or two quarters ahead. Longer-run paths provide the Committee with a means for focusing on the emerging trend of growth in the money supply or in bank credit, while recognizing that, over very short-run periods of a week or a month or so, there may be irregular movements in rates of change in monetary aggregates because of erratic shifts in the public’s demand for deposits and such factors as Treasury financings, a large change in U.S. Government deposits, or move ments of funds between the U.S. and foreign countries. ROLE OF MONETARY AGGREGATES The somewhat greater use of monetary aggregates in the formulation and conduct of open market policy during the past year represents for the most part an extension of the trend of policy over the previous several years. It has always been recognized that monetary policy achieves its effects through its influence on bank credit, money supply, interest rates, and financial flows generally. But the benefits that might be expected from an in creased degree of emphasis on monetary aggregates in the con duct of open market operations relate to the question of mone tary control under conditions of uncertainty. Greater emphasis on aggregates is consistent with a variety of economic theories, and it does not necessarily imply any par ticular judgment as to the importance for the economy of mone tary flows relative to interest rates and credit conditions or relative to other influences such as fiscal policy and technological innovation. Operationally, however, by placing more emphasis MONETARY AGGREGATES AND MARKET CONDITIONS on monetary aggregates in the instructions to the Account Manager, the FOM C has a greater assurance that unexpected and undesired shortfalls or excesses in the demands for goods and services in the economy, and hence in the demands for credit and money, will not lead more or less automatically to too little or too much expansion in bank reserves, bank credit, and money. Giving more weight to monetary aggregates means, for example, that if there were an unexpected and undesired short fall in business and consumer demand for goods and services, the Federal Reserve would continue to provide reserves to try to keep growth in money and bank credit from weakening unduly at a time when the public, with transactions demand for cash reduced, was seeking to invest excess funds in various financial assets. In the process, there would be a greater short-run decline in interest rates than would otherwise be the case. The drop in interest rates and the easing of credit conditions would help to provide financial incentives that would encourage a strengthen ing of demands for goods and services. While increasing the emphasis on monetary and credit aggre gates tends to increase the protection against undesired shifts in demands for goods and services, it at the same time runs the risk of reducing protection against unexpected shifts in the pub lic’s demand for cash and liquidity. Thus, for example, if the public decides to hold more liquidity relative to income than had been earlier assumed, failure to permit a faster rise in the money supply to accommodate this desire would lead to higher interest rates and tighter credit conditions as the public seeks to sell other assets to acquire cash. The tightening of credit conditions would tend to lead to a weaker GNP than desired. In contrast, the tendency toward tighter conditions could be averted if the Federal Reserve helped to meet the desire for greater liquidity by increasing its purchases of financial assets (through open market acquisitions of U.S. Government securities)— thereby providing more bank reserves to support an increase in bank deposits and in the money supply and to keep interest rates from rising. In practice, allowance has to be made for uncertainties about both the demand for goods and services and the demand for money and liquidity. Opinions differ among professional econo mists as to the relative degrees of stability of these types of demand, and practical experience over the past several years suggests that there is a good deal of variation in both. There have been periods when large increases in Federal Government 201 purchases of goods and services and/or in private sector de mands for capital goods and inventories have caused marked shifts in over-all demands for goods and services at given inter est rates. But there have also been periods when liquidity strains, greatly increased financial transactions, and various international uncertainties have resulted in a sizable upward shift in the demand for cash and closely related assets at given interest rates. Furthermore, open market policy not only needs to dis tinguish between, and take account of, shifts in both the demand for goods and services and the demand for money and liquidity at given interest rates, but also must evaluate the extent to which such shifts are transitory or more permanent. The late spring and the summer of 1970 are an example of a period when liquidity strains in the economy— typified by rising long-term interest rates at a time when economic activity was sluggish, by the bankruptcy of a major railroad, and by a gen erally cautionary attitude on the part of investors toward securi ties, particularly commercial paper— were giving rise to consider able uncertainty and were threatening a marked erosion in confidence. Under those circumstances Federal Reserve policy stressed the need to moderate pressures on financial markets and to accommodate liquidity needs. In late June the Board of Governors suspended maximum ceiling rates on large CD’s maturing in 30- to 89-days as part of the effort to reliquify the economy. This action made it possible for banks to compete for funds and to accommodate borrowers who were not able, in the conditions of the time, to refinance their borrowings in the commercial paper market, or were not able to do so without a bank loan commitment as back-up. And open market operations during the period were conducted in such a way as to provide the reserves to sustain the very large increase in bank credit resulting from renewed ability of banks to obtain funds through issuance of certain large CD’s. The FO M C’s policy directives in that period (see directives of M ay 26 and July 21, 1970, on pp. 99 and 100) tended to subordinate, tem porarily, longer-run objectives for monetary aggregates to the shorter-run liquidity needs of the economy. In general, in evaluating the appropriateness of particular operating guidelines at a particular time, the FO M C has to make judgments about the nature of the fundamental influences that are affecting the domestic economy and the international posi tion of the dollar. If, for example, it developed that interest rates MONETARY AGGREGATES AND MARKET CONDITIONS were higher, and over-all credit conditions tighter, than expected for a given rate of increase in bank credit or money, the FOM C would have to make a judgment as to whether GNP was stronger than anticipated, whether inflationary expectations were affect ing interest rates, or whether the demands for money and closely related assets had shifted at given levels of income and interest rates. Or, as another example, interest rate movements might be undesirably affecting capital flows between the United States and foreign countries; in this case judgments might have to be made as to how the various policy instruments could be adapted to such a development. Judgments made with respect to interrelationships among policy objectives would affect not only the open m arket policy instrument but also other monetary policy instruments. With respect to open market policy, types of adjustments called for in operating instructions would include, for instance, whether to change the targets for aggregates an d /o r whether to put more stress on money or credit m arket conditions. Or adjustments might be called for in other policy instruments— such as the discount rate or reserve requirements, including provisions such as those recently made affecting Euro-dollar borrowings of U.S. banks— in order to achieve a variety of policy objectives more effectively. In looking toward a desired longer-run growth rate in mone tary aggregates, the FOM C has focused on money and bank credit in its operating instructions. The concept of money used for these purposes has generally been the so-called narrowly defined money supply— currency in circulation outside the bank ing system plus demand deposits other than U.S. Government and domestic interbank deposits— but broader definitions have also been taken into account. The determination of what rates of growth may be desired for money takes into account not only what is happening in credit markets but also the rates of growth in certain types of assets held by the public that are closely related to narrowly defined money and that the public holds as a store of value and as a source of immediate liquidity. A number of broader concepts of the money supply and of liquidity have been utilized by economic analyst? in relating money supply to economic activity. These include, in addition to the narrowly defined money supply, a concept— here termed M ,— that adds time and savings deposits other than large CD’s at commercial banks to narrowly defined money; and a concept, 203 termed Ms, that adds deposits at both mutual savings banks and savings and loan associations. And even these concepts can be broadened by adding other money-like assets, such as large marketable negotiable CD’s issued by banks and other short term marketable securities. Annual, quarterly, and monthly rates of change over the past year in the three concepts of money noted above are shown in the table below. VARIOUS MEASURES OF MONEY: RATES OF CHANGE Seasonally adjusted annual rates, in per cent Period Mi (Currency plus demand deposits1) m2 (M i plus coml. bank time deposits other than large CD’s) Mi (M 2 plus deposits at S&L’s and mutual savings banks) 1969....................... 1970....................... 3.1 5.4 2.4 8.2 2.8 7.9 1970—Q1............... Q2............... Q3............... Q4............... 5.9 5.8 6.1 3.4 3.4 8.4 11.0 9.2 2.6 7.9 10.5 9.7 1970—January___ February. . . March........ April........... May............ June............ July............. August........ September.. October___ November.. December. . 9.4 “ 4.1 12.3 9.9 5.2 2.3 5.7 6.8 5.7 1.1 2.8 6.2 2.2 - 1 .5 9.6 10.8 7.6 6.7 9.9 12.5 10.3 7.3 7.0 13.0 1.2 - 1 .2 7.8 9.7 7.2 6.6 9.9 11.4 10.0 8.1 8.1 12.6 i D em and deposits other than interbank and U .S . G ovt. N o t e . — M onthly rates o f change based o n the daily-average levels o u tsta n d in g . Q uarterly and a n n u a l rates o f changes m easured from daily*average levels outstand ing in en d -of-p eriod m o n th s. As may be seen, the rates of change for the various measures may diverge noticeably, and they may show a high degree of fluctuation over the short run. Differential tendencies in the various measures of money and liquidity have been the result in large part of sharp shifts of funds by the public between deposits and market securities when market interest rates moved above and then back below ceiling rates on deposits at banks and thrift institutions. But divergent movements, particularly in the short run, may develop even when ceiling rates are not a disturbing element. This highlights the need to evaluate a variety of money and liquidity measures, among other things, in gauging the impact of monetary policy on the economy. Moreover, the MONETARY AGGREGATES AND MARKET CONDITIONS relatively large month-to-month variations in growth for any particular money measure— and variations are even larger from week to week— emphasize the need to evaluate data over some long period of time in judging the underlying tendency of the series. As noted earlier, in addition to the money supply, the second paragraph of the directive has emphasized bank credit. A current measure of bank credit for the guidance of the Account Manager was provided by measuring bank credit from the liability side, since liability data are available more quickly and can be used to construct a series on a daily-average basis. This daily-average measure does not encompass all bank liabilities (it excludes non member bank deposits and bank capital, for example) but it includes the most volatile ones. It encompasses not only the member bank component of deposits included in M s above, but also funds obtained by banks through large time CD’s, U.S. Government deposits, and interbank deposits and through non deposit sources such as Euro-dollars and commercial paper issued by bank-related affiliates. The sum of these deposits and nondeposit sources is called the adjusted credit proxy. Inclusion of bank credit in the directive might be considered as recognition of a broader concept of money, since time and savings deposits at commercial banks are a key source of bank credit. In addition, however, the inclusion recognizes that bank credit is a key component of total credit availability and one that is immediately sensitive to open m arket operations. The amount of bank credit that the FO M C is willing to encourage or to countenance depends, like the money supply, on over-all economic and financial conditions. When, for exam ple, banks have been unable for an extended period to increase time and savings deposits because interest rate ceilings on time deposits were unrealistically low relative to market rates, it was to be expected that outstanding bank credit would grow rapidly for a time after ceiling rates again became competitive. This growth would represent mainly a shifting of credit flows from market to banking channels as banks sought to restore their previous competitive position and as the public restructured its financial asset portfolios to reflect the changed yield relation ships. Federal Reserve open m arket operations could provide the reserves necessary to sustain the shift in the public’s ability and willingness to hold time deposits relative to other assets. The accompanying chart shows monthly changes in bank credit, as CREDIT PROXY ADJUSTED; TOTAL RESERVES 206 BI L LI ONS OF DO LL AR S 1967______________ 1968_____________ 1969 1970 B ank "cred it proxy ad ju sted ” is to tal m em ber b an k deposits plus funds p ro v id e d by E u ro dollar borrow ings and b an k -related com m ercial pap er. T h ro u g h th e first h a lf of 1969, n o d ata on b ank-related com m ercial p aper w ere available, b u t am o u n ts o u tsta n d in g were n o t thought to be grow ing significantly in those p eriods. measured by the adjusted credit proxy, along with total bank reserves. DAY-TO -DAY OPEN M A R K E T O PER A TIO N S The day-to-day operations in the m arket by the System Account M anager have continued to be guided m ainly by money m arket conditions, in part because the inform ation that is available daily and continuously as to the state of the money m arket— for ex ample, the Federal funds rate and dealer loan rates— reflects the interaction of the dem and for and existing supply of bank reserves and hence provides a basis for m aking daily decisions as to whether the System should be in the m arket providing addi tional or absorbing existing reserves; and if so, by how m uch and through what means. But the degree to which the M anager seeks to influence money m arket conditions has been affected by the relationship that is presum ed to exist at any given time among money m arket conditions, reserves, and the m onetary aggregates and by the Com m ittee’s desires with respect to m one tary aggregates and over-all conditions in the credit m arket. Changes in money m arket conditions, of course, m ay reflect factors other than efforts to influence reserve flows in accordance MONETARY AGGREGATES AND MARKET CONDITIONS with longer-run targets for monetary aggregates. Some changes in money market conditions reflect no more than shifts in the distribution of reserves among banks. Others represent the shortrun effects of bulges in demand for day-to-day credit at times of Treasury financings or in tax payment periods. Yet others repre sent unanticipated, virtually random changes in technical factors — such as float or currency in circulation— that supply to or absorb from the market more reserves than was either expected or seemed likely to be sustained. And as in the summer of 1970, open m arket operations in relation to money market conditions may sometimes reflect primarily a concern with liquidity pres sures in the economy. Although recognizing that money market conditions are sub ject to a number of influences, the System Account Manager takes into consideration the relationship between money market conditions and the trends in bank credit and money that has pre vailed in the recent past and the relationship that is expected to develop in the future in making decisions concerning reserve provision or absorption through open market operations. At the beginning of a statement week, for example, his operations may be aimed at a condition of tightness or ease in the money market roughly similar to that of previous weeks. This would mean that such variables as the Federal funds rate, dealer loan rates, the net reserve position of member banks, and borrowings by member banks from the Federal Reserve would generally tend to fluctuate within the range of recent experience— although there might be special, sometimes unforeseen developments (such as a mail strike) that could cause marked short-run changes in money m arket conditions. If and as it becomes evident that monetary aggregates are running above or below the desired path, however, the Account M anager may aim at correspondingly tighter or easier money market conditions. Also, if it should turn out that the apparent new relationship was not long-lasting, the Account Manager would subsequently have to reverse the direction of his opera tions. Thus, to the extent that monetary aggregates are given more emphasis in the operating paragraph of the directive, money market conditions may be subject to a somewhat greater degree of fluctuation. While the counterpart of greater sensitivity to monetary aggre gates would be a somewhat greater tendency for actual money market conditions to change more frequently than otherwise, 207 208 sharp short-run shifts in money market conditions are not likely to develop, in part because the FOMC is concerned with the state of money and credit markets as well as with tendencies in monetary aggregates. There are a number of reasons for the continuing role of money market conditions as a day-to-day guide for open market operations. First, the money market reflects the pressure of demand for liquidity, and the nation’s central bank has a unique responsi bility for maintenance of orderly conditions in such a market. Second, there are large and often unpredictable week-to-week and month-to-month swings in the economy’s demand for money and bank credit. These demands are often self-correcting, and as a result there is little purpose in permitting the sharp fluctuations in money market conditions, and perhaps in credit markets gen erally, that would be likely to develop should the flow and ebb of these demands not be accommodated in Federal Reserve opera tions affecting bank reserves. Third, because of the key role of the money m arket in quickly reflecting shifts in the need for and availability of liquid funds, presumably in large part as a result of the interaction of the public’s spending decisions and monetary policy, sharp shifts in money market conditions may be interpreted by market par ticipants as a harbinger of relatively permanent changes in credit demand or monetary policy. Investors, businessmen, and con sumers may vary their credit outlook, and perhaps their economic outlook too, in response to the money market to the extent that they regard changes in the market as a signal of events to come. This prospect itself counsels caution in undertaking open market operations that lead to large short-run changes in money market conditions until it becomes fairly certain that longer-run ten dencies in money supply, bank credit, and over-all credit conditions require such changes. While there are reasons for emphasis on money m arket condi tions, it should be stressed that money market conditions are only instrumental to the attainment of the main financial ob jectives of policy— flows of monetary aggregates and over-all credit conditions— that are appropriate to achievement of over all economic goals. For the Account Manager, the day-to-day operations of the Account and the effect of these operations on the money market are made even more complex because he is aware that the FOM C generally has in mind not only some view concerning the desired longer-run trend in various monetary MONETARY AGGREGATES AND MARKET CONDITIONS aggregates but also a view concerning what should be sought in the way of associated credit conditions. These desires may sometimes turn out to be in conflict; for example, monetary aggregates as a group may be rising more rapidly than desirable while credit conditions may be tightening more than desirable. Meeting one desire by holding back on the provision of reserves in order to restrain growth in bank credit and money would tend, at least temporarily, to thwart the other desire by leading to even more tightening of credit conditions. Under such circumstances, the Account M anager would have to adjust his operations— thereby affecting day-to-day money market conditions— in line with the sense of priority among op erating objectives given by the FOMC. While the whole set of objectives would be reconsidered at the next FOM C meeting, the Account M anager’s operations are monitored daily through a morning telephone conference call. This call involves the Trading Desk in New York, senior officials on the staff of the Board of Governors in Washington, and one of the Reserve Bank Presidents (serving in rotation) who is a voting member of the FOM C (other than the President of the Federal Reserve Bank of New York). Individual Board members may also participate in the call from time to time, as may the President of the New York Reserve Bank. In this call the M anager explains his program for the day, and that program, or possible alternative approaches, are discussed. As part of this process, not only are current figures on bank reserve posi tions, money market conditions, and broader credit conditions reported, but also information on the latest deposit and bank credit figures and how these compare with FOM C desires is appraised. In general, as the FO M C’s objectives with respect to monetary aggregates, and also over-all credit conditions, have been given increased stress in the directive to the Account M anager, the timing and extent of the System’s day-to-day open m arket opera tions have, of course, been altered, with consequent effects on day-to-day money m arket conditions. At the same time, the M anager still takes account of the emerging tightness or ease in the money m arket as a factor affecting the timing and extent of day-to-day open m arket operations. But this emerging tightness or ease is evaluated against trends in money, bank credit, and over-all credit conditions, which are, and always have been, among the basic financial objectives of monetary policy. 209 210 APPENDIX: Current Economic Policy Directives Issued by the FOMC M eeting held on December 1 6 ,1 9 6 9 The information reviewed at this meeting indicates that real econom ic activity has expanded only moderately in recent quarters and that a further slowing o f growth appears to be in process. Prices and costs, however, are continuing to rise at a rapid pace. Most market interest rates have advanced further in recent weeks partly as a result of expectational factors, including concern about the outlook for fiscal policy. Bank credit rose rapidly in November after declining on average in October, while the money supply increased moderately over the 2-month period; in the third quarter, bank credit had declined on balance and the money supply was about un changed. The net contraction of outstanding large-denomination C D ’s has slowed markedly since late summer, apparently reflecting mainly an in crease in foreign official time deposits. However, flows o f consumer-type time and savings funds at banks and nonbank thrift institutions have remained weak, and there is considerable market concern about the poten tial size o f net outflows expected around the year-end. In N ovem ber the balance of payments deficit on the liquidity basis diminished further and the official settlements balance reverted to surplus, mainly as a result o f return flows out of the German mark and renewed borrowing by U.S. banks from their foreign branches. In light o f the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to the reduction of inflationary pressures, with a view to encouraging sustainable economic growth and attaining reasonable equi librium in the country’s balance of payments. To implement this policy, System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining the prevailing firm conditions in the money market; provided, however, that operations shall be modified if bank credit appears to be deviating significantly from current projections or if unusual liquidity pressures should develop. M eeting held on January 1 5 ,1 9 7 0 The information reviewed at this meeting suggests that real econom ic activity leveled off in the fourth quarter of 1969 and that little change is in prospect for the early part o f 1970. Prices and costs, however, are con tinuing to rise at a rapid pace. Most market interest rates have receded from highs reached during December. Bank credit and the money supply increased slightly on average in December and also over the fourth quarter as a whole. Outstanding large-denomination C D ’s held by domestic deposi tors have continued to contract in recent months while foreign official time deposits have expanded considerably. Flows o f consumer-type time and savings funds at banks and nonbank thrift institutions have remained weak, and there apparently were sizable net outflows after year-end interest cred iting. U.S. imports and exports have both grown further in recent months but through November the trade balance showed little or no further im provement from the third-quarter level. At the year-end the over-all balance o f payments statistics were buoyed by large temporary inflows o f U .S. cor porate funds. In light o f the foregoing developments, it is the policy o f the Federal Open Market Committee to foster financial conditions conducive MONETARY AGGREGATES AND MARKET CONDITIONS to the orderly reduction of inflationary pressures, with a view to encourag ing sustainable econom ic growth and attaining reasonable equilibrium in the country’s balance of payments. To implement this policy, while taking account of the forthcoming Treas ury refunding, possible bank regulatory changes and the Committee’s desire to see a modest growth in money and bank credit, System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining firm conditions in the money market; provided, how ever, that operations shall be modified if money and bank credit appear to be deviating significantly from current projections. Meeting held on February 1 0 ,1 9 7 0 The information reviewed at this meeting suggests that real economic activity, which leveled off in the fourth quarter of 1969, may be weakening further in early 1970, Prices and costs, however, are continuing to rise at a rapid pace. Long-term market interest rates recently have fluctuated under the competing influences of heavy demands for funds and shifts in investor attitudes regarding the outlook for monetary policy. Bank credit declined in January but the money supply increased substantially on aver age; both had risen slightly in the fourth quarter. Flows of time and savings funds at banks and nonbank thrift institutions have remained generally weak since year-end, and they apparently have been affected little thus far by the recent increases in maximum rates payable for such funds. The U.S. foreign trade balance improved somewhat in December, as imports fell off. The over-all balance of payments has been in substantial deficit in recent weeks. In light of the foregoing developments, it is the policy o f the Federal Open Market Committee to foster financial conditions conducive to the orderly reduction of inflationary pressures, with a view to encouraging sustainable economic growth and attaining reasonable equilibrium in the country’s balance o f payments. To implement this policy, while taking account o f the current Treasury refunding, possible bank regulatory changes and the Committee’s desire to see moderate growth in money and bank credit over the months ahead, System open market operations until the next meeting of the Committee shall be conducted with a view to moving gradually toward somewhat less firm conditions in the money market; provided, however, that operations shall be modified promptly to resist any tendency for m oney and bank credit to deviate significantly from a moderate growth pattern. M eeting held on March 1 0 ,1 9 7 0 The information reviewed at this meeting suggests that real economic activity, which leveled off in the fourth quarter of 1969, is weakening fur ther in early 1970. Prices and costs, however, are continuing to rise at a rapid pace. Market interest rates have declined considerably in recent weeks, partly as a result of changing investor attitudes regarding the outlook for econom ic activity and monetary policy. Both bank credit and the m oney supply declined on average in February, but both were tending upward in the latter part of the month. Outflows o f time and savings funds at banks and nonbank thrift institutions, which had been sizable in January, appar ently ceased in February, reflecting advances in rates offered on such funds following the recent increases in regulatory ceilings, together with declines in short-term market interest rates. The U .S. foreign trade surplus narrowed 210 APPENDIX: Current Economic Policy Directives Issued by the FOMC M eeting held on December 1 6 ,1 9 6 9 The information reviewed at this meeting indicates that real econom ic activity has expanded only moderately in recent quarters and that a further slowing o f growth appears to be in process. Prices and costs, however, are continuing to rise at a rapid pace. M ost market interest rates have advanced further in recent weeks partly as a result of expectational factors, including concern about the outlook for fiscal policy. Bank credit rose rapidly in November after declining on average in October, while the m oney supply increased moderately over the 2-month period; in the third quarter, bank credit had declined on balance and the money supply was about un changed. The net contraction o f outstanding large-denomination C D ’s has slowed markedly since late summer, apparently reflecting mainly an in crease in foreign official time deposits. However, flows o f consumer-type time and savings funds at banks and nonbank thrift institutions have remained weak, and there is considerable market concern about the poten tial size of net outflows expected around the year-end. In N ovem ber the balance o f payments deficit on the liquidity basis diminished further and the official settlements balance reverted to surplus, mainly as a result o f return flows out o f the German mark and renewed borrowing by U.S. banks from their foreign branches. In light o f the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to the reduction o f inflationary pressures, with a view to encouraging sustainable economic growth and attaining reasonable equi librium in the country’s balance of payments. To implement this policy, System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining the prevailing firm conditions in the money market; provided, however, that operations shall be modified if bank credit appears to be deviating significantly from current projections or if unusual liquidity pressures should develop. M eeting held on January 1 5 ,1 9 7 0 The information reviewed at this meeting suggests that real econom ic activity leveled off in the fourth quarter o f 1969 and that little change is in prospect for the early part of 1970. Prices and costs, however, are con tinuing to rise at a rapid pace. Most market interest rates have receded from highs reached during December. Bank credit and the m oney supply increased slightly on average in Decem ber and also over the fourth quarter as a whole. Outstanding large-denomination C D ’s held by domestic deposi tors have continued to contract in recent months while foreign official time deposits have expanded considerably. Flows o f consumer-type time and savings funds at banks and nonbank thrift institutions have remained weak, and there apparently were sizable net outflows after year-end interest cred iting. U.S. imports and exports have both grown further in recent months but through November the trade balance showed little or no further im provement from the third-quarter level. A t the year-end the over-all balance o f payments statistics were buoyed by large temporary inflows o f U .S. cor porate funds. In light o f the foregoing developments, it is the policy o f the Federal Open Market Committee to foster financial conditions conducive MONETARY AGGREGATES AND MARKET CONDITIONS to the orderly reduction o f inflationary pressures, with a view to encourag ing sustainable econom ic growth and attaining reasonable equilibrium in the country’s balance o f payments. To implement this policy, while taking account of the forthcoming Treas ury refunding, possible bank regulatory changes and the Committee’s desire to see a modest growth in money and bank credit, System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining firm conditions in the money market; provided, how ever, that operations shall be modified if money and bank credit appear to be deviating significantly from current projections. Meeting held on February 10, 1970 The information reviewed at this meeting suggests that real economic activity, which leveled off in the fourth quarter of 1969, may be weakening further in early 1970. Prices and costs, however, are continuing to rise at a rapid pace. Long-term market interest rates recently have fluctuated under the competing influences of heavy demands for funds and shifts in investor attitudes regarding the outlook for monetary policy. Bank credit declined in January but the money supply increased substantially on aver age; both had risen slightly in the fourth quarter. Flows o f time and savings funds at banks and nonbank thrift institutions have remained generally weak since year-end, and they apparently have been affected little thus far by the recent increases in maximum rates payable for such funds. The U.S. foreign trade balance improved somewhat in December, as imports fell off. The over-all balance of payments has been in substantial deficit in recent weeks. In light of the foregoing developments, it is the policy o f the Federal Open Market Committee to foster financial conditions conducive to the orderly reduction of inflationary pressures, with a view to encouraging sustainable economic growth and attaining reasonable equilibrium in the country’s balance of payments. To implement this policy, while taking account of the current Treasury refunding, possible bank regulatory changes and the Committee’s desire to see moderate growth in money and bank credit over the months ahead, System open market operations until the next meeting of the Committee shall be conducted with a view to moving gradually toward somewhat less firm conditions in the money market; provided, however, that operations shall be modified promptly to resist any tendency for m oney and bank credit to deviate significantly from a moderate growth pattern. Meeting held on March 10,1970 The information reviewed at this meeting suggests that real economic activity, which leveled off in the fourth quarter o f 1969, is weakening fur ther in early 1970. Prices and costs, however, are continuing to rise at a rapid pace. Market interest rates have declined considerably in recent weeks, partly as a result of changing investor attitudes regarding the outlook for econom ic activity and monetary policy. Both bank credit and the money supply declined on average in February, but both were tending upward in the latter part of the month. Outflows of time and savings funds at banks and nonbank thrift institutions, which had been sizable in January, appar ently ceased in February, reflecting advances in rates offered on such funds follow ing the recent increases in regulatory ceilings, together with declines in short-term market interest rates. The U .S. foreign trade surplus narrowed in January and the over-all balance of payments deficit has remained large in recent weeks. In light of the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions condu cive to orderly reduction in the rate of inflation, while encouraging the resumption of sustainable economic growth and the attainment of reason able equilibrium in the country’s balance of payments. To implement this policy, the Committee desires to see moderate growth in money and bank credit over the months ahead. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining money market conditions consistent with that objective. Meeting held on April 7,1970 The information reviewed at this meeting suggests that real economic activity weakened further in early 1970, while prices and costs continued to rise at a rapid pace. Fiscal stimulus, of dimensions that are still uncer tain, will strengthen income expansion in. the near term. Most long-term interest rates backed up during much of March under the pressure of heavy demands for funds, but then turned down in response to indications o f some relaxation of monetary policy and to the reduction in the prime lending rate of banks. Short-term rates declined further on balance in recent weeks, contributing to the ability of banks and other thrift institutions to attract time and savings funds. Both bank credit and the money supply rose on average in March; over the first quarter as a whole bank credit was about unchanged on balance and the money supply increased somewhat. The U.S. foreign trade surplus increased in February, but the over-all balance o f payments appears to have been in considerable deficit during the first quarter. In light of the foregoing developments, it is the policy of the Fed eral Open Market Committee to foster financial conditions conducive to orderly reduction in the rate of inflation, while encouraging the resumption of sustainable economic growth and the attainment o f reasonable equilib rium in the country’s balance of payments. To implement this policy, the Committee desires to see moderate growth in money and bank credit over the months ahead. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining money market conditions consistent with that objec tive, taking account o f the forthcoming Treasury financing. Meeting held on May 5,1970 The information reviewed at this meeting indicates that real economic activity weakened further in the first quarter of 1970. Growth in personal income, however, is being stimulated in the second quarter by the enlarge ment of social security benefit payments and the Federal pay raise. Prices and costs generally are continuing to rise at a rapid pace, although some components of major price indexes recently have shown moderating ten dencies. M ost market interest rates have risen sharply in recent weeks as a result o f heavy demands for funds, possible shifts in liquidity preferences, and the disappointment o f earlier expectations regarding easing o f credit market conditions. Prices of common stocks have declined markedly since early April. Attitudes in financial markets generally are being affected by the expansion of military operations in Southeast A sia and by concern about the success of the Government’s anti-inflationary program. Both bank credit MONETARY AGGREGATES AND MARKET CONDITIONS and the money supply rose substantially from March to April on average, although during the course of April bank credit leveled off and the money supply receded sharply from the end-of-March bulge. The over-all balance of payments was in considerable deficit during the first quarter. In light of the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate of inflation, while encouraging the resumption of sustainable eco nomic growth and the attainment of reasonable equilibrium in the country’s balance o f payments. To implement this policy, the Committee desires to see moderate growth in money and bank credit over the months ahead. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consistent with that objective, taking account of the current Treasury financing; pro vided, however, that operations shall be modified as needed to moderate excessive pressures in financial markets, should they develop. Meeting held on May 26,1970 The information reviewed at this meeting indicates that real economic activity declined more than previously estimated in the first quarter of 1970, but little further change is projected in the second quarter. Prices and costs generally are continuing to rise at a rapid pace, although some components of major price indexes recently have shown moderating tendencies. Since early May most long-term interest rates have remained under upward pres sure, partly as a result o f continued heavy demands for funds and possible shifts in liquidity preferences, and prices of common stocks have declined further. Attitudes in financial markets generally are being affected by the widespread uncertainties arising from recent international and domestic events, including doubts about the success of the Government’s anti-infla tionary program. Both bank credit and the money supply rose substantially from March to April on average; in May bank credit appears to be changing little while the m oney supply appears to be expanding rapidly. The over-all balance of payments continued in considerable deficit in April and early May. In light o f the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate of inflation, while encouraging the resumption of sus tainable economic growth and the attainment of reasonable equilibrium in the country’s balance of payments. T o implement this policy, in view o f current market uncertainties and liquidity strains, open market operations until the next meeting of the Com mittee shall be conducted with a view to moderating pressures on financial markets, while, to the extent compatible therewith, maintaining bank re serves and m oney market conditions consistent with the Committee’s longerrun objectives o f moderate growth in money and bank credit. Meeting held on June 23,1970 The information reviewed at this meeting suggests that real economic activity is changing little in the current quarter after declining appreciably earlier in the year. Prices and costs generally are continuing to rise at a rapid pace, although some components of major price indexes recently have shown moderating tendencies. Since late May market interest rates have shown mixed changes following earlier sharp advances, and prices of com 213 mon stocks have recovered part of the large decline of preceding weeks. Attitudes in financial markets continue to be affected by uncertainties and conditions remain sensitive, particularly in light of the insolvency o f a major railroad. In May bank credit changed little and the money supply rose m od erately on average, following substantial increases in both measures in March and April. Inflows of consumer-type time and savings funds at banks and nonbank thrift institutions have been sizable in recent months, but the brief spring upturn in large-denomination C D ’s outstanding at banks has ceased. The over-all balance of payments was in heavy deficit in April and May. In light of the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate of inflation, while encouraging the resumption of sustainable economic growth and the attainment of reasonable equilibrium in the country’s balance of payments. To implement this policy, in view of persisting market uncertainties and liquidity strains, open market operations until the next meeting of the Committee shall continue to be conducted with a view to moderating pres sures on financial markets. To the extent compatible therewith, the bank reserves and money market conditions maintained shall be consistent with the Committee’s longer-run objective of moderate growth in money and bank credit, taking account of the Board’s regulatory action effective June 24 and some possible consequent shifting of credit flows from market to banking channels. Meeting held on July 21,1970 The information reviewed at this meeting indicates that real economic activity changed little in the second quarter after declining appreciably earlier in the year. Prices and wage rates generally are continuing to rise at a rapid pace. However, improvements in productivity appear to be slow ing the rise in costs, and some major price measures are showing moderat ing tendencies. Since mid-June long-term interest rates have declined con siderably, and prices of common stocks have fluctuated above their recent lows. Although conditions in financial markets have improved in recent weeks uncertainties persist, particularly in the commercial paper market where the volume of outstanding paper has contracted sharply. A large proportion of the funds so freed apparently was rechanneled through the banking system, as suggested by sharp increases in bank loans and in largedenomination C D ’s of short maturity— for which rate ceilings were sus pended in late June. Consequently, in early July bank credit grew rapidly; there was also a sharp increase in the money supply. Over the second quar ter as a whole both bank credit and money supply rose moderately. The over-all balance of payments remained in heavy deficit in the second quarter. In light of the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to orderly reduc tion in the rate of inflation, while encouraging the resumption o f sustainable economic growth and the attainment of reasonable equilibrium in the coun try’s balance of payments. T o implement this policy, while taking account of persisting market un certainties, liquidity strains, and the forthcoming Treasury financing, the Committee seeks to promote moderate growth in money and bank credit over the months ahead, allowing for a possible continued shift o f credit flows from market to banking channels. System open market operations MONETARY AGGREGATES AND MARKET CONDITIONS until the next meeting of the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consistent with that objective; provided, however, that operations shall be modified as needed to counter excessive pressures in financial markets should they develop. Meeting held on August 18,1970 The information reviewed at this meeting suggests that real economic activity, which edged up slightly in the Second quarter after declining appre ciably earlier in the year, may be expanding somewhat further. Prices and wage rates generally are continuing to rise at a rapid pace. However, im provements in productivity appear to be slowing the rise in costs, and som e major price measures are showing moderating tendencies. Credit demands in securities markets have continued heavy, and interest rates have shown mixed changes since mid-July after declining considerably in preceding weeks. Some uncertainties persist in financial markets, particularly in con nection with market instruments o f less than prime grade. In July the money supply rose moderately on average and bank credit expanded substantially. Banks increased holdings o f securities and loans to finance companies, some of which were experiencing difficulty in refinancing maturing commercial paper. Banks sharply expanded their outstanding large-denomination C D ’s of short maturity, for which rate ceilings had been suspended in late June, and both banks and nonbank thrift institutions experienced large net inflows o f consumer-type time and savings funds. The over-all balance of payments remained in heavy deficit in the second quarter, despite a sizable increase in the export surplus. In July the official settlements deficit continued large, but there apparently was a marked shrinkage in the liquidity deficit. In light of the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate o f inflation, while encouraging the resumption o f sustainable eco nom ic growth and the attainment of reasonable equilibrium in the country’s balance of payments. To implement this policy, the Committee seeks to promote some easing o f conditions in credit markets and somewhat greater growth in money over the months ahead than occurred in the second quarter, while taking account of possible liquidity problems and allowing bank credit growth to reflect any continued shift of credit flows from market to banking channels. System open market operations until the next meeting o f the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consistent with that objective, taking account o f the effects o f other monetary policy actions. Meeting held on September 15,1970 The information reviewed at this meeting suggests that real econom ic activity, which edged up slightly in the second quarter, is expanding som e what further in the third quarter, led by an upturn in residential construc tion. W age rates generally are continuing to rise at a rapid pace, but im provements in productivity appear to be slowing the rise in costs, and some major price measures are rising less rapidly than before. Interest rates de clined in the last half o f August, but most yields turned up in early Septem ber, as credit demands in securities markets have continued heavy; existing yield spreads continue to suggest concern with credit quality. The m oney supply rose rapidly in the first half of August but moved back down through early September. Bank credit expanded sharply further in August as banks continued to issue large-denomination C D ’s at a relatively rapid rate, while reducing their reliance on the commercial paper market after the Board of Governors acted to impose reserve requirements on bank funds obtained from that source. The balance of payments deficit on the liquidity basis diminished somewhat in July and August from the very large secondquarter rate, but the deficit on the official settlements basis remained high as banks repaid Huro-dollar liabilities. In light of the foregoing develop ments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate o f inflation, while encouraging the resumption of sustainable economic growth and the attainment of reasonable equilibrium in the country’s balance of payments. To implement this policy, the Committee seeks to promote some easing of conditions in credit markets and moderate growth in money and attend ant bank credit expansion over the months ahead. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consistent with that objective. Meeting held on October 20,1970 The information reviewed at this meeting suggests that real output of goods and services increased slightly further in the third quarter but that employment declined and unemployment continued to rise; activity in the current quarter is being adversely affected by a major strike in the auto mobile industry. Wage rates generally are continuing to rise at a rapid pace, but improvements in productivity appear to be slowing the increase in costs, and some major price measures are rising less rapidly than before. Most interest rates have declined since mid-September, although yields on corporate and municipal bonds have been sustained by the continuing heavy demands for funds in capital markets. The money supply rose slightly on average in September and increased moderately over the third quarter as a whole. Bank credit expanded further in September but at a rate consider ably less than the fast pace of the two preceding months. Banks continued to issue large-denomination C D ’s at a relatively rapid rate and experienced heavy inflows of consumer-type time and savings funds, while making sub stantial further reductions in their use of nondeposit sources o f funds. The balance of payments deficit on the liquidity basis diminished in the third quarter from the very large second-quarter rate, but the deficit on the official settlements basis remained high as banks repaid Euro-dollar liabili ties. In light of the foregoing developments, it is the policy o f the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate of inflation, while encouraging the resumption o f sus tainable economic growth and the attainment of reasonable equilibrium in the country’s balance of payments. To implement this policy, the Committee seeks to promote som e easing of conditions in credit markets and moderate growth in money and attendant bank credit expansion over the months ahead. System open market opera tions until the next meeting of the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consistent with those objectives, taking account of the forthcoming Treasury financings. MONETARY AGGREGATES AND MARKET CONDITIONS M eeting held on Novem ber 17, 1970 The information reviewed at this meeting suggests that real output of goods and services is changing little in the current quarter and that un employment has increased. Part but not all of the weakness in over-all activity is attributable to the strike in the automobile industry which ap parently is now coming to an end. Wage rates generally are continuing to rise at a rapid pace, but gains in productivity appear to be slowing the increase in unit labor costs. Recent movements in major price measures have been erratic but the general pace of advance in these measures has tended to slow. M ost interest rates declined considerably in the past few weeks, and Federal Reserve discount rates were reduced by one-quarter of a percentage point in the week o f N ovem ber 9. Demands for funds in capital markets have continued heavy, but business loan demands at banks have weakened. The m oney supply changed little on average in October for the second consecutive month; bank credit also was about unchanged, following a slowing of growth in September. The balance of payments deficit on the liquidity basis was at a lower rate in the third quarter and in October than the very high second-quarter rate, but the deficit on the official settlements basis remained high as banks repaid Euro-dollar lia bilities. In light o f the foregoing developments, it is the policy o f the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate o f inflation, while encouraging the resump tion o f sustainable econom ic growth and the attainment o f reasonable equilibrium in the country’s balance of payments. To implement this policy, the Committee seeks to promote some easing o f conditions in credit markets and moderate growth in money and attend ant bank credit expansion over the months ahead, with allowance for temporary shifts in m oney and credit demands related to the auto strike. System open market operations until the next meeting of the Committee shall be conducted with a view to maintaining bank reserves and money market conditions consistent with those objectives. M eeting held on D ecem ber 15, 1970 The information reviewed at this meeting suggests that real output of goods and services has declined since the third quarter, largely as a consequence of the recent strike in the automobile industry, and that unemployment has increased. Resumption of higher autom obile produc tion is expected to result in a bulge in activity in early 1971. W age rates generally are continuing to rise at a rapid pace, but gains in productivity appear to be slowing the increase in unit labor costs. M ovements in major price measures have been diverse; most recently, wholesale prices have shown little change while consumer prices have advanced substantially. Mar ket interest rates declined considerably further in the past few weeks, and Federal Reserve discount rates were reduced by an additional one-quarter o f a percentage point. Demands for funds in capital markets have contin ued heavy, but business loan demands at banks have been weak. Growth in the money supply was somewhat more rapid on average in N ovem ber than in October, although it remained below the rate prevailing in the first three quarters o f the year. Banks acquired a substantial volum e of securities in N ovem ber, and bank credit increased moderately after chang ing little in October. The foreign trade balance in September and October was smaller than in any other 2-m onth period this year. The over-all 217 218 balance of payments deficit on the liquidity basis remained in October and November at about its third-quarter rate. The deficit on the official settle ments basis was very large as banks continued to repay Euro-dollar lia bilities. In light of the foregoing developments, it is the policy of the Federal Open Market Committee to foster financial conditions conducive to orderly reduction in the rate of inflation, while encouraging the resump tion of sustainable economic growth and the attainment of reasonable equilibrium in the country’s balance of payments. To implement this policy, System open market operations shall be conducted with a view to maintaining the recently attained money market conditions until the next meeting of the Committee, provided that the ex pected rates of growth in money and bank credit will at least be achieved. 3 0313 00DH1 334=1