The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
The Review is published 10 times per year by the Research Department o f the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public free o f charge. Mail requests for subscriptions, back issues, or address changes to: Research Department, Federal Reserve Bank of St. Louis, P.O. Box 442, St. Louis, Missouri 63166. Articles herein may be reprinted provided the source is credited. Please provide the Bank’s Research Department with a copy o f reprinted material. The Discount Rate and Market Interest Rates: What’s the Connection? DANIEL L. THORNTON A 3 lS C O U N T rate changes invariably send new s paper reporters to the phone to call their favorite econom ist to ask the inevitable question: W hat will this do to m arket interest rates? The im pact of dis count rate changes on m arket interest rates appar ently is the source of m uch public confusion and m isunderstanding. This confusion arises from a variety of factors. First, the discount rate is an adm inistered rate set by the Federal Reserve. Second, high interest rates often occur w hen the discount rate is high, w hile low interest rates often occur when the discount rate is low. Finally, discount rate changes often are asso ciated with changes in other interest rates in the same direction. T hese factors have led to a m is understanding about the pre-em inence of the dis count rate in credit m arkets.1 The idea of the pre-em inence of the discount rate steins, in part, from a failure to understand the m echanism through which changes in the discount rate are transm itted to m arket interest rates. The purpose of this article is to analyze the theoretical basis of the link betw een the discount rate and m arket interest rates, and to review the recently ob served relationship betw een these rates in light of the theoretical discussion. THE THEORETICAL CONNECTION BETWEEN THE DISCOUNT RATE AND MARKET INTEREST RATES T he discount rate is the interest rate at which Federal Reserve banks lend reserves to depository institutions, prim arily to enable these institutions to m eet their reserve requirem ents.2 T he relationship 1For a recent statement on the importance of the discount rate, see Saul H. Hymans, et al., “The U.S. Outlook for 1982,” Economic Outlook USA (Winter 1982), p. 3. F ora statement about the dis count rate as a pivotal rate in the market, see George McKenney, The Federal Reserve Discount Window (Rutgers University Press, 1960), p. 6. 2As a result ofthe Monetary Control Actofl980, enacted on March 31, 1980, all depository institutions will have the same reserve betw een the discount rate and m arket interest rates can be illustrated using a simple, static m odel of interest rates called the loanable funds theory. Ac cording to the loanable funds theory, interest rates are determ ined by the intersection of the dem and for and supply of credit, as illustrated in figure 1. The dem and for credit consists of investm ent dem and, governm ent dem and (deficits) and changes in the dem and for m oney.3 The supply of credit is com posed of public and private savings and changes in the supply of m oney. Changes in the discount rate affect m arket interest rates only to the extent that they alter the dem and for or the supply of credit. The Discount Rate and the Supply of Credit Changes in the discount rate directly affect the supply of credit through their im pact on the money supply. To illustrate this, consider the sim ple m odel of the money supply given by: (1) MS = m . B. The supply of nom inal m oney (Ms) is determ ined by the product of the m onetary base (B) and the m oney m ultiplier (m). The m onetary base consists of the total reserves of depository institutions plus cur rency held by the nonbank public. The m oney m ulti plier sum m arizes the effect of all other factors on the m oney supply and, for the purpose of our analysis, is requirem ents. The uniform reserve requirem ents will be phased in over a num ber of years. For more details, see “The Federal Reserve Requirem ents” (Board of Governors of the Federal Reserve System, 1981). The M onetary Control Act also has given thrift institutions access to the discount window through “extended credit borrowing.” For more details, see “The Federal Reserve Discount W indow” (Board of Governors of the Federal Reserve System, 1980). 3The supply curve is sloped positively on the assumption that higher interest rates encourage more savings and because the money supply may be positively related to the interest rate (see footnote 4 below). The demand for loanable funds is downward sloping due to the downward sloping marginal efficiency of investm ent and the inverse relationship betw een the demand for money and interest rates. 3 FEDERAL RESERVE BANK OF ST. LOUIS assum ed to be constant and independent of m arket interest rates.4 Total reserves supplied by the Federal Reserve can be broken dow n into those supplied at the dis count window, called borrow ed reserves (BR), and those supplied through open m arket operations, called nonborrow ed reserves (NBR). The monetary base, therefore, can be w ritten as the sum of BR, NBR and currency held by the nonbank public (C). Thus, equation 1 can be rew ritten as: (2) MS = m . (BR + NBR + C). Changes in the discount rate affect m arket interest rates through their im pact on borrowing from the Federal Reserve. For example, an incrca.se in the discount rate w ill reduce the level of borrowing, ceteris paribus, reducing both the m onetary base and the m oney supply. As a result, the supply-oferedit schedule in figure 1 will shift to the left and m arket interest rates will rise. Reducing the discount rate will have the opposite effect. Discount Rate Changes and Depository Institution Borrowing JUNE/JULY 1982 ing the discount rate.6 Given the nonpecuniary costs associated with discount window adm inistration, an increase in the discount rate w ould reduce the level of borrowing; reductions in the discount rate would have the opposite effect. Later, it was recognized that the relationship betw een the discount rate and borrow ing at the dis count w indow was not quite so sim ple. Borrowing from the Federal Reserve is only one of several m ethods depository institutions use to adjust their reserve positions. They can borrow from the Federal Reserve, buy federal funds in the federal funds market, or sell earning assets, such as short-term Treasury securities.7 It is not simply the level of the discount rate that influences a depository institu tion’s decision to borrow, but the level of the dis count rate relative to rates on alternative adjustm ent assets. A financial institution confronted w ith a reserve deficiency will adjust its reserve position in the least costly m anner. Thus, the im portant variable in the decision to borrow is the so-called least-cost spread betw een the rate on the next best reserve adjustm ent asset and the discount rate. In the aggregate, borrow ing is usually represented by an equation like (3) below, in which (ici) denotes the discount rate and (ia) denotes the interest rate on next best reserve adjustm ent asset.8 (3) BR = aH + a! (ia - id), a„ s= 0, > 0 In this equation, a0 denotes a “ frictional” level of The crucial link betw een the discount rate and m arket interest rates is the connection betw een the discount rate and borrowing from the Federal Re serve. W hen the discount m echanism originally was form ulated, it was assum ed that banks w ould be re luctant to be in debt to the Federal Reserve and 6It is still thought that depository institutions are reluctant to w ould endeavor to repay their indebtedness as soon borrow from the Federal Reserve; however, it has been a long as possible.5 It was thought that the Federal Reserve standing question w hether the reluctance is inherent or induced. rationing at the could control the level of bank borrowing by rein The use of nonpriceClay Andersen, Adiscount window fbegan as early as 1918. See Half-Century o Federal forcing banks’ reluctance to borrow, through the Reserve Policymaking: 1914-1964 (Federal Reserve Bank of adm inistration of the discount window, and by alter- Philadelphia, 1965). ’Prior to Septem ber 1968, depository institutions could adjust their reserve position by reducing the level of their deposits reserves. In September 1968, 4It is sometimes argued that the money supply is positively re and, hence, required lagged reserve accounting, inthe Federal introduced which re lated to interest rates due to changes in the public’s desire to hold Reservereserves in the current week are based on deposit levels various assets in response to interest rate changes. For an quired weeks previous. analysis of the monetary base approach to the money supply of two process, see Jerry L. Jordan, “Elem ents of the Money Stock At the same time, the Federal Reserve changed Regulation D to Determ ination,” this Review (October 1969), pp. 10-19. permit a reserve deficiency carryover equal to 2 percent of re Depository institutions can also adjust 5Winfield Riefler noted that “the reluctance of m em ber banks to quired reserves. by carrying over the deficiency into the their next borrow is not based solely upon the philosophy of reserve banks, reserve positionCarryovers in excess of 2 percent of required reserve week. however. Indeed, that philosophy merely expresses the desire reserves are charged a rate 2 percentage points above the lowest of the great majority of the member banks themselves to remain month in out of d e b t.. . and a feeling on their part that borrowing for profit discount rate in effect on the first day of the calendar that only which the deficiency occurs. It should be noted is unsound. . . . Long before the establishm ent of the reserve system, it was one of the fundamental traditions of sound bank borrowing from the Federal Reserve adds reserves to the system as a whole. ing practice in this country, that a bank’s operations should be confined to the resources which it derives from its stockholders 8The borrowing equation usually includes variables to measure and depositors and interbank borrow ing was at all tim es the degree of reserve pressure of depository institutions, such as lim ited.” Winfield Riefler, Money Rates and Money Markets in the level oforthe change in nonborrowed reserves. Because they the United States (Harper and Bros., 1930), p. 29. have no significance for our purpose, they were ignored here. 4 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 thus the m onetary base. As a result, the supply of credit schedule shifts to the left and m arket interest rates rise until the least-cost spread is restored. Thus, increasing the discount rate will, ceteris pari bus, cause m arket rates to increase. The extent of the increase in the m arket interest rate is determ ined by the sensitivity of borrowing to the least-cost spread (ai) and by the interest sensitivity of the dem and for credit. The more bor rowing is interest-sensitive to the least-cost spread (i.e., the larger ai), the greater will be the shift in the supply of credit for any change in the discount rate. The larger the shift in the supply of credit, the greater the change in the m arket interest rate, for any given cred it dem and curve. Also, the less interest-sensitive the dem and for credit (i.e., the steeper the dem and curve), the greater the change in the m arket interest rate for any given shift in the supply schedule resulting from a change in the discount rate. F ig u re 1 C re d it M arket E quilibrium In te re s t ra te Q u a lit y o f c r e d it The Discount Rate, Interest Rates and Monetary Policy Unfortunately, the above analysis is overly sim ple in that it ignores the role of m onetary policy in influencing the link betw een the discount rate and m arket interest rates. Specifically, the relationship borrowing (i.e., borrowing that occurs even if the betw een the discount rate and m arket interest rates discount rate is not the least costly alternative).9 Given equations 2 and 3, the connection betw een depends on other m onetary policy actions and, in the d isco u n t rate and m arket in te re st rates is particular, on the operating procedure of the Federal apparent. Increases in the discount rate reduce the Reserve. For exam ple, if the Federal Reserve were least-cost spread, w hich reduces borrow ing and to pursue a policy of controlling the level of interest rates, changes in the discount rate w ould have no independent impact on m arket rates. The reason For a discussion of the various theories of depository institutions’ for this is straightforward. U nder an interest rate borrowing, see Rieffer, Money Rates and Money Markets in the targ etin g p ro ced u re, the T rad in g D esk of the United States; Lauchlin Currie, The Supply and Control o f Federal Reserve Bank of New York w ould offset any Money (Harvard U niversity Press, 1934); R obert Turner, m ovem ent in m arket rates by changing the level of Memher-Bank Borrowing (Ohio State University Press, 1938); Murray E. Polakoff “Reluctance Elasticity, Least-Cost, and nonborrow ed reserves through open m arket opera M ember Rank Borrowing: A Suggested Integration,” Journal tions; that is, the leftward shift in the credit supply o f Finance (March 1960), pp. 1-18; Murray Polakoff and William schedule due to an increase in the discount rate Silber, “Reluctance and Member-Bank Borrowing: Additional Evidence,” Journal o f Finance (March 1967), pp. 88-92; and would be offset by a rightward shift resulting from Stephen Goldfeld and Edward Kane, “The Determinants of F ed eral R eserve open m arket o peratio n s. T he M ember Bank Borrowing: An Econometric Study "Journal of im pact of the change in the discount rate on the Finance (September 1966), pp. 499-514. m arket rate would be nil.10 9The fact that there is usually some level of borrowing even when the discount rate is above most other short-term market interest A sim ilar result w ould hold if the Federal Reserve rates is usually construed as prima facie evidence of the inade chose to control the level or growth of the m oney quacy of the alternative mechanisms in providing the reserve adjustment needs of all depository institutions. At the other ex treme, borrowing takes the form of a subsidy if the discount rate is substantially below market rates. See R. Alton Gilbert, "Bene fits of Borrowing from the Federal Reserve when the Discount Rate is Below Market Interest Rates,” this Review (March 1979), pp. 25-32. 10It should be noted that the Federal Reserve cannot “peg” interest rates in an inflationary environment without continually accelerating the growth rate of money. See Milton Friedman, “The Role of Monetary Policy,” American Economic Revieiv (March 1968), pp. 1-17. 5 FEDERAL RESERVE BANK OF ST. LOUIS supply, and if it effected its control through m onetary base (or total reserve) targeting. In this instance, an increase in the discount rate w ould low er the level of borrow ing and, hence, the m onetary base. If this change caused the base to deviate from its desired path, given a m oney growth objective, the Federal Reserve w ould increase nonborrow ed reserves via open-m arket operations in order to return the m one tary base to its desired path. Changes in the discount rate w ould have no independent effect on either the m oney supply or m arket interest rates. The effect of a discount rate change on m arket rates could be significant w hen the Federal Reserve targets on nonborrow ed reserves as it currently does. In this instance, changes in the discount rate alter aggregate borrowing, the m onetary base and the m oney supply as before. The m ovem ent in the base w ould not necessarily be offset through open m arket operations. As long as nonborrow ed reserves are on path, the Federal Reserve m ight choose not to offset changes in borrowings associated with changes in the discount rate.11 U nder the present system of lagged reserve accounting (LRA), how ever, the effect of a discount rate change on aggregate borrow ing, the monetary base and the money supply will be m uch smaller. The Role of Lagged Reserve Accounting The present system of lagged reserve accounting, w hich was introduced in Septem ber 1968, has m ade depository institutions’ dem and for reserves less responsive to in terest rate changes.12 Thus, any change in the supply of reserves, either through changes in NBR or the discount rate, produces a larger change in the rates on reserve adjustm ent assets, such as federal funds and Treasury bills. lr The reader might legitimately inquire as to why the Federal Reserve would not offset all changes in aggregate borrowing if it did not desire a change in the money supply. Unfortunately, there is no simple answ erto this question. Recently the Federal Reserve has attem pted to offset changes in borrowing only if they are viewed to be perm anent in some sense. See David E. Lindsey, “ Nonborrowed Reserve Targeting and M onetary C ontrol” in Improving Money Stock Control: Problems, Solutions and Consequences, conference cosponsored by the Federal Reserve Rank of St. Louis and the C enter for the Study of Am erican Rusiness, W ashington University, October 30-31, 1981 (forthcoming). It should be noted, however, that if the Federal Reserve were to offset all changes in borrowings that move them off their nonborrowed reserve path, they would essentially be targeting on total reserves or the base. 12Since this article was completed, the Federal Reserve Board adopted a resolution to return to contemporaneous reserve accounting. 6 JUNE/JULY 1982 In order to see this point, consider the following sim ple m odel of the m arket for reserves. Reserves are supplied by the Federal Reserve either through open m arket operations or at the discount window. NBR are determ ined solely by Federal R eserve actions and are independent of m arket interest rates. In contrast, BR are related to interest rates via equa tion 3. Depository institutions’ dem and for reserves is com posed of their dem and for required reserves (as determ ined by their deposit levels) and their dem and for excess reserves. U nder a system of con tem p oran eo us reserv e acco un tin g (CRA), both required reserves and excess reserves are assum ed to be negatively related to the rate on reserve adjust m ent assets.13 This equilibrium is illustrated in figure 2a by the intersection of Rs and R^. U nder a system of LRA, current required reserves are determ ined by depository institutions’ deposits of the prior two weeks. The dem and for current required reserves is com pletely insensitive to the interest rates on reserve adjustm ent assets. T he interest responsiveness of the dem and for reserves is d eterm ined solely by the dem and for excess reserves. Thus, dem and for reserves under LRA is less interest-sensitive (steeper), as illustrated by Rj in figure 2b.14 The im pact of a change in the discount rate under CRA and LRA is illustrated in figure 2. An increase in the discount rate reduces the am ount of reserves supplied at each m arket rate, shifting the reserve supply curve to R^. G iven that the dem and for 13Under CRA, depositoiy institutions must weigh the marginal costs of having to adjust their reserve position either at the discount window or in the market with the marginal gain from making additional loans and investm ent and, thereby, creating additional deposits. Thus, when either the discount rate or the rates on alternative adjustm ent assets increase relative to depository institutions’ lending rates, they respond by curtail ing their lending and investm ent activities, which reduces their deposit liabilities and their dem and for required reserves. Thus, the demand for required reserves would be interestsensitive under CRA. Under LRA, the demand for required reserves is determ ined by deposit levels two weeks previous and, hence, is independent of current interest rates. Excess reserves are thought to be held as a source of liquidity for the depository institution. As such, the opportuntiy cost of holding excess reserves is income forgone by not investing them in some income-generating asset, like federal funds. Thus, the dem and for excess reserves is thought to be responsive to changes in market interest rates. The dem and for excess reserves, however, is generally not thought to be responsive to interest rates. 14The equilibrium market rate is shown the same for both CRA and LRA for ease of illustration. This accommodation to con venience does not affect the conclusions. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 F ig u re 2 The Effect of Discount Rate Change Under Contemporaneous and Lagged Reserve Accounting reserves is less interest-sensitive under LRA, inter est rates m ust rise by more in order to restore m arket equilibrium . Thus, a change in the discount rate will result in a larger change in the interest rates on reserve adjustm ent assets, and a sm aller change in aggregate borrowing, the m onetary base and the m oney supply. The Effect on Other Market Rates A change in the discount rate has its initial effect on the m arket interest rate of reserve adjustm ent assets. The extent to w hich a change in the m arket rates of these assets spills over to other m arket in terest rates depends on the substitutability of assets in the portfolios of financial interm ediaries and the public. To illustrate this point, assum e for sim plicity that depository institutions use only one asset as an alternative to borrow ing from the Federal Reserve, and that this asset is not held in the port folios of the rest of the private sector of the economy (e.g., federal funds). Thus, there are no close substi tutes for this asset in the portfolios of nondepository institutions. In this case, the initial im pact of a change in the discount rate w ould be reflected prim arily in the m arket rate of this asset. The effect on other m arket interest rates w ould m aterialize only as depository institutions m odified their lend ing and investm ent activities in light of the higher m arginal cost of reserve adjustm ent funds. The Discount Rate and the Demand for Credit The discount rate also affects m arket interest rates via the dem and for credit through the so-called announcem ent effect. According to this view, the business and financial com m unities regard discount rate changes as signals of the future direction of monetary policy. D iscount rate changes are thus said to alter expectations about the future of business profits and the direction of interest rates. Unfortunately, the im pact of the announcem ent effect depends on the exact nature of these expecta tion effects.15 To illustrate this, consider the follow ing: If the Federal Reserve increased the discount rate, individuals m ight interpret this action as an indication that a slow er rate of m onetary growth, a low er rate of inflation and, hence, low er interest rates will soon follow. If this w ere the case, they m ight 15Warren Smith has argued that the exact impact of the an nouncement effect depends on the market perception of the efficacy of monetary policy, the elasticity of interest rate expec tations and the distributions of these expectations among bor rowers and lenders in the market. See W arren Smith, “ Instru ments of General Monetary Control,” National Banking Review (September 1963), pp. 47-76; “The Discount Rate as a Credit Control W eapon,” Journal o f Political Economy (April 1958), pp. 171-77; and “On the Effectiveness of Monetary Policy,” American Economic Review (September 1956), pp. 588-606. 7 JUNE/JULY 1982 FEDERAL RESERVE BANK OF ST. LOUIS F ig u re 3 Discount Rate Change and Expectations Effects In te re s t ra te reduce their current dem and for credit in anticipa tion of low er future interest rates. The dem and for credit w ould shift to the left and, ceteris paribus, current interest rates w ould fall. T he com bined effects of a discount rate increase on the supply of and the dem and for credit in this instance, under nonborrow ed reserve targeting, are illustrated in figure 3a. An increase in the discount rate shifts both th e supply-of-credit and th e dem and-for-credit schedules to the left. M arket interest rates w ould rise or fall depending on w hether the shift in the dem and curve is small or large, relative to the shift in the supply curve. Conversely, individuals m ight interpret the dis count rate increase as an indication that m arket interest rates will tem porarily rise. In this case, the current dem and for credit w ould increase. U nder these circum stances, an increase in the discount rate w ould shift the supply of credit to the left and the de m and for credit to the right as illustrated in figure 3b. M arket in terest rates w ould then have risen in response to a discount rate change.16 16Warren Smith has commented that, rather than changing the demand for credit in the short run, a discount rate increase may merely induce market participants to shift to different term assets in response to expectations of higher or lower future interest rates. If this were the case, the yield curve would shift with changes in the discount rate. See Smith, “The Discount Rate as a Credit Control W eapon.” 8 In te re s t ra te It should be noted, how ever, that there are those who question w hether there should be any signifi cant expectational effect associated w ith a discount rate change. They argue that a discount rate change is only one of a m yriad of signals that individuals receive concerning the direction of econom ic activ ity and interest rates; therefore, it is doubtful that changes in the discount rate alone have any signifi cant im pact on the dem and for credit. Furtherm ore, it has been noted that changes in the discount rate are som etim es m erely technical adjustments, designed to bring the discount rate in line with changes in m arket interest rates. Thus, if discount rate changes are com monly interpreted as signals of policy change, they may be m isinter preted. It has even been suggested that, given the Federal Reserve Banks’ tendency to m ake these technical adjustm ents, a failure to change the dis count rate w hen m arket rates are changing could be construed as a change in Federal Reserve policy.17 17For a recent interpretation of discount rate changes as technical adjustments, see Hymans, et. al., “The U.S. Economic Outlook for 1982.” For an interesting look at various interpretations of a discount rate change, see Charles Walker, “ Discount Policy in Light of Recent Experience,” Journal o f Finance (May 1957), pp. 223-37; Milton Friedm an, A Program for Monetary Stabil ity (Fordham University Press, 1959); and Ralph A. Young, “Tools and Processes of Monetary Policy,” in Neil H. Jacoby, ed., United States Monetary Policy (Fredrick A. Proeger, 1964), pp. 24-72. FEDERAL RESERVE BANK OF ST. LOUIS The Discount Rate and the Level of Market Interest Rates Up to this point, the discussion has been solely in term s of the effect of changes in the discount rate on m arket interest rates. Nothing has been said about the relationship betw een the level of the discount rate and the level of m arket interest rates. Thus, one additional point m ust be m ade before proceeding to the em pirical analysis. The point is that there are num erous factors that affect the supply of and the dem and for credit besides the discount rate. Thus, there is no one level of m arket interest rates that necessarily corresponds to any given level of the discount rate. It would not be surprising, then, to find th at other factors dom inate m ovem ents in m arket interest rates in the longer run. This is especially true w hen one recognizes that the dis count rate is an adm inistered rate that is changed infrequently. THE DISCOUNT RATE AND MARKET INTEREST RATES: THE RECENT EXPERIENCE Now consider the em pirical evidence on the rela tionship betw een the discount rate and m arket interest rates. The data analyzed is from January 1978 to April 1982, a period chosen because it is tim ely and because it is characterized by markedly different Federal R eserve operating procedures. Until O ctober 6,1979, the Federal Reserve followed a procedure of federal funds rate targeting; that is, it conducted open m arket operations in such a way as to keep the federal funds rate in a narrow range established by the Federal O pen M arket Com m ittee (FOMC). Also, the Federal Reserve followed a policy of changing the discount rate frequently to m aintain a fairly constant federal funds rate/ discount rate differential. Since O ctober 1979, the Federal R eserve has pursued a policy of controlling the m onetary aggre gates through a nonborrow ed reserve targeting procedure.18 Thus, the announced federal funds 18For a discussion of the Federal Reserve’s operating procedure since October 6, 1979, see Stephen Axilrod and David E. Lind sey, “ Federal Reserve System Implementation of Monetary Policy: Analytical Foundations of the New Approach,” Ameri can Economic Review (May 1981), pp. 246-52; R. Alton Gilbert and Michael E. Trebing, “The FOMC in 1980: A Year of Reserve Targeting,” this Review (August/September 1981), pp. 2-22; Richard W. Lang, “The FOMC in 1979: Introducing Reserve Targeting,” this Review (March 1980), pp. 2-25; and Lindsey, “ Nonborrowed Reserve Targeting and M onetary Control.” JUNE/JULY 1982 rate range has been m uch w ider since O ctober 6, and the federal funds rate has exhibited more dayto-day variability. M oreover, the average daily spread betw een this rate and the discount rate has been much w ider.19 Establishing the precise relationship betw een the discount rate and m arket interest rates is extrem ely difficult. Ideally, sets of equations representing the dem and for credit, the supply of credit and a m arket-clearing condition should be specified. In this way, one could not only estim ate the extent of the im pact of a discount rate change on various m arket interest rates, but also identify the m ost sig nificant source of the change (i.e., its effect through the supply of or the dem and for credit).20 In practice, however, this is difficult. As a result, the im pact of a discount rate change on m arket interest rates is usually estim ated w ith a reduced-form m odel, which 19For a discussion of the relationship betw een the federal funds rate and the FOM C’s announced federal funds rate range, see Lang, “The FOMC in 1979: Introducing Reserve Targeting”; and Gilbert and Trebing, “The FOMC in 1980: A Year of Reserve Targeting.” 20One possible way to identify a separate announcem ent effect is to specify a general model of the supply of and the demand formoney. This could be done by simply includingthe discount rate as a separate variable in the demand for money and supply of money functions, and testing to see w hether it has a signifi cant effect on either or both. However, the correspondence betw een the discount rate and market interest rates, due to the fact that discount rate changes tend to follow market interest rate changes, biases this test toward the rejection of the an nouncement effect unless one has precise knowledge of the Federal Reserve’s discountrate reaction function. This problem could be overcome by sim ply estim ating a reduced-form , equilibrium money stock equation. This equation would have the money stock a function of the exogenous variables of the system: aggregate income, the monetary base and the discount rate. A significant discount rate effect would be clear evidence of an announcem ent effect, since the impact of a discount rate change on the money supply would be incorporated in the base. Unfortunately, an insignificant discount rate will not neces sarily im ply the absence of an announcem ent effect; this result could also be obtained if the money supply is relatively interest-inelastic.Thus, one would have to show both that the money supply schedule is interest-clastic and an insignifi cant discount rate in such a reduced-form equation to argue convincingly that there is no announcem ent effect. Regretablv, practical problems make this virtually impossible. It is possible to show that the discount rate is insignificant in a reduced-form equation, employing seasonally adjusted data, for the 10/1979 — 10/1981 period. The money supply equation exhibits some interest elasticity, however, only if seasonally unadjusted data is used. Because personal income (the only available monthly income series) is available only on a season ally adjusted basis, it is impossible to estimate the reducedform equation using seasonally unadjusted data. Thus, the insignificant discount rate variable in the seasonally adjusted, reduced-form equation is not conclusive evidence against an announcem ent effect. 9 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 does not perm it one to differentiate betw een supplyside and dem and-side effects.21 The Discount Rate and Market Rates To determ ine the effect of discount rate changes on m arket in terest rates, the follow ing equation was estim ated using both the federal funds and the 3-month Treasury bills to represent alternative ad justm ent assets: (4) A iat = 10 1 SjAiat.j j= l + t>2 A D R t + et This equation was estim ated using daily data for the period from January 10, 1978, to April 13, 1982, and for subperiods of federal funds rate targeting and NBR targeting.22 The 10-day distributed lag of the m arket rate was included to capture the effect of other factors on the m arket rate before the discount rate change. Table 1 presents estim ates of equation 4.23 The change in the discount rate, denoted by ADR, equals the change only on the day that it becam e effective. T he ADR variable was partitioned into technical ch an g es—A D R T — and n o n tech n ical ch an g es— ADRNT—to test w hether there is a different effect if discount rate changes are m ade solely for tech nical reasons (i.e., to keep the discount rate in line w ith m arket interest rates [see insert, page 12]).24 21Among the studies that have attem pted to test for an an nouncem ent effect using a reduced-form model are: H. Kent Baker and James M. Meyer, “ Impact of Discount Rate Changes on Treasury Bills,” Journal o f Economics and Business (Fall 1980), pp. 43-48; Douglas R. M udd, “ Did D iscount Rate Changes Affect the Foreign Exchange Value of the Dollar During 1978?” this Review (April 1979), pp. 20-26; Rodger Waud, “ Public Interpretation of Federal Reserve Discount Rate Changes: Evidence on the ‘Announcement Effect,’” Econometrica (March 1970), pp. 231-50; and Raymond Lombra and Raymond Torto, “ Discount Rate Changes and Announcement Effects,” Quarterly Journal o f Economics (February 1977), pp. 171-76. 22The data were partitioned on Septem ber 19, 1979, the effective date of the last discount rate change prior to the implementation of the new operating procedures on October 6, 1979. 23The equations were estimated with ordinary least squares (OLS) and with a maximum likelihood procedure that adjusts for first-order autocorrelation. OLS results are reported if the estimate of the coefficient of autocorrelation was not significant ly different from zero. The results, however, were essentially invariant to the estimation technique. 24Discount rate changes were made for purely technical reasons on May 11 and July 3, 1978, and on May 30, June 13, July 28, 1980, and Decem ber 4, 1981. Digitized for10 FRASER Also, a discount rate surcharge variable, ASC, was included in some of the regressions in the NBR tar geting period to capture any effect of the Federal R eserve’s surcharge on large, frequent borrow ers.25 The results for the entire period indicate that a discount rate change has a significant positive effect on both the federal funds and the Treasury bill rates. W hen the equation is estim ated for subperiods of federal funds rate and NBR targeting, how ever, the results change. The coefficient on A DR is not sig nificantly different from zero for the Treasury bill rate during the period of federal funds rate targeting. In contrast, the coefficient on A D R is significant for both m arket rates during the period of NBR targeting. Furtherm ore, the coefficient estim ates on ADR are larger during the latter period. The preceding section noted that discount rate changes would not affect m arket interest rates if the F ed eral R eserve targ eted on them , b u t w ould affect m arket rates under NBR targeting. The results for the Treasury bill rate equation correspond with this analysis, b ut the results from the federal funds rate equation do not. If depository institutions pri m arily rely on the federal funds m arket to adjust their reserve positions, how ever, it is conceivable that m ost of the im pact of a discount rate change could be absorbed by the federal funds rate w ith virtually no spillover to other m arket rates. This even seems likely w hen one recognizes that the F ederal R eserve has never follow ed a policy of rigidly pegging the level of the federal funds rate. In addition, discount rate changes generally w ere m ade in order to keep the rate spread betw een the discount rate and the federal funds rate in a fairly narro w b an d d u rin g th e fu n d s rate ta rg e tin g period.26 Thus, during this period, discount rate changes may have been anticipated and fully re flected in m arket rates before the discount rate change. The Federal Reserve allowed the spread betw een the discount and the federal funds rates to be much larger and variable during the NBR target- 25The Federal Reserve first introduced a surcharge of 3 percent to the basic discount rate for large and frequent borrowers on March 17, 1980. The effective surcharges and dates are: 3 per cent on March 17, 1980, removed May 7, 1980; 2 percent on November 17, 1980; 3 percent on D ecem ber 5, 1980; 4 percent on May 5, 1981; 3 percent on Septem ber 22, 1981; 2 percent on October 13, 1981, removed November 17, 1981. 26The average spread betw'een the discount and the federal funds rates betw een discount rate changes ranged from 50 to 100 basis points. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 Table 1 Estimates of Equation 4 Period Constant ADR ADRNT ADRT ASC Sum of lags P R2/SEE -.2 6 (8.84) .117 .648 -.41 .12 1 Federal Funds Rate 1/10/784/13/82 .007* (.476) .487 (2.831) .003* (.185) 1/10/789/19/79 .023 (2.892) .006* (.217) -.2 5 3 (14.76) .312 (2.247) -1 .4 6 2 .431 -.583 * (1.564) (3.023) 1.378 .323 .252 -.6 8 .332 .414* (1.679) -.1 9 (4.92) .110 -.2 8 3 -.092 * (.239) .250 -.3 2 6 .884 (3.006) .007* (.237) .646 -.61 (16.00) (19.28) .553 (2.300) .0 0 1 * (0.057) .0 0 1 * (.105) -.134 * (.462) .736 (3.678) .023 (2.963) 9/20/794/13/82 -.5 3 6 -.3 2 .114 .803 (8.58) .423 (2.823) -.4 5 3 .375 -.4 2 6 .800 -.1 4 (3.59) .119 .792 -.2 2 .12 0 .797 N.A. .046 .230 .308 -.0 8 (2.64) .051 .228 -.2 2 6 N.A. .067 .095 -.3 8 4 N.A. .068 .095 .286 -.038* (.098) (5.73) .261 .687 (2.206) N.A. .050 .286 -.12 (3.07) .059 .282 (2.501) Treasury Bill Rate 1/10/784/13/82 .0 02 * (.307) .357 (5.655) .000 * (.028) 1/10/789/19/79 .010 (2.103) .473 (6.234) .028* (.454) .0 10 .054* (2.13) 9/20/794/13/82 (.839) .0 0 2 * (.181) .0 0 1 * (.067) -.167* (.970) .434 (4.979) - . 0 0 2* (.157) .003* (.199) .104* (.942) .613 (5.687) .1 1 0 * .349 (.778) .396 (4.546) .094* (1.762) .573 (5.056) .139* (.967) .124 .13 (3.33) .053 .283 .064* .262 N.A. .059 .285 (1.153) The absolute value of the "t-ra tio s " are in parentheses below each coefficient. 'Indicates the coefficient is not sig nifica nt at the .05 level. N.A. indicates the equation was estimated with ordinary least squares. ing period. H ence, discount rate changes may not latter period w ere 100 basis points in absolute value. have been anticipated as w ell during this period, Thus, one could argue that only larger discount rate resulting in a more significant announcem ent effect changes have a significant effect on m arket interest rates. on the dem and side. Furtherm ore, the absolute value of discount rate To further investigate the relationship betw een changes w ere larger in the latter period. The nine discount rate changes and m arket interest rates, the discount rate changes in the early period averaged equations were re-estim ated using both ADRNT 50 basis points, w hile each of the 11 changes in the and ADRT, w hich reflect nontechnical and technical 11 JUNE/JULY 1982 FEDERAL RESERVE BANK OF ST. LOUIS Reasons for Changes in the Discount Rate Date Change Reason May 11, 1978 6V2 to 7% July 3, 1978 7 to 7 V4% August 21, 1978 7'/4 to 7%% Action taken in view o f recent disorderly c on ditions in foreign exchange markets, as well as the continuation of serious dom estic inflation. S ep tem be r22 ,1978 7% to 8% Action taken to bring discount rate in closer alignm ent w ith short-term interest rates, and as a furthe r step to strengthen the dollar. O ctober 16, 1978 8 81/ 2% Action taken to bring the discount rate in closer alignm ent w ith short-term interest rates, and in recognition of the continued high inflation rate and of the current international financial condition. November 1, 1978 81/2 to 91 % /2 Action taken to strengthen the dolla ran d to counter con tinu ing dom estic inflationary pressures. July 20, 1979 9’/2 to 10% Action taken in view of the recent rapid expansion of the m onetary aggregates, to strengthen the do lla r on foreign exchange m arkets and to bring the discount rate into alignm ent w ith short-term interest rates. August 17, 1979 1 0 to 1 0 1/ 2% A ction taken in view of the con tinu ing strong inflationary forces and the relatively rapid expansion in the monetary aggregates. Septem ber 19,1979 1 0 V2 to 1 1 % A ction taken to bring the discount rate into alignm ent w ith short-term interest rates, and to discourage excessive borrow ing from the discou nt w indow . O ctober 9, 1979 11 to 12% Action taken to bring discount rate into closer alignm ent with short-term rates, and to discourage excessive borrow ing. February 15, 1980 12 to 13% Concern about the increased price of im ported o il adding to inflationary pressures underscored the need to raise the discount rate and maintain firm con tro l over the growth of money and credit. May 30, 1980 13 to 12% Action taken entirely in recognition of recent substantial declines in short-term market interest rates to levels below the discount rate. June 13, 1980 to 1 1% 10% 1 1% Essentially the same as above. to Septem ber 26,1980 12 11 10 November 17, 1980 11 to 12% Action taken in view of the current level of short-term interest rates and the recent rapid grow th in the m onetary aggregates and bank credit. December 5, 1980 12 to 13% Action taken in light of the level of market rates and consistent with the existing policy to restrain excessive grow th in money and credit. May 5, 1981 13 to 14% Action taken in light of the current levels in short-term market interest rates and the need to maintain restraint in the monetary and credit aggregates. November 2, 1981 14 to 13% Action taken against the background of recent declines in short-term interest rates and the reduced level of adjustm ent borrow ing at the discou nt w indow . It is consistent w ith a pattern of continued restraint on the growth of money and credit. December 4, 1981 13 to 12% Action taken to bring the discount rate into better alignm ent w ith short-term interest rates that were prevailing recently in the market. July 28, 1980 to to Action taken to bring discount rate in closer alignm ent with short-term interest rates. Essentially the same as above. Essentially the same as above. Action taken as part of a continuing policy to discourage excessive growth in the monetary aggregates. Source: Federal Reserve Bulletins released the m onth of o r one month after the announced change in the discount rate. changes in the discount rate, respectively. D iscount rate changes that are m ade purely for technical reasons m ight have less of an im pact on m arket rates in that either (1) the Federal Reserve offsets their effect on the supply of credit through open m arket operations because they w ere not intended as a change in policy, or (2) the announcem ent effect was w eaker because m arket participants do not view such changes as indications of a change in Federal Digitized for12 FRASER Reserve policy.27 If either of these is true, the coef ficient on ADRNT will be larger than the coefficient on ADR, and the coefficient in A D RT will not be statistically significant. Table 1 shows that these 27Under LRA a change in the discount rate produces a much smaller change in aggregate borrowing than under contempo raneous reserve accounting. Thus, the level of open market operations required to offset the effect of this change on money is much smaller. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 would be reflected in m arket rates rather quickly, so that m ovem ent in these rates betw een discount rate changes w ould be dom inated by other factors.30 This is borne out in a casual observation of the rela tionship betw een the discount rate and m arket rates over this period as shown in chart 1. It is clear from this chart that m arket interest rates varied from levels substantially above the discount rate to levels substantially below it over this period. This m erely reflects the previously noted fact that The Effects of the Surcharge there is no level o f market interest rates that necessarily T he effects of the discount rate surcharge on count rate. corresponds to a given level of the dis m arket in terest rates during the NBR targeting period are mixed. W hen the discount rate surcharge Furtherm ore, there w ere at least three occasions variable is added to the federal funds rate equation, when discount rate changes w ere closely followed the coefficients on the discount rate variables b e by m ovem ents in the 3-month Treasury bill rate in come smaller. Furtherm ore, the coefficients on the the opposite direction (June 13, 1980, D ecem ber 5, surcharge variables are statistically significant. 1980, and May 5, 1981). In the last instance, the T hese results indicate a significant positive sur federal funds rate and the Treasury bill rates m oved charge effect on the federal funds rate. In addition, in opposite directions. The federal funds rate rose they indicate that the estim ates of the discount rate from early May to m id-July 1981, then declined. In effect alone are unduly large w hen the surcharge contrast, the bill rate fell from early May to early variable is ignored. This is likely because of the July, then rose until late August. Thus, it is difficult interaction of discount rate and surcharge effects.29 to find any co n sisten t longer-term relatio n sh ip een of the W hen the surcharge variable is included in the betwarketthe level rates. discount rate and the level of m interest Treasury bill rate equation, the coefficients on the discount rate variables are essentially unaffected. The coefficients on the ASC variable are insignifi CONCLUSIONS cant and small. Thus, it appears that the surcharge M arket interest rates are influenced by num erous has no appreciable im pact on the Treasury bill rate. factors that affect the supply of and dem and for credit. One of these factors is the discount rate. The im pact of the discount rate on m arket rates varies The Levels o f the Discount Rate and with the Federal R eserve’s operating procedures. If Market Rates the Federal Reserve is controlling interest rates, the The fact that discount rate changes have a signifi m onetary base or total reserves, changes in the dis cant im m ediate effect on m arket interest rates does count rate have no effect on interest rates indepen not mean that there is a significant relationship b e dent of the general tenor of monetary policy; the tw een the level of the discount rates and the level Federal Reserve simply w ould offset the effect of of m arket rates. One w ould anticipate that any effect discount rate changes through open m arket opera of a discount rate change on m arket interest rates tions. If the Federal Reserve is targeting on non borrowed reserves, changes in the discount rate are more likely to have an im pact on m arket rates, espe 28The results presented in this section appear to be robust. They cially under lagged reserve accounting. are essentially unchanged if the equation is estimated in level results w ere obtained in every instance. Thus, it appears that only discount rate changes that are m ade for nontechnical reasons have a significant im pact on m arket interest rates. The coefficient on ADRNT in the Treasury bill rate equation, however, was not significant during the early period. D iscount rate changes appear to have had no im pact on the 3-month Treasury bill rate under interest rate target ing, regardless of the reason for the change.28 form, although the R2s are much larger. Also, essentially the same results are obtained by a statistical comparison of the onedav percentage changes in the market rates on the day the dis count rate change became effective with the 10-day and 20-day growth rates prior to the discount rate change. 29It is important to include the surcharge variable in the latter period because some of the changes in the discount rate and the surcharge overlap. The overlapping dates are: November 17, 1980, D ecem ber 5,1980, and May 5, 1981. Failure to include the surcharge could result in a spurious estimate of the discount rate effect. 30In an effort to uncover a possible lagged response of the federal funds rate to discount rate changes, equation 4 was estimated with a 20-day distributed lag of the ADR variable. None of the lagged variables, however, was significant except for the seventh day. It is interesting to note that, since most of the discount rate changes became effective on a Monday, the seventh-day lag would be W ednesday, the close of the “reserve week.” This result, however, is perhaps too tentative to assign any significance to it. 13 FEDERAL RESERVE BANK OF ST. LOUIS C h art JUNE/JULY 1982 I Selected Interest Rates Percent J F W eekly a v e r a g e s of da ily rates M A M J J A 1978 S O N D J F M A M J J A S O N D J 1979 Percent F M A M J J ‘ R a t e c h a n g e s w h e n o n e o f t h e t w e lv e R e s e r v e B a n k s h a s t h e a p p r o v a l o f t h e F e d e r a l R e s e r v e B o a r d to c h a n g e Data indicate that changes in the discount rate have produced a significant, albeit varied, im m e diate im pact on both the federal funds rate and the 3-month Treasury bill rate since January 1978. The effect of a discount rate change on the federal funds rate was significant for periods of both federal funds rate targeting and nonborrow ed reserve targeting. D iscount rate changes significantly affected the Treasury bill rate, however, only in the period of n o n b o rro w ed re serv e targ etin g . F u rth erm o re, changes in the discount rate that w ere m ade for purely technical reasons had no effect on either m arket interest rate, w hile changes in the Federal 14 A S O N 1980 D J F M A M J J 1981 A S O N D J F M A 1982 its d is c o u n t r a t e . R eserve’s surcharge on large, frequent borrowers during the nonborrow ed reserve targeting period had a significant effect only on the federal funds rate. T here is virtually no evidence, how ever, that discount rate changes have had a significant, inde pendent effect on m arket rates in the longer run. Therefore, w hile changes in the discount rate do produce changes in m arket interest rates in the short run, they do not appear to be the most significant factor affecting the level of m arket interest rates in the longer run. Inflation Misinformation and Monetary Policy LAWRENCE S. DAVIDSON C onsum er prices, held back by the recession and an other drop in gasoline and car prices, rose only two-tenths of one percent in February from January’s level, contin uing the sharp decline in the inflation rate. . . . It shows a steady decline in inflation over the past several m onths.1 T H E above excerpt is a perfect exam ple of m is information, a problem that stems from confusing the m easurem ent o f price change with the m easurem ent and causes of inflation. The failure to distinguish the symptoms — like changing gasoline prices — from the causes of inflation can lead to serious policy errors. Th is article presents evidence to support the hypothesis w hich states that efforts to counteract short-term price changes generally are unnecessary and counterproductive.2 We begin by analyzing the behavior of the individual com ponents of the per sonal consum ption expenditures index to determ ine the “causes” of observed quarterly changes in the Lawrence S. Davidson, an associate professor of business eco nomics and public policy at Indiana University, is a visiting scholar at the Federal Reserve Bank of St. Louis. 'New York Times, March 24, 1982. 2This does not imply, however, that such price changes do not impose costs on certain groups. Policymakers may wish to enact legislation to address these problems. It is argued here only that such increases do not warrant macroeconomic remedial policy. Alan Blinder comes to the same conclusion: “ From the macro perspective, the volatility of the CPI often distracts attention from the economy’s underlying or ‘baseline’ rate of inflation. I speculate that extreme swings in the CPI inflation rate occa sionally contribute to extreme swings in national economic policy.” Alan Blinder, “The Consumer Price Index and the M easurem ent of Recent Inflation,” Brookings Papers on Eco nomic Activity (February 1980), p. 564. average price level. We then analyze the perform ance of a variable series constructed to approxim ate the cyclical or nontrend m ovem ents in the m easured inflation rate. An analysis of this series reveals why the public should be reluctant to pressure policy makers into reacting quickly to even large short-run changes in the m easured inflation rate. Finally we present data which suggest that m onetary policies to com bat short-run changes in the inflation rate raise the risk of increasing the underlying or long-term trend of inflation. Two Views of Inflation: Arithmetic vs. Monetary The m easurem ent of inflation necessarily begins with a price index. The most w idely known and used index is the consum er price index (CPI), an index of the average price of a fixed basket of goods and serv ices chosen by a typical urban family. The fixedw eight personal consum ption expenditures price index (PCEI), though sim ilar in most respects to the CPI, is preferable to it in one particular asp ect— its treatm ent of the w eight of housing costs.3 The im portant points for our discussion are: (1) The PC EI is a w eighted average of individual goods prices, (2) T he value of the PC EI in any given m onth can be greatly influenced by changes in the price of indi vidual com m odities. The m easured inflation rate is a sim ple m athe m atical transfonnation of the above price index. For exam ple, instead of saying that the value of the PC EI rose from 100 to 104, the inflation rate expresses this 3For more on this problem, see Blinder, “The Consumer Price Index and the M easurem ent of Recent Inflation,” pp. 539-65. 15 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 price rise as a percentage change. In the above ex am ple, we w ould say that the inflation rate was 4 .104 - 100, inn percent, or (---- --------- ) x 100 percent. change in relative dem and does not cause sustained inflation, though it does cause perm anent changes in relative prices and may cause a tem porary change in the price level. Calculating the inflation rate in this way leads one Relative price changes also occur w hen there are to the valid conclusion that a large increase in the changes in supply conditions.5 T hese include rela price of one good (e.g., food) can cause a large change tive changes in labor productivity, wages or other in the value of the PC EI and, therefore, in the m ea costs associated with the production process. Such sured inflation rate. It is incorrect, how ever, to say changes in a given individual m arket can cause the that food prices cause inflation. cost-per-unit to rise, which in turn causes its relative price to rise. W ith a given income, people who con This is because the arithm etic view tells only part o f the story. Individual prices rise and fall, often in tinue to buy the higher-priced item w ill b e forced to seem ingly random and unpredictable ways. Econ spend less on other goods, w hich puts dow nw ard omists call these relative price changes (since indi pressure on these prices. This “cost-push” example vidual prices are changing relative to one another). has the same outcom e as the xelative dem and ex M onetary and fiscal policy are not designed to be ample: relative prices are perm anently changed, the effective in changing relative prices. T hese and price level may change tem porarily, but inflation is other macro stabilization policies are better suited to unaffected. affect the joint m ovem ent of all prices, or inflation. In the case of increases in the price of inputs like oil, which are used to produce many goods, the in To understand inflation, we m ust first distinguish betw een inflation and relative price changes. Rela creases in the price level may be more pervasive and tive prices are determ ined by the supply and d e sustained. If increases in the price of oil are “pushed m and conditions in the markets for individual goods. through,” causing the retail price of m ost goods to For exam ple, suppose that there w ere a change in rise, individuals whose income has not sim ilarly people’s tastes that caused them to spend more of risen are able to buy few er goods and services at the th eir incom e on recreation and less on durable higher prices. Roth the quantity dem anded and goods, w hile other saving and spending plans re supplied are, therefore, low ered. This low er rate of m ained the same. This change in relative dem and output is perm anent unless incomes rise. A tax re should raise the relative price of recreational goods bate accom panied by an increase in the growth rate and services w hile lowering that of durables. Since of m oney could tem porarily raise incomes enough to total spending rem ains unchanged, the total dem and restore dem and to the earlier rate of production, but for all goods and services is unchanged; only the w ill lead to another increase in the price level as allocation of d em and across m arkets has b een individuals attem pt to buy more of all goods. altered. Therefore, the overall price level is the The point of these exam ples is that a variety of same; only relative prices have changed. factors affecting the cost and relative dem and struc If individuals tem porarily reduced saving so they tures in individual markets can cause relative prices could continue purchasing the sam e am ount of to change. T he co nstrain t th at b in d s th e price durable goods while purchasing more recreational changes in all the m arkets is total spending, or in services, then the total dollar dem and and the price come. W ithout a com m ensurate increase in spend level w ould be higher.4 Individuals w ould be acting ing, none of these factors can cause all prices to rise, as if they were given more income, causing them to that is, none can lead to a perm anent rise in the spend m ore. O nce they replenish their savings, price index. how ever, total dem and and the price level w ill re turn to their original lower levels. Thus, a perm anent The Relationship Retween Inflation and 4If all individuals reduced their savings, there would be less loanable funds available for business investment. Therefore, the increase in consum er spending facilitated by the temporary re duction in saving would be offset by a decline in business spend ing on investm ent goods. Although the consum er price index is temporarily increased, an investm ent deflator would be lower. A combined measure of overall consumer and business prices would be unaffected by this change in saving. 16 Individual Price Changes A rise in the m easured inflation rate always hides a great deal of information. The increase may result 5Foram ore detailed explanation of cost-push inflation, see Dallas S. Batten, “ Inflation: The Cost-Push Myth,” this Review (June/ July 1981), pp. 20-26. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 Table 1 Means and Standard Deviations of Percentage Changes in the PCEI and Its 18 Major Components1 11/1959 - IV/1967 Category W eight Mean Standard deviation 1/1968 - 1/1981 Mean Standard deviation M otor vehicles .052 1.13% 3.85% 5.06% 4.83% Furniture .045 0.30 1.16 3.69 2.56 O ther durables .017 1.27 1.69 5.01 3.36 Food .261 1.82 2.37 6.96 4.58 Clothing .082 1.66 1.55 3.81 2.26 Gas & oil .031 1.62 4.81 10.58 17.29 Fuel oil & coal .012 1.01 4.33 14.72 20.05 O ther nondurables .081 1.78 1.29 5.71 3.31 Housing services .137 1.53 0.45 5.54 1.84 Housing operations .060 1.73 1.69 6.57 3.24 T ransportation services .037 2.32 1.97 7.33 4.82 Personal care services .019 2.76 1.76 7.15 3.02 4.09 Medical services .058 3.76 1.88 7.64 Personal business services .054 3.39 3.44 7.11 3.23 Education & research .013 2.87 1.66 7.50 2.67 Recreation services .022 3.53 1.90 5.11 1.88 Religious & welfare .015 1.61 3.09 7.31 3.66 Net foreign travel .003 1.62 5.29 7.67 14.96 1.000 1.85 0.98 6.34 2.39 PCEI ’ Figures are averages of annualized quarterly rates of change. from all prices rising together, or m erely one price rising by itself. Furtherm ore, this change may prove to be either tem porary or perm anent. Policymakers concerned with the causes of and cure for inflation would find this hidden inform ation highly relevant. Consider the behavior of the individual prices of goods and services included in the PC EI over the past 23 years. Table 1 lists various information about the 18 major categories that make up this index. Because inflation generally has been higher since 1968, the table can be conveniently divided into two periods: a nine-year period before 1968 and a 14year period afterward. The table shows the m ean and the standard deviation for the PC EI and each of its 18 com ponents over both periods. This PC EI is afixedw eight version, which retains the weights from the first quarter of 1959.6 The w eights are the p er 6A fixed-weight index is used because variable-weight indices, when used to compare quarter-to-quarter changes, mix together centages of total ex p en d itu re allocated to each com ponent. The m easured average yearly inflation rate more than tripled from 1.85 percent in the initial period to 6.34 percent in the latter. The standard deviation, a m easure of dispersion around the average, more than doubled. In the 1968-81 period, the annualized quarterly inflation rate averaged 6.34 percent per year, but the average deviation in any particular quarter was about 2.4 percent. This im plies that the inflation rate was betw een 1.5 percent and 11.1 per cent, 95 percent of the tim e. D uring this period price and quantity change. The fixed-weight index is a measure of pure quarter-to-quarter price change. Once fixed, no set of weights perfectly captures the buying patterns of the average household over a long period of time. We arbitrarily chose to use weights from the beginning of the sample period. Using weights from the end of the period would not measurably alter the results here. This is because the weights have not changed enough on individual price components to change the behavior of the over all measured inflation rate. 17 FEDERAL RESERVE BANK OF ST. LOUIS (1968-81), selected categories averaged betw een:7 Housing services: 1.9% to 9.1% M otor vehicles: —1.0% to 14.6% Fuel oil and coal: -24.6% to 54.02% Fuel oil and coal prices, the fastest-growing con sum er prices, averaged over 14 percent per year, fol lowed closely by gas and oil at about 10.6 percent per year. Furniture (3.7 percent) and clothing (3.8 per cent) were the most slowly growing consum er prices. The evidence from table 1 suggests that the m ea sured inflation of the recent past is not the result of all prices rising at the same rate each quarter. T hese figures, how ever, say very little about the role of particular relative prices as causes ofsustained price change. For exam ple, fuel oil and coal prices rose, on average, faster than any of the other prices. But these increases w ere anything but gradual or persistent. O f the 88 quarters from 11/1959 to 1/1981, the inflation rate of fuel oil and coal exceeded the rate ofthe PCEI only 45 tim es. That means during 43 ofthe quarters, fuel oil and coal prices rose more slowly than overall inflation. In 22 of these quarters, the absolute price of fuel oil and coal fell (a negative inflation rate for this category). D uring these 88 quarters, there was not a single episode w hen the inflation rate on fuel oil and coal increased for more than four consecutive quarters. This pattern (though not necessarily the m agnitude) of volatility is typical of m ost price com ponents. Chart 1, w hich presents the growth rates of the PC EI and two of its com ponents, reveals the oscillatory behavior of the PCEI. Note that there has been only one episode since 1959 w hen the overall PC EI inflation rate clim bed consecutively for more than three quarters. More will be said about that episode below. It is cum bersom e to discuss each individual price change and its im plications for the m easured overall inflation rate. Therefore, we introduce a summary m easu re of n o n p ro p o rtio n al or re lativ e p rice 7These confidence intervals assume that quarterly inflation rate changes are normally distributed. A normal distribution roughly means that quarterly inflation rate values fall equally above and below the mean and that most ofthe values are close to the mean. The standard deviation of a random variable measures how much these quarterly inflation rate changes differ from the mean value on the average. The 95 percent confidence interval contains any observations of the quarterly inflation rate that are within two standard deviations of the mean. Since the mean and standard deviation are respectively 6.34 percent and 2.39 percent, there is a 95 percent probability that the quarterly inflation rate is be tween 1.5 percent (= 6.34 percent - 2 (2.39 percent)) and 11.1 percent (= 6.34 percent + 2 (2.39 percent)). Similar confidence intervals can be constructed for any of the inflation rate series. 18 JUNE/JULY 1982 changes (RELP). The RELP series is constructed as follows: For each quarter, subtract the rate of change of the overall PC EI (which is, by definition, the average inflation rate of all com ponents) from each of the 18 com ponent inflation rates. Then m ultiply the absolute value of each of these 18 deviations for this quarter by its w eight and add them .8 This gives the value of RELP for each quarter. If all prices grow at the same rate, RELP w ill equal zero. If, how ever, a few prices rise significantly faster during the quarter than the rest, the value of RELP will rise. If these prices then decelerate (and/ or ifth e others accelerate), so that all prices are again rising more equally, RELP will fall. As chart 2 shows, the RELP m easure has a num ber of interesting features: (1) The greatest increases in RELP cam e in 1972 and 1973 during food-price shocks, during wage and price decontrol and after oil prices quadrupled. (2) W hile the value of RELP fell from the end of 197.3 until 1978, it generally averaged a higher value than before 197.3. (3) W hile RELP show ed no obvious trend before 1970, its average value has been rising since then (from about 1.62 before 1971 to 3.46 thereafter).9 In summary, inflation has been anything but a smooth, upw ard transition in all prices. It is typified by a few prices racing ahead of the others, then falling back relatively quickly. In one episode, RELP accelerated for seven consecutive quarters, but this was an unusual period, typified by a series of food supply shortfalls, wage and price decontrol and, finally, the oil crisis. One im plication of this evidence is that individual price changes have a significant — albeit tem porary 8The same category weights used to construct the overall PCEI are used here. 9While we have noted how RELP arithmetically “causes” price change, others have argued that increases in the inflation rate have caused higher levels of relative price change. One can see from chart 2 that there is a correlation betw een the average percentage change in the PCEI and the average value of RELP. The implication of this finding is that higher average inflation rates, which raise the value of RELP, increasingly confuse eco nomic agents and raise the likelihood of reduced output and higher unemployment rates. See, for example, Mario I. Blejer and Leonardo Leiderm an, “On the Real Effects oflnflation and Relative-Price Variability: Some Empirical Evidence,” Review o f Economics and Statistics (November 1980), pp. 539-44; and Milton Friedman, “Nobel Lecture: Inflation and Unemploy m ent,” Jonmal o f Political Economy (June 1977), pp. 451-72. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 1 G ro w th Rates o f the PCEI a n d Tw o C om ponents 19 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 2 R e la tiv e Price C h a n g e a n d the A v e ra g e In fla tio n R ate 1959 60 61 62 63 64 65 66 67 68 69 — im pact upon overall changes in the m easured inflation rate. This finding has im portant policy con tent. M acroeconomic policies, which are designed to affect incomes or spending, are not efficient devices for com bating the frequent and quickly reversible relative price changes. T herefore, policy aim ed exclusively at stabilizing all changes in the inflation rate will be unproductive. It may even be counter productive if the relative price changes are both highly unpredictable and transient. 20 70 71 72 73 74 75 76 77 78 79 80 1981 Nonmonetary Price Change M onetarists have argued that the dom inant deter m inant of sustained sp ending change is m oney growth. Therefore, they say, it is prim arily sustained m oney growth that produces inflation (a sustained increase in the prices of all goods and services). Past studies have found that the underlying infla tion rate is significantly related to past growth rates FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 of the money supply.10 Carlson finds that, since the ones. T herefore, the average value of A w ould 1970s, about 12 quarters of past m onetary growth be zero. translate into an equal sustained change in the infla It is that discussion does not tion rate. Thus, we assum e that a sim ple 12-quarter imply im portant to note valuethis PDEV is zero. The that the average of moving average of money growth rates approximates for two the monetary influence on sustained inflation.11 For average value of PDEV need not equal zerosuggests reasons. First, the theory discussed here exam ple, if this moving average rate equals 4 per that m onetary grow th affects the average of all cent, then we assum e that m oney is responsible for prices. This does not m ean that m oney growth is the an underlying inflation rate of 4 percent in a given source of all changes in consumer goods prices as quarter. If the inflation rate actually is 6 percent in m easured PCEI. Second, that quarter, then the residual 2 percent can be at affect the by theof inflation for there are factors that rate some tim e w ithout tributed to nonm onetary causes of price change. being a constant source of its variability. For ex M onetarists also believe that there are num erous am ple, the trend rate of growth of labor force pro sources of price change, yet only changes in money ductivity may keep the inflation rate above or below growth can perm anently alter the rate of inflation. any given sustained m onetary growth rate for some Therefore, we expect that nonm onetary factors will period of tim e.12 som etim es affect short-term m easured inflation (3) Even if revers rates. If these nonm onetary sources of m easured ible, there could A w ere transient and totally were be room for policy action if it inflation arise unexpectedly over tim e, and if they This only tem porarily affect the inflation rate, then the predictable.policy. would givetopolicymakers tim e to formulate a According the m onetarist only lasting, predictable and controllable source of negative As will follow positive ones. This view, rela inflation w ould be m onetary growth. tionship, how ever, should not allow for reliable One way to determ ine if the monetary explanation predictions of A over time. of inflation is valid is to exam ine the im pact of non Chart 3 presents and its change, A. m onetary influences on price changes to see if they 1959 to 1981, PDEV PDEV averaged -0 .0 9 and From and A 0.01, have any long-run influences on inflation. To do this, respectively. Prior to 1973, PDEV was generally we define nonm onetary price change as the m ea negative; thereafter it was positive. The overall and sured inflation rate of a given quarter, m inus the subperiod averages are shown in table 2. 12-quarter moving average of m oney growth rates. We then examine the behavior of this series (referred Judging from the average value of PD EV in the to as PDEV) and the changes in it (henceforth called two subperiods, m oney growth does not fully ex A). The m onetarist view of inflation w ould be sup plain the average inflation rate in either period. In ported by a variety of evidence about PDEV and A: the earlier period, inflation was 0.87 percent below growth to (1) If changes in nonm onetary inflation, A, are the 3.56 percent inflationrate of money. From 1973the 1981, however, was 1.21 percent above tem porary, then positive values of A soon would be 6.42 growth rate of m oney.13 follow ed by negative ones. Accordingly, PD EV would rise and then fall toward its original value. 12One measure ol labor productivity is output per hour ot all persons in the private business sector. After increasing at a 2.9 (2) If the increases in A are totally reversible, then percent annual rate from 1961 to 1971, it rose at only a 1.2 over the sample period the sum of the negative As percent annual rate from 1971 to 1980. would be exactly equal to the sum of the positive 13As a check on these results, an alternative proxy for PDEV was 10Keith M. Carlson, “The Lag from Money to Prices,” this Review (October 1980), pp. 3-10; and Denis S. Kamosky, “The Link Between Money and Prices: 1970-76,” this Review (June 1976), pp. 17-23. ''T h ese studies of money and prices use econometric methods and employ distributed lag functions. Furthermore, these rela tionships have been found using the overall gross national product deflator. Therefore, this 12-quarter moving average is only a rough approximation of the influence of money on the trend rate of inflation. However, this moving average as well as longer moving averages and econometric proxies behave quite similarly and therefore the qualitative findings here would not be seriously changed by using these other measures. See foot notes 13 and 16 for more details on one econometric variant. developed. In this case, the monetary contribution to inflation is estimated from an econometric price equation. This equation relates the percentage change in the PCEI to a 12-quarter Almon lag on growth rates of M l, contemporaneous and two lag values of relative energy prices, and two dummy variables for the control and decontrol phases of the Nixon wage-price con trols. PDEV' is calculated by subtracting from the actual rate of change of the deflator its predicted value based only on the monetary part of the estimated equation. The average value of PDEV from 1959 to 1981 is .097, veiy close to the .090 value of the variant reported in the text. The values of PDEV' over the early and later subperiods are -.5 4 and .50, respectively. This version of PDEV suggests a smaller, but still evident, contribution of nonmonetary factors to the measured inflation rate over the two subperiods. 21 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 3 Measures of N onm onetary Inflation NO TE: S h a d e d a re a re p re se n ts th e 95 p e rc e n t c o n fid e n c e in te rv a l. In contrast, the small average values of A in both periods reveal that the average change in PDEV was nearly zero. This suggests that, although factors other than m oney help to detennine the average level of the inflation rate, short-run changes in these nonm onetary factors tend to offset one another over time. O ut of 88 quarters, PDEV fell (A was negative) 45 tim es. Further, there w ere 56 times when a rise in PDEV was followed by a fall, or vice versa. Using a statistical test designed to m easure the regularity of these changes, we find no significant relationship betw een A values over tim e.14 This m eans that changes in the rate of nonm onetary price change are not correlated with past changes. Thus, persistent nonm onetary effects on changes in the inflation rate are not evident, and past values of A are not reliable predictors of future ones. This sim ple test says nothing about the size of changes in PDEV, especially over specific episodes w ithin the sample period. We can use a standard statistical procedure to indicate w hether any given PDEV or A is worth worrying about (large enough to be considered a statistically im portant deviation from zero). For exam ple, in chart 3, note that PDEV is less than zero during most quarters prior to 1973. Is this evidence that nonm onetary factors w ere holding inflation substantially below the rate dictated by money? 14See Edward J. Kane, Economic Statistics and Econometrics (Harper & Row, 1968), especially pages 364-65, for a description of this runs test. To answ er this question, we analyze w hat m ight be called “large” values of PDEV. Values of PDEV 22 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 Table 2 Nonmonetary Price Change 1959-1972 PDEV A 1973-1981 1959-1981 -0.87 % 1.21% -0 .0 9 % -0.06% 0.15% 0.01% or A in chart 3 that fall outside the shaded area are evidence that nonm onetary factors caused large price changes.15 A num ber (say three or four) of consecutive quarters of large and rising values of PD EV or rising As w ould be considered evidence of the persistent effect of nonm onetary factors on price change. Chart 3 reveals that the only run of large PDEV values occurred over the four-quarter period from 1/ 1974 to IV/1974.16 Here, nonm onetary factors con tributed to inflation rising significantly faster than mortey for one year. Another episode, from 11/1972 to IV/1972, which lies near the rejection region, com prises three quarters w hen inflation grew slower than m oney. These episodes deserve additional con sideration since it could be argued that system atic nonm onetary factors caused sustained inflation above and below the m oney growth rate. W hat happened during 1974 had its beginning in IV/1973 w hen the prices of fuel oil and coal rose at an annualized rate of 63 percent, and gas and oil prices 15O ur sample yields only one estimate of the true mean of PDEV. The shaded area in chart 3 is called a confidence interval. This shows by how much the mean could vary in repeated samples without refuting that the population mean is zero. Thus, if we took another independent sample and found a non-zero value for the mean that was inside the confidence interval, it would not refute the hypothesis that the population mean is zero. The area outside the confidence interval is called the rejection re gion. If a sample mean lies in this zone, it rejects the hypothesis that the mean value of nonmonetary inflation is zero. By choos ing a level of confidence higher than 95 percent, say 99 percent, the area in chart 3 would be wider and there would be no runs of PDEV values in the rejection area. Lowering the confidence level to 90 percent does not change the results, though there are two episodes that nearly fall into the rejection region: I/1980-IV/ 1980 and II/1972-IV/1972. The former period witnessed severe oil price shocks while the latter, which is discussed more in the text, occurred during wage and price controls. 16The econometric variant of PDEV discussed in footnote 13 yields the same general conclusion: the largest values of PDEV occur during 1974. Using this variant of PDEV, however, there is no series of consecutive values of PDEV in the rejection area. This is even stronger evidence than that presented in the text for the transitory nature of changes in nonmonetary inflation. increased by 33 percent. In 1/1974 both energy groups again had large annualized rate increases of 91 percent and 63 percent, respectively. T hese in creases, though very large, accounted for only about half of the increase in the m easured inflation rate of the first quarter in 1974. In fact 17 of the 18 com ponent prices accelerated — an historical rarity. By 11/1974 the inflation rate of energy item s, though still high, was falling dram atically. Judging from food and energy prices alone, the overall inflation rate could have fallen as low as 7.4 percent (from 12.4 percent in 1/1974) had it not been for an increase in the relative price of motor vehicles and nondurables (other than food and energy). T he overall inflation rate stayed at 9.6 percent in III/1974 and inched up to 9.7 percent in IV/1974 despite the fact that energy prices had leveled off. In the last quarter, the problem appears to be the 12 percent increase in food prices. Given the large w eight on food prices, m easured inflation could have been down to about 8 percent or less had it not been for this single event. To sum m arize, this historical period found non m onetary sources of inflation persistently greater than zero. It followed, however, on the heels of an unprecedented jum p in the rate of increase of energy prices. It appears that w ithin six m onths the peak nonm onetary effect had been reached.17 Further, it appears that events beyond the second quarter of 1974 w ere separate but adjacent periods of equally bad luck. In the first quarter of 1974, m ost prices responded to the oil crisis. If the subsequent in creases in m otor vehicles, nondurables and food prices at various tim es in the next nine m onths were related to earlier energy price increases, then we do have a single episode. Even in this interpretation, the bulk of the effect of PD EV occurred w ithin six m onths, and traces of it w ere scarce within 12.18 The other interesting episode occurred in 1972 w hen inflation was below the tren d grow th of money. This episode shows that the more stringent 17Using very different methods, John A. Tatom, “Energy Prices and Short-Run Economic Performance,” this Review (January 1981), pp. 3-17, also found a very short peak in the inflation rate attributable to energy prices. His econometric model of the price level used the GNP implicit price deflator and found it to peak within four quarters after the rise in energy prices. 18The Labor D epartm ent attributed the large increases in food prices over the last half of 1974 to poor w eather and crop fail ures. See Toshika Nakayama, Lloyd E. W igren and Paul Monsen, “ Price Changes in 1974 — An Analysis,” Monthly Labor Review (February 1975), especially page 15. 23 JUNE/JULY 1982 FEDERAL RESERVE BANK OF ST. LOUIS C h a rt 4 D eviations from Trend of Inflation and M o n e y G ro w th S h a d e d a re a s re p re s e n t p e rio d s o f la rg e p ric e in c re a s e la s tin g tw o o r m ore q u a rte rs in w h ic h th e m e a s u re d in fla tio n ra te g re w fa s te r th a n its tre n d . NO TE: The c o lo re d lin e h ig h lig h ts the b e lo w -tre n d o r re d u c tio n in m o n e y g ro w th th a t g e n e r a lly fo llo w e d la rg e , a b o v e - tre n d in fla tio n in c re a s e s . phases of the Nixon wage-price controls effectively kept m easured inflation from catching up to trend m oney growth (w hich accelerated from about 5 percent at the end of 1971 to 6.5 percent by the last quarter of 1972). It is interesting that w hen the less restrictive Phase III of the controls began in January 1973, PD EV quickly turned positive as prices began to make up for lost ground. Money Growth and Inflation M isi nfo rmat ion The previous sections suggest that the m ain cause of sustained increases in m easured inflation is not 24 changes in relative prices. The data presented in this section show that the trend growth rate of m oney rose from about 2 percent in the early 1960s to 7 percent in the early 1980s. This section suggests that this rising trend stems from an information problem . We already have shown that the m easured inflation rate often accelerates w hen relative prices change. If policymakers m isread such tem porary increases as perm anent changes in the inflation rate, they may em ploy a contractionary m onetary policy. We show below that tight m oney periods have usually fol lowed large increases in the m easured inflation rate but have been followed by periods of m onetary expansion. At the end of each cycle, the trend growth rate of both m oney and prices has been higher. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 5 Trend G ro w th R ate of M l Percent Percent 8 8 1959 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 1981 S h a d e d a re a s r e p r e s e n t p e r io d s d u r in g w h ic h th e m e a s u re d in f la t io n ra te w a s g re a te r th a n its tre n d . Chart 4 plots the deviations from trend for both the annualized quarterly rates of growth of the CPI and M l.19 The shaded vertical bars represent episodes of large price increases, lasting two or more quarters, in w hich the m easured inflation rate grew faster than its trend. In each case, we find these above-trend price increases accom panied by large reductions in the growth rate of m oney and/or below -trend m onetary growth.20 19Above we argued that the PCEI is a better measure of price change, and therefore the CPI is not used throughout this article. In this section, however, it is important to use the CPI because it is announced more regularly (monthly instead of quarterly) and probably is used more widely. The results in chart 4 are not greatly altered when the PCEI is used instead of CPI, since the two generally move together. One important exception occurred during the first two quarters of 1979. The rate of change of the CPI increased in both quarters; the rate of change of the PCEI fell. Therefore, if the PCEI were used in the analysis in the text, there would be one less historical episode w hen m easured inflation rose in two or more consecutive quarters. T hese reductions, how ever, w ere generally of short duration. C hart 5 presents the 12-quarter m oving average o f th e an n u alized p ercen tag e change in M l. The shaded vertical bars refer to the same periods of large price increase as those in chart 4. C hart 5 shows that the contractions in money fol 20The theme of this article is that all short-term changes in pub lished indices of prices do not dem and policy responses. The evidence, however, suggests that monetary growth has fallen after large short-term m easured price increases. This does not imply that monetary policy is solely determ ined by price changes or that it always responds to them. The behavior of money is determ ined by several factors, and to argue that all monetary changes are attributable to price change would be incorrect. The evidence does suggest, however, that large short term increases in measured inflation above its 12-quarter trend have been associated with subsequent large short-term de creases in the rate of growth of money below its 12-quarter trend. Stanley Fischer, “Relative Shocks, Relative Price Vari ability, and Inflation,” Brookings Papers on Economic Activity (February 1981), pp. 381-431, in an econometric investigation, also finds evidence that monetary contractions trail inflation surges following relative price shocks. See especially page 408. 25 FEDERAL RESERVE BANK OF ST. LOUIS lowing these large price increases generally had only tem porary effects on the trend growth rate of m oney and therefore on a variety of m easures of inflation. These abrupt contractions in m onetary growth generally have been offset by subsequent m onetary expansions. Furtherm ore, these variations in m one tary growth have had severe side effects. Poole finds that monetary decelerations generated recessionary conditions in the U nited States.21 Batten and H aler come to the same conclusion in their analysis of the im pact of short-run m oney growth in the U nited States, Britain, W est Germ any and Italy.22 SUMMARY AND CONCLUSIONS This article provides evidence of an information problem inherent in policies that respond to ob served changes in the m easured inflation rate. The evidence is not inconsistent w ith the theory that short-run bouts of tight m oney follow short periods 21William Poole, “The Relationship of Monetary Decelerations To Business Cycle Peaks: Another Look at the Evidence,” journal o f Finance (June 1975), pp. 697-712. 22Dallas S. Batten and R. W. Hafer, "Short-Run Money Growth Fluctuations and Real Economic Activity: Some Implications for Monetaiy Targeting, this Review (May 1982), pp. 15-20. Digitized for26 FRASER JUNE/JULY 1982 of rising inflation, help to quickly generate reces sionary co n d itio n s, lead to su b seq u en t longer periods of expansionaiy m onetary policy and result in a rising trend growth rate of the m oney supply. The information problem that sets off these cycles is the m isinterpretation of increases in m easured price change as sustained inflation. We have provided evidence that nonm onetary sources of m easured inflation are frequent, highly variable and quickly self-reversible. Therefore, em ploying policy to off set these individual shocks is difficult to accomplish or to justify. This analysis has broad im plications for policy makers. First, short-term changes in m easured infla tion do not call for an activist m onetary policy. Second, a policy of steadily declining m onetary growth will contribute to m ore econom ic stability, while it reduces the underlying rate of inflation. Finally, there is a need to distinguish the nature of the causes of individual bouts of price change as the first step in policy formulation. A sustained increase in the rate of change of all prices, once uncovered, is im portant information w hich policym akers can use to guide monetary and fiscal policies. O f course, the evidence reported here suggests that policym akers could ignore short-run m easurem ents of inflation altogether by simply concentrating on the appro priate long-term m onetary target. Short-Run Money Growth Volatility: Evidence of Misbehaving Money Demand? SCOTT E. HEIN T HE A last tw o years have b een anything b u t tranquil for the U.S. economy. Interest rates, for example, have been high and volatile. Twice during this period they rose to record levels: the prim e rate hit20 percent in April 1980, then rose to 21.5 percent in January 1981. Two recessions have occurred during this brief period, one of w hich apparently still lingers. Significant financial changes have taken place w ith an influx of deposits into m oney m arket m utual funds and an outflow from small tim e and savings deposits. T he nationw ide legalization of NOW accounts in early 1981 also resulted in a siz able reallocation of funds. Amid all of these devel opm ents, m oney growth also has been quite volatile. Should the volatility of short-run m oney growth be a m atter of concern? There appear to be two distinct schools of thought with regard to this question. One school argues that such volatility is not really a problem . It holds that “the need for precise short-run m oney supply control is technically questionable.” 1 The other school argues that such volatility damages the economy. For exam ple, M ilton Friedm an, in evaluating m onetary policy over the last couple of years has w ritten that “the yo-yo swings in monetary growth affected the econom y directly, as w ell as through in terest rates. Each surge in m onetary growth was followed after some m onths by an ac celeration in spendable income, output and em 1Stephen H. Axilrod and David E. Lindsey, “Federal Reserve System Im plem entation of Monetary Policy: Analytical Foun dations of the New Approach,” American Economic Review Papers and Proceedings (May 1981), p. 252. Also, see George W. McKinney, Jr., “The Name of the Game,” Economic View from One Wall Street (February 26, 1982). ploym ent; and each decline in monetary growth, by a retardation.”2 Som ewhat surprisingly, the two schools do not disagree about theoretical issues. Both schools agree that, in theory, the desirability of stabilizing shortrun money growth depends on the stability of the public’s dem and for money. Achieving stable money growth benefits the economy only if the public’s dem and for m oney does not change unexpectedly. The issue that separates the two schools of thought is chiefly an em pirical one: has money dem and been reasonably stable? Those who argue that the vola tility of short-run m oney growth in the past has not been a problem hold that m oney dem and has been subjected to a series of unpredictable shifts. Ac cording to this reasoning, holding the rate of money growth in a tight band w ould have im posed sig nificant costs on the economy. Suppose, for example, the public wants to hold larger money balances. If such a preference is thw arted by an adherence to p re-estab lish ed m onetary targets, the econom y w ould be subjected to unnecessary restraint. Indi viduals seeking to build their m oney balances will reduce their dem and for goods and services and financial assets, resulting in an econom ic slowdown. The other school argues that m oney dem and has been basically stable. In this view, as Friedm an contends, rapid money growth overstim ulates the economy, ultim ately causing inflation, w hile slug gish money growth imposes undue restraint. 2Milton Friedman, “The Yo-Yo Economy,” Newstceek (February 15, 1982). Also, see Milton Friedman, "The Federal Reserve and Monetary Instability,” Wall Street Journal, February 1, 1982. 27 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 This article examines the evidence to determ ine w hether m oney dem and behavior over the last two years has been erratic enough to justify the observed volatility in money growth. ities “ can ‘create’ a product w ithout necessarily being lim ited by the dem and for it.”3 Thus, one should not necessarily interpret changes in money growth as shifts in money dem and. MONEY GROWTH AND THE DEMAND FOR MONEY A CONVENTIONAL MONEY DEMAND EQUATION AND THE EVIDENCE OF SHIFTS Chart 1 provides evidence on short-run (quarterly) money growth volatility. The chart plots, for each quarter since 11/1962, quarterly money growth (at an annual rate) less the average of money growth over the prior 12 quarters. Thus, for exam ple, the —2.0 percent reading for III/1962 shows that m oney grew 2 percentage points less in that quarter than its average growth rate in the previous three years. The volatility shown in this chart has two different dim ensions. One dim ension is simply the magni tude of the deviation from trend. For example, in the third quarter of 1980, m oney grew at a rate 8 per centage points above trend, the largest positive deviation in the last 20 years. In the second quarter of 1980, money grew at a rate over 10 percentage points below trend, the largest negative deviation in the last 20 years. Thus, according to such a m easure, m oney growth has been quite volatile over the last two years. T he second dim ension is the frequency w ith which deviations of money growth relative to trend ch an g e signs. The chart show s that m o n e y g ro w th relative to trend frequently has changed sign from positive to negative, and vice versa, over the last two years. This fluctuation stands in sharp contrast to the historical norm w hereby money growth usually is above or below trend for several quarters in a row. Thus, the increased frequency of change of quarterly m oney growth relative to trend also supports the view that m oney growth over the last two years has been volatile. One can analyze money dem and on a m ore so phisticated basis by using econom etric techniques. This article provides no new analysis on this topic; instead it describes how such evidence can be evaluated. Econom ic theory holds that nom inal m oney bal ances relative to the general price level (generally called “real” m oney balances) are the relevant quan tity m easure for dem and analysis (just as standard dem and theory explains the dem and for physical goods and services, not the dollar value of those goods and services). Thus, w hen one focuses on real m oney, one recognizes that the usefulness of m oney clearly depends on the price of goods and services. For exam ple, if the quantity of m oney that people hold remains unchanged w hile the average price of goods and services fall, a given stock of m oney will have greater value; that is, it will perm it the purchase of more goods and services. Thus, the econom ically m eaningful m easure is the m oney stock relative to the average price of goods and services.4 Analysts com monly hypothesize that real m oney balances move opposite to a change in m arket in terest rates and in tandem with a change in real income. A change in m arket interest rates negatively affects the dem and for real balances, because it represents the opportunity cost of holding money. If m arket interest rates rise, individuals forgo more interest incom e by holding m oney and thus are ex pected to desire less m oney balances. As real income rises, how ever, individuals will w ant larger real m oney balances to purchase more goods and serv ices. Thus, a change in real income is expected to have a sim ilar effect on desired real m oney balances. The increased volatility in m oney growth alone does not dem onstrate that the dem and for money was unstable. Such a conclusion im plicitly holds that the growth of the nom inal m oney stock is com pletely dem and-determ ined, ignoring com pletely the ac tions taken by monetary authorities. Since monetary 3Stephen H. Axilrod, “ Monetary Policy, Money Supply, and the authorities can change bank reserves, reserve re Federal Reserve’s O perating Procedures,” Federal Reserve quirem ents or the discount rate, it is entirely pos Bulletin (January 1982), p. 13. a interpretation of changes in real balances, sible that changes in nom inal m oney growth reflect 4For A.discussion of theDenis S. Karnosky, “Real Money Balances: see B. Balbach and their actions, instead of shifts in the public’s desired A Good Forecasting Device and a Good Policy Target?” this m oney holdings. In other words, m onetary author Review (September 1975), pp. 11-15. 28 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 1 Quarterly M l G ro w th Relative to Trend Percent A Typical Empirical Money Demand Equation Percent the com bination of the estim ated regression coeffi cients and the values of the independent variables. To em pirically investigate the dem and for money, the relationship betw een real m oney balances (M/ P)t and current interest rates (it), real income (yt), and lagged real balances (M/P)t.i, is estim ated using m ultiple regression analysis. T he equation to be estim ated is typically w ritten as: Last period’s real balances are usually included in em pirical estim ations of m oney dem and to capture an assum ed adjustm ent process. Because of relevant transaction costs of adjusting real money balances, it is usually presum ed that actual balances only slowly adjust to desired levels. The lagged value of real balances is included to capture such an adjustm ent (1) (M/P), = ft, + fti, + f t y, + f t (M/P),.! + et. process. By including lagged real m oney balances in The coefficients j8o Pi, P 2 an<3 P3 show how desired the equation, we are assum ing actual real balances real m oney balances respond to changes in the re only partially adjust to current changes in interest spective independent variables. The residual, et, is rates or real income. assum ed to be a random variable that fluctuates about zero. It represents the unexplained variation of A com m on procedure used in evaluating the actual real m oney balances from that predicted by behavior of m oney dem and is to consider how well 29 FEDERAL RESERVE BANK OF ST. LOUIS an em pirical relationship such as equation 1 sim u lates or predicts actual real m oney balances beyond the estim ation period.5 Chart 2 plots the level of real m oney balances sim ulated with equation 1 and the actual real m oney balances for the out-of-sample interval I/1980-I/1982.6 Table 1 sum m arizes these results using a variety of statistical m easures.7 5This procedure apparently dates back to Stephen M. Goldfeld, “The Case of the Missing Money,” Brookings Papers on Eco nomic Activity (3:1976), pp. 683-730. One crucial difference between Goldfeld's evidence and more recent interpretations is that Goldfeld provided evidence of sustained one-sided simu lation error. Logically, Goldfeld’s findings suggest a shift. More recent discussions incorrectly deduce a shift from a single pe riod’s simulation error. This point is subsequently more fully developed. For a more recent application, see Brian Motley, “Innovation and Money Dem and,” Federal Reserve Bank of San Francisco Weekly Letter (January 1, 1982). E stim ating equation 1 in natural log (In) form yields the fol lowing coefficient estimates and summary statistics for the 1/ 1960-IV/1979 sample period (absolute value of t-statistics in parentheses): (1') In (M/P)t = 0.34 + 0.07 In yt - 0.01 In RCP, (1.41) (3.94) (3.27) + 0.85 In (M/P)t.j - 0.02 D1 (13.69) (3.89) R2 = 0.94 where M is M l, P is the GNP deflator, yt is real GNP, and RCP is the commercial paper rate.5 The estimated coefficient on In yt (0.07) indicates that a 1 percent increase in real income this quarter is usually associated with a 0.07 percent increase in real money balances. In a similar vein, the interest rate coefficient suggests that a 1 percent increase in interest rates (for example, from 10.0 percent to 10.1 percent) will lead to a 0.01 percent decline in real balances. Finally, the coefficient on lagged real balances (0.85) indicates that real balances will adjust to desired levels at a rate of 15 percent (1.00-0.85) per quarter. Thus, the long-run response to changes in interest rates and real income is much higher than the short-run response. In the out-of-sample simulations reported below, these coefficients along with actual values of the right-hand side variables are used to project the dependent variable. This relationship is similar to that in R. W. Hafer and Scott E. Hein, “The Shift in Money Demand: What Really H appened?” this Review (February 1982), pp. 11-16. However, the passbook rate variable is excluded since its coefficient was insignificant. The equation was estimated using the Hatanaka two-step pro cedure to correct for first-order serial correlation in the residuals. D1 is a dummy variable that takes on a value of 1 after 1/1974, capturing a one-time shift in the demand for money. The stan dard error of the estimated regression is 0.0045 and the estimate of the serial correlation coefficient is 0.35. 7The equation simulates the natural log of real M l balances. Table 1 presents the antilog of these simulated values, that is, levels of real money balances. Such a transformation, being nonlinear, will not yield optimal predictions. However, it does yield a better “feel” for the size of errors. These simulations are static (when actual values of the lagged dependent variable are used) rather than dynamic (when pre dicted values of the lagged dependent variable are used). See Scott E. Hein, “ Dynamic Forecasting and the Dem and for 30 JUNE/JULY 1982 THE SECOND QUARTER OF 1980 Much hoopla has been made of the difference betw een the sim ulated real balances in the second quarter of 1980 and the actual balances at that time. Real money balances in that period turned out to be alm ost $7 billion below w hat equation 1' predicted. Such a finding has been interpreted as evidence that m oney dem and shifted dow nw ard significantly in 11/1980. Simulation Errors and Shifts Equating a “ shift” with a sim ulation error, how ever, is clearly inappropriate. D eviations of real balances from predicted or sim ulated values do not provide evidence of a behavioral shift in the rela tionship. Recall that w hen the equation is estim ated, it is assum ed that actual real m oney balances will fluctuate randomly around its predicted or sim u lated level. By assum ption, the actual and sim ulated real m oney balances will usually deviate from each other by some unknown random value. Thus, we should expect sim ilar fluctuations to occur out-ofsam ple. W hen considering only one sim ulation error, it is im possible to ascertain w hether one is observing a shift (as represented by a change in one ol the coefficients), or sim ply a large random fluctuation.8 W hen the deviations are consistently one-sided, how ever, one can conclude that a “ shift” in the behavioral relationship has occurred (i.e., one of the coefficients, /30, /3i, /8o or j83, has changed). Chart 2, however, shows no evidence of consistent one-sided errors. Thus, there is little evidence from these sim ulations to indicate a “ shift” in the behavioral relationship. M oreover, recognize that if policym akers incor rectly equate prediction errors with shifts in money dem and, then they will view any observed behavior in real m oney balances as correct. Thus, in either the case of rapid or slow m oney growth, no corrective action w ould be called for. However, if these dis turbances are not true shifts in m oney dem and, policymakers will actually allow m oney growth to fluctuate more than necessary. Money,” this Review (June/July 1980), pp. 13-23, where it is argued that static forecast errors provide a better foundation from which to judge shifts in the dem and for money. 8This is true regardless of the size of the error, because there is always a positive probability of drawing from the extreme tails of a normal probability distribution. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 2 Actual and Simulated Real M l IV 1980 IV 1981 1982 Other Evidence o f a Money Demand Shift These techniques, once in place, lead to permanent decreases in desired real m oney balances relative to Few who argue that a shift occurred in 11/1980 a given level of real income and interest rates. In base their case on die one sim ulation error ofchart 2, other words, m oney dem and shifts dow nw ard fol however. Two auxiliary argum ents also are used to lowing a sharp rise in interest rates. Such an argu support the notion that there was a dow nshift in m ent has been used to explain the abnorm al b e m oney dem and. One argum ent is that a dow nshift havior of m oney dem and since 1974 and is used now occurred “ in response to the very high and record to bolster the evidence of another downshift. levels of short-term interest rates reached in early Chart 2 proves false this explanation of the 11/1980 spring.”9 This argum ent holds that a sharp rise in in terest rates, especially one th at pushes rates decline in real balances. W ere there actually a beyond previous peaks, causes firms and individuals decline in the dem and for real cash balances caused to institute new cash m anagem ent techniques.10 by individuals and firms instituting new cash m an agem ent techniques in response to high interest rates, one should observe a level of real money balances that is consistently below sim ulated levels 9Axilrod and Lindsey, “Federal Reserve System Implementation following the “dow nshift.” of Monetary Policy,” p. 251. 10One of the earliest espousals of this hypothesis can be found in Richard D. Porter, Thomas D. Simpson, and Eileen Mauskopf, “ Financial Innovation and the Monetary Aggregates,” Brook- tugs Papers on Economic Activity (1:1979), pp. 213-29. 31 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 Table 1 Out-of-Sample Simulations of a Money Demand Equation (billions of dollars, seasonally adjusted) Date Actual (Mt/Pt) Simulated (M,/Pt) E rror1 1/1980 $230.1 $230.1 0.0 11/1980 223.0 229.8 - 6 .8 3.9 111/1980 225.8 221.9 IV/1980 226.2 226.0 0.2 1/1981 223.5 226.4 -2 .9 11/1981 225.2 222.7 2.5 111/1981 220.1 225.5 - 5 .4 IV/1981 218.3 219.5 -1 .2 1/1982 221.9 218.2 3.7 Summary Statistics Mean error: 0.6717 Mean absolute error: 2.9621 Root-mean-squared error: 3.6635 T heil’s inequality coefficient: 0.0164 Fraction of error due to (A) Bias: 0.03 (B) Variation: 0.03 (C) Co-variation: 0.94 ''Actual less simulated If this shift w ere perm anent, as this argum ent suggests, the prediction error should rem ain nega tive for all quarters after 11/1980. Chart 2 shows, how ever, that the equation does not consistently overpredict real balances after 11/1980. Actual real balances in II1/1980, instead, w ere slightly higher than the relationship would suggest. Further, real balances were slightly higher, on average, than the equation im plies for the full III/1980-I/1982 period. Thus, one cannot em pirically support the argum ent that a persistent, sizable dow nshift in money d e m and was precipitated by record interest rates in 11/1980. T he second argum ent in support of a m oney dem and dow nshift in 11/1980 contends that the im position of credit controls in M arch 1980 was re sponsible for a decrease in desired real balances. Such an argum ent contradicts econom ic theory, however. W ith credit controls explicitly lim iting the extension of bank credit, individuals and business 32 firms w ould desire larger m oney balances for antic ipated transactions or precautionary purposes. Thus, theory suggests an in crease in m oney dem and during this period, not a decrease. Thus, both auxiliary argum ents in favor of a b e havioral shift in m oney dem and in 11/1980 lack either logical foundation or supportive em pirical evidence. M oreover, if there was a behavioral shift in m oney dem and, the excess supply (supply ex ceeding dem and) of m oney m ust have been offset by an increase in dem and elsew here. In other words, if econom ic participants actually w anted less m oney balances, they m ust have desired m ore of som ething else in exchange. There is little evidence, however, of increased dem and for labor, goods and services, or financial assets in the economy. Further, the generally declining interest rates in this period do not necessarily suggest a behavioral dow nshift in money dem and as many insist. D e clining interest rates do suggest an excess supply of credit, which can come about either because of an increase in credit supply or a decrease in credit dem and. O nly an increase in the supply of credit (as individuals becom e more w illing to give up m oney today in exchange for a prom ise of m oney in the future) w ould be consistent w ith the notion of a dow nshift in m oney dem and in 11/1980, since there is no evidence of an increased dem and elsew here which w ould be required to offset the decreased dem and for both credit and money. Yet, there ap pears little evidence of an increased supply of credit in th is period. Chart 3 s h o w s that the total funds raised by nonfinancial sectors declined m arkedly in 11/1980. Thus, the fall in rates in the second quarter of 1980 is b etter explained by w eakening credit dem ands associated with the recession, rather than the increased supply of credit. If No Shift, Then What? If m oney dem and did not shift in 11/1980, why w ere real money balances low relative to predicted levels? Perhaps the irregular behavior occurred on the “ supply side.” Robert W eintraub has suggested, for example, that slow m oney growth resulted from an unexpected decline in the m oney m ultiplier (the ratio of M l balances to the m onetary base), in re sponse to a sizable shift in the desired currency holdings, as consum ers becam e wary about the acceptability of credit cards during the control pe FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 3 Credit M arket Funds Raised by Nonfinancial Sectors 1976 1977 1978 1979 1980 1981 _______ S ource: B o a rd o f G o v e rn o rs o f the F ed era l Reserve System riod.11 Such a change would drive up die currencydeposit ratio and reduce the m oney m ultiplier. If the m oney m ultiplier declines, banks have to reduce the am ount of deposits they create for a given am ount of source base (or bank reserves). According to W eintraub’s hypothesis, M l balances declined because monetary authorities did not anticipate the increased dem and for currency and offset it by in creasing the base. Therefore, the observed decline in real m oney balances was due, not to a reduction in the demand for real balances, but to this unantici pated change in the supply of m oney caused by an increased dem and for currency as a result of the credit controls. 'Robert W eintraub, The Impact o f the Federal Reserve System’s Monetary Policies on the Nation’s Economy (Second Report), Staff Report of the Subcom m ittee on Domestic M onetary Policy, House Committee on Banking, Finance and Urban Affairs, 96 Cong. 2 Sess. (Government Printing Office, 1980), p. 17. Although individuals w anted to hold as much, if not more, M l balances following the imposition of the credit controls, the banking system precluded these dem ands from being satisfied. Once credit controls were rem oved, the W eintraub hypothesis suggests, the m ultiplier w ould come back w ithin its historical ranges (see chart 4). Thus, real money balances could be expected to return to more his torical levels as well. This is indeed w hat happened: actual real balances rose to about $226 billion in III/1980 (see chart 2). Therefore, one can interpret the behavior of real balances in 11/1980 as evidence of a supply-side lim itation, not a decrease in the dem and for money. In this light, the large sim ulation error is m erely evidence of tem porary disequilibrium . Real money balances deviated from predicted levels, not b e cause individuals desired less money, but because m onetary authorities did not anticipate the effect of 33 FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 C h a rt 4 M l M ultiplier and Ratio of Currency to Total Checkable Deposits credit controls on the way people decided to hold their money. John Judd and John Scadding also argue that “the rapid m onetary deceleration in the second quarter of 1980 (as w ell as the rapid growth in the first and third quarters) was caused, not by a m oney -demand shift, but by a money-supply ‘shock’.” 12 W hile disagree ing w ith W eintraub about the m echanics of the supply shock (Judd and Scadding trace the supply shock to the contraction in bank loans that followed 12John P. Judd and John L. Scadding, “Liability Management, Bank Loans, and Deposit ‘Market’ Disequilibrium ,” Federal Reserve Bank of San Francisco, Economic Review (Summer 1981), p. 21. Digitized for 34 FRASER the Special C redit Control Program of 1980), Judd and Scadding, like W ein trau b , recog n ize th at “changes in the.ssupply ofm oney can dom inate shortrun m ovem ents in the m onetary aggregates.”13 The im portant point here is not to differentiate betw een the W eintraub and Judd-Scadding hypotheses, but to recognize that both views explain the contraction in m oney growth by supply-side occurrences. Thus, deviations of actual real balances from those sim u lated by a m oney dem and equation may be evidence of supply shocks, rather than dem and shifts as many suggest. 13Ibid., p. 22. FEDERAL RESERVE BANK OF ST. LOUIS JUNE/JULY 1982 THE NATIONWIDE NOW ’ EXPERIENCE IN 1981: ANOTHER SHIFT? larger-than-expected balances. In chart 2, w here observed (not shift-adjusted) real m oney balances are shown, however, no consistent underprediction occurred during the last five quarters. In fact, the The sim ulated values of real m oney balances also equation slightly overpreclicts real money balances. allow an evaluation of the im pact of the nationw ide Thus, it does not appear that the nationw ide legal legalization of NOW accounts on the dem and for ization of NOW accounts increased desired M l money. It has been argued that the introduction of balances in any im portant way.16 NOW accounts m ight result in an increased dem and for M l balances, supposedly because of the explicit interest paid on such balances.14 The Federal O pen M arket Com m ittee (FOMC) apparently believed such a result likely. In the first place, the FOM C increased the targeted growth ranges for M l balances in 1981. In addition, the staff of the Federal Reserve Board of Governors devel oped a “ shift-adjusted” M l m easure that w ould subtract the “artificially induced” dem and resulting from the nationw ide introduction of NOW accounts. This adjustm ent was determ ined, in large part, by surveying new NOW account depositors about the original source of the funds they deposited into these accounts. Asking such a question, however, provides little, if any, inform ation about d esired m oney holdings.15 An analysis of a conventional m oney dem and relationship should be a b etter vehicle to address this issue. CONCLUSION Many analysts of m onetary policy have used the recent financial innovations and the volatility of m oney growth as am m unition against pre-established m onetary growth targets. These innovations supposedly have caused unpredictable swings in m oney dem and. T he behavior of actual m oney growth has been taken as evidence of such swings. This article offers a counter argum ent. To begin with, swings in m oney growth are reliable indicators of m oney dem and only to the extent that the supply of m oney has not itself been shocked. In the face of such shocks, large fluctuations in m oney growth cannot be interpreted as evidence of m oney dem and shifts. The second quarter of 1980 was an episode of unusual m oney grow th caused, not by shifting If the nationw ide legalization of NOW accounts m oney dem and, but rather by supply-side occur had actually resulted in an increased desire to hold rences. \1 1 balances fell because the banking system M l balances, the conventional money dem and rela was unable to support the public’s desired deposit tionship should have consistently underpredicted levels. The lesson learned from this episode is that real balances after the nationw ide introduction of these accounts. In other words, actual (real) M l balances should have been consistently above the 1GW hile no apparent irregularities exist when M l is used, this is level sim ulated by the equation, as individuals held not the case when the shift-adjusted measure is employed. 14Much of the discussion about the impact of NOW accounts lias centered on the minimum balance requirem ents of such ac counts. Since minimum balance requirem ents are higher on NOW accounts than on conventional demand deposits, it has been argued that M l will grow. David E. Lindsey, “Nonbor rowed Reserve Targeting and Monetary Control,” paper pre sented at Economic Policy Conference on “Improving MoneyStock Control: Problems, Solutions, and Consequences,” has correctly pointed out, however, that the issue is one of money demand. No adjustment need be made if the demand for M l remains unchanged. 15See John A. Tatom, “Recent Financial Innovations: Have They Distorted the M eaning of M l?” this Review (April 1982), pp. 23-35. Some have argued that the shift adjustm ent was devel oped to capture thesourees of NOW inflows rather than the uses. Such an adjustment should not have been incorporated in the targeting of the money aggregates then! Many have recognized this fact. See, for example, Motley, “ Innovation and Money D em and;” and John W enninger, Lawrence Radecki and Elizabeth Hammond, “Recent Insta bility in the Demand for Money,” Federal Reserve Bank of New York Quarterly Review (Summer 1981), pp. 1-9, where many explanations of such anomalous behavior are provided. The point of the present article, however, is that such explanations are not required. A puzzle exists only when the questionable shift-adjusted measure is used. Just because individuals are moving funds from savings to NOW accounts does not indicate, as the shift-adjustment procedure suggests, that more M l bal ances are desired. There are always people moving funds from savings accounts to dem and deposits. Such movement of funds, however, have never before been taken to suggest that the demand deposit measure should be adjusted. Why should such movements of funds now provide any more useful information? While it is clearly possible that the introduction of explicit interest payments on checkable deposits did result in an in creased demand for M l balances, surveying individuals to find out where funds for new NOW accounts came from is not going to be useful in addressing such an issue. Examining a money demand equation, which is a useful procedure, shows no evi dence of an increased demand. 35 one-tim e deviations of real m oney balances from predicted levels do not necessarily indicate a shift in m oney dem and. Such a deviation could just as well denote a tem porary money m arket disequilibrium , caused by the growth of the m oney supply or a ran dom fluctuation. ship. A conventional m oney dem and equation, how ever, shows evidence of neither sustained p e riods of overprediction (a downshift) nor sustained p eriods of u n d erp red ictio n (an upshift) in the underlying em pirical relationship. Thus, w hile sig nificant financial innovations have occurred in the last two years, there is little evidence that these One precondition for a “ shift” in m oney dem and is innovations resulted in m oney dem and shifts. The a set of consistent, one-sided prediction errors, de M l m easure continues to have significant econom ic rived from an estim ated m oney dem and relation and policy content. Digitized for 36 FRASER