The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Review ___________ Vol. 69, No. 7 August/September 1987 5 Solving the 1980s’ Velocity Puzzle: A Progress Report 24 A Revision in the M onetary Base The Review is published 10 times p e r year by the Research and Public Inform ation Department o f the Federal Reserve Bank o f St. Louis. Single-copy subscriptions are available to the public fre e o f charge. Mail requests f o r subscriptions, back issues, o r address changes to: Research and Public Inform ation Department, Federal Reserve Bank o f St. Louis, P.O. Bo\ 442, St. Louis, M issouri 63166. The views expressed are those o f the individual authors and do not necessarily reflect official positions o f the Federal Reserve Bank o f St. Louis o r the Federal Reserve System. Articles herein may be reprinted provided the source is credited. Please provide the Bank’s Research and Public Inform ation Department with a copy o f reprinted material. Federal R eserve Bank o f St. Louis Review August/Septem ber 1987 In This Issue . . . For more than a third o f a century, the velocity o f money — the ratio o f GNP to M l — grew at a relatively stable rate o f slightly more than 3 percent per year. This stable relationship contributed to the rise o f monetarism and the adoption of m onetaiy aggregate targets by the Federal Reserve. Since 1982, however, there has been a dramatic change in the behavior o f velocity: it has grown more variably and, on average, has declined by more than 2 percent per year. This dramatic and unanticipated turnaround has produced a myriad o f wouldbe explanations. In the first article in this Review, “ Solving the 1980s’ Velocity Puzzle: A Progress Report,” Courtenay C. Stone and Daniel L. Thornton evaluate the validity o f the major theories about velocity’s puzzling behavior in recent years. After grouping the various theories into three categories — misspecification, structural shifts and cyclical factors — the authors show that, with the possible exception of one variant o f the financial innovations explanation, no single theory successfully solves the velocity puzzle. Moreover, w hile certain explanations in tandem appear to offer some insight into velocity’s behavior, they, too, prove insufficient to solve the complete puzzle. The adjusted m onetaiy base, a series published by the Federal Reserve Rank of St. Louis, is a measure o f the Federal Reserve’s influence on the money stock. In the second article in this issue, “A Revision in the M onetaiy Rase,” R. Alton Gilbert describes how this series was recently revised to incorporate the final phase-in of the reserve requirement structure specified in the Monetary Control Act o f 1980. The new structure o f reserve requirements is used in deriving the adjusted m onetaiy base from Novem ber 1980, when the phase-in began, to the present. Data for the prior series, which are used for periods before November 1980, are linked to the post-November 1980 data to create a continuous adjusted m onetaiy base series. 3 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 Solving the 1980s’ Velocity Puzzle: A Progress Report Courtenay C. Stone and Daniel L. Thornton T M . HE velocity o f m oney measures the relationship between nominal incom e and the m oney stock. In its simplest form, the quantity theory o f money states that nominal incom e is equal to the money stock m ultiplied by its velocity. If velocity is reasonably stable, changes in the m oney stock have predictable consequences on nominal income; if the money stock is controllable as well, the quantity theoiy has useful implications for econom ic policy.' The relationship between m oney growth and inflation can be derived from the quantity th eoiy framework by “breaking u p ” nominal incom e into its two components — the price level and real output. Thus, the stability of the monevprice link, holding real output constant, is also related closely to the stability o f velocity. For over a third o f a century — from 1946 to 1981 — the growth o f the velocity o f money, measured as the ratio o f gross national product (GNP) to the narrow money stock (M l), was stable. Its stability contributed Courtenay C. Stone is a senior econom ist and Daniel L. Thornton is a research officer at the Federal Resen/e Bank o f St. Louis. Rosemarie V. M ueller provided research assistance. The authors w ould also like to thank M ichael Darby for helpful suggestions on an earlier draft. ’The money stock need not be perfectly controllable; neither, for that matter, must velocity be constant. M ovem ents in velocity (or its growth), however, must be explainable by the behavior of the variables that influence it. This idea, fundam ental to macroeconom ic policy, was developed by Friedman (1956). See Thornton (1983) for a discussion of the role of velocity for policy purposes. to the rise o f monetarism and the adoption o f m one tary aggregate targets by the Federal Reserve and other central banks around the world. Its stability also re sulted in two em pirically based rules o f thumb that came to be used fairly successfully as guides to m oney grow th’s effects on incom e and inflation. Now, how ever, analysts believe that these rules have failed to explain the course o f incom e and inflation during the 1980s, due to a relatively sudden and unanticipated drop in velocity. Given the important role that velocity plays in eco nomic and policy analysis, it is not surprising that considerable effort has been devoted to solving this velocity puzzle. Unfortunately, these efforts have pro duced a w elter o f com peting and occasionally confus ing explanations. To bring some order to this disarray, this article highlights the problems that have resulted from the puzzling behavior o f velocity in recent years and examines the more prominent explanations o f the velocity puzzle. Because the concept o f velocity stems directly from the theoiy o f the dem and for money, anything that affects velocity can be related to some aspect o f the demand for money. (See shaded insert on the follow ing page.) Because the demand-for-money approach is likely to be less intuitive to the general reader, however, w e w ill discuss the various explanations o f the velocity puzzle in terms o f velocity itself. 5 AUGUST/SEPTEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS Velocity and the Demand for Money The demand for money is usually expressed as de pending on GNP and various other factors, denoted bv Z. That is, the demand for money can he written as: (1) M l = ftGNP, Z). Under certain technical conditions it is possible to re write (1) as: (2) Ml = GNP f(Z), or (31 Ml/GNP = f(Z). Since velocity is simply the reciprocal of equation 3, it can be written as: (4) Velocity = GNP/M1 = —1— = giZ). fIZl it is possible to characterize all "explanations" of the velocity puzzle in terms of equation 4, which is a slightly rewritten version of the demand for money. For example, suppose that equation 1 is not the correct specification of the demand for money, that instead, money demand depends on some broad transactions measure, T, rather than GNP. The true measure of velocity would not be equation 4: instead it would be: 15) T/Ml = k(Z). If velocity as defined in equation 4 were stable for a number of years it would imply that GNP was roughly proportional to T, or GNP = aT, where a is a constant. Consequently, the usual measure of velocity is just a times the "tme" measure, that is, IGNP/M1) = ql(T/M1I. While the level of the usual measure of velocity would be wrong, its movements would mimic movements in the true velocity measure. If this characterization were cor rect, the crucial question is not "Why did velocity de cline?” ; instead, it is "What caused the break in the relationship between T and GNP?" Alternatively, GNP may be proportional to T in the long run but there may be short-run, cyclical variations in GNP relative to T. This characterizes the argument that GDFD is a better measure of transactions than GNP when there are sizable changes in net exports and inventories. The tax-cut explanation of the decline in velocity is analogous, except that a reduction in marginal lax rates 6 can increase the demand for money, permanently re ducing velocity. It argues that the demand for money depends on after-tax income, not on GNP. A cut in the marginal tax rates increases after-tax income and, hence, the demand for money relative to GNP. The usual mea sure of velocity falls, though the "correct measure," based on after-tax income, does not. The arguments based upon the incorrect measure of money are analogous to those that contend the use of GNP is inappropriate. For example, let M* denote the theoretically correct measure of money. If GNP is- the correct scale variable, the correct measure of velocity would be: (61 GNP/M* = w(Z). The extent to which equation 4 is a good proxy for equation 6 depends on the relationship between M* and M. Again, the interesting questions are “Why was equa tion 4 stable for so long?” and "What caused the recent shift in the relationship between M l and M*?” Structur al shift arguments imply that there has been a change in the functional relationship determining veloc ity, that is, a change in g(-). Such a change could be due to a number of factors. The key point is that the former relationship no longer explains velocity. The important issues are to identify the factor(s) that produced this shift and to identify the new relationship. In some sense, specification problems can be thought of as structural shifts because they are presumed to result from some shift in the underlying relationships, for example, be tween GNP and T or between Ml and M*. The cyclical explanation can be characterized by unu sual movements in the factors that determine the de mand for money, Z. Unusual movements in these vari ables can produce the appearance of unusual behavior in velocity. For example, one element of Z is the nominal interest rate. Because the demand for money is inversely related to the nominal rate of interest, a decrease in the interest rate could increase the demand for money rela tive to GNP, causing measured velocity to decline. For this explanation to be valid, however, there should have been a similar rise in velocity when the interest rate was rising. AUGUST/SEPTEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS WHAT WENT WRONG AND WHEN? Tw o fundamental relationships between M l and specific econom ic measures have been supported em pirically for decades. One relationship is the link be tween m oney and GNP, a measure o f total incom e in the economy. The second relationship is the link be tween money and prices. Charts 1 and 2 show the dramatic changes in these relationships that occurred during the 1980s. Chart 1 depicts the behavior o f the incom e velocity (GNP divided by M l ) for the past 40 years; as the chart suggests, something unusual occurred to velocity around 1982. From 1946 through 1981, it rose fairly steadily at about 3.6 percent per year; since then, it has declined at an annual rate o f about 2.4 percent. Chart 2 shows the relationship since 1948 between annual inflation (as measured by the growth o f the GNP deflator! and the average growth in M l over a three-year period; use o f M l ’s trend growth is de signed to capture the long-run impact o f m oney on prices. W hile the rate o f inflation deviated from the trend growth o f M l, sometimes substantially, from 1948 to 1981, the deviations generally w ere temporary. More importantly, the larger deviations w ere attribut able to non-monetary events (for example, govern ment mandated wage-price controls, OPEC oil price actions and the like). Since 1982, however, inflation has been substantially and persistently below the trend growth in M l. These deviations are not easily attribut able to a specific non-monetary event. Numerous attempts have been made to explain the recent changes in velocity. In this paper, these expla nations are grouped loosely into three categories: misspecification, a portmanteau category w e call "struc tural shifts” and cyclical factors.2 MISSPECIFICATION The most w idely used velocity measure, the income velocity o f M l, is calculated bv dividing nominal GNP bv the nominal stock o f M l. Both GNP and M l are empirical counterparts to theoretical concepts that appear in various theories o f the demand for money. One explanation for the shift in velocity is that GNP or 2A number of these are considered in studies by Rasche (1986), Darby et. al. (1987), Hetzel (1987), Trehan and Walsh (1987) and Kretzmer and Porter (1987). The categories considered here are somewhat more general than those considered by Trehan and Walsh. M l or both have becom e less reliable proxies for their corresponding theoretical concepts. This problem is called a specification problem.3 GNP Vs. Transactions Measures One specification problem could arise if m oney is held primarily to make daily transactions.4 If these include intermediate and financial transactions, the usual velocity measure could vary with changes in the proportion of such transactions relative to transac tions on final goods and services. Because GNP mea sures only final output, it w ill differ w idely from the level o f expenditures on all transactions. In this case, GNP is a useful proxy for total transactions only if the proportion o f GNP to total transactions remains rela tively constant. This problem can manifest itself in several wavs. For example, suppose consumers purchase more goods and, as a result, increase their m oney holdings in proportion to their increased desire to spend. If these newlv purchased goods are imported or drawn from domestic inventories o f previously produced goods, GNP w ill remain unchanged w hile the demand for m oney rises. Consequently, the usual measure o f ve locity w ould decline, w hile an alternative measure based on total transactions w ould remain unchanged. Thus, using GNP as the transactions measure to calcu late velocity may produce sizable swings in velocity w henever there are large swings in inventories or net exports. Some analysts have argued that gross domes tic final dem and (GDFD), which equals GNP minus inventory adjustments and net exports, is preferable to GNP as the transactions proxy.5 Unfortunately, the substitution o f GDFD for GNP does not explain the velocity pu zzle o f the 1980s. As chart 3 indicates, this velocity measure performs essentially the same as the usual measure both before and after 1981. Conse quently, simply replacing GNP with GDFD does not explain the protracted velocity decline during the 1980s.K 3See the appendix to Thornton (1983) for an illustration of the specification problem involved in finding the appropriate m easure of “ incom e.'’ “There are two distinct, though not mutually exclusive theories of the demand for money: the transactions approach and the asset ap proach. The asset approach em phasizes the role of m oney as an asset and, hence, as an alternative way of holding wealth. The transactions approach em phasizes the role of money as a medium of exchange. For a useful discussion of this distinction in relation to the velocity issue, see Spindt (1985). 5Radecki and W enninger (1985). 6Rasche (1986) also rejects this explanation for much the same reason. 7 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 C h a rt 1 Velocity 2 \r I 1 1__ 1946 48 50 52 54 56 58 60 62 64 C h a rt 2 Inflation and Ml-Trend Growth J— L J __ 2 1 1__ I__ __ I__ I__ I__ L J __ I__ I__ __ I__I__ I__ __ I__ I__ L -L.J— I— I__ __ I__ 1 — I— — I— I— 1 66 68 70 72 74 76 78 80 82 84 1986 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 C h a rt 3 Velocities of GNP and GDFD Ratio 7.5 Ratio 7.5 Q u a r t e r ly D a t a 7.0 7.0 Wj \ J \\ 6.5 \\ \\ \\ U - j r 6.0 n s 5 .0 6.0 y _ / 5.5 6 .5 5 .5 ' 5 .0 6DF D/ M1 GNP / M1 4.5 4.5 y * 4.0 4.0 3 .5 3.5 1 1 1 3 .0 1961 63 65 1 1 1 1 67 69 71 1 73 Recently, McGibany and Nourzad (1985) have of fered another variant o f the specification problem. They too argue that the dem and for m oney is based on expenditures instead o f current incom e or GNP. In their view, the 1980 tax cut initially increased disposa ble personal and business incom e relative to GNP and, hence, raised desired expenditures relative to GNP; consequently, the tax cut increased the demand for money, resulting in a fall in velocity.7 One way to evaluate this explanation is to look at fhe ratio o f disposable personal incom e to GNP. If their explanation is valid, this ratio should increase when velocity is falling and decrease when velocity is rising. As chart 4 indicates, however, this has not generally happened during the 1980s. While there was an initial expansion in disposable incom e following the tax cut, the ratio o f disposable incom e to GNP has generally declined since 1982." 7Recently, M cGibany and Nourzad (1986) have provided estimates indicating that the demand for m oney is inversely related to the average tax rate. 8Rasche (forthcoming) rejects the tax cut hypothesis by arguing that, for it to explain the velocity decline, marginal tax rates would have had to have fallen continuously over the 1980s. 1 i 75 i 77 1 i 79 1 81 i 1 83 i 85 i 3.0 1 987 Others have argued that the recent velocity decline is related to a sharp rise in financial transactions relative to total output. According to this view, the rise in financial transactions caused an increase in the demand for m oney relative to GNP. One way to assess this claim is to compare velocity measures using broad measures o f financial and non-financial trans actions in place o f GNP.’ These alternatives are pre sented in chart 5. The non-financial transactions ve locity measure shows the same pattern as the GNP velocity measure. Consequently, explanations o f the velocity puzzle that rely on the recent slowing o f GNP growth relative to the growth o f more general nonfinancial transactions measures are implausible. The financial transactions velocity measure does not show the downturn in the 1980s that characterizes the non-financial and GNP-based velocity measures. Nor, however, does it show substantial increases dur ing the 1980s which w ould be required if the rise in financial transactions is to account for the decline in M l velocity. In fact, the annual growth rate o f the financial transactions velocity measure has averaged 9These data were obtained from the Board of G overnors of the Federal Reserve System. 9 AUGUST/SEPTEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS C h a rt 4 Ratios of G N P /M 1 and Disposable Incom e/G NP Ratio 7.5 Rat io .7 4 Q u a r t e r ly D a ta ry\ . 7.0 .73 1.5 .72 A 6.0 /Vs A t \ 5 .0 f \ i ; ( } L 5.5 h / \ i \ a " * V ' V .70 1 1i V 1 \ » 4.5 .71 !\ 1 fi 1 1 I M\ / h > w v 1 \>j .69 i Disposable income/GNP ^ SCA LE I H .68 .67 4 .0 GNP/M1 ^ S C A IE .66 3.5 3 .0 _ L - 1 ..... 1 - . 1961 63 1 1 65 1 i 67 i 69 i i 71 i i 73 i 75 i 77 i 1 79 1 1 1 81 83 1 1 85 .65 1 987 C h a rt 5 Velocities of Financial and N onfin ancial Transaction Debits Ratio 280 Ratio 23.0 Q u a r te r ly D a ta Nonfim ncial transactio n debts/M I s c a le | 250 21.5 n r 220 \ ______ ^ f V 19 0 20.0 \ / 18.5 s\ \ \ \ 160 / /\ 130 1 7. 0 \ \ \ / 15.5 / / r 100 / 14.0 12.5 70 ^ Rn ancial transactio n debts/MI fsc UE 1 40 1970 10FRASER Digitized for 71 72 73 1 1 74 75 76 1 77 1 78 79 1 1 1 80 81 82 83 84 1 85 1 86 11.0 1 987 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 C h a rt 6 Ratios of GNP/M 1 and NYSE/GNP about 10 percent since 1981, somewhat below its 12 percent annual growth rate from 1970 to 1981. If this measure accurately represents total financial transac tions, its velocity movement does not support the view that the velocity problem resulted from a shift from non-financial transactions to financial transactions. A somewhat different way to assess w hether a rise in financial transactions produced the fall in velocity is shown in chart 6; it compares the movement of veloc ity with that o f the annual ratio o f the value o f shares sold on the N ew York Stock Exchange (NYSE) to GNP since 1926.'" While the ratio o f NYSE sales to GNP has risen somewhat during the 1980s, there has been no consistent relationship between this ratio and velocity over the past 60 years. GNP Vs. Wealth Another potential specification problem arises from the use o f GNP to calculate velocity instead o f using a ,0lt has been argued that the recent decline in velocity can be explained by the rise in stock m arket transactions, see Morgan Guarantee (1986). measure o f “permanent incom e” or wealth. The per manent incom e theory o f consumer dem and suggests that individuals primarily base their consumption de cisions on their permanent incom e or wealth, rather than on current income. Analogously, the demand for m oney may be more closely related to permanent incom e or wealth." Panel A in figure 1 illustrates the theoretical relationship between permanent incom e and measured incom e during cyclical fluctuations. If the demand for m oney depends upon permanent income, it will fluctuate less than w ill current income over the business cycle. Thus, measured velocity will rise (fall) as measured incom e increases (decreases) relative to permanent incom e because the amount of m oney held w ill change less than measured income. Chart 7 displays both the usual velocity measure and one based on permanent incom e estimates.12 Once again, it does not appear that the velocity d e cline in the 1980s is explained by movements in cur s o r example, see Friedman and Schwartz (1982, p. 38). ,2The m easure of perm anent incom e used here was suggested by Darby (1972). 11 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 F ig u r e 1 Cyclical M o vem en t in A ctu al GNP and M easu red Velocity and the Effect of a O ne-Tim e Increase in Perm anent Income (B) (A) C h a rt 7 Velocities of G N P /M 1 and Permanent In c o m e /M l Ratio Ratio Q u a r t e r ly D a ta 7 .5 7 .5 7 .0 7 .0 6 .5 6 .5 / V i / 6.0 i* ' 6.0 Permanent in c o m e /M I 5 .5 5 .5 ___ '■ 5 .0 5 .0 ?^GM P/M 1 4 .5 4 .5 4 .0 4 .0 3 .5 3 .0 3 .5 / . I ... I 1961 Digitized for 12FRASER I I 63 I 65 I 1 67 69 1 1 1 71 73 1 1 1 75 1 77 1 ___L 79 1 81 J___ 83 1 1 85 ! 1987 3 .0 FEDERAL RESERVE BANK OF ST. LOUIS rent relative to permanent income. Although the downturns in the permanent incom e velocity m ea sure are less pronounced than those in the current incom e velocity measure, the general downward shift in velocity during the 1980s shows up clearly in the permanent incom e velocity measure. There is an explanation consistent with the perma nent income or wealth approach to the demand for m oney and the observed decline in the incom e veloc ity o f m oney in recent years. Suppose that a rise in permanent incom e or wealth relative to current in come produced a sharp rise in the dem and for money.13In this event, depicted in panel B in figure 1, there w ould be an associated drop in current income velocity. Because wealth is the present value o f the expected future net income, it w ill increase either if expected incom e increases or the expected real interest rate used to discount future incom e declines. If there was a rise in expected incom e without a corresponding in crease in measured incom e during the 1980s, velocity w ould have fallen as the demand for money increased relative to GNP. Eventually, measured incom e w ill rise or expected incom e will decline as individuals realize that their expectations w ill be unfulfilled.14 Conse quently, after sufficient time has elapsed, velocity will return to its former path. If the rise in wealth is due solely to a sharp fall in society’s preference for current relative to future con sumption, however, the path o f measured income w ould be unaffected and the level o f velocity w ould be permanently below its form er path. This possibility seems unlikely, because it implies a permanent fall in the real interest rate.13 ,3Rasche (1986), Santoni (1987) and Kopcke (1986) also consider the wealth explanation. Though their approaches are different, both Rasche and Santoni reject the wealth explanation for the velocity puzzle. Kopcke, on the other hand, finds evidence to support it. His wealth measure, however, includes financial assets that have off setting liabilities; consequently, at best, it represents a proxy for financial transactions. 14Since wealth is the discounted present value of the stream of expected future income, an exogenous increase in wealth relative to current income can result only from a fall in the “ real” interest rate or an increase in the expected future income stream. If these latter expectations are correct, m easured incom e will eventually increase, and velocity will eventually return to its long-run level as either the nominal m oney stock expands or the price level falls. If the expecta tions prove to be wrong, this too will be discovered and velocity will rise subsequently. 15The perm anent fall in the real interest rate necessary to explain the fall in velocity is inconsistent with recent estim ates of the ex ante real interest rates during the 1980s. See Holland (1984). AUGUST/SEPTEMBER 1987 Potential Problems with Using M l Some have suggested that using M l as the m oney stock measure w hen calculating velocity causes sig nificant problems. They argue that the relevant m one tary measure cannot be obtained simply by adding together the stocks o f various “m onetary” assets (cur rency, checkable deposits, and so on), because each component may provide different quantities o f m one tary services per unit. Consequently, critics have sug gested that an index o f the monetary “ services” pro vided by the stock o f all relevant financial assets is preferable to the use o f M l for evaluating the relation ship between m oney and spending or prices.16If this criticism is valid, changes in “ simple-sum” monetary aggregates like M l and M2 may deviate markedly from changes in their underlying monetary services w hen ever substantial shifts among various m onetaiy assets occur. In such cases, the usual measure o f velocity may show sizable variations, w hile those based on the underlying m onetary services measures should be relatively stable.17 Various monetary services indices (MSI) and the MQ measure have been developed; they are currently com piled and maintained by the Federal Reserve Board on an experimental basis.18The MSI1 measure is an index o f the monetary services associated with components o f the M l m oney stock. The M Q measure is an index o f all financial assets that can be directly used in transactions; it incorporates the components of M l plus telephone transfers, m oney market mutual fund balances and m oney market deposit accounts. Chart 8 shows velocity measures based on the MSI1 and MQ.19 These velocity measures show the same general pattern for recent years as the usual M l veloc ity measure. Similar results hold for broader m onetaiy services indices. Consequently, despite their theoreti cal appeal, substituting monetary service flows for M l in measures o f velocity does not explain the recent behavior of velocity.20 16See Batten and Thornton (1985) for a discussion of these issues. ,7This need not be the case, however. See Milbourne (1986). 18The m onetary services indices originally were called Divisia m one tary aggregates; they were developed by W illiam Barnett (1980). The MQ m easure was developed by Paul Spindt (1985). The current monetary services indices differ from the original Divisia m easures in several respects; see Farr and Johnson (1985). 19These alternative m oney measures are only available since 1/1970. “ This interpretation is invariant to alternative measures of income (perm anent income or GDFD). 13 AUGUST/SEPTEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS C h a rt 8 Velocities of M onetary Indexes M Q , MSI 1 and M l 1 97 0 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 1987 N O T E : D a ta n o r m a liz e d f ir s t q u a r t e r 1 9 7 0= 1. STRUCTURAL SH IFTS AND THE VELOCITY PUZZLE Some analysts have suggested that there have been one or more structural shifts in the money/income relationship. Unlike the specification problems previ ously discussed, this explanation presumes that the fundamental relationship between m oney and in com e has changed even if the demand for m oney is correctly specified in terms o f M l and GNP.’ ’ (For a 2'One structural shift argum ent not considered explicitly in the text was presented recently by Roley (1985). He suggested that the velocity puzzle of the 1980s was actually caused by the welldocumented, albeit still unexplained, structural shift in the demand for m oney that took place in 1974. He argues that the downward shift in velocity in the 19 8 2-8 3 period is consistent with the behavior of M1 velocity from 1974 through 1981; it is inconsistent, however, with M1 velocity before 1974. Roley’s observation does not solve the velocity puzzle — although about 13 years have passed, we still don't know w hy m oney demand shifted in the m id-1970s. Furthermore, if his suggestion were valid, the mid-1970s' velocity increase should have been as dram atic as its drop in the 1980s. A glance at chart 1 shows that this is not the case. Moreover, Roley’s M1 series was derived from the flow of funds accounts. When conventional money stock and m oney demand equations are used instead, his results are not confirmed. Digitized for 14FRASER different structural shift argument, see shaded insert on the opposite page. Financial Innovation and Deregulation Several analysts have suggested that the introduc tion o f NOWs, Super NOWs and m oney market deposit accounts (MMDAs) and the removal o f regulation Q interest rate ceilings in recent years have produced a shift in the relationships between M l and both spend ing and inflation. In particular, the redefinition o f M l to include interest-bearing checkable deposits (NOWs and Super NOWs) as w ell as non-interest-bearing de mand deposits and currency is alleged to have altered significantly its “moneyness;” n ow M l is presumed to include a significant amount o f savings balances.22 Consequently, changes in M l resulting from changes in these savings balances are likely to have a smaller 22jh e reader should note the sim ilarity between this and the specifica tion problem. The argum ent here is that savings balances are now effectively hidden among transactions balances so that a given level of interest-bearing checking account balances effectively can repre sent different am ounts of “ transactions m oney.” This is a specifica tion problem, and results from a fundam ental change in the institu tional structure. FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 A Time Series Explanation of the Velocity Puzzle Some analysts have argued that the decline in velocity can be explained by analyzing the statistical properties of economic time-series data. To understand this expla nation for the seemingly abrupt change in velocity in recent years requires a brief discussion of the time series properties of economic variables. The time series properties of a variable describe how it behaves over time. A variable that tends to return to its mean (average) level through time is said to be “station ary." Many economic time series are not stationary in this sense; instead, they show positive or negative growth over time. Time series that display such growth patterns are often said to be “trend stationary” and many economic series, including GNP, prices, the money stock and velocity, have long been viewed as trend stationary, orTS processes. In addition to the TS process, there is an alternative stationary time series process that describes variables whose first- or higher-order differences are stationary; these are called "difference stationary (DS) processes.” The simplest DS process is the well-known random walk. Unlike TS variables, a variable whose behavior fits the random walk has no trend to which it returns as it moves over time.1 Recently, Nelson and Plosser (1982) and Haraf (1986) have concluded that velocity is better represented by the DS than the TS model. According to this view, the recent change in velocity could have been produced bv a large impact on output and prices than previously.-3Specifi cally, there may be extended periods when significant increases in M l produce little or no associated growth in spending or inflation; on these occasions, velocity w ould decline substantially.24Moreover, if the savings portion o f M l is related to GNP differently than its 23From another perspective, the growth rate of old M2 velocity had a trend growth rate of zero; see O tt (1982). Some have argued that new M1 is close to old M2 — old M1 plus tim e and savings deposits, so perhaps the trend growth rate of its velocity, too, will be about zero. W hile the period since 1981 is too short to establish a trend, the growth rate of the new M1 velocity over this period has been about - 2 . 4 percent. 24W hile the experim ental m onetary aggregates should reduce or elim inate such problems, this does not seem to be the case. See Batten and Thornton (1985, pp. 3 2 -3 3 ) for a discussion of this point. shock that caused velocity to wander off in a new direc tion; the probability it will wander back is very small. As Rasche (1986), McCallum (1986) and others have pointed out, however, it is extremely difficult to deter mine whether velocity is a TS or DS process; the test used to discriminate between the competing representations is not powerful when used on time series data like velocity. Consequently, there can be reasonable dis agreement about which representation is more accurate. Unfortunately, knowing whether velocity is better characterized by a TS or DS process does not tell us what we would like to know. Suppose velocity has a DS time series. While this might indicate that an unusually large shock caused velocity to wander off in a new direction in recent years, it leaves a more interesting and more im portant longer-run velocity puzzle: Why was velocity’s growth rate stable for the past 30 years? Is it plausible that, for such a long period, there were no shocks large enough to make velocity walk away from its apparent trend before 1982? Moreover, and more importantly, what economic factors determine velocity’s DS process and what was the nature of the shock that caused veloc ity to walk off in a new direction? Even if economists could be completely certain about the time series pro cess that generates velocity, these fundamental eco nomic questions would remain. 'A random walk may be said to drift; however, this drift param eter is not related to tim e as it would be in a trend stationary process. transaction components, the relationship between the growth rates o f M l and GNP may be permanently altered. These savings balances appear only in the “other checkable deposits” (OCD) component o f M l. Thus, the validity o f this explanation can be examined by comparing the behavior o f velocity measures using M IA (which consists o f currency and non-interestbearing checkable deposits) or currency alone with that o f the M l velocity measure during the 1980s. By increasing the cost o f holding currency and demand deposits, the introduction o f interest-bearing checka ble deposits (NOWs and Super NOWs) should have induced a relative shift from demand deposits and currency into these new accounts; this, in turn, should produce a significant rise in currency and M IA velocity measures. Once individuals’ portfolios are 15 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 C h a rt 9 Velocities of G N P /M 1 and G N P /M 1 A realigned, however, the prior currency and M IA veloc ity relationships should be restored. interest rate w ill cause the dem and for m oney to rise relative to GNP and, hence, velocity w ill decline. Charts 9 and 10 show the M IA - and currencyvelocity measures. The M lA-velocity measure and, to a lesser extent, the currency-velocity measure rose sharply in the first quarter o f 1981 when NOWs were introduced nationwide. Contrary to this structural shift explanation, however, both measures subse quently declined.25 The theoretical basis for this argument stems from basic consumer dem and theory, which argues that the responsiveness o f the demand for a com m odity to changes in its price increases with the number and closeness o f substitute goods. The financial innova tions of the 1980s produced new and close substitutes for traditional dem and deposit and currency com p o nents o f M l. While the interest rate is not the price of money, it represents a significant opportunity cost for holding it. Consequently, the financial innovations o f the 1980s should have increased the responsiveness o f some o f the components o f M l to changes in the interest rate. The “other-checkable-deposit” com p o nent o f M l bears interest, and the interest rate paid on these deposits is n ow free to change with market rates.26Consequently, this com ponent o f M l should be Another explanation for the change in M l velocity is an increased responsiveness o f various M l com p o nents to changes in the interest rate. According to this explanation, the financial innovations o f the 1980s did not necessarily cause a downward shift in velocity due to a shift o f savings balances into transactions ac counts; instead, they altered the sensitivity o f M l bal ances to interest rates. Since the demand for m oney is inversely related to the interest rate, a decline in the 25This is the basis for R asche’s (1986) rejection of this explanation. The introduction of these new accounts, however, m ay have increased the interest elasticity of the demand for the M1A components. Digitized 16 for FRASER “ Businesses cannot hold interest-bearing checking accounts. See G ilbert and Holland (1984) for a sum m ary of the major innovations and deregulations of the 1980s. Also, the currency com ponent of M1 generally is m ore closely tied to real income than to interest rate movements. AUGUST/SEPTEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS C h a r t 10 Velocities of G N P /M 1 and GNP/Currency Ratio Q u a r t e r ly D a ta Ratio P/currency SC ALE^ relatively unresponsive to interest rate movements. This could be mitigated by the fact that rates on these deposits appear to have been slow to adjust to changes in other market interest rates. This view suggests that the relationship between velocity and interest rates should have strengthened since the financial innovations o f the 1980s. Indeed, this pattern is reflected in Chart 11, which shows 1VI1 velocity and the three-month Treasury bill rate. Prior to 1981, velocity appears to be unrelated to m ove ments in the T-bill rate. Since 1981, however, the two have similar patterns. This is consistent with a num ber o f studies which report an increased interest sen sitivity o f M l balances during the 1980s.27 (Additional analysis is provided in the appendix.) It remains to be seen whether the apparent change in M l ’s interest sensitivity alone can account for the aberrant behavior o f M l velocity. 27For example, Hetzel (1987), Trehan and Walsh (1987) and Rasche (1986). Rasche reports mixed results and concludes that this argu ment needs further study and analysis. CYCLICAL EXPLANATIONS OF THE VELOCITY PUZZLE Until now, w e have assumed im plicitly that the supply o f m oney passively expands to m eet society’s demand. Another interpretation argues that substan tial exogenous changes in the supply o f M l can induce cyclical swings in measured velocity because o f their lagged effect on the economy. For example, an acceler ation in the growth rate o f M l initially may produce a less than proportionate rise in the level o f nominal GNP, and, thus, an initial decline in velocity. Eventu ally, however, when the m onetary change has w orked its w ay throughout the econom y fully, the longer-run relationship between M l growth and the rate o f spending is reestablished, and velocity returns to its long-run path. This analysis can explain a continuous fall in veloc ity relative to its underlying trend only if M l growth is continuously accelerating. The “ever-and-ever-faster M l grow th” explanation for the velocity decline in the 17 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 C h a r t 11 Velocity of G N P /M 1 and the Three-Month Treasury Bill Rate 1980s is examined in chart 12. Although m oney growth has been rapid since 1982, it does not appear to have been accelerating fast enough relative to previous years to produce the recent sharp decline in velocity.28 Expected Inflation and Velocity Another explanation is that velocity’s recent behav ior results from changes in the public’s expectations of inflation. According to this view, the demand for m oney is inversely related to the expected rate of “ There is a 4 percentage point spread between peak trend-M1 growth in the 1980s and the late 1970s. Hence, even if there were no nominal output response to the more rapid M1 growth over the entire period, the acceleration in M1 growth, at most, could account for a 4 percentage point decline in trend velocity growth; that is, from about 3 percent to about - 1 percent. In addition, this explanation im plies a significant lengthening in the estimated lag on m oney growth in the St. Louis equation during the 1980s, which has not been confirmed. Another cyclical explanation not considered explicitly in the text has been suggested by Friedman (1983), M ascaro and M eltzer (1983) and Tatom (1983a, 1983b); in their view, an im portant influence on the demand for m oney is m onetary uncertainty. Sup pose that people increase their m oney holdings relative to their current income when they become more uncertain about their future incomes. If m onetary uncertainty increased sufficiently in recent years, this could explain the velocity puzzle. Digitized 18 for FRASER inflation. Thus, when inflation (and presumably in flationary expectations as well) is declining, the d e mand for m oney should rise, and the velocity of m oney should fall. Since the nominal interest rate can be thought o f as com posed o f the real rate plus a premium for the expected rate o f inflation, this expla nation is closely aligned to the interest sensitivity argument. The principal difference between them is that proponents o f the expected-inflation explanation do not argue that the relationship has undergone a structural change.29Judd 11983), Tatom (1983a, 1983b) and Friedman (1983) have argued that the decline in velocity in the 1981-83 period can be attributed pri marily to disinflation and the associated decline in market interest rates that substantially low ered the opportunity costs o f holding m oney relative to GNP. In one sense, this explanation is specious or, at the very least, suspicious if extended to velocity m ove ments in more recent years. If inflationary expecta- “ The expected rate of inflation also could have an independent effect on the dem and for m oney, e.g., m d = f (i, ir e), where ir° is the expected rate of inflation. This issue has not been resolved. FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 C h a r t 12 Short-Run and Trend G row th of M l tions have fallen over the past five years, they must have done so for non-monetary reasons; as chart 2 shows, trend M l growth has risen rapidly since 1983. These non-m onetaiy factors must have been suf ficiently powerful to have swamped the usual in fluence that rapid trend m oney growth has on in flation and inflationary expectations. the rise in the real exchange value o f the dollar during the early 1980s made holding dollars relatively more attractive, increasing the dem and for m oney relative to incom e and reducing velocity.30 Since the real ex change value o f the dollar has generally m oved with changes in the U.S. inflation rate, this argument is closely related to the inflation argument. Furthermore, if disinflation and declining nominal interest rates caused velocity to decline, then, by the same argument, velocity should have risen sharply when inflation accelerated and nominal interest rates rose during the 1970s. Unfortunately, this is not the case. Chart 13 shows M l-velocity and the e?c post inflation rate. While velocity moves with the inflation rate after 1981, it does not appear to be affected sub stantially by the inflation rate over the pre-financialinnovations period. Velocity growth during the 1970s is not rapid enough to support this explanation. This explanation is examined in chart 14, which shows the movements in velocity and the nominal trade-weighted exchange rate since 1973. The nominal rather than the real exchange rate is used for two reasons. First, movements in the nominal exchange rate are more appropriate in assessing the relative returns on two different monies. Second, movements in the nominal and real trade-weighted exchange rates have been highly correlated since 1973. Thus, the Hetzel and Mehra (1985) suggest that the demand for m oney balances varies positively with the real value o f the dollar in foreign exchange markets. Their explanation is based on the currency-substitution hy pothesis, which states that different currencies are close substitutes for each other. In this explanation, “ This argum ent does not seem firm ly based in either the transactions or asset approaches to the dem and for m oney. Except for some border situations, there is very limited substitutability between two currencies for transactions purposes. On the other hand, money balances, even interest-bearing checking accounts, are dominated on a risk-adjusted return criterion by other non-m oney assets. Consequently, it is unlikely that foreign m oney is held as an asset in portfolios. 19 FEDERAL RESERVE BANK OF ST. LOUIS C h a r t 13 Velocity of G N P /M 1 and Inflation C h a r t 14 Velocity of G N P /M 1 and Trade-W eighted Exchange Rate Digitized 20 for FRASER AUGUST/SEPTEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 C h a r t 15 Velocities of G N P /M 1 A and Permanent In co m e /M IA general pattern o f exchange rate movements is the same whether the nominal or real exchange rate is used. Chart 14 shows that the exchange rate explanation does not provide a satisfactory answer to the velocity puzzle. From 1973 to 1981, exchange rate movements appear to have no influence on velocity. W hile velocity did decline from 1981 to 1983, w hen the exchange rate was rising, it also fell sharply in 1985 and 1986 when the exchange rate was plummeting. TWO EXPLANATIONS MAY BE BETTER THAN ONE Darby, Mascaro and M arlow (1987) have recently suggested that the velocity puzzle o f the 1980s is a product o f financial innovation and cyclical effects in measured velocity. Incorporating both effects, chart 15 compares the usual velocity measure with a measure derived by dividing permanent incom e by M IA. There is a sharp rise in the permanent income/MlA velocity measure beginning with the nationwide introduction o f NOW accounts. The movement in this measure following that event is consistent with a gradual ad justment to the initial and subsequent innovations that increased the cost o f holding M IA, such as the introduction o f Super NOWs in January 1984 and the reduction o f the minimum balance requirements on these accounts in January 1985. The permanent income/MlA velocity measure, un like virtually all velocity measures shown in the pre vious charts, does not decline during the bulk o f the 1980s. This measure does not decline until the last three quarters o f 1986; however, it turns up again during the first half o f 1987. Darby, Mascaro and M arlow suggest that the 1986 decline can be explained by the extremely rapid M IA growth during the last three quarters o f the year. Consequently, a combina tion o f the effects o f financial innovations, cyclical movements in GNP and sharp acceleration in M IA growth could account for much o f the velocity puzzle o f the 1980s. SUMMARY AND CONCLUSIONS This article reviews a number o f suggested explana tions o f the puzzling downturn in M l velocity during 21 FEDERAL RESERVE BANK OF ST. LOUIS the 1980s and attempts to assess the credibility of each. Alone, none o f these explanations can account for the behavior o f M l velocity. Perhaps, instead, sev eral influences have com bined to produce the anomalous velocity behavior that has pu zzled many researchers. If there are several influences at work, financial innovations and cyclical variations in measured in come seem to be among the best candidates. This combination works w ell in explaining the velocity puzzle through the first quarter o f 1986. W hen com bined with cyclical variation in velocity induced by rapid m oney growth, it may explain the behavior of velocity through last year. Another explanation that deserves further scrutiny is the possible increased interest sensitivity of M l balances as a result o f m one tary innovations during the 1980s. AUGUST/SEPTEMBER 1987 Judd, John P. “ The Recent Decline in Velocity: Instability in the Demand for M oney or Inflation?” Federal Reserve Bank of San Francisco Economic Review (Sum m er 1983), pp. 12-19. Kopcke, Richard W. “ How Erratic is Money G row th?” Federal Re serve Bank of Boston New England Economic Review (May/June 1986), pp. 3 -2 0 . Kretzmer, Peter E., and Richard D. Porter. “ Total Transaction M ea sures and M1 G row th,” Contemporary Policy Issues (January 1987), pp. 6 4 -7 5 . Mascaro, Angelo, and Allan H. Meltzer. “ Long- and Short-Term Interest Rates in a Risky W orld,” Journal o f Monetary Economics (Novem ber 1983), pp. 4 8 5-518. McCallum, Bennett T. “ On ‘Real’ and ‘Sticky-Price’ Theories of the Business C ycle,” Journal o f Money, Credit and Banking (Novem ber 1986), pp. 3 9 7-414. McGibany, Jam es M., and Farrokh Nourzad. “ Interest Rate Volatil ity and the Demand for M oney,” The Quarterly Review o f Eco nomics a nd Business (Autumn 1986), pp. 7 3 -8 3 . _________ _ “ Income Taxes and the Income Velocity of M oney: An Empirical Analysis,” Journal o f M acroeconom ics (Fall 1985), pp. 5 23-35. REFERENCES Milbourne, Ross. “ Financial Innovations and the Demand for Liquid Assets,” Journal o f Money, Credit, and Banking (Novem ber 1986), pp. 506-11. Barnett, W illiam A. “ Econom ic M onetary Aggregation: An Applica tion of Index Num ber and Aggregation Theory,” Journal o f Econo metrics (Septem ber 1980), pp. 11-48. Morgan Guarantee. “ Demystifying M oney’s Explosive G row th,” M organ Economic Quarterly (March 1986), pp. 10-13. Batten, Dallas S., and Daniel L. Thornton. “ Are W eighted Monetary Aggregates Better Than Sim ple-Sum M1 ?” this Review (June/July 1985), pp. 2 9-40. Nelson, Charles, and Charles Plosser. “ Trends and Random W alks in M acroeconom ic Tim e Series: Some Evidence and Im plica tions,” Journal o f M onetary Economics (Septem ber 1982), pp. 139-62. Darby, Michael R. “ The Allocation of Transitory Income Among C onsum ers' Assets,” American Economic Review (Decem ber 1972), pp. 9 28-41. Ott, Mack. “ Money, Credit and Velocity,” this Review (May 1982), pp. 2 1 -3 4 . Darby, Michael R., Angelo R. Mascaro, and Michael L. Marlow. “ The Empirical Reliability of M onetary Aggregates as Indicators: 19 8 3-1 9 8 6 ,” U.S. Departm ent of Treasury Research Paper No. 8701 (1987). Darby, Michael R., W illiam Poole, David E. Lindsey, Milton Friedman, and Michael J. Bazdarich. “ Recent Behavior of the Velocity of M oney,” Contemporary Policy Issues (January 1987), pp. 1-33. Farr, Helen T., and Deborah Johnson. “ Revisions in the Monetary Services (Divisia) Indexes of M onetary Aggregates,” mimeo, Board of Governors of the Federal Reserve System (1985). Friedman, Milton. “ W hy a Surge in Inflation is Likely Next Year,” Wall Street Journal (Septem ber 1,1983). --------------- - “ The Quantity Theory of M oney — A Restatem ent,” in Studies in the Quantity Theory o f M oney (The University of Chicago Press, 1956), pp. 3 -2 1 . Friedman, Milton, and Anna J. Schwartz. M onetary Trends in the United States and the United Kingdom, 1 867-1975 (University of Chicago Press, 1982). G ilbert, R. Alton, and A. Steven Holland. “ Has the Deregulation of Deposit Interest Rates Raised M ortgage Rates?” this Review (May 1984), pp. 5 -1 5 . Haraf, W illiam S. “ The Recent Behavior of Velocity: Implications for Alternative Policy Rules,” m anuscript (1986). Hetzel, Robert L. “ W ill Recent High Growth Rates of M oney Revive Inflation?” Contemporary Policy Issues (January 1987), pp. 4 1 -5 3 . Hetzel, Robert L., and Yash Mehra. “ U.S. Inflation and the Real Exchange R ate,” manuscript, July 1985. Holland, A. Steven. “ Real Interest Rates: W hat Accounts for Their Recent R ise?” this Review (Decem ber 1984), pp. 18-29. Digitized 22 for FRASER Radecki, Lawrence J., and John W enninger. “ Recent Instability in M 1’s Velocity,” Federal Reserve Bank of New York Quarterly Review (Autumn 1985), pp. 16-22. Rasche, Robert H. “ M1 -Velocity and M oney Demand Functions: Do Stable Relationships Exist?” , Journal o f Monetary Economics (forthcoming). _________ _ “ Velocity and the Choice of Policy Regim es,” The Cato Journal (Fall 1986), pp. 6 6 3-66. Roley, Vance V. “ The Demand for M1 by Households: An Evalua tion of Its Stability,” Federal Reserve Bank of Kansas City Eco nomic Review (April 1985), pp. 17-27. Santoni, G. J. “ Changes in W ealth and the Velocity of M oney,” this Review (March 1987), pp. 16-26. Spindt, Paul A. “ M oney is W hat M oney Does: M onetary Aggrega tion and the Equation of Exchange,” Journal o f Political Economy (February 1985), pp. 175-204. Tatom, John A. “ W as the 1982 Velocity Decline Unusual?” this Review (August/Septem ber 1983a), pp. 5 -1 5 . __________ “ Alternative Explanations of the 1982-83 Decline in V elocity,” in M onetary Targeting and Velocity, C onference Pro ceedings, Federal Reserve Bank of San Francisco, 1983b, pp. 2 2 56. Thornton, Daniel L. “ W hy Does Velocity M atter?” this Review (De cem ber 1983), pp. 5 -1 3 . Thornton, Daniel L., and Dallas S. Batten. “ Lag Length Selection and Tests of G ranger Causality Between M oney and Incom e,” Journal o f Money, Credit and Banking (M ay 1985), pp. 164-78. Trehan, Bharat, and Carl E. Walsh. “ Portfolio Substitution and Recent M1 Behavior,” Contemporary Policy Issues (January 1987), pp. 5 4-63. FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 Appendix To examine whether velocity has become more interest sensitive in the 1980s, the growth rate of M l velocity was regressed on distributed lags of its own past growth rate and changes in the three-month Treasury bill rate for three alternative periods from 1/1960 to II/1987. The results are presented in table 1. The lag length was determined sepa rately for each period using the final prediction error crite rion; see Thornton and Batten (1985). The maximum lag length considered was 12 for the two longer periods and four for the shorter one. The pre-1980 results indicate that neither its own past growth nor that of short-term interest rates significantly influenced M l velocity growth. The lag lengths selected were zero for velocity growth and the contemporaneous and first lag for the change in the Trea sury bill rate. However, even though the lag coefficient on the change in the T-bill rate is both positive as expected and statistically significant at the 5 percent level, the hypothesis that the contemporaneous and lag coefficients are jointly insignificant cannot be rejected at the 5 percent level. A considerably different result emerges when the regres sion is extended to include the 1980s. The lag-length selec tion procedure now chose a sixth-order lag for velocity growth and a fourth-order lag for the change in the T-bill rate. Moreover, the hypothesis that these coefficients are jointly insignificant is rejected at the 5 percent level; con temporaneous and past changes in the Treasuiy bill rate exert a significant influence on current M l velocity growth. When the equation is estimated only for the period of the 1980s, there is again evidence of a statistically significant effect of interest rates on M l velocity. Indeed, the sum of the distributed lag coefficients on the Treasury-bill rate is posi tive and significant, indicating a longer-run positive rela tionship between M l velocity and interest rates that does not appear to have existed in the prior period. Hence, these results are consistent with the hypothesis that the interest sensitivity of M l balances changed significantly following the monetary deregulation and financial innovations of the 1980s. It will take more research, however, to determine how much of the velocity puzzle can be attributed to this factor. Table 1 Estimates of a Velocity Growth Rate Equation1 I/1960-IV/1979 Intercept 1/1960-11/1987 3.138* (7.36) 0.378 (0.71) 0.262* (2.70) VDOT-1 VDOT-2 - 0 .0 1 9 (0.20) VDOT-3 0.132 (1.32) VDOT-4 0.214* (2.34) VDOT-5 - 0 .0 9 5 VDOT-6 ATBR-0 ATBR-1 1/1980-11/1987 - 0 .2 1 7 (1.06) 0.202* (2.31) - 0 .0 2 5 (0.04) 0.740 1.579* (2.07) ATBR-2 0.889 (1.73) 3.066* (5.88) - 1 .3 5 4 * (2.28) - 1 .7 7 5 * (2.35) 0.774 ATBR-4 (1.58) 2.999’ (6.16) ATBR-3 (1.35) 0.806 (1.65) - 1 .1 1 7 * (2.01) ZATBR 1.554 R2 0.033 SEE (1.80) 2.041 (1.83) 2.987* (2.25) 0.418 0.711 3.759 4.014 3.747 — 4.204* 9.437* F-TBR 2.359 11.315* 15.710* DW 1.998 2.029 2.096 F-VDOT (0.31) 0.512* (3.07) ' 111 statistics in parentheses. •Indicates statistical significance at the 5 percent level. 23 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 A Revision in the Monetary Base li. Alton Gilbert T J L HE Monetary Control Act o f 1980 mandated a substantial change in the structure o f reserve require ments faced by depository institutions. The reserve requirement structure was phased in over a sevenyear period from Novem ber 1980 to September 1987. The adjusted monetary base, a measure o f the Fed eral Reserve’s influence on the m oney stock, has been revised to reflect this new structure. This article ex plains w hy the base series was revised and describes the difference between the previous and revised series. THE A DJUSTED MONETARY BASE: PURPO SE AND COMPOSITION The adjusted monetary base (AMB) is designed to be a single measure o f all Federal Reserve actions that influence the m oney stock, including changes in re serve requirements. It is equal to the source base plus the reserve adjustment magnitude (RAM). The source base consists o f total currency outstand ing (held by the public and in the vaults o f depository institutions) plus the reserve balances o f depository institutions at Federal Reserve Banks.1The level o f the m oney stock (currency in the hands o f the public plus checkable deposits) that can be supported with a given level o f the source base depends on reserve requirements. I f required reserve ratios are reduced, R. Alton Gilbert is an assistant vice president o f the Federal Reserve Bank o f St. Louis. Nancy D. Juen provided research assistance. 'R eserve balances of depository institutions included in the source base are net of required clearing balances and balances held to com pensate for float. 24FRASER Digitized for fo r example, a given level o f the source base can support a higher level o f the m oney stock. RAM is specified in term s o f the reserve requ ire ments in effect in a base period. It equals the reserves that w ould be required (given current deposit liabili ties) if the reserve requirem ents o f a base period w ere in effect minus the reserves that are actually required. RAM rises (falls) i f reserve requirem ents are low ered (raised). Including RAM in the AMB rem oves the e f fects o f reserve requirem ent changes fro m the rela tionship between the m oney stock and the AMB, even though such changes affect the relationship between the m oney stock and the source base. THE IMPLICATIONS OF RESERVE ACCOUNTING FOR AN APPROPRIATE MEASURE OF RAM The m oney stock is the product o f the monetary base m ultiplier and the AMB. The issues involved in developing an appropriate measure o f RAM can be analyzed in terms of the determinants of the monetary base multiplier. This section discusses the relation ships between the structure o f reserve requirements, the eauations for measuring RAM. and the determ i nants o f the AMB multiplier. The appendix presents the specific equations used for measuring RAM and derives the determinants o f the AMB m ultiplier associ ated with each specification. The AMB measures all three policy actions that influence m oney growth: open market operations, dis count w in dow lending and changes in reserve re quirements. The monetary base multiplier reflects the effects that choices o f both depository institutions and the public have on the m oney stock. The determinants FEDERAL RESERVE BANK OF ST. LOUIS o f the m ultiplier include the ratio o f currency in the hands of the public to checkable deposits, the com po sition o f deposits and the excess reserves held by depository institutions. These determinants depend on h ow RAM is mea sured. The appropriate specification o f RAM, in turn, depends on the structure o f reserve accounting in effect. This principle can be illustrated for two features o f reserve accounting: the reserve requirements of members and nonmembers and those on time and savings deposits. Prior to 1980, only banks that w ere members o f the Federal Reserve w ere subject to the Fed’s reserve re quirements; nonmem ber institutions w ere exempt from these requirements.2Thus, shifts o f deposits be tween members and nonmembers affected the level o f deposits that could be supported by a given level of bank reserves. Also, since there were reserve require ments on the time and savings deposits o f member banks, shifts o f deposits between demand deposits and time and savings deposits at member banks af fected the amount o f checkable deposits that could be supported by a given level o f bank reserves. Because these deposit shifts represented the public’s rather than the Federal Reserve’s actions, the AMB series was constructed so that the deposit shifts affected the AMB multiplier; the effects o f these shifts are dem on strated algebraically in the appendix. This AMB series was appropriate for periods before 1980.3 The M onetaiy Control Act o f 1980, however, im posed identical reserve requirements on both member and nonmem ber institutions. With the new structure o f reserve requirements fully phased in, as o f Septem ber 1987, a deposit shift between members and non 2Before 1980, state-chartered nonm em ber banks were subject to the reserve requirements of the state in which they were chartered. For information on the levels of the state reserve requirem ents and their effects, see G ilbert and Lovati (1978) and Gilbert (1978). 3Given the nature of the prior m easure of RAM, some actions of the public, such as shifts of deposits among banks, did not affect the m ultiplier. The structure of reserve requirements on m em ber bank deposits in effect prior to Novem ber 1972 was based on the location of m em ber banks. Shifts of deposits among m em ber banks in cities of different size changed the average reserve requirement on m em ber bank deposits, but did not affect the AM B m ultiplier. Under the structure of reserve requirem ents adopted in Novem ber 1972, there was a graduated structure of reserve requirements on demand deposits at m em ber banks. Shifts of demand deposits between large and small m em ber banks changed the average reserve requirem ent on m em ber bank dem and deposits. Changes in the average reserve requirement on m em ber bank demand deposits did not affect the multiplier. AUGUST/SEPTEMBER 1987 members no longer affects the amount o f checkable deposits that can be supported with a given amount of reserves.4 Maintaining the old RAM measure w ould continue to make the m oney m ultiplier a function o f the distribution o f deposits between m em ber and nonmem ber institutions; this distinction, however, has no relevance under the current system o f reserve accounting. Thus, the current measure o f RAM must be changed to make the AMB m ultiplier invariant to these deposit shifts. Under the new structure, the only categories o f time and savings deposits subject to positive reserve re quirements are Eurodollar liabilities and nonpersonal time and savings deposits with initial maturities o f 18 months or less."’ With these exceptions, shifts o f d e posits between checkable deposits and time and sav ings deposits do not affect the amount o f checkable deposits that can be supported with a given amount of reserves. In the new equation for RAM, the base period reserve requirement on all time and savings deposits is zero. This feature removes the ratio o f time and savings deposits to checkable deposits as a determi nant o f the m oney multiplier. There is a problem, however, with the use o f this new equation for RAM in measuring the AMB before 1980. The new equation eliminates as determinants o f the AMB m ultiplier the distribution o f deposits be tween members and nonmembers and the distribu tion o f deposits at m em ber banks between demand deposits and time and savings deposits. This pro duces an undesirable revision in the time series rela tionship between the m oney stock and the AMB prior to 1980. THE SOLUTION: LINK TWO DIFFERENT AMB SERIES TOGETHER The major challenge in revising the AMB series is creating a continuous series w hile maintaining the determinants o f the AMB multiplier that are appropri ate for periods both before and after Novem ber 1980. The solution is to link together, at the week ending November 19,1980, two series based on different equa tions for RAM. (November 19,1980, was the first reserve 4One exception involves nonm em ber institutions in Hawaii that were in operation on or before August 1 ,1978; their reserve requirements will be phased in through January 1993. 5Also subject to reserve requirements are nonpersonal ineligible acceptances and obligations of affiliates with initial m aturity greater than seven days. 25 FEDERAL RESERVE BANK OF ST. LOUIS settlement week under the reserve requirements spe cified in the Monetary Control Act.) Using seasonally unadjusted observations for that week, the value o f the AMB derived from the new equation for RAM is di vided bv the value based on the prior equation for RAM; that ratio equals 0.9704. The AMB for each period through Novem ber 12, 1980, based on the prior equa tion for RAM, is then m ultiplied by that ratio.6 This adjustment leaves unchanged the growth rates o f the AMB series between any two points in time prior to November 1980; it also adjusts the level o f the AMB series prior to Novem ber 1980 to avoid a break in the series on that date due to the change in the equation for RAM. The new measure o f RAM alters the seasonal pat terns in the AMB. The revised series is not seasonally adjusted as one continuous series. Instead, the data through October 1980 are seasonally adjusted without incorporating data with the new measure of RAM, and the data since November 1980 are seasonally adjusted with observations based entirely on the new measure o f RAM. THE DATA Table 1 presents quarterly growth rates o f these series from 1981. As the table shows, the growth rates of these series generally rise and fall together. On average, the new series grew slightly faster than the old series over this period. Data are not presented for periods prior to Novem ber 1980, since the construc tion of the revised series keeps the growth rates unchanged. AUGUST/SEPTEMBER 1987 Table 1 Quarterly Growth Rates of the Adjusted Monetary Base (compounded annual rates of change, seasonally adjusted) Q uarter Old series New series 1981 1 2 3 4 2.9% 6.5 4.7 3.9 2.4% 7.9 3.8 3.2 1982 1 2 3 4 8.6 8.5 6.8 8.8 8.4 6.7 5.8 9.6 1983 1 2 3 4 11.7 11.3 7.3 7.9 12.0 12.2 8.3 8.5 1984 1 2 3 4 9.0 8.1 6.8 5.4 9.8 7.1 5.6 4.6 1985 1 2 3 4 8.0 7.3 9.0 7.3 7.7 7.5 10.4 8.9 1986 1 2 3 4 1987 1 7.3 8.2 8.8 9.7 10.5 7.0 9.7 10.6 11.2 11.3 CONCLUSIONS The revision o f the adjusted monetary base AM B) involves a new equation for the reserve adjustment magnitude (RAM), the component of the AMB that reflects the effects o f changes in reserve requirements. The new measure o f RAM reflects the structure of reserve requirements specified in the Monetary Con trol Act o f 1980, which w ere phased in between N o vember 1980 and September 1987. 6For a discussion of this method of linking together distinct m easures of the AMB, see Tatom (1980). Digitized 26 for FRASER Data prior to Novem ber 1980 are calculated using the prior measure o f RAM. In this revision o f the AMB, therefore, the series through October 1980 is distinct from the series from Novem ber 1980 to the present. The two distinct series are linked together in N ovem ber in a w ay that makes the revised AMB one continu ous series. The prior measure o f RAM is used for periods prior to Novem ber 1980 to retain the determ i nants o f the monetary base m ultiplier (M l -f- AMB) that are appropriate for the reserve requirement struc ture then in effect. FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 REFERENCES Gilbert, R. Alton. “ New Seasonal Factors for the Adjusted Monetary Base," this Review (Decem ber 1985), pp. 29-3 3 . -----------------“ Calculating the Adjusted Monetary Base under Con temporaneous Reserve Requirem ents,” this Review (February 1984), pp. 27-32. _________ _ “ Revision of the St. Louis Federal R eserve’s Adjusted Monetary Base,” this Review (Decem ber 1980), pp. 3 -1 0 . _________ _ “ Effectiveness of State Reserve Requirem ents,” this Review (Septem ber 1978), pp. 16-28. _________ , and Jean M. Lovati, “ Bank Reserve Requirem ents and Their Enforcement: A Com parison Across States,” this Review (March 1978), pp. 2 2 -3 2 . _________ , and Michael E. Trebing. “ The New System of C ontem poraneous Reserve Requirem ents,” this Review (Decem ber 1982), pp. 3 -7 . Tatom , John A. “ Issues in M easuring An Adjusted M onetary Base,” this Review (Decem ber 1980), pp. 11-29. 27 FEDERAL RESERVE BANK OF ST. LOUIS AUGUST/SEPTEMBER 1987 Appendix Two Equations for the AMB and the Corresponding Monetary Base Multipliers This appendix presents the equation for the AMB adopted in 1980 and the n ew equation that is n o w used to measure the AMB for the period from N ovem ber 1980 to the present. The determinants o f the AMB m ultiplier are d e rived for each measure o f the AMB. See table A1 for defini tions o f the terms used in specifying the AMB and its m oney multiplier. Table A1 Terms Used in Specifying the Adjusted Monetary Base and the Monetary Base Multiplier Description Term OLD MEASURE OF THE AMB In a revision o f the m onetary base in 1980, the AMB was m easured as follows: (A l) AMB, = SB, + 0.12664 (TDM!,.,, AMB Adjusted m onetary base SB Source base RAM Reserve adjustm ent magnitude CDM Checkable deposits of m em ber banks FM Demand deposits of m em ber banks due to foreign banks and official institutions GM Demand deposits of m em ber banks due to the U.S. Treasury TDM Transaction deposits of m em ber banks TSM Tim e and savings deposits of m em ber banks RR Required reserves CP Currency in the hands of the public E E xcess reserves, in cluding the va u lt cash of nonm em ber banks n Share of total checkable deposits at nonmem ber banks CD Total checkable deposits — those at m em bers and nonmem bers + 0.031964 (TSM),_14 - HR, The deposit data, w h ich are for m em ber banks only, are lagged 14 days to reflect the fact that the required reserves for each w eek w ere based on deposits o f tw o weeks earlier. The weights on the transaction deposits o f m em ber banks (0.12664) and the time and savings deposits o f m em ber banks (0.031964) are the average reserve requirem ents on these categories o f deposits in the period from January 1976 through August 1980.' In deriving the m ultiplier associated w ith the AMB series specified in equation A l, the tim e lags on the deposit data are ignored to sim plify the equation. The first step in deriv ing the m ultiplier involves expressing the source base as the sum o f its com ponents. (A2) SB = CP + RR + E AMB = SB + RAM = CP + RR + E + 0.12664 (TDM) + 0.031964 (TSM) - RR (A3) = CP + E + 0.12664 (CDM + FM + GM) = CDM + FM + GM k CP + CD e E - CD fm FM 4- CD gm GM -s- CD fm TSM -r CD F Demand deposits of all depository institutions due to foreign banks and official institutions G Demand deposits of all depository institutions due to the U.S. Treasury TD Transaction deposits at all depository institutions f F ^ CD g G - CD + 0.031964 (TSM) Total checkable deposits, the deposit com ponent o f M l, equals the checkable deposits o f m embers plus those o f nonm em bers. Using low er case “ n ” as the fraction o f check able deposits at nonmembers, the com ponents o f the AMB in equation A3 can be respecified as follows: 1See Gilbert (1980) for a description of this measure of the AMB. Digitized 28 for FRASER = CD + F + G AUGUST/SEPTEMBER 1987 FEDERAL RESERVE BANK OF ST. LOUIS (A4) AMB = CD (k + e + 0.12664 (1 —n + fm + gm) + 0.031964(tm) ) The AMB m ultiplier can be specified as follows: Ml 1+ k (A 5 )------ = ----------------------------------------------AMB k + e + 0.12664(l-n + fm + gm) + 0.031964(tm) Thus, given the equation for the AMB adopted in 1980, the AMB m ultiplier is a function of: 1. the ratio o f currency in the hands o f the public to checka ble deposits (k), 2. the ratio o f excess reserves to checkable deposits (e), 3. the fraction o f checkable deposits at nonm em ber institu tions (nl, 4. the ratio o f the dem and deposits o f m em ber banks due to foreign banks and official institutions divided by total checkable deposits (fm), 5. the ratio o f the dem and deposits o f m em ber banks due to the U.S. Treasury divided by total checkable deposits (gm), and 6. the ratio o f time and savings deposits at m em ber banks to checkable deposits (tm). The revised measure o f the AMB prior to N ovem ber 1980 is obtained by m ultiplying the measure described above by a specific ratio; this ratio is the level o f the n ew measure o f the AMB divided by the level o f the prior measure for the w eek ending Novem ber 19, 1980. M ultiplying the AMB spe cified above by this fixed ratio alters the level o f the AMB m ultiplier for periods prior to Novem ber 1980; however, this procedure leaves both its determinants and its grow th rate unchanged. (A6) AMB, = SB, + (0.12) TD ,_„ - RR, The base period reserve requirem ent on transaction d e posits, 12 percent, is the marginal reserve requirem ent on most o f the transaction deposits o f depository institutions under the n ew structure o f reserve requirements. Contem poraneous reserve requirem ents becam e effec tive the w eek ending February 8, 1984. The m ethod forcalculating the AMB in this period is presented in equation A7.3 (A7I AMB, = SB, + 10.12) TD,_, - RR, Seasonal Adjustment Contem poraneous reserve requirements altered the sea sonal patterns o f the AMB. In a previous revision o f the AMB, Gilbert (1985) described a m ethod for deriving seasonal factors for the period after February 1984. That m ethod is applied to this new series on the AMB. It involves develop ing a counter-factual series for weeks prior to February 1984 that reflects estimates o f the seasonal patterns in the AMB if contem poraneous reserve requirements had been in effect. T he counter-factual series is calculated for the period Janu ary 1975 through January 1984. Observations for that series are com bined with actual values o f the AMB for the period since February 1984 to derive seasonal factors that are used for seasonally adjusting the AMB data for the period since February 1984. The New AMR Multiplier Using steps similar to those in equations A2 and A3, the new measure o f the AMB can be specified as follows: IA8I AMB = CP + E + 0.12 IC D + F + Gl THE NEW MEASURE OF THE AMB Reserve Accounting The tim ing o f data in the n ew equation for calculating the AMB is different for the periods under lagged and contem poraneous reserve requirements. For the periods under lagged reserve requirements, that is, for the w eekly reserve maintenance periods through the w eek ending February 1, 1984, the AMB is calculated as indicated in equation A6.3 2The deposit component of equation A6 is transaction deposits of all deposi tory institutions for the week ending on a Wednesday, 14 days before the end of the current maintenance period. Historical data are no longer available on transaction deposits for weeks ending on Wednesdays. When the Federal Reserve adopted contemporaneous reserve requirements in February 1984, the weekly data series on deposits, currency and vault cash of depository institutions were converted from averages for weeks ending on Wednesdays to averages for weeks ending on Mondays. It is possible to derive a series for deposits in weeks ending on Wednesdays (TD,_14) from the data on transac tion deposits for weeks ending on the surrounding Mondays (TDt_16 and TD,_9), as indicated in the following equation: TD,_)4 = (5/7) TD,_16 + (2/7) TD,_9. Data on transaction deposits derived in this manner are used in the calcula tion of the AMB through February 1, 1984. The AMB multiplier- can be expressed as follows: Ml 1 + k (A 9 )------ = ----------------------------AMB k + e + 0.12 (1 + f+ g ) The AMB multiplier, based on the n ew equation for the AMB (equations A6 and A7), is a function of: 1. the ratio o f currency in the hands o f the public to checka ble deposits (kl, 2. the ratio o f excess reserves to checkable deposits (e), 3. the ratio o f dem and deposits o f depository institutions due to foreign com m ercial banks and official institutions to checkable deposits (f), and 4. the ratio o f U.S. Treasury deposits at depository institu tions to checkable deposits (g). 3For a description of contemporaneous reserve requirements, see Gilbert and Trebing (1982). For an earlier discussion of the implications of contempora neous reserve requirements for the measurement of RAM, see Gilbert (1984). 29