Consequences of the Federal Reserve's Reattachment

Transcription

1 Carnegie Mellon University Research CMU Tepper School of Business Consequences of the Federal Reserve's Reattachment to Free Reserves Allan H. Meltzer Carnegie Mellon University, am05@andrew.cmu.edu Follow this and additional works at: Part of the Economic Policy Commons, and the Industrial Organization Commons This Response or Comment is brought to you for free and open access by Research CMU. It has been accepted for inclusion in Tepper School of Business by an authorized administrator of Research CMU. For more information, please contact researchshowcase@andrew.cmu.edu.

2 Consequences of (he Federal Reserve's Re-attachment to Free Reserves by Allan H. Meitzer To be presented at the Western Economic Association meetings, July 1981 rr^ ^ K-y > i ^ ^

3 Revised and Corrected Draft June 1981 Consequences of the Federal Reserve's Re-attachment to Free Reserves by Allan H. Meltzer* Eighteen years ago, at these meetings, Karl Brunner and I presented a preliminary version of our study of the Federal Reserve system that was then in process. 1 At the time, there was much less public and professional scrutiny of Federal Reserve decisions and actions. Minutes were not released; money targets were not announced; there were no regularly scheduled oversight hearings; and the weekly or monthly data on the money stock were not the subject of widespread speculation and comment. It was a different time. Consumer prices rose that year by less than 1 Three per cent was still considered a relatively high rate of inflation. Federal government spending had just passed $100 billion and was less than 20% of GNP. A $5 billion deficit (about $14 billion at 1980 prices for government services) in the Federal budget was considered large. The dollar exchange rate was fixed by the Bretton Woods agreement, and the Federal Reserve, the Treasury, and the President (Kennedy) were publicly committed to maintenance of the fixed rate system. Brunner and I had been asked by Congressman Wright Patman, Chairman of the House Committee on Banking, to do a study of the dealer market for government securities.^ We convinced him - or perhaps it would be correct to say we assured him - that the main problems of monetary policy lay with the Federal Reserve, not with the dealer market. I do not know whether he was convinced, but he authorized us to undertake a study of the Federal Reserve's procedures for conducting monetary policy. One section of the report, called The Federal Reserve's Attachment to Free Reserves, documented the central role of free reserves in Federal Reserve decisions and operations and showed that control of free reserves did not achieve effective control of the stocks of money and credit. Earlier work by Meigs (1962) and Dewald (1963) reached a similar conclusion.

4 2 The most surprising result of the study was the nearly complete absence of a systematic connection between policy decisions and money, prices and output. There was talk about prices, balance of payments and unemployment. Policy discussions made many references to these and other measures of performance, but there was little evidence of a systematic framework relating these goals of policy to policy actions. Although our claim that the Federal Reserve concentrated on the higgling and jiggling of the money market-later called money market myopia - seemed contentious at the time, subsequent release of the minutes suggests it was accurate at the time. Our main criticisms of policy up to that time can be summarized in four related points. First, we showed that money growth was pro-cyclical. Money growth rose in periods of economic expansion and declined (or turned negative) during recessions. Consequently, we said, expansions were followed by inflation and recessions were made more severe. Second, the Federal Reserve did not distinguish between market rates and real rates of interest. During contractions, interest rates fell. The decline was interpreted as an easing of monetary policy even if money and prices fell. On many occasions, and particularly during the contraction from 1929 to 1933, the Federal Reserve described monetary policy as expansive despite rising real rates of interest. Third, aggregate member bank borrowing rose during expansions and declined during recessions. These cyclical changes in borrowing were one of the causes of pro-cyclical movements in the Federal Reserve f s asset portfolio, in the monetary base and in the money stock. Under the free reserve conception, an increase in member bank borrowing from the Federal Reserve raises interest rates. Lower free reserves and higher interest rates were interpreted by the Federal Reserve as a sign that policy had become more restrictive. Larger free reserves and lower interest rates were interpreted as greater ease, or less restriction. Fourth, concentration on free reserves and interest rates produced systematic misinterpretations of the future effects of policy. We called this the "indicator problem" and claimed that reliance on an inappropriate indicator, free reserves, led the Federal Reserve to misinterpret the thrust, or future effect, of policy on economic activity and prices. We claimed that the Federal Reserve f s case for reliance on free reserves was based on the fallacy of composition. In our view, an increase in borrowing (or a decline in banks 1 desired excess reserves) expands the monetary base and the money stock, so it must be considered expansive; a decline in borrowing is restrictive. The reserves supplied at the discount window are indistinguishable from other reserves or base money once they are issued. There is no logic behind the claim that reserves supplied by an open market purchase expand bank credit and money while reserves supplied at the discount window contract bank credit and money. The fact that individual banks are not permitted to borrow repeatedly or for long periods is not sufficient to prevent the aggregate volume of borrowing from rising during expansion and falling in recession.

5 3 There have been many changes in practice and in the aims and techniques of monetary policy during the past two decades. The quality and quantity of research done within the Federal Reserve System - in St. Louis, Boston, Minneapolis, in Washington and elsewhere - has shown greater commitment than in the past to development of an understanding of the role of money and the operation of the financial system. Our criticism of the System for failing to study and learn about the properties of the monetary system and the effects of policy is no longer valid. Our criticisms of policy operations and decisions remain. The Federal Reserve now permits larger daily or monthly fluctuations in market interest rates on Federal funds and other short-term instruments. The staff and some of the members of the open market committee distinguish between market rates and real rates of interest and rates of exchange. But official spokesmen, including Chairman Volcker, continue to interpret increases in borrowing as restrictive and defend this position with the same arguments used in the twenties.3 And monetary policy remains pro-cyclical; money growth rises in periods of economic expansion and falls in recessions. The disadvantages of pro-cyclical growth of money seem obvious. Increased money growth during periods of economic expansion eventually increases inflation; slower money growth during recession reinforces the decline in output. Repetition of this pattern generates expectations that the pattern will be repeated.3 a Fewer and fewer people anticipate that recessions will have a lasting effect on inflation. Contracts for wages and prices of a wide range of goods and services are signed reflecting the belief that the average rate of increase in prices and wages is positive. The much discussed "sticky" prices become even "stickier". Inflation is built into the system, and the belief that inflation will not be reduced becomes a belief that inflation cannot be reduced. The cost of any attempt to reduce inflation rises. All of this is familiar. The problem is to find advantages of a pro-cyclical policy to policymakers at the Federal Reserve, at the Bank of England and elsewhere that cause them to persist in these policies. What private or social gain might they see that offsets the more obvious costs of pro-cyclical money growth? I have not found an answer to that question and have concluded, instead, that pro-cyclical money growth is an unintended effect of policy decisions taken for other reasons. Two reasons for pro-cyclical policy occur. One is that central bankers 1 concentration on free reserves or interest rates, particularly short-term rates, is a vestige of the gold standard. The gold standard, or more generally the classical price-specie flow mechanism, implies that money growth rises in expansions and falls in contractions. The second reason, not entirely unrelated, is the failure of central banks (and others) to distinguish between permanent and persistent changes in money growth. In the following section, I discuss both reasons but concentrate on the latter as a cause of pro-cyclical changes in money growth.

6 The decisions taken by the Federal Reserve on October 6,1979 were interpreted as a major break with long-standing tradition. Concentration or targetting of non-borrowed reserves appears to shift their attention from interest rates to monetary aggregates. The shift is smaller than appears at first sight, however. The use of lagged reserve requirements, slow adjustment of the discount rate to changes in market rates and reliance on non-borrowed reserves as the target for Federal Reserve operations, when taken together, restore the prominence once given to free reserves or member bank borrowing. In a later section, I analyze current Federal Reserve procedures and compare the effect of controlling the monetary base to the effect of controlling market interest rates or non-borrowed reserves. The Federal Reserve has now accepted, and frequently reaffirms, many of the propositions once associated with "monetarism". These include the dominant influence of sustained money growth on inflation, the necessity of reducing money growth to reduce the maintained rate of inflation, the distinction between real and market rates of interest and the positive effect of inflation on market rates. Some of the principal remaining issues, or differences, concern the effects of different control procedures in a world in which there are both permanent and transitory shocks. Therefore, I limit the scope of the analysis, and its complexity, by neglecting the feedback from policy actions through the credit, money, output and labor markets. This permits concentration on the effect of different control procedures on the impulses that are sent to the various markets. These impulses become the bases for anticipations and responses to policy and for the pro-cyclicality of monetary policy. Reasons for Pro-cyclical Policy No better general statement of central bank policy can be found than the words of one of the great 19th century economists, Henry Thornton. Thornton (1802, p. 259) advised central bankers: "To limit the amount of paper issued and to resort for this purpose, whenever the temptation to borrow is strong to some effectual principle of restriction; in no case, however, materially to diminish the sum in circulation, but to let it vibrate only within certain limits; to afford a slow and cautious extension of it, as the general trade of the kingdom enlarges itself; to allow of some special, though temporary, increase in the event of any extraordinary alarm or difficulty...this seems to be the true policy of an institution circumstanced like that of the Bank of England. To suffer either the solicitations of merchants, or the wishes of government, to determine the measure of the bank issues, is unquestionably to adopt a very false principle of conduct." Thornton's prescription for the Bank of England, or any central bank, is to allow the money stock to rise at the maintained growth rate of real expenditure, to function as lender of last

7 5 resort, to prevent an absolute decline in the money stock and to avoid both the real bills fallacy and the excessive financing of government expenditure. One can only add that had Thornton's strategy been followed, there would have been fewer monetary crises, and the effects of the remaining disturbances would have been smaller. Money growth would have been mildly pro-cyclical, however. Central banks have not followed Thornton's prescription. They paid much less attention to money growth than he advocated and much more attention to interest rates and exchange rates. Their use of interest rates (or exchange rates) as an indicator of monetary ease and restraint antedates Keynes and the Keynesians, so Keynesian analysis is, at most, a modern justification or rationalization of an older central banking practice. To find the reason central bankers chose interest rates as an indicator of monetary policy, we can put aside recent controversies about Keynesian analysis. Central bank practice developed under the gold standard, so we look first, at the influence of the gold standard on traditional central bank practice and on pro-cyclicality. Under a classical gold standard, exchange rates are fixed and central banks are committed to increase the money stock when their gold holdings increase and to reduce the money stock when the gold stock falls. Thornton recognized clearly and stated without qualification, that wages changed much less than prices as the economy expands and contracts so real wages were expected to fall in periods of monetary and economic expansion and rise in recessions. See Thornton (1802, p. 119). In his analyses, output, income and money rise and fall together; real wages change in the direction opposite to the changes in money and output. A modern economist might reason that it is difficult for workers and employers to agree on a forecast of real wages, so they adjust wages when changes in market conditions are expected to persist and absorb transitory fluctuations. Under a classical gold standard, cyclical changes in rates of price change do not persist. Neither Thornton nor other classical economists mention this explanation of "sticky" wages, or to my knowledge any other, but a belief that households do not adjust to transient changes is consistent with the emphasis given in classical economics to persistent, longterm effects. The reasons for "sticky" wages are less important for this discussion than three implications for monetary policy. First, money growth is expected to be pro-cyclical. Second, the long-run price level is fixed by the level and distribution of the world gold stock, techniques of mining and similar factors that change slowly. Third, the money stock is an endogenous variable. Under the classical gold standard, the main functions of the central bank were to serve as lender of last resort in a financial panic and to moderate fluctuations in interest rates and money growth

8 6 arising from seasonal and transitory changes in borrowing and lending by domestic and foreign residents. Central banks carried out their day-to-day functions mainly by lending to financial institutions. To control borrowing they moved their lending rate relative to market rates. Hence they attached great significance to movements of market interest relative to the discount rate. Any persistent difference between the two required the central bank to change the discount rate - a decision about monetary policy - or to change the exchange rate. Market rates became an indicator of changes in the discount rate, that is to say of changes in monetary policy. These procedures provide a basis for two traditional Federal Reserve claims: Bank borrowing must be temporary in the aggregate, and interest rates often lead discount rates and money. Persistent increases in aggregate borrowing occur, under the gold standard, principally as the economy expands and the demand for bank credit rises. Interest rates rise relative to the discount rate in response to the increased demand. Later, when prices rise and gold flows out, the economy contracts; the demand for bank credit declines; and market rates lead the discount rate to lower levels; bank borrowing and money growth fall. The classical gold standard does not make the interest rate an obvious indicator of monetary policy. A central bank with sufficient gold and foreign exchange reserves might choose to dampen fluctuations in output and the price level by varying the domestic source component of the monetary base so as to offset changes in the foreign sources of the base. 4 A central bank operating in this way, would seek to control the monetary base. Discount rate changes or open market operations would be timed to moderate the size of changes in the base when gold flowed in and out. A policy of this kind eliminates or reduces the pro-cyclicality resulting from unanticipated gold flows. The main problem that I see in this procedure arises from the difficulty of distinguishing between persistent and transitory changes in the growth of aggregate output. Suppose there is a one-time change in the growth rate of productivity that raises or lowers the sustained growth of real output in a single country. The change in real growth will be followed by a persistent change in the growth or gold and foreign exchange that must induce a permanent change in the exchange rate or a higher rate of inflation. The policy of offsetting changes in foreign reserves by movements of the domestic source component of the monetary base would break down. The problem of separating persistent and transitory changes arises under any monetary arrangement. Suppose exchange rates are free to fluctuate, instead of fixed by a gold standard. Consider two alternatives. In the first, the central bank sets a market rate of interest and only varies

9 7 the interest rate when there are perceived deviations of the growth rate of money from the announced growth path. In the second, the central bank sets a target rate of money growth and allows the interest rate to adjust as frequently and as much as required to keep the growth rate of money as close to the target as possible. If all shocks are transitory, the difference between the two control strategies is reduced. The former produces less variability in interest rates and greater variability of money growth; the latter produces less variability of money growth and greater variability of market interest rates from day to day. Greater variability of interest rates or of money growth imposes different costs and provides different tasks for markets. The principal problem of central bank strategy - providing the growth rate of money that achieves the target rate of inflation or that maintains the desired exchange rate at lowest social cost - can be achieved either way. The choice depends on the relevant costs and is a tactical rather than a strategic problem. The principal reason is that, by assumption, all shocks are transitory, so changes in money growth or interest rates do not persist. Transitory shocks are not likely to produce the observed pattern of pro-cyclical money growth. Choice of strategy becomes more important, when there are persistent changes to real and nominal variables that cannot be distinguished from transitory changes promptly by either the market or the central bank. A policy of controlling money allows persistent real shocks to affect real rates of interest and exchange, but prevents permanent changes in money growth. Permanent increases in productivity cause the level of output to rise and the price level to fall, and permanent reductions in productivity lower the level of output and raise the price level. These changes are one time, or once-and-for-all, effects. The price level rises or falls around the maintained rate of price change, and the exchange rate adjusts. Since the economy's long-term growth is unaffected by the change in productivity, maintaining the growth of money maintains the average rate of inflation. Permanent changes in real variables affecting the rate of growth of real output and the rate of growth of the demand for money cause the rate of inflation to change. Usually, neither the central bank, nor the market, can distinguish persistent or permanent changes in the rate of change when they occur, so these changes cannot be avoided. Once the changes are perceived to be permanent, the central bank must either change the target rate of growth of money or allow the rate of inflation to change. A policy of controlling interest rates requires the central bank to change the money stock whenever the public changes desired money balances relative to output. Central bank policy assures that every increase or decrease in the desired money balances, whether permanent or transitory, is matched by an increase or decrease in the money stock. The increase in money prevents any effect of such changes in the demand for money on prices and output. Shifts in the demand for money cannot explain the pro-cyclicality of monetary policy under an interest rate target.

10 8 A policy of controlling interest rates requires the central bank to change the stock of money when there are changes in aggregate real demand. If the real changes are permanent, the central bank reinforces the effect of the change on prices and output by increasing money when aggregate demand increases and reducing money when aggregate demand falls. Monetary policy is pro-cyclical, as it is under the gold standard and for similar reasons. In both cases neither the degree to which the change is permanent or persistent nor the timing of the change in demand is fully anticipated, so the rate of inflation does not change sufficiently and the exchange rate does not change sufficiently to prevent an effect on output. Suppose that the real growth of the public sector increases. The increase in the growth of real expenditure raises interest rates. Central bank policy prevents or delays the increases in interest rates by increasing the stock of base money and its rate of change. The higher growth of the monetary base finances the growth of public spending. As long as the public and the central bank believe that the changes in government spending and money are temporary, they do not adjust anticipations of inflation fully; the exchange rate does not depreciate at the equilibrium rate. Policy is pro-cyclical. The increase in money growth reinforces the effects of increased government spending on output. The central bank policy of maintaining interest rates in the presence of a real increase in the growth of aggregate demand heightens the real expansion in the economy and the subsequent inflation. Neither the central bank, nor anyone else, can separate real from nominal changes with enough accuracy to read from market values information about how much to raise interest rates to slow inflation. If the central bank raises the market rate too little, money growth remains procyclical. If the central bank increases interest rates too much, the anticipated rate of expansion is revised downwards. A sudden reduction in the anticipated rate of expansion reduces the growth of real expenditure and often brings on a recession. Money growth and real growth fall. There is no way for the central bank to know precisely where to set the interest rate to minimize variability and avoid incorrect inferences. To find the interest rate that slows the growth of base money, the central bank must have information about anticipations, about the relative size of real and nominal shocks, about the persistence of changes and about the way in which beliefs about the persistent effects of policies adjust to changes in interest rates and other variables. By setting an interest rate, the central bank permits permanent changes in the growth of real expenditure to change the growth of money. Unless the central bank correctly separates permanent and transitory changes, it cannot set interest rates to prevent persistent changes in the rate of growth of money and the rate of inflation. If the change in real expenditure is a change in level, there is a bulge in the money stock, rather than a permanent change in the growth of money. People cannot quickly separate changes in the rate of growth from changes in the nominal stock, so they under-

11 9 estimate or overestimate the sustained rate of inflation. Central bank policy increases uncertainty about future prices and rates of price change. Let me summarize this section by stating the consequences of alternative central bank procedures. (1) If unanticipated changes in the growth rate of the demand for money relative to output are the principal cause of fluctuations in the economy, the central bank damps fluctuations in economic activity by controlling interest rates. Changes in the demand for money relative to output call forth changes in the money stock that dampen changes in real rates of interest. There is no problem of pro-cyclicality. Evidence of pro-cyclicality is inconsistent with this explanation of fluctuations. (2) If the dominant causes of fluctuations are changes in the growth of spending, controlling the growth of money prevents changes in the rate of inflation. Controlling money maintains inflation at some average rate and reduces the severity of recessions by preventing counter-cyclical policy. Controlling interest rates induces pro-cyclical changes in money growth. (3) If there are permanent changes in the growth of productivity of financial or non-financial firms, the central bank cannot maintain zero (or constant) average inflation by setting the rate of monetary base growth permanently. To keep prices stable, on average, the rate of monetary base growth must adjust periodically to reflect these changes in productivity growth. If the central bank is slow to respond to permanent changes in productivity growth, permanent changes in productivity growth cause changes in the rate of inflation. A policy of controlling interest rates cannot induce pro-cyclicality in this case. Moreover, the.historical record makes it hard to sustain a belief that the dominant cause of changes in inflation in the United States, the United Kingdom or much of Western Europe has been the result of permanent changes in the growth of productivity. A policy of controlling the growth of money does not permit the central bank to offset the effect of transitory real shocks to aggregate demand or to productivity. Consequently, prices will vary around the maintained rate of inflation, and output will vary. Recessions and expansions induced by changes in productivity will not be avoided. What will be avoided is the pro-cyclical effect of changes in money introduced by the central bank to offset the effect of changes in real demand and supply - real shocks. No one will mistake one-time or temporary changes in money for persistent changes in the rate of money growth, however, so the strategy of controlling money reduces a main source of uncertainty about the future price level and the maintained rate of inflation. Implementing Monetary Policy^ The decision to control interest rates or money growth does not dictate an answer to the tactical issue. This section compares three alternative tactical procedures for controlling money. The first uses a market interest rate, the second uses non-borrowed reserves, and the third uses the mone-

12 10 tary base as the control variable. The focus of this section is on the differences in impulses sent from the financial markets to the real sector under different control procedures. Net wealth at market prices, W, consists of base money, B, government securities, S, and private capital, K. The nominal value of the capital stock is PK. Government bonds bear interest at a market determined rate, i. I ignore the additional complexity introduced by residents 1 ownership of foreign securities or foreign ownership of domestic securities. W = B + S + PK (1) Banks are required to hold reserves and are permitted to borrow from the central bank at the discount rate, p. The monetary base consists of non-borrowed reserves NR, bank borrowing, A, and currency, C. I ignore excess reserves.^ B = NR + A + C (2) The demand to hold base money depends on interest rates, asset prices and the cost of borrowing from the central bank - the discount rate - and on other variables treated as pre-determined, such as income, the price level and their expected values. B = L(i, P, p, y...) Li <0; Lp,L p,ly>0 (3) The demand for government securities is S = S(i, P, y...) Sp, Sj, Sy > 0 (4) The (partial) response of the demand for government securities to an increase in the price of capital is assumed, as in Keynesian analysis, to be positive. Equilibrium in the asset markets determines the allocation of assets into (base) money, bonds and capital and the equilibrium values of i and P. Permanent and transitory shocks affect the markets for money and securities by changing government spending, taxes and the random component of aggregate demand. Differentiating the equations for B and S, holding output, expectations and other variables fixed, we obtain 0 = Ljdi + LpdP (5) 0 = Sjdi + Sp dp (6)

13 11 Equations (5) and (6) define the slopes of the MM and CM curves; the MM curve is the locus of money market equilibrium positions, and the CM curve is the locus of bond market equilibrium positions. The curves are shown in Figure 1. [Insert Figure 1 here] Compare, first, a policy of fixing a market interest rate, i, to the policy of fixing the monetary base. The former means that the central bank chooses i = ig. The latter means that the central bank fixes the stock of base money and does not adjust the stock when there are changes in the demand for money. Suppose tax collections decline or government spending increases by more than is anticipated. The unanticipated deficit is financed by selling bonds, so CM shifts to the right, in Figure 1; the interest rate rises and P rises, as shown by intersection of CMj and MMg. The change in interest rates and asset prices affects spending and output. Aggregate demand, d, depends on i and P and on the price level, p, output, y, and other variables. d = d( i, P, p, y...) (7) di,d p <0; dp,d y >0 The increases in i and P have offsetting effects on aggregate demand, so the change in aggregate demand is relatively small. A policy of controlling interest rates amplifies the response of aggregate demand. The monetary base increases; the MM line shifts to the right until it intersects CMj at the fixed rate of interest, ig. The interest rate control policy not only requires the central bank to finance part of the unanticipated government deficit, but also increases the response of aggregate demand. The reasons are that the increase in P is larger and the effect of P on aggregate demand is not offset, partly or wholly, by an increase in i. Monetary policy is, therefore, pro-cyclical. An unanticipated increase in money reinforces the response to the unanticipated deficit. If the unanticipated government deficit is transitory, the economy returns to the initial equilibrium at CMg and MMg. A persistent change in the deficit, however, amplifies the initial responses. The stock of securities increases, so the CM line shifts further to the right. Asset prices and interest rates rise, under a policy of base control, damping the effect on aggregate demand. A policy of interest rate control heightens the effect on aggregate demand and the pro-cyclical response of money. Sooner or later the central bank must change the interest rate target. Where should the new interest rate be set? No one can be certain that his knowledge of the slopes of CM and MM,

14 12 Figure 1

15 13 of the prevailing state of anticipations, and the magnitude of permanent and transitory components in the deficit, aggregate demand and output is sufficiently precise to give a correct answer. A policy of controlling the base does not require this type of information. Control of the base maintains the average or inherited rate of inflation but does not prevent all changes in prices and output. Base velocity rises, as spending rises and falls as spending falls. This is shown by the movement along MMg. Growth of the base is constant, however, not pro-cyclical. One of the principal reasons given for using an interest rate target is to reduce variability of interest rates. Presumably this reasoning applies to transitory changes, but even for transitory changes the argument is incomplete. The discussion shows that the reduced variability of the interest rate is achieved by increasing the variability of asset prices and real rates of return to capital assets. The increased variability of real interest rates is accompanied by increased variability of aggregate demand. This finding is reversed if the monetary base is controlled. The variability of aggregate demand and real rates of return is lower when there are unanticipated changes in government spending. If the principal shocks in the economy are shocks to aggregate demand, a policy of controlling the base is preferable to a policy of interest rate control. The latter introduces variability into aggregate demand and real returns that can be avoided by controlling the base. Incorporating uncertainty about permanent and transitory effects reinforces this conclusion. For it must be clear that, whenever the increased spending is permanent or persistent, the expansion of the base must be reversed to prevent an increase in inflation. The initial effort to keep interest rates from changing leads, under reasonable assumptions, to greater variability once the effort to stop the increase in prices and anticipations of future price changes is sustained. A similar conclusion is reached if we consider an unanticipated change in aggregate demand financed by buying or selling (existing) securities. Suppose people in the aggregate wish to sell securities and increase spending. With a fixed stock of outstanding debt, the reduced demand for securities appears as a shift of CM to the right. By controlling the base instead of interest rates, the central bank increases the change in i and reduces the change in P, thereby dampening instead of reinforcing the impact of the change in aggregate demand. If instead, the public finances the change in aggregate demand by increasing (or reducing) desired money balances, interest rate control stabilizes both asset prices and interest rates. The reason is that the change in the demand for money is matched by a change in the stock of base money. The desired increase in spending is financed by an increase in nominal money under the interest rate control policy. A policy of monetary base control induces changes in i and P, in this case and, ultimately, requires a change in the price level to restore equilibrium.

16 14 A policy of setting interest rates causes the central bank to produce pro-cyclical monetary growth in response to changes in aggregate demand. To the degree that the increases in aggregate demand are financed by reductions in the demand for money, or conversely, pro-cyclicality is reduced. The fact that we observe pro-cyclical monetary growth suggests that changes in aggregate demand are not financed entirely by changes in desired money holdings. Shocks to the level of aggregate output, for example, an unanticipated reduction in the level of productivity, leave a pattern of responses that are more ambiguous than the responses to changes in aggregate demand. A transitory reduction in productivity causes an unanticipated loss of output and rise in the price level. The reduction of output reduces the demand for nominal money and debt; the unanticipated increase in the price level raises the demands for nominal money and debt. The instantaneous net effect of the productivity shock on P and i is ambiguous, so comparison of base control and interest rate control is inconclusive in this case. It is unlikely that productivity shocks would induce a persistent pro-cyclical pattern of money growth. Free Reserves and Non-borrowed Reserves Currently, the Federal Reserve does not control the monetary base or interest rates. Their choice of control variable, non-borrowed reserves, is defined as non-borrowed reserves, NR, plus any excess reserves that member banks hold. Current practice, lag reserve accounting, fixes the volume of required reserves in relation to deposits held two weeks earlier. Banks borrow or repay to meet required reserves. The higher is the market rate relative to the discount rate, the larger the amount banks in the aggregate choose to borrow (and the smaller the amount of excess reserves they hold). Current practice also sets the discount rate. Changes in the discount rate are infrequent, and the discount rate is a given at any time. Banks choose the volume of free reserves - excess reserves minus borrowing - in response to interest rates. Excess reserve holdings are small, so changes in free reserves largely reflect changes in banks' borrowing from the Federal Reserve. I ignore excess reserves. To control money by controlling non-borrowed reserves, the Federal Reserve must estimate the volume of borrowing (free reserves). Using equations (2) and (3), we have A(i, p) + NR + C = L(i, P...), (g) in place of equation (3), with Ajdi = Ljdi + Lp dp (9)

17 15 in place of equation (5). The money market curve is now flatter. Figure 2 shows the money market equation from Figure 1, based on control of the monetary base and denoted MM, and equation (9) denoted NN. [Insert Figure 2 here] As before, we can illustrate the working of the system by supposing there is an unanticipated increase in government spending. The CM curve shifts to CM\. Interest rates rise less, and asset prices rise more than under a policy of base control; interest rates change more, and asset prices change less than under a policy of interest rate control. The differences depend on the rules affecting borrowing. The market rate, now at i j, exceeds the discount rate, pg. If banks are permitted to borrow as much as they choose, than Aj is very large. The NN curve is very flat and can be approximated by a horizontal line at pg. At the opposite pole, the more borrowing is restricted by non-pecuniary means, the smaller is the actual Aj and the closer is the slope of NN to MM. Suppose the unanticipated change in government spending and deficit finance raises the interest rate and asset price to the position shown at the intersection of NNg and CMj. The induced change in aggregate demand is smaller than under interest rate control because the unanticipated change in money is smaller. The shift from interest rates to non-borrowed or free reserves as the means of implementing control reduces the effect of transitory changes in the deficit on aggregate demand. Comparison of the effect on aggregate demand under free reserves and base control leads to a clear conclusion also. Control of free reserves permits a smaller response of interest rates and a larger response of asset prices than a policy of controlling the base, so the induced change in aggregate demand is smaller when the base is controlled. Transitory shocks to aggregate demand induce more variability under free reserve control than under base control. Current Federal Reserve procedures, therefore, reinforce the effects of transitory changes in government and maintain pro-cyclicality. The pro-cyclical effects of permanent shocks were analyzed in the previous section. A policy of controlling free or non-borrowed reserves permits a greater initial change in P and a smaller initial change in i in response to changes in government spending (or in aggregate demand). The change in aggregate demand is reinforced. If the shock is positive, banks borrow, increasing the monetary base and adding to the expansive influences. If the shock is negative, member bank borrowing and the base decline. The Federal Reserve must offset the change in bank borrowing by shifting the NN curve. To the extent that anticipated changes in money growth affect interest rates before the shift is made, failure to control the base increases the variability of interest rates and asset prices.

18 16 Figure 2

19 17 Conclusion One of the major complaints about monetary control is that monetary control increases the variability of interest rates. The truth of this statement is not obvious. To make the statement true requires two very strong assumptions. First, we must confine analysis to a world in which there is only one rate of interest. Once we introduce a term structure - or an asset price level and interest rate, the simple truth vanishes. If the dominant shocks in the economy are the result of changes.in aggregate demand, interest rate targets stabilize one asset price and increase the variability of all others relative to a policy of controlling the monetary base. In the model used here, we can treat i as a short-term rate and P as the price of durable capital. The conclusion follows directly that control of one rate, in the face of shocks to spending, increases the response of other rates. Second, the statement that interest rate control reduces the variability of interest rates ignores the feedback from policy changes to inflation, output and interest rates. If all shocks are transitory, the feedback is small. If shocks have permanent, or persistent, components, these elements increase the variability of all interest rates under an interest target policy. There is, moreover, little reason to concentrate on the variability of interest rates except insofar as the variability of market interest rates affects such real variables, as current and future consumption. The object of economic policy, including monetary policy, is to reduce variability to the minimum inherent in nature. The analysis here, shows that interest rate control - and control of free reserve or non-borrowed reserves - does not minimize the variance of income or consumption particularly when there are unforseen changes in spending. Control of interest rates, free or nonborrowed reserves is a cause of pro-cyclical money growth. Control of the monetary base reduces the response of aggregate demand to fiscal shocks and to other shocks to aggregate demand under current conditions. The gain from controlling the base (or total reserves), instead of interest rates or non-borrowed reserves, increases if some of the shocks are permanent or persistent, while others are transitory. The gain from controlling the base is less clear if the economy is affected mainly by shocks to productivity, and the case for interest rate control is strongest if the principal cause of variability is shocks to the growth of productivity or in the growth rate of money relative to output. History gives some evidence that aids in discriminating between these alternatives. Usually, money and real output rise and fall together. This pattern is consistent with the interpretation of cyclical movements principally as events started by shocks to aggregate demand, not shocks to aggregate

20 18 supply or to the growth of productivity. Control of interest rates reinforces the shocks and, therefore, induces money to rise and fall cyclically. Cyclical expansions are followed by price increases or higher inflation. Under a classical gold standard, monetary policy is pro-cyclical. Changes in money are mainly a response to payments on trade account. Increased demand for exports brings gold that reinforces the expansion in demand; later, prices rise. The rise sets in motion a train of events reversing the expansion of money. The use of free reserves or borrowing as a target of monetary policy began when the U.S. was on a gold standard, so it is not surprising that monetary policy was pro-cyclical. The Federal Reserve continued the practice of controlling free reserves, and later short-term interest rates, under the Bretton Woods agreement, so monetary policy remained pro-cyclical. The recent shift from an interest rate target to a non-borrowed reserves target is a step away from the pro-cyclicality that arises when aggregate demand changes and interest rates are controlled. The size of the step is limited by the constraints that the Federal Reserve imposes on itself. Maintaining lag reserve requirements and holding the discount rate fixed for long periods permits, or perhaps forces, bank borrowing from the Federal Reserve to respond to shocks. Shocks to aggregate demand, from fiscal policy for example, raise interest rates and induce banks to borrow and repay. These shocks are, therefore, partly financed by changes in aggregate bank borrowing; money growth rises and falls with borrowing. Monetary policy remains pro-cyclical. Continued pro-cyclicality is a consequence of the re-attachment to free reserves as a target of monetary policy. The new reliance on free reserves, like the old, permits interest rates to adjust to some transitory random fluctuations but requires the Federal Reserve to respond to persistent changes in borrowing that change total reserves and money. The new procedure, like the old, requires the Federal Reserve to estimate or forecast bank borrowing and to estimate (or guess at) the relation between interest rates and borrowing. Systematic under - or over-estimates of borrowing formerly induced changes in the free reserve target and now induce changes in the target for non-borrowed reserves. Many of the errors of forecast, perhaps most of them, arise because of the inability to separate persistent and transitory changes in aggregate demand or bank borrowing. The evidence that these errors continue is clear: money rose slowly, or declined, during the unexpectedly sharp decline in output in the spring of 1980, and rose faster than anticipated during the unexpectedly strong recovery in 1980 and 1981.

21 19 The Federal Reserve's re-attachment to free reserves continued to cause money growth to rise faster in periods of expansion than in recessions. Although there have been many desirable changes in the intervening years, the reasons for pro-cyclicality remain much the same as they were eighteen years ago.

22 20 Footnotes * Much of this paper was written while I was a Visiting Professor at the Banking Centre, City University, London. I cannot absolve Karl Brunner from responsibility for what is written, since part is based on work that we have done together and the rest on material we have discussed together. Part of the paper appeared earlier in Meltzer (1981). William Dewald made helpful suggestions on an earlier draft. 1 The study was published in 1964 in three parts by the (then) House Committee on Banking and Currency. See Brunner and Meltzer (1964a), (1964b) and (1964c). 2 His apparent aim was to expand the study that Gert von der Linde and I had done earlier. (1960) 3 See Paul Volcker's statement to the House Committee, November The claim is that banks 1 borrowing is restricted by their "reluctance" to borrow. Earlier versions of the argument recognized that the relation between discount rate and market rate would affect "reluctance" to borrow. 3 a The fact that the cyclical pattern is anticipated does not remove the cyclical effect. The timing of changes is unanticipated, and so are the magnitudes. 4 Keynes praised the Federal Reserve in the 1920's because, he believed, they acted in this way. 5 The model in this section is taken from Brunner and Meltzer (1976a). In (1976b), we show that demand deposits, time deposits and other financial assets are not neglected. If banks set the rates paid on deposits, money, intermediation and bank credit can be accommodated. 6 Excess reserves respond to interest rates and the discount rate in a direction opposite to borrowing. Including excess reserves makes non-borrowed reserves depend on the market rate but introduces no new element.

23 21 Bibliography Brunner, Karl and Meitzer, A.H. (1964a). The Federal Reserve 's Approach to Policymaking. Washington: House Committee on Banking and Currency. and (1964b), The Federal Reserve's Attachment to the Free Reserves Concept. Washington: House Committee on Banking and Currency. and (1964c). An Alternative Approach to the Monetary Mechanism. Washington: House Committee on Banking and Currency. and _ (1976a). "An Aggregative Theory for a Closed Economy" in J.Stein (ed.) Monetarism, pp and (1976b). "Reply, Monetarism: The Principal Issues, Areas of Agreement and the Work Remaining", in J. Stein (ed). Monetarism, pp Dewald, W.G. (1963). "Free Reserves, Total Reserves and Monetary Control". JPE 71 (April) pp Meigs, A.J. (1962). Free Reserves and the Money Supply. Chicago. Meitzer, A.H. (1981). "Central Bank Policy: Some First Principles" Monetary Review of the City University, London, pp. and von der Linde, G. (1960). The Dealer Market for US. Government Securities. Washington: Joint Economic Committee. Thornton, Henry (1802). An Enquiry into the Nature and Effects of the Paper Credit of Great Britain London, reprinted New York, Voleker, Paul A. (1980). Washington: House Committee on Banking, Currency and Urban Affairs. (November).