The Behavior of Monetary Aggregates in Major Industrialized Countries

International Finance Discussion Papers Number 207

May 1982

THE BEHAVIOR OF MONETARY AGGREGATES IN MAJOR INDUSTRIALIZED COUNTRIES

by

David H. Howard and Karen H. Johnson

NOTE: International Finance Discussion Papers are preliminary materials circuleted to stimulate discussion and critical comment. References in publicetions to International Finance Discussion Papers (other than an acknowledgment by a writer that he has had access to unpublished material) should be cleared with the author or authors.

THE. BEHAVIOR OF MONETARY AGGREGATES IN MAJOR INDUSTRIALIZED COUNTRIES

by

David H. Howard and Karen H. Johnson*/

In recent years, most of the major industrialized countries have adopted some form of monetary growth target as a guide or intermediate objective of monetary policy. For all but the United States, this approach has involved picking one aggregate from the two or more "money supply" totals regularly compiled, and focusing policy primarily on that measure. The aggregates thus selected range from very narrow (Swiss monetary base) to very broad (U.K. sterling M3). A puzzling feature common to most industrialized countries in recent years (1979-1981) is the markedly divergent behavior of the various monetary aggregates in each country, which made the setting, implementation, and interpretation of a growth target for a particular monetary aggregate extremely difficult. In general, narrow monetary aggregates grew much more slowly than did broader monetary aggregates. The U.S. experience in 1981 is an example of this crhenomenon: in the twelve months to December, M1-B (shift adjusted) increased only 3-1/2 percent while M2 rose 10 percent. There are much more dramatic examples in foreign countries. Eetween December 1979 and December 1980, Ml in the United Kingdom grew 6-1/2 percent and sterling M3 increased nearly 20 percent; during the same period, Swiss Ml actually fell slightly while M2

rose 13-1/2 percent.

In this paper, recent monetary growth in the United States, Japan, Germany, the United Kingdom, Canada, and Switzerland is examined. The implications of the general pattern of divergent monetary growth rates are then discussed. Perhaps the two most dramatic instances of divergent monetary growth ‘in recent years occurred in the United Kingdom and Switzerland. The U.K. and Swiss cases are particularily interesting because these two countries have taken their monetary growth targets quite seriously in recent years, and because in the U.K. case the targeted aggregate was the one showing buoyant growth. In this paper, disaggregated demand for money functions derived in separate studies (Howard and Johnson) are used to examine in detail the U.K. and Swiss experiences. It is concluded that although the behavior of interest rates played an important role in the determination of the paths of the monetary aggregates ini the United Kingdom and Switzerland, several other factors also were operating. The disaggregated demand functions, by allowing differential adjustment speeds, impact effects, and structural shifts, enable the analyst to assess more accurately the forces at

work during recent years.

I. Historical Experience

In Table 1, annual monetary growth rates in the United States, Japan, Germany, the United Kingdom, Canada, and Switzerland are presented for the 1979-1981 period. The general pattern mentioned earlier in this paper is clear: in every country included in the table, growth of the narrow monetary aggregate (Ml) is exceeded by that of the wider aggregate. The differences in growth rates vary and at times are substantial.

Divergent monetary growth is not necessarily remarkable; different trend growth rates can, of course, produce such a result. However, when monetary aggregates are used as indicators and/or targets of monetary policy, divergent monetary growth across aggregates may create problems for the setting, implementation, and interpretation of monetary policy. For example, in Canada, Germany, and Switzerland, consumer prices rose in each of the twelve-month periods reported in Table 1 at a rate in between the rates of growth of narrow and broad money. Thus if one used the rate of growth of the "real" money supply as a criterion, one could characterize recent monetary policy in those three countries as being either restrictive -- using a narrow monetary aggregates measure -- or expansionary -- using a wider measure. For Japan, the United Kingdom, and the United States, such a criterion is ambiguous in one or two of the years reported in Table 1.1/ Alternatively, one might look at the direction of the change in money growth from one interval to the next as an

indication of tightening (if growth slows) or easing (if growth

Table 1

' Monetary Growth in Major Industrialized Countries

Percentage Change over Twelve Monzhs Earlier December 1979 December 1980 December 1981

U.S.: M1-B (shift adjusted) 7.1 6.6 3.4

M2 8.2 9.0 10.0

Japan: Ml 3.0 -1.7 10.8 M2 9.2 7.5 11.1

Germany: M1 3.9 4.4 -0.9 M2 8.2 8.8 8.4

U.K.: Ml 8.7 6.4 8.1 STGM3 12.4 19.9 13.7

Canada: Ml 3.3 10.2 -2.4 M2 17.7 16.0 16.9 Switzerland: Ml -7.0 -0.2 -7.3 M2 12.2 13.7 12.6

increases). . By this criterion, monetary policy in the United States and the United Kingdom was tighter in 1980 than it was in 1979 if the narrow aggregate is used, but easier if the broad measure is the standard. This is again the case for the United States in comparing 1981 to 1980. Just the opposite is true for Canada for 1979 to 1980 and for the United Kingdom from 1980 to 1981.

For the most part, Ml excludes interest-bearing deposits while the broader aggregates include such deposits. This fact suggests that the general increases in nominal interest rates in recent years may be largely responsible for the general ly sluggish growth of (primarily non-interest-bearing) Ml relative to the growth of the broader monetary aggregates. However, many other factors are also likely to have been at work, and it is necessary to examine these possible influences systematically before any conclusions can be drawn. In this paper, the monetary experiences of two countries -- the United Kingdom and Switzerland -- are studied, and the roles played by the various factors are ascertained. The experiences of the other four countries discussed in this section are beyond the scope of the present

paper.

-6-

II. Divergent Monetary Growth

One implication of divergent monetary growth is that the decision concerning which particular monetary aggregate to use as an intermediate target is an important one. Extensive analysis is required before the aggregate and its target are chosen. Furthermore, after the aggregates policy is underway, it is necessary to monitor developments for any behavioral changes induced by the new policy which may have a bearing on the choice of target(s).2/ In fact, when rates of growth diverge for the various measured aggregates as much as they have in recent years, the implementation, as well as justification, of a monetary policy focused on the path of the money stock may be called into question. For both the policymaker and the outside observer the usefulness of the rate of money growth, or changes in the rate of money growth, as an indicator of monetary policy is lessened when different measures of money growth provide different signals.

The 1980 experience of the United Kingdom is an example of the policy problem that can result when one measure (sterling M3) is growing rapidly while another (M1) is growing slowly. The fact that sterling M3 was expanding so rapidly was used as evidence by some critics that the Bank of England was not pursuing a tight monetary policy. The slow growth of M1 led others to claim that the Bank of England was tightening monetary policy.

The British experience in 1980 is particularly interesting because

in 1980 the United Kingdom was in a severe and prolonged

-7recession. Since sterling M3 was the U.K.'s targeted aggregate

and its 1980 growth exceeded the target by a wide margin,

achievement of the monetary growth target presumably would have

implied a significant tightening of policy during a recession. In fact, the British authorities accepted the overshooting and

justified this decision in part by referring to the much slower

growth of Ml. A similar policy dilemma confronted the U.S.

monetary authorities in 1981. David Lindsey, of the Federal

Reserve Board's staff, states (p. 12) the problem succintly:

The other major episode in which M1-B has strayed from

its annual range has been in 1981, when it has generally

been below its lower bound. However, M2 has been near

or above its upper bound in every month since February.

Is one to conclude that the [monetary authorities]

should have expanded nonborrowed reserves more rapidly

over this period in an attempt to attain the lower

bound for M1-B sooner, but in so doing drive M2 above

its upper bound? Or should growth of nonborrowed

reserves have been more restrained so as to bring M2

well within its range, but in the process forcing M1-B

further below its lower bound?

The fact that the monetary aggregates of a particular country are growing at widely differing rates means that as the aggregation proceeds from the narrowest to the broadest measure, at least some of the assets that are included at each step are behaving very differently from those already included. One possible explanation for this divergent behavior is that different economic processes are

determining the paths of the individual components of the monetary

-8aggregates. This explanation suggests that the components are not homogeneous, that the aggregates defined over them are somewhat arbitrary, and, therefore, that one would not necessarily expect such aggregates, per se, to play a stable role in the transmission mechanism through which monetary policy affects the economy or to be suitable as intermediate targets for policy. Alternatively, widely divergent behavior among the assets may result fran failure to recognize that two assets are very close substitutes and that a small change in the own rate of return on one asset relative to that of the other induces very large shifts in portfolios. If the process of aygyregation at some point includes one of these assets and not the other, the aygregate so defined may vary a great deal without that variation having any implications for the future path of nominal income. In order to know whether this is the case, one must have detailed information about the demand for money function and the mechanisin that links monetary developments to nominal income.

In order to resolve the ambiguities resulting from the disparate behavior of the various aggregates, it is necessary first to understand what is causing this behavior, and then to examine the questions of appropriate definitions for the aggregates and the proper role of the aggregates in monetary policy formulation. Since divergent behavior over time of the aggregates clearly must indicate even more divergent behavior for at least some of the component assets, and since the current definitions of the aggregates also are in question, it would seem

that the appropriate way to proceed is to explore the properties

-9of the individual component assets. In general, these assets are currency in the hands of the public, demand deposits, and one or more categories of time deposits.3/

In order to compare the properties of these assets and to understand how a given set of market conditions can induce very different responses in different assets, it is necessary to . estimate the demand function for each asset. These demand functions have interacted over time with the money supply process (including policy actions) and external economic forces to produce the observed path of the outstanding stocks of monetary assets. With such estimated functions, one can compare interest rate responses across assets in order to see if and in wnat way interest rate changes are a suitable explanation for the recently observed divergence of the aggregates. On the basis of the resulting estimated coefficients, it is possible to infer whether two assets are similar in their response to changes in the rate of return on some third asset (and are therefore somewhat homogeneous). It is also possible to look directly for the very high cross-rate elasticities that can signal the presence of very close substitution between two assets.4/ In addition, diséggregated demand functions allow the analyst to identify the roles of various other factors such as differential adjustment speeds, structural changes, and rates of return on other assets, including assets denominated in foreign currencies, in determining the behavior of the components of the monetary aggregates.

Elsewhere this disaggregated approach is used to

estimate demand functions for monetary assets in the United

-10-

Kingdon and Switzerland (see the papers by Howard and by Johnson). The experiences of these two countries are especially relevant to the issues being raised in this paper because growth rates of the monetary aggregates in each country have diverged sharply in recent years (see Table 1) and because during the 1970s both the British and Swiss monetary authorities adopted monetary aggregate growth rates as the intermediate target for monetary policy. Moreover, these two examples should provide insights into the general problen of explaining divergent growth among monetary

aggregates.

-ll-

III. The Demand for Monetary Assets A. Methodology

In the general case, there are n assets, as, i = 1, ..., n, where a, is the real value of tne stock of asset i, and each asset has an expected rate of return, Rie There are n demand functions:

* n

where the asterisk denotes what demand for a: would be if adjustment to changes in the various right-hand-side variables were complete in one period, y is a scale variable such as real income, and the u, are error terms. In order to allow for

partial adjustment, it is specified that:

where the subscript denotes a lag of one period, the d.

are the adjustment parameters, and the W; are error terms.

Equations (1) and (2) can be combined to yield:

-12-

n

A, + zr a5 AGR, + 5 AGY + (1 - dj )a j

i “0 j=l -]

+ -uU. +W. AGU i’

i = l, eee, ne

Thus in general the demand functions are of the fom:

; B45; + Biyy + (1 - ai + Vo, T= 1, veo, Me

+ tins

Equation (4) is a standard specification of a demand for money function and is the specification used in this study for the

demand for monetary assets in the United Kingdom and Switzerland.

In some instances, it is necessary to supplement this equation with a trend term and variables representing various structural changes.

In this study, two of the alternatives to domestic currency denominated monetary assets are considered to be foreign currency denominated notes and coin, that is, foreign currency, per se, and foreign currency denominated Eurodeposits. It is useful to examine the nature of the expected nominal rates of return in these cases: the

expected nominal return on foreign notes and coin is simply the

-13-

expected rate of appreciation of the foreiyn currency, and that on Eurodeposits is the sum of their nominal rate of interest and the expected appreciation. Thus, if asset n- 1 is a foreign currency Eurodeposit and asset n is foreign notes and coin,

Riel =rte,

Ry ~ @s where r is the nominal rate of interest on the Eurodeposit and e is the expected rate of appreciation of the foreiyn currency. Equation (4)

can be rewritten as:

n-2 (5) 4 Big * B Big8s Fina” * Fi sna * ind? + Ba yy + (1 - dj) a, tv, if 1, oes, Ne

-1

In equation (5), the coefficients of r and e are expected to be negative for all assets denominated in domestic currency.

Since e is not observable, a proxy for it must be used. Several alternative measures for e are used in this study. One possible method is a calculation of e based on the actual path of the exchange rate, which enables equation (5) to be estimated directly. Another specification for such a proxy is e = Rio? where

R,-2 iS the interest rate on a bank deposit similar to

-14-

a but denominated in domestic currency. In this case equation

n-]

(5) becomes:

n-3 (6) 4 = Bio * Bal * (65 nz * 8isn-1 * PinRn-2 Bi," + Bi yy + (1 - A; aj + Vis i=l, ..., n.

-1

In equation (6), the coefficients of Rae? and r are predicted

to be negative and positive, respectively, for all assets denominated in domestic currency other than ano? and indeterminate and positive, respectively, for that asset. Note that in equation (6) the coefficient on r refers to the effect of the rate of return on foreign notes and coin (asset n), not to that of a foreign currency denominated Eurodeposit (asset n-1). Alternatively, Richard Meese and Kenneth Royoff argue that the best predictive model of the exchange rate is a random walk time series model. If such a model is used to yenerate exchange rate expectations, then e = 0

and equation (5) becomes:

n-3 (7) 4 “Big * 2

Bik. + r

1 jj Bi n-2%n-2 * Bi n-1

y

In contrast to equation (6), in equation (7) the coefficients of

Rig and r are both negative for domestic currency assets other

than ayo» and positive and negative, respectively, for a, _os

In addition, in equation (7), the coefficient of r refers to the effect of the rate of return on foreign currency Eurodeposits rather than that

of foreign notes and coin as in equation (6).

-15-

The two main British monetary aggregates -- Ml and sterling M3 -- can be separated into three components: currency, sight deposits (M41 less currency), and other bank deposits (sterling M3 less Ml); the two main Swiss aggregates -- Ml and M2 -- similarly can be decomposed into currency, demand deposits (Ml less currency), and term deposits (M2 less M1).2/ In this study, nonbank demand functions for these six monetary assets are estimated and analyzed. Demand functions based on equation (4), in which a and y are specified to be natural logarithms unless otherwise stated, are estimated using a two-stage least-squares technique. The data used in this study are not seasonally adjusted; tnus seasonal binary variables are included in the equations. The estimated equations reported in this paper are selected on the basis of the behavior of the error terms and the signs and significance levels of the coefficients, although variables that figure prominently in most demand for money formulations are included regardless of the signs and t-ratios of their estimated coefficients.6/ (See the papers by Howard and by Johnson for more extensive discussions of the estimating

procedures used to obtain the equations reported in this study.)

-16- B. The United Kingdom

Tne "basic" U.K. demand for monetary assets function is specified to include a constant term, the lagged dependent variable, a representative three-month nominal interest rate, the natural logarithm of real GDP, three binary seasonal variables, and, in the demands for Siyht deposits and other bank deposits, a binary variable reflecting a major change in the method of classifying bank deposits which was instituted in the second quarter of 1975. The dependent variable is the natural logarithm of the real value of the monetary asset. There. are several additional variables that might be expected to belong in : these demand functions, including other rates of return and various binary variables representing the operation of the Bank of England's suppl ementary-special-deposits (SSD) scheme (a systen of quantitative limits on the expansion of banks' interest-bearing liabilities). An SSD variable in a demand for money function can be interpreted as | representing changes in rates of return (implicit as well as explicit) -- other than those specifically included in the function -- which are attributable to the SSD scheme's effects on bank behavior.

As reported elsewhere (Howard), the only supplementary variable added to the basic demand for currency equation is a linear trend term. In the sight deposit equation, the supplementary variables are a binary variable (SSD4) representing each of the three SSD episodes (1974:Q1-1975:Q1, 1976:Q4-1977:Q3, and 1978:Q2-1980:Q2) and the expected rate of change of the exchange rate, proxied by the actual rate of change of the dollar-sterling exchange rate during the subsequent three-month period.// For the demand for sterling bank

deposits other than sight deposits, a binary variable (SSD5)

-17-

representing the entire period during which the SSD scheme was available as a policy instrument (1974:Q1-1980:Q2) and the expected nominal yield on equities (including expected capital gains as proxied by the actual outcome) are added to the basic demand function. Table 2 presents the estimated equations, excluding the coefficients of the seasonal variables.

For present purposes, there are three interesting features of the equations reported in Table 2.8/ First, the demands for the two components of Ml display negative interest rate coefficients, and, although the coefficient for currency is not statistically significant, that for sight deposits has a very large t-ratio in absolute value -- 5.440. On the other hand, the demand for other bank deposits, that is, the demand for those deposits that are included in sterling M3 but not in Ml, has a positive interest rate coefficient, although its t-ratio is only 0.891. The neyative interest sensitivity of the Ml components as well as the positive interest sensitivity of the demand for those assets in sterling M3 but not in Ml strongly suggest that interest rate developments can have an important role in explaining divergent monetary growth, particularly during a period of substantial interest raze changes. Another interesting feature of the Table 2 estimates is the presence of the SSD4 and SSD5 variables. For the second quarter of 1980, these variables equal unity but in the third quarter, that is, afer termination of the SSD scheme, they become zero. The negative coefficients reported in the table indicate that the removal of the

Bank of England's "corset" -- that is, the SSD scheme -- had important

M1 Variable Currency Sight Deposits Other Bank Deposits Constant -4.001 -0.004 9.735 (1.882) (1.614) (4.939) Lagged Dependent Variable 0.811 0.897 C.677 (0.117) (0.062) (C.110) Local Authority 3-month -0.192 -0.697 ('.262 Interest Rate (0.140) (0.137) (C'.198) Real GDP 0.557 0.100 -('.638 (0.232) (0.179) (C'.392) Deposit Classification Variable 0.004 -0'.069 (0.015) (0.025) Expected Rate of Dollar -0.078 Appreciation (0.036) Expected Rate of Return on -().031 Equities (0,014) SSD4 -0.035 (0.011) SSD5 -0.045 (0.020) Trend -0.003 (0.001) Re 0.912 0.939 0.972 SE 0.017 0.020 ().029 h -0.696 -1.012 -1.496 n 32 32 32 Note: Figures in parentheses are the standard errors. The above equations also include

- 18 -

Table 2

Estimated U.K. Demand for Monetary Assets Equations: 1973-1980

Sterling M3

seasonal binary variables. Sight deposits include some interest-bearing bank liabilities, such as overnight money, as well as non-interest-bearing deriand deposits; other bank deposits are interest bearing and include certificates of deposit. The rates of return used for the U.K. estimates are expressed as decimals; in the Swiss estimates they are expressed in percentage points. Therefore the

U.K. rate-of-return coefficients are 100 times greater than comparable Swiss coefficients.

-19-

implications for the demand for bank deposits. Therefore, any explanation of the U.K.'s monetary performance in 1980 must take the effects of the termination of the SSD scheme into account. Finally, the significant exchange rate expectations effect in the sight deposit equation is quite interesting because of its implication that U.K. nonbanks appear to engage in some form of "currency" substitution in the sense that when there is a change in expectations about exchange rate changes, U.K. nonbank residents adjust their demand for sterling denominated sight deposits .9/ Following the discussion of equation (5), the lack of significance of the Eurodollar interest rate and the significance of the exchange rate expectations term would appear to indicate that the relevant substitute for sterling denominated sight deposits is foreign notes and coin only, not Eurodollar deposits .10/ (It should be noted that a binary variable representing the operation of British exchange and capital controls has no impact on the estimated - equations.) Furthermore, the significance of the exchange rate effect in the sight deposit equation means that the behavior of the dollar-sterling exchange rate is another potentially important factor contributing to the recent divergence in U.K. monetary growth.

In Charts 1 and 2 the actual and predicted paths of Ml and sterling M3 are plotted; the predicted values are those implied by the estimated currency, sight deposits, and other bank deposits equations reported in Table 2. The closeness of the fit, especially in the

turbulent 1979-1980 period, is heartening and means that the

- 20-

Ratio Scale CHART 1: British M1 (sterling billions)

Solid Line = Actual | 07 28

Dashed Line = Predicted 4 fi jee | 24 20 i 16 / A 4, a. / Ls V/ 12

1974 1976 1978 “1980

-?21-

Ratio Scale CHART 2: British Sterling M3 (sterling billions) Solid Line = Actual Dashed Line = Predicted Oe ft 60 4 Z 4 r / 4 50 . fe! / / / y jf +A ~~ , 40 {7 Z ee 4 4 / X / a 30

1974 1976 1978 1980

-22-

explanatory variables included in the Table 2 equations constitute a reasonably complete list of the factors contributing to the pattern of

British monetary growth in 1979-1980.

-23- C. Switzerland

In the Swiss case, the volume of retail sales is used as the scale variable because Swiss GNP is available on an annual basis only. There is little to be gained from converting the monthly retail sales data to quarterly averages and so monthly observations are used to estimate the Swiss equations. The volume of retail sales would seem to be a good choice for the scale variable in the currency demand equation, but perhaps is less suitable in the demands for the other assets. Except for the currency equation, the dependent variable and retail sales variable are natural logarithms of the real value of the variable; in the currency equation, these variables are levels of the real value. Each of the Swiss equations includes eleven binary seasonal variables.

In Table 3, the estimated equations -- derived and discussed in Johnson -- for the Swiss nonbank demands for currency, demand deposits, and term deposits. are presented. Two domestic short-term bank interest rates, the rate on thrift accounts and the rate on three-month term deposits in Zurich banks, are used as measures of the opportunity cost involved in holding currency and demand deposits, respectively. The term deposit rate is, of course, used as the own rate in the demand for term deposits. The three-month Eurodollar interest rate is included in all three equations and in the term deposit equation the expected rate of dollar appreciation -proxied by the annualized actual rate of change in the dollar-Swiss franc exchange rate in the subsequent three months -- is included also.

A possible source of the divergent growth of the narrow and

broad Swiss monetary agyregates is a structural change in one or more

Estimated Swiss Demand for Monetary Assets Equations:

Variable

Constant

Lagged Dependent Variable

Thrift deposit interest rate

3-month term deposit interest rate

3-month term deposit rate 1978-1980

3-month Eurodollar rate Expected rate of dollar appreciation

Real Retail Sales

Trend

Ke S.E. h

n

Note: monthly binary variables. is not in logarithmic form.

- 24 -

Table 3

M2 M1 —Deamand Currency Deposits 16.973 0.858 (6.753) (0.236) 0.778 0.843 (0.063) (0.043) -0.791 (0.194) -0.012 | (0.003) 0.106 - 0.0016 (0.057) (0.0015) 4.055 0.159 (2.262) (0.082) | 0.0004 (0.0002) 0.969 0.974 0.864 0.022 0.291 0.952 96 96

Figures in parentheses are standard errors.

1973-1980

Term Deposits

2.369 (0.429)

0.519 (0.082)

().037 (0.009)

0.048

.-(0,0046 ~ (0,0034)

1.0006 (0.0003)

-().049 (0.220)

~().0007 (0.0007)

().978

(0.036

1.109 66

The above equations also include Unlike the other equations, the currency demand equation

Term deposit equation ts estimated over the interval

June 1975-December 1980 because monthly data on term deposits. are not available for

January 1973-April 1975,

These components are as defined in the Swiss National

Bank Monthly Bulletin, except that the label "term deposits" is used to refer to the component "quasi-money", which includes in addition to term deposits foreign

currency deposits held by Swiss residents at Swiss banks.

are expressed in percentage points; see note to Table 2.

The rates of return

-25of the underlying behavioral functions. The extraordinary appreciation of the Swiss franc during 1978 suggests the possibility that at about that time some major shift in the demand for Swiss franc. assets occurred. However, of the three equations reported in Table 3, there isa significant’ shift in an interest rate coefficient in the term deposit equation only. Finally, trend terms are included in both deposit equations.

In Table 3 both currency and demand deposits are _ significantly inversely related to short-term Swiss interest rates, but “superior results are obtained when different rates are used in each LL/ Tern deposits are positively related to the (own) short-tenn rate, and: there is strong evidence of a shift to even greater interest elasticity in early 1978. These three assets display quite different speeds of ‘adjustment, with term deposits completing approximately one half of the response to a change in interest rates in | the first month. Charts 3 dnd 4 show the paths of actual M1 and M2 and _ the fitted values over the sample period that are implied by the : equations reported in Table Bec “the. closeness of the fit suggests that the forces identified in the estimated eqyations can account for the divergent behavior of Ml and M2 during the sample period. The Eurodollar rate appears positively in both the currency - and demanc deposit equations and in each its coefficient is about one-eighth the magnitude of the Swiss interest rate coefficient. . The t-ratic for the Eurodollar rate is much higher in the currency equation. (Recall that only the currency equation reflects the response of nonresicent as well as resident asset-holders to the opportunity ‘to

substitute foreign currency assets for Swiss currency and vice

CHART 3: Swiss Ml (SF Billions

Solid Line = Actual Dashed Line = Predicted

- 2 -

Ratio Scale 65 qa 4 | 4 ! t " f it ‘ f] a i t | 55 rT) ‘ 45 1976 1978 . 1980

CHART 4: Swiss M2 (SF Billions)

Solid Line = Actual Dashed Line = Predicted

1976

-27-

1978

1980

100

90

80

70

-28-

versa.) Following the discussion of equations (5) - (7), the positive signs of the coefficients for the Eurodollar rate in the currency and

' demand deposit equations suggest that the relevant foreign currency

- substitute is foreign currency, per se, rather than foreign currency dénominated interest-bearing bank deposits. In contrast, the negative coefficient for the Eurodollar rate in the term deposit equation along with the negative coefficient for the exchange rate expectations variable indicates that Eurodollar deposits, and perhaps dollar currency as well, are relevant substitutes for term deposits. This interpretation of the coefficients is interesting but, unfortunately, it implies that exchange rate expectations proxies differ across equations. Nevertheless, the roles of the external variables in all three of the Swiss monetary assets equations suggest that currency substitution between'Swiss francs and foreign currency denominated monetary assets cannot be dismissed. Further work on this question in both the Swiss and U.K. cases is warranted, particularly with regard to measuring the expected rate of exchange rate appreciation; expansion of

the set of possible currencies might also be useful 12/

CHART 4: Swiss M2 (SF Billions)

Solid Line Dashed Line

Actual = Predicted

1976

- 27 -

1978

1980

100

90

80

70

-28-

versa.) Following the discussion of equations (5) - (7), the positive signs of the coefficients for the Eurodollar rate in the currency and .demand deposit equations suggest that the relevant foreign currency

- substitute is foreign currency, per se, rather than foreign currency denominated interest-bearing bank deposits. In contrast, the negative coefficient for the Eurodollar rate in the term deposit equation along . with the negative coefficient for the exchange rate expectations variable indicates that Eurodollar deposits, and perhaps dollar currency as well, are relevant substitutes for term deposits. This interpretation of the coefficients is interesting but, unfortunately, it implies that exchange rate expectations proxies differ across equations. Nevértheless, the roles of the external variables in all three of the Swiss monetary assets equations suggest that currency substitution between Swiss francs and foreign currency denominated monetary assets cannot be dismissed. Further work on this question in both the Swiss and U.K. cases is warranted, particularly with regard to > measuring the expected rate of exchange rate appreciation; expansion of:

the set of possible currencies might also be useful 12/

-29-

D. Summary and Comparison

The results shown in Tables 2 and 3 confirm the appropriateness of estimating separate equations for each component asset. The two components of Ml in each country are inversely related to short-term interest rates, but to differing extents. While in the U.<. currency demand equation the interest rate coefficient is not significant, it is impressively so in the Swiss equation; in both cases, demand or sight deposits are Significantly negatively related to interest rates. In each country the Ml component assets combine with assets that are positively related to the short-term interest rate to form the broader aggregates of sterling M3 and Swiss M2. Since the speed of adjustment (as measured by the coefficient for the lagged dependent variable) differs across the components as well, the interest rate response of the broader aggregates results from the complex interaction of all these factors in both the United Kingdom and Switzerland.

In addition to the differences in estimated interest elasticities, the demand functions for the component assets in each country reflect different structural shifts during this period (as measured by the differential effects of the SSD variables for the United Kingdom and by the significant shift of the interest rate coefficient in one equation in the Swiss case). A trend variable is present significantly in one U.K. function and one Swiss function. For both countries the effects of other variables, such as the expected rate of return on equities, the expected exchange rate change, and the

interest rate on foreign-currency deposits, differ importantly across

~30-

the estimated demand functions as well. The degree of disaggregation in this study is limited by the availability of data.13/ It may be that even further disaggregation, if possible, would reveal additional

differences in the structure of the underlying demand functions.

-31-

IV. The Responsiveness Over Time of British and Swiss Money Demands to Changes in Interest Rates

The demand for monetary assets functions reported in the preceding section do not provide directly a simple measure of the interest rate semi-elasticity of the demands for narrow and broad money. This is the case primarily because the monetary aggregates are the sum of the component assets for which demand functions are estimated in this study. Since percentage change (i.e., change in natural logarithms) is not a linear operator, the percentage effect of a change in the interest rate on Ml, for example, does not equal the sum of the estimated coefficients of the effects on currency and demand (sight) deposits. In the Swiss case, the fact that the currency equation is estimated in level form while the deposit equations are in logarithmic form further complicates the calculation of an elasticity measure from the estimated coefficients reported in Tables 2 and 3. Furthermore, because of the lagged adjustment specification, the effect of a change in the domestic interest rate is distributed over time. There is an initial impact effect and then continued additional effects until the long run is reached and adjustment is complete.

Altogether these various factors mean that the equations reported above imply that there is no single constant interest rate seni-elasticity. In this paper, in order to obtain some sense of the size of the net impact of a change in the interest rate on the narrow and broad aggregates, the estimated equations are used to calculate the effects over time on the component assets of a change in the

interest rate. The implied effects on the monetary aggregates are

-32-

then computed and expressed as a percentage deviation from the actual level so that an elasticity-like number is obtained.

In the U.K. case, the actual path of money is compared to that implied by the equations if the U.K. local authority rate (and, implicity, foreign interest rates 14/) had been 1 percentage point lower than the actual path starting in 1979:Q3.15/ The effect of the change -- expressed as a percentage deviation -- is traced through to the end of 1980 in Table 4. The effects by the end of the six quarters are an Ml some 2-1/2 percent higher than actual Ml and a sterling M3 about 1/2 percent higher than the actual sterling M3 outcome. Thus interest rates have a substantial effect on the U.K. demand for Ml, but only a small effect on sterling M3 demand. The modest interest sensitivity of sterling M3 is attributable to the positive interest elasticity in the demand for the bank deposits that are included in sterling M3 but not in Ml.

For Swiss Ml and M2, the hypothetical path is calculated starting in January 1979. In this calculation, the thrift deposit rate, term deposit rate, and Eurodollar rate are postulated to have values 1 percentage point lower than their actual historical values. Table 4 also presents the percentage deviations implied by the estimated equations for such an interest rate path. The lower level of term deposits implied by the lower interest rates dominates the higher level of Ml and produces an absolute decline in M2 (from the historical path) throughout the 24-month period. Because the full effect o* the

lower rates is approached much sooner for term deposits than for the

- 33 -

Table 4

Effect of al Percentage Point Decrease in Interest Rates

United Kingdom:

ee

Quarter

1 (1979:93) 3 6

Switzerland:

Month

1 (January 1979) 5

10

15

20

24

Percentage Deviation in Ml

0.5 1.4 2.4

Percentage Deviation in Ml

0.9 3.3 4.6— 5.0 5.2 5.2

Percentage Deviation in Sterling M3

0.1 0.4 0.7

Percentage Deviation

in M2

-1.1 -1.7 -1.8

-4.3 ~4.3

73.9

-34-

other components (owing to the differences in adjustment parameters) , the continued increases in Ml during the last third of the period cause the percentage effect of the lower interest rates on M2 to decline

toward the end of the period.

-35-

V. Conclusion

In this paper, the recent behavior of the monetary aggregates in several of the major industrialized countries is examined, and the pattern of markedly slower growth of narrow money relative to the growth of broad money in each country is discussed. Two. interesting examples -- the United Kingdom and Switzerland -- are chosen for closer analysis. It is found that the behavior of interest rates is an important factor in determining the behavior of narrow and broad monetary growth. Between mid-1979 and mid-1980, short-term nominal interest rates in both the United Kingdom and Switzerland rose some 5 percentage points. Thus it is clear that an important part of the explanation of the recent divergence in monetary growth rates has been the behavior of interest rates. The disaggregated demand for monetary assets functions that are used in this study enable the analyst to model some of the complexity of the process by allowing adjustment speeds and interest rate elasticities to differ across monetary assets. The implicit demand for money equations thus exhibit a more detailed transmission mechanism between interest rates and the monetary aggregates than would equations estimated using the narrow or broad monet:ary aggregates directly. It is found that variables other than interest rates also contributed importantly to the determination of recent monetary growth in the United Kingdom and Switzerland; it is noteworthy that in both countries these additional explanatory variables include structural shifts (the removal of the Bank of England's supplementary-special-deposits scheme in the U.K. case) and

external variables such as exchange rate expectations.

-36-

One of the interesting features of recent monetary experience in the major industrialized countries is the problem confronting policymakers when divergent monetary growth transmits conflicting signals about the restrictiveness of monetary policy. The empirical results reported in this paper indicate that the demand functions for the narrow monetary aggregates in the United Kingdom and Switzerland are quite sensitive to interest rate changes. Whether this interest sensitivity constitutes a problem for the use of Ml as a monetary target or indicator depends on the nature of the interest rate effect involved as well as on the relationship between Ml and ultimate policy goals. If the interest rate effect is simply reflecting a shift between close substitutes that happen to fall on different sides of the dividing line between narrow and broad money, then perhaps a more useful monetary aggregate would include the substitutes; that is, a measure broader than Ml would be appropriate. In present circumstances, however, the broader aggregates include more than ‘ust very close substitutes for sight deposits: in the U.K. case, certificates of deposit and time deposits of all sizes are included in sterling M3; in Switzerland, term deposits of all sizes and maturities are included in M2. Thus, it might well be the case, in the United Kingdom and Switzerland, that Ml is too narrow and the broader aggregates are too broad to be useful as monetary targets or indicators. In this regard, it is interesting to note that the Bank of

England is now collecting data on "retail deposits" in order to be

-37-

able eventually to construct and publish a monetary aggregate somewhere

in between M1 and sterling M3,16/ and that the Swiss National Bank

switched from an Ml target to a monetary base target in part because of

the problems with Ml that are discussed in this paper.

-38- Literature Cited

. A. Allen, “Recent Developments in Monetary Control in the United Kingdom," paper presented at “Improving Money Stock Control," a conference cosponsored by the Federal Reserve Bank of St. Louis and Washington University, St. Louis, Missouri, October 30-31, 1981.

. D. Bordo and E. U. Choudhri, "Currency Substitution and the Demand for Money," Journal of Money, Credit, and Banking, February 1982, 14, 48-57.

. M. Boughton, "Recent Instability of the Demand for Money: An International Perspective," Southern Economic Journal, January 1981, 47, 579-597. |

. Brittain, "International Currency Substitution and the Apparent Instability of Velocity in Some Western European Economies and the United States," Journal of Money, Credit, and Banking, May 1981, 13, 135-155.

. Dornbusch, and others, "The Black Market for Dollars in Brazil," unpublished paper, 1981.

. T. Farr, "The Monthly Money Market Model," unpublished paper, 1981.

. Girton and D. Roper, "Theory and Implications of Currency Substitution," Journal of Money, Credit, and Banking, February 1981, 13, 12-30.

. J. Hamburger, "Tne Demand for Money in an Open Economy: Germany and the United Kingdom," Journal of Monetary Economics, January

1973, 3, 25-40.

-39-

. H. Howard, “The Demand for Monetary Assets in the United Kingdom," unpublished paper, 1982. |

. H. Johnson, "The Demand for Swiss Monetary Assets," unpublished paper, 1982.

. E. Lindsey, "Nonborrowed Reserve Targeting and Monetary Control," paper presented at "Improving Money Stock Control," a conference cosponsored by the Federal Reserve Bank of St. Louis and Washington University, St. Louis, Missouri, October 30-31, 1981.

. E. Lucas, Jr., "Econometric Policy Evaluation: A Critique,"

in K. Brunner and A. H. Meltzer, eds., The Phillips Curve and Labor Markets, Amsterdam, 1976.

. I. McKinnon, "Currency Substitution and Instability in the World Dollar Standard," American Economic Review, forthcoming.

- Meese and K. Rogoff, “Empirical Exchange Rate Models of the Seventies: Are Any Fit to Survive?" International Finance Discussion Papers, no. 184, June 1981.

- A. Miles "Currency Substitution, Flexible Exchange Rates, and Monetary Independence," American Economic Review, June 1978, 68, 428-436.

. B. Slovin, "On the Relationships Among Monetary Aggregates,"

Journal of Money, Credit, and Banking, August 1974, 6, 353-366. . D. Thomson, J. L. Pierce, and R. T. Parry, "A Monthly Money Market

Model," Journal of Money, Credit, and Banking, November 1975, 7, 411-431.

-40- Footnotes |

International Finance Division, Federal Reserve Board. This paper represents the views of the authors and should not be interpreted as reflecting the views of the Board of Governors of the Federal Reserve systen or other members of its staff. We would like to thank our colleagues in the International Finance Division for their many helpful comments and suggestions. The sign of the rate of growth of real money is not a satisfactory criterion for monetary policy because, among other things, it does not take trend rates of monetary growth into consideration and ignores income growth. The real money criterion can be misleading -- e.g., during a hyperinflation, when velocity is increasing rapidly, negative real money growth does not necessarily imply a restrictive monetary policy -- and is used here for illustrative purposes only. On the general point that significant changes in behavioral relationships can be induced by changes in policy, see Robert

Lucas. This disaggregated approach to the estimation of demand for money equations is used in the Federal Reserve Board's inonthly model. See Helen Farr for a summary of the Federal Reserve Board's model. An early version of a disaggreyated monthly model is reported in Thomas Thomson, James Pierce, and Robert Parry. For an alternative approach see Myron Slovin.

However, measurement and collinearity problems often make it difficult to estimate these cross-rate elasticities with much

reliability or precision.

-41-

In the United Kingdom, Ml consists of sterling denominated

notes and coin in circulation with nonbanks (including, presumably, some nonresidents) plus sterling sight deposit liabilities of U.K. banks held by the U.K. nonbank private sector (that is, U.K. residents only). In addition to demand deposits, sight deposits include money at call and money placed overnight, which are interest bearing. Sterling M3 consists of Ml plus

all other sterling deposit liabilities (including certificates of deposit) of U.K. banks held by U.K. nonbank residents, including the public sector. In Switzerland, Ml consists of Swiss franc

notes and coin in circulation (including that held by

- nonresidents) plus Swiss franc sight deposits held by residents at

Swiss banks. M2 adds to these Swiss franc term deposits plus foreign currency deposits held by residents at Swiss banks.

In most cases, a t-ratio of 2 or more is used as the criterion for statistical significance, although there are some exceptions made to this rule of thumb.

Under rational expectations, this measure should differ from the true expected rate of exchange rate change by an error term

that is independent of information available at the time of the observation. The error will not, however, be independent of the asset demands and rates of return being determined simultaneously. The expected rate of exchange rate change (as measured by the actual change) is treated in this study as an endogenous variable.

See Howard for a more detailed discussion of the equations.

10/

1y/

12/

13/

14/

15/

16/

-42-

Recent papers on currency substitution include Marc Miles, Lance Girton and Don Roper, Bruce Brittain, Michael Bordo and Ehsan Choudhri, and Ronald McKinnon. See also Rudiger Dornbusch, and others.

The Eurodollar interest rate has a t-ratio of less than unity in

_ absolute value when added to the equations reported in Table 2.

See Howard for further discussion of the evidence of currency substitution as well as possible alternative explanations for these results.

Note that the Swiss currency equation is in levels, not natural logarithms, and that the estimated coefficients must be interpreted accordingly.

Other empirical studies have included external variables in the demand for money in some countries. See, for example, the papers by Michael Hamburger and by James Boughton.

The U.K. data now separate interest-bearing and non-interestbearing sight deposits, but these data are not available for the entire sample period.

This implicit assumption is made so that it is reasonable to leave the path of the exchange rate term unchanged.

Because of the nature of the calculation, the result depends on the starting points as well as the subsequent actual paths of the component assets over the interval. This, of course, is also true for the Swiss case.

William Allen presents the definition being used for retail

deposits.