III A Monetary Impulse Measure for Medium-Term Policy Analysis

International Monetary Fund
Published Date:
September 1995
  • ShareShare
Show Summary Details
Bennett T. McCallum and Monica Hargraves1 

For a number of years, the biannual issues of the World Economic Outlook have regularly featured a measure of “fiscal impulse” for each of the seven major industrial economies in the discussion of conditions and prospects of the industrial countries. Although it has recently been given less emphasis, as attention has shifted to the concepts of structural and cyclical fiscal balances, the fiscal impulse measure continues to be reported. In contrast, there has been no presentation of a comparable common or standardized measure of monetary impulse for the major industrial countries. As monetary policy is unquestionably crucial for the success or failure of macroeconomic policies, it would seem that some additional attempt at systematization could be useful; in particular, it could provide a complement to the ranges of monetary indicators relied upon by individual monetary authorities and to the analysis of the world economic outlook itself. Consequently, the present paper is devoted to the development and investigation of one potential measure of monetary policy stance.

Organizationally, the paper proceeds as follows. The design and rationale of the proposed monetary impulse measure are discussed in the first section, followed by a presentation of the values of the proposed measure for the past 25 years. These annual values illustrate the construction of the measure and provide the basis for an evaluation, presented in the succeeding section, of its historical reliability. Next, values of the measure based on quarterly observations are reported, and some brief comparisons with other potential impulse measures are provided. Concluding remarks are then presented. Information concerning monetary base data is included as an appendix.

Design and Rationale

Two main criteria have guided the design of the proposed measure of monetary impulse. First, the measure should clearly reflect the medium-term inflationary implications of the current stance of monetary policy. Second, the measure should pertain to actions taken by the central bank itself—indeed, should pertain primarily to a variable over which the central bank could exert accurate control if it chose to do so. The first of these criteria suggests the use of some aggregate spending variable, such as the growth rate of nominal GDP, with other possibilities provided by the growth rate of nominal domestic demand, final sales, or personal income, for example.2 The second criterion, by contrast, points toward use of a measure such as the growth rate of the (adjusted) monetary base. Other controllable reserve aggregates could be considered instead, but the monetary base has the desirable property of reflecting the effect of open market purchases on both reserves and currency. Thus, if adjusted for changes in reserve requirements, the base provides a reasonably comprehensive summary measure of the actions of the central bank.3 Short-term nominal interest rates are also controllable, of course, and are emphasized by central banks, but these rates are less satisfactory indicators of monetary policy actions for reasons that will be discussed below.

Together, the two main criteria can be reasonably well satisfied by the proposed measure of monetary impulse, which is defined as the growth rate of the (adjusted) monetary base plus the expected medium-term growth rate of base velocity. This expected rate will be represented provisionally in this study by the actual change in base velocity over the most recent four years.4 To be specific, if DLX and DLB denote growth rates (that is, changes in the logs) of nominal GDP and the monetary base, respectively, the growth rate of velocity over a single period—either a quarter or a year—will be DLV=DLX - DLB. Its average over the past four years will then be DLVBAR = (1/N) (DLV + DLV(-1) + … + DLV(-N+1)), where N equals 16 or 4, depending on the time interval adopted.

With quarterly data, then, the impulse measure to be studied below is

where DLVBAR is lagged one quarter so as to reflect data available to the central bank when setting DLB. When using annual data, however, the choice has been made to represent the impulse measure as

with no time lag for the average velocity term. This choice represents a compromise in terms of the controllability criterion, but to impose a full-year lag would seem to be excessively conservative.

In the sections that follow, values of the impulse measures in equations (1) and (2) will be reported for each of the major industrial economies. As the measures are rates of base growth adjusted by GDP velocity, they represent nominal growth rates of GDP that are implied—on average, from a medium-term perspective—by current monetary policy actions. Consequently, they are appropriately compared with “target” paths of nominal GDP growth that would be consistent with low inflation rates (that is, approximate price stability) in terms of GDP deflators. For presentational purposes, 2 percent has been adopted as the common annual inflation target, and 1 percent to 3 percent will be treated as the range of low inflation.5 Average growth rates of real output must be accounted for, of course. Accordingly, the charts that present the impulse measures include low-inflation bands that represent the relevant economy’s estimated long-run growth rate of real GDP plus 1-3 percent. For several economies, the estimate of long-run real growth is 3 percent, so in those cases the low-inflation bands extend from 4 percent to 6 percent.6

It is important to recognize that the impulse measure presented in this paper builds upon growth rates of the monetary base for the reasons mentioned above, not because of a belief that the base (or any other monetary aggregate) has any mystical properties. It should also be noted that the case for using this measure does not rely on a presumption that technological and regulatory change in the payments and financial industries will be insignificant. Indeed, the purpose of the DLVBAR term is in large part to account for precisely such changes, which have been important over recent decades in several countries and which may be important in the future.

As was mentioned above, the monetary base is not currently used as an instrument or intermediate target variable in any of the countries under study—although a variant (termed “central bank money”) served as a target in Germany prior to 1987. That fact does not negate the reasons for focusing upon the base, however: that it could be accurately controlled7 and that it represents a reasonably comprehensive measure of the impact of central bank actions (however governed) on aggregate demand. It can therefore be argued that the precision that is lost by not using a measure that reflects more specifically individual countries’ policy mechanisms is outweighed by the advantages of using one that can be calculated in the same way for different countries and that can be used to study long periods of monetary history.

It is also well known that the base in many countries is largely composed of currency, and that currency is demand determined, in the sense that demand deposits can be redeemed for currency at the wish of the deposit holder. However, that fact does not imply that the base could not be controlled by the central bank or that manipulation of the base (rather than of reserves alone) would not be more effective for influencing total spending. Somewhat more troublesome is the likelihood that a large fraction of recent currency issues in the United States and Germany have gone abroad. In that regard, it should be kept in mind that such movements will eventually be reflected in the velocity growth term DLVBAR, For the first year or two after a major change, however, the impulse measures in equations (1) and (2) will reflect their effects only partially and will accordingly yield signals that are somewhat distorted.

The more traditional monetary growth rate measures—of M1 or M2, for example—and interest rates fare poorly in comparison with both of the adopted criteria. Aggregates of the M1 and M2 variety cannot be controlled with a high degree of accuracy, compared with the monetary base, and the growth rate of M1 or M2 that is necessary to yield a target inflation rate (over a decade or two, on average) is much more uncertain than the necessary growth rale for nominal GDP.8 Even if one added predicted velocity growth to an M1 or M2 measure, as equations (1) and (2) do for the base, this would still fail to produce a variable that is accurately controllable by the central bank.

With regard to interest rates, the most basic point is that the level of any short-term nominal interest rate is an ambiguous indicator of monetary policy stance. Specifically, high interest rates are associated with tight monetary policy from a short-term perspective but with easy monetary policy from a longer-term point of view. Accordingly, the interest rate effects of a monetary tightening—for example, an open market sale of securities—are in opposite directions from the short-term and long-term perspectives. Moreover, whether a particular interest rate level corresponds to a tight or easy monetary stance depends on economic conditions at the time, which would seem to be an extremely undesirable feature for a monetary indicator.

Conceivably, other variables involving interest rates—changes or spreads, for example—could be found that would perform more satisfactorily.9 A preliminary examination of the performance of one interest spread measure is reported below in the discussion on alternative monetary indicators. A more extensive consideration of other measures is a possible topic for future research.

Annual Measures

The purpose of this section is to calculate and present annual values of the impulse measure IM, defined in equation (2), for the major industrial economies.10 In a fundamental sense, only two basic variables are utilized—nominal GDP and the adjusted monetary base—which is one attractive feature of the proposed measure. The basic source of data for the nominal GDP series is quarterly (seasonally adjusted) observations; for annual series, the quarterly observations are averaged. Natural logarithms of these averages are calculated and first differenced to yield growth rates (denoted DLX).

One economy for which more needs to be said is Germany, with the difficulty arising from the east-west reunification of 1990. The procedure adopted in this paper for handling that event is determined in large measure by data availability. Specifically, the monetary statistics are those of the Bundesbank and pertain to west Germany before July 1, 1990, and to unified Germany thereafter. Official quarterly GDP figures since 1990 are available, unfortunately, only for west Germany. (Annual GDP figures are available, however, for west and east Germany separately.) An IMF staff estimate of quarterly GDP for unified Germany has been used for 1990-93.

More discussion is also needed on the monetary base series. As mentioned above, it is important to have a measure of the base—currency held by banks and the private nonbank sector plus banks’ reserves at the central bank—that is adjusted for reserve requirement changes.11 For four of the seven countries, such series have been obtained from national statistics.12 For Japan, Germany, and France, however, adjustments have been made owing to the unavailability of these series. The basic idea of the procedure is as follows. If only one type of bank deposit were subject to reserve requirements—and if the requirement ratio were a single flat rate of, for example, 10 percent of deposits—the magnitude of deposits would be approximately equal to the ratio of the volume of reserves (R) to the reserve requirement ratio (rr). Consequently, an appropriate measure of adjusted reserves would be R multiplied by the factor rr0/rr, where rr0 is the value of rr at some reference date. In practice, more than one type of deposit is subject to requirements in these countries, and, in Japan and Germany, the rr values depend on bank size.13 However, the calculations have been performed as if these complications do not exist by using the rr values pertaining to demand deposits at the largest banks.14 In the case of France, the calculation is nevertheless more complicated than just described because of the zero or very low rr values occurring in parts of the sample. For details, the reader is referred to the appendix.

Once adjusted reserve series have been obtained, they are added to currency magnitudes to yield an adjusted base series. These (quarterly) series are then seasonally adjusted.15 Finally, averages of the quarterly values—which are used below in the section on quarterly measures—are calculated to yield the annual data used in the present section. Logarithmic differences, denoted DLB, are used as growth rates.

It is a straightforward step to calculate series for DLV = DLX - DLB, the rate of change of base velocity, and DLVBAR =(¼) (DLV + DLV(-1) + DLV(-2) + DLV(—3)). The annual impulse measure IM = DLB + DLVBAR is then readily obtained. With observations for GDP and the monetary base beginning in 1960 in the data set, the earliest year for which DLVBAR and IM values are available is 1964. For some countries, one or both of the basic series begins after 1960, so the starting date for the IM values is correspondingly delayed. The charts in the next section present values of the IM measure for each economy, together with a pair of dashed horizontal lines marking the upper and lower limits of the low-inflation band described above. For reference, the charts also present data on the growth of the adjusted monetary base alone (DLB), corresponding to the impulse measure without the velocity adjustment component.

The interpretation of the IM values and low-inflation bands in the charts is as follows. When the IM magnitude is above (or below) the band, current monetary policy as reflected in base growth rates is too rapid (or slow) to be consistent on a sustained basis with low inflation. In that sense, the impulse measure suggests that monetary policy is overly expansionary (or restrictive) in the case at hand. The emphasized qualification is important, however, because real-time decision making must recognize the existence of substantial time lags in the inflationary process. In addition, there may be other policy goals besides the attainment of low inflation. Consequently, a value of IM above (or below) the low inflation band does not necessarily imply that policy should be tighter (or easier) from the perspective of that moment.

Evaluation of Annual Measures

This section discusses the interpretations of monetary policy over the years 1965-93 that are suggested by the annual impulse measure plots appearing in Charts 1-8. A leading purpose of the discussion is to consider whether the measure has accurately signaled episodes now generally regarded to have featured monetary policy stances that were inappropriate—either excessively lenient or stringent—from a medium-term perspective.

Before focusing on IM plots for individual economies, accordingly, it should be emphasized that the impulse measures indicate that monetary policy was extremely lenient for a period of several years in the early 1970s in all seven major industrial countries. In each of Charts 1-8, that is, the IM measure remained substantially above the low-inflation band for three or more years during the interval 1970-75. In most of the plots, furthermore, this period of leniency extended until 1979 or beyond. However, the plots for two countries, Japan and Germany, stand out by indicating a distinctly earlier movement—about 1974 or 1975—toward the low-inflation range. Thus, taken as a group, the charts very clearly suggest that a widespread inflationary stance in the early 1970s was eliminated during the 1980s by most countries but much earlier in Japan and Germany. That general feature of the charts corresponds quite accurately with views concerning the 1970s and 1980s that are widely accepted today. Although, to some readers, this feature might not seem to represent a major achievement, most alternative measures of monetary policy stance would probably fail to satisfy this basic criterion. For instance, as is demonstrated in the following section, plots of long- and short-term interest rate spreads do not show a contrast between the 1970s and 1980s and do not indicate markedly different policy patterns for Japan and Germany in comparison with the other major industrial economies.

The IM plot for the United States (Chart 1) shows a gradually increasing movement in the inflationary direction before 1980, a tendency that shows up even more strongly in Chart 2 (which begins in 1954 rather than 1964). A sharp drop in the IM measure occurs between 1979 and 1982 and another one between 1987 and 1989. The first of these drops represents the major deflationary episode that began with the Federal Reserve’s famous actions of October 6, 1979 and ended in the third quarter of 1982. The second drop pertains to a period of tightening that has not been so widely publicized, but which was arguably quite significant: an attempt by the Federal Reserve to reduce the “trend” rate of inflation from about 4.5 percent to something closer to 2 percent. Whether that attempt was partially responsible for the recession of 1990-91 is debatable, but Chart 1 would certainly provide support to anyone inclined to make such an argument.16 The sharp upward movement of IM over 1990 and 1991 is probably overstated to some extent because of an increase in the fraction of currency outstanding that is held abroad.17 Changes such as these will eventually be reflected in the IM measure, as is pointed out above, but not promptly.

Chart 1.United States: Monetary Impulse Measure, Monetary Base Growth, and Low-Inflation Band

(In percent)

Sources: See appendix.

Chart 2.United States: Monetary Impulse Measure, 1954-93

(In percent)

Sources: See appendix.

To clarify the contribution of the DLVBAR term to the impulse measure, Chart 1 and Charts 3-8 include plots of the growth of the adjusted base alone, that is, just the first component of the impulse measure. In Chart 1, the importance of the velocity adjustment is apparent in the sustained difference between the two series from 1971 through 1981 and in the trend movement of DLB from below to above the impulse measure. Base growth during this ten-year period is not much different from its average in the following six years; yet, as the discussion below will make clear, U.S. inflation performance did differ during these two periods, and the impulse measure captures the change in inflationary pressures much more closely.

Chart 3 plots the monetary impulse measure for Japan, with the first value pertaining to 1967 because of the absence of monetary base data for 1960-62. The low-inflation band limits are not shown before 1972 because average real growth was considerably higher than 4 percent during the 1960s. The Japanese record is dominated by a major tightening of monetary policy over the years 1973-81, with a relatively brief interruption during 1977-79. Since 1975, the IM measure has stayed fairly close to the low-inflation band, a pattern that is entirely consistent with the excellent record actually compiled in terms of realized inflation. However, the IM measure moves from below the band in 1985 to above it in 1987, thereby representing a tendency toward more lenient monetary conditions. This shift probably reflects actions, prompted by the Louvre Accord and the U.S.-Japan bilateral agreement of October 1986, to prevent major changes in the yen-dollar exchange rate.18 The ensuing period of easier monetary conditions may have contributed to the asset price “bubble” of the late 1980s.19 Quite notable in the chart is the degree of monetary stringency indicated for 1991 and 1992. The slight easing of monetary conditions evident in 1993—to a point still well below the low-inflation range—captures the Bank of Japan’s response to the prolonged recession.

Chart 3Japan: Monetary Impulse Measure, Monetary Base Growth, and Low-Inflation Band

(In percent)

Sources: See appendix.

Note: The low-inflation band for Japan is presented only from 1972 onward because average real growth in Japan was significantly higher during the 1960s than in the following decades.

Chart 4 pertains to Germany. The proximity of IM to the low-inflation band over 1974–89 has already been mentioned. The Bundesbank’s sharp tightening during 1973 and 1974 occurred after a five-year buildup of inflationary tendencies had threatened to get out of hand during the final years of the Bretton Woods regime. The other prominent feature of Chart 4 is its suggestion of unusually tight monetary policy in 1990 and distinctly loose policy in 1992-93.20 With regard to these plotted values, it is important to understand an adjustment that was made to the impulse measure in light of the major event of recent German experience, the unification of mid-1990. In principle, unification might be expected to have had some impact on one component of the impulse measure—velocity—both because of possible differences in money demand relationships in the former east and west Germany, and possibly in reaction to unification itself or to the associated uncertainty. As such, DLVBAR would have been affected, and adjustments would be needed. Estimates of such effects are, however, very difficult to make. In this paper, adjustments have been made only for the effects of unification on the other component of the impulse measure, the base growth term. The monetary base expanded very rapidly during 1990 and 1991 to accommodate the jump in the size of the German economy represented by unification. The appropriate adjustment would be to subtract from the unadjusted impulse measure a magnitude equal to the fractional increase in the size of the economy brought about by unification. But the appropriate distribution over time of this adjustment is not obvious, and the exact magnitude is itself open to question. The adjustment that was implemented, which is reflected in Chart 4, involves a subtraction of terms summing to 0.12 from the raw IM measure,21 with half of the magnitude of this adjustment assigned to 1990 and half to 1991.22

Chart 4.Germany: Monetary Impulse Measure, Monetary Base Growth, and Low-Inflation Band

(In percent)

Sources: See appendix.

From the foregoing discussion it should be apparent that one cannot be confident about the accuracy of the adjustments made in the case of Germany or, therefore, about Chart 4’s suggestion that monetary policy was quite tight in 1990, about right in 1991, and rather loose in 1992. Those indications could be altered significantly by a different adjustment magnitude pertaining to unification or a different distribution over time. An additional problem, as mentioned previously, is the increased use of deutsche mark currency outside Germany.

The French IM values in Chart 5 show over 1974-80 a beginning of the usual movement away from the inflationary stance of the early 1970s and then a sharp jump upward in 1981 and 1982, clearly representing the so-called Mitterrand Experiment. Since that time, however, the impulse measure has moved toward and finally through the low-inflation zone, with values for 1991–93 that represent a tight policy stance. The contribution of the velocity adjustment term is apparent for much of the period of study, particularly in the early 1970s and again in the early 1990s. Here again, the impulse measure corresponds more closely to subsequent inflation than does base growth by itself.

Chart 5.France: Monetary Impulse Measure, Monetary Base Growth, and Low-Inflation Band

(In percent)

Sources: See appendix.

The Italian impulse measure in Chart 6 indicates a continuation of substantial monetary leniency until the mid-1980s. Its peak value is quite high, and the plot shows a substantially greater cumulative magnitude of excessive monetary impulse for Italy than for any of the other six countries. A sustained movement toward a low-inflation stance has dominated the record since the mid-1980s, however, despite a slight interruption in 1989-90.

Chart 6.Italy: Monetary Impulse Measure, Monetary Base Growth, and Low-Inflation Band

(In percent)

Sources: See appendix.

The U.K. measure in Chart 7, by contrast, shows a sharp tightening as early as 1979-1982, but one that reached the low-inflation zone only in 1991.23 The upward surge in 1987-88 presumably reflects a monetary loosening that was the consequence of Chancellor of the Exchequer Lawson’s controversial device of “shadowing the deutsche mark” at a time that turned out, in retrospect, to have been inappropriate.

Chart 7.United Kingdom: Monetary Impulse Measure, Monetary Base Growth, and Low-Inflation Band

(In percent)

Sources: See appendix.

Finally, Chart 8 shows a predominantly downward path for Canada since 1974, with temporary upward blips in 1983 and 1988. The velocity adjustment term contributes to make this downward path less Steep through 1982, indicating a smaller disinflationary impulse than would have been indicated by base growth alone. The drive toward a “zero inflation” target that began about a year after John Crow became Governor of the Bank of Canada in 1987 appears clearly in the diagram. For 1992, the Canadian monetary policy stance appears to have been quite tight, but policy shifted toward a more neutral stance in 1993.

Chart 8.Canada: Monetary Impulse Measure, Monetary Base Growth, and Low-Inflation Band

(In percent)

Sources: See appendix.

It should perhaps be mentioned explicitly that, with one exception, Charts 1-8 use the same vertical axis range rather than modifying the range to reflect the IM values experienced by the country in question. This convention was chosen, despite a resulting slight loss of detail for the countries with lower inflation peaks, to reveal more clearly the comparative performance of the various countries in keeping their monetary policies close to a low-inflation range. The one exception is the chart for France, which uses the same scale as the other countries but a different range of values in order to accommodate the very low IM values in the 1990s.

Quarterly Measures and Alternative Monetary Indicators

This section examines impulse measures based on quarterly data. As quarterly values of the impulse measure IMI defined in equation (1) are very choppy in appearance, plots of a smoothed series are presented.24 This series, denoted IMIAVG, is a four-quarter moving average of IMI, defined as (¼) (IMI + IMI(-1) + IMI(-2) + IMI(-3)).25 Plots are presented in the top panels of Charts 9-15, with magnitudes scaled to represent annual growth rates comparable to those in Charts 1-8.26 The story told by the new charts is basically consistent with that of the plots utilizing annual data (Charts 1-8); however, the details are not identical because the two measures involve different timing.

A question that arises naturally in response to the foregoing presentation is whether there are other controllable variables that might be superior—in terms of reflecting more accurately the medium-term inflationary implications of current monetary policy—to the measures of monetary impulse used in this paper. Here, two such measures are considered, namely, the spread or difference between long- and short-term nominal interest rates, and growth in a broad monetary aggregate. Before looking at these series, however, it will be useful to consider time plots of the actual inflationary experiences of the countries in the sample. The second panels of Charts 9-15 show inflation rates in terms of consumer price indices for each of the countries in the sample. The charts are expressed in percent on an annualized basis and reflect four-quarter changes (rather than year-to-year values based on annual averages of price level indices).

In these inflation plots, several important aspects of the trajectories match parallel movements in the monetary impulse plots in the top panels and in Charts 1-8. For the United States, to begin with, Chart 9 reveals a fairly regular upward trend persisting from 1960 until 1980, with dips in 1971-72 and 1975-76 that show up two years earlier for the IM measure. The sharp decline in inflation in 1981-83 also has a corresponding decline in the IM measure roughly two years earlier. For Japan, a period of very high inflation rates in the mid-1970s is followed by an abrupt decline and a low-inflation record (Chart 10). For Germany, the inflation rate has stayed reasonably low throughout most of the period but surged about 1972, corresponding to the 1969-73 buildup evident in the top panel of Chart 11 and in Chart 4.27 In France, by contrast, inflation was more severe; despite a temporary slowdown in the late 1970s, inflation did not begin to fall steadily until after 1982 (Chart 12). The case of Italy is somewhat similar, but with even more severe inflation experienced in the 1970s (Chart 13). In the United Kingdom, inflation was also rather severe in the 1970s, but its elimination began somewhat earlier than in Italy. This downward trend was interrupted by an increase over 1987-89 (that follows the upward IM jump in Chart 7 and the top panel of Chart 14). In Canada, the main movement away from the moderate inflation of the 1970s came only after 1982 (Chart 15).

Chart 9.United States: Inflation, Monetary Impulse, and Alternative Monetary Indicators

(Percent change from four quarters earlier, unless otherwise noted)

Sources: See appendix.

1Four-quarter moving average of quarterly impulse measure.

2Difference between ten-year government bond yield and three-month treasury bill rate.

Chart 10.Chart 10. Japan: Inflation, Monetary Impulse, and Alternative Monetary Indicators

(Percent change from four quarters earlier, unless otherwise noted)

Sources: See appendix.

1Four-quarter moving average of quarterly impulse measure. The low-inflation band is presented only from 1972 onward because average real growth was significantly higher during the 1960s than in the following decades.

2Difference between ten-year government bond yield and three-month repurchase (Gensaki) rate.

Chart 11.Germany: Inflation, Monetary Impulse, and Alternative Monetary Indicators

(Percent change from four quarters earlier, unless otherwise noted)

Sources: See appendix.

1Four-quarter moving average of quarterly impulse measure.

2Difference between ten-year government bond yield and three-month interbank rate

Chart 12.France: Inflation, Monetary Impulse, and Alternative Monetary Indicators

(Percent change from four quarters earlier, unless otherwise noted)

Sources: See appendix.

1Four-quarter moving average of quarterly impulse measure.

2Difference between ten-year government bond yield and three-month interbank rate.

Chart 13.Italy: Inflation, Monetary Impulse, and Alternative Monetary Indicators

(Percent change from four quarters earlier, unless otherwise noted)

Sources: See appendix.

1Four-quarter moving average of quarterly impulse measure.

2Before June 1991, difference between average yield on government bonds with two- to four-year residual maturity and three-month treasury bill rate; thereafter, yield on ten-year government bonds.

Chart 14.United Kingdom: Inflation, Moneatry Impulse, and Alternative Monetary Indicators

(Percent change from four quarters earlier, unless otherwise noted)

Sources: See appendix.

1Four-quarter moving average of quarterly impulse measure.

2Retail price inflation excluding mortgage interest.

3Difference between ten-year government bond yield and three-month interbank rate.

Chart 15.Canada: Inflation, Monetary Impulse, and Alternative Monetary Indicators

(Percent change from four quarters earlier, unless otherwise noted)

Sources: See appendix.

1Four-quarter moving average of quarterly impulse measure.

2Difference between average yield on government bonds with ten or more years residual maturity and three month corporate commercial paper rate.

More generally, these plots of inflation rates based on consumer price indices are encouragingly similar to those of die monetary impulse measures. There are disagreements, of course, pertaining to timing and to brief movements. But in terms of broad tendencies, the profiles have many features in common.

With respect to the interest rate spread, the third panels of Charts 9-15 show values of the spread between long- and short-term interest rates. The precise series utilized are those listed in Table A18 of the October 1994 World Economic Outlook; ten-year government bonds and three-month treasury bills are fairly representative. Thus the plotted spread—like the impulse measure—is supposed to serve as an indicator of monetary ease rather than tightness. But, in general, these profiles do not agree to any significant extent with the inflation rate profiles. Most of these panels do show substantial dips in 1980-81, presumably reflecting policy movements toward an anti-inflationary posture. However, the story told by the pictures does not accord nearly as well with the inflationary experience as does the impulse measure.

Other measures could be examined, of course, especially if the criterion of controllability is sacrificed. Consideration of other measures is limited in this paper to growth rates of a broad monetary aggregate, plots of which appear as the final panels of Charts 9-15.28 For the United States,Chart 9 shows a steady but irregular inflationary buildup from 1960 into the 1980s, but the date of the tightening of policy is not as clearly located as in Chart 1 (or the top panel of Chart 9). In particular, there is little evidence of sustained anti-inflationary efforts before 1986.

For Japan, the behavior of the M2 + CD growth rate does provide a satisfactory picture of policy stance, one that is quite similar to that based on the impulse measure developed in this paper. For Germany, however, the inflationary stance of the early 1970s is not adequately signaled by the broad aggregate’s growth rate. Broad money growth in France does not accord well with inflation performance in the 1960s; moreover, it seems to lag behind the downturn in inflation in the 1980s. In the case of Italy, the short time span covered by the data makes evaluation difficult; there appears to be some correspondence in the movements of M2 growth and inflation, although, in many instances, the lags appear to be extremely short.

For the United Kingdom, the growth of the broad aggregate indicates a generally increasing inflationary tendency from 1975 through the 1980s, followed by a sharp dip beginning in 1990, The impulse measure, by contrast, turns down in 1979 and remains comparatively low through the 1980s—a pattern that is significantly closer to that of the consumer price inflation rate. In the case of Canada, however, the impulse measure does not outperform the broad monetary aggregate as an indicator of inflationary pressures.

Overall, the impulse measure performs somewhat better than the broad monetary aggregates, even though the aggregates have been selected as the most reliable of various available candidates on a country-by-country basis. Besides having the same definition for all countries, the impulse measure has the additional advantage of being based on a variable that—unlike broad aggregates—is potentially controllable by central banks.


In this paper, a measure of monetary impulse has been proposed that is intended to reflect the medium-term inflationary implications of a country’s current monetary policy. The measure consists of the growth rate of the monetary base, adjusted for reserve requirement changes and augmented by an implicit forecast of future growth rates of base velocity. As this forecast is based on past velocity growth—its average value over the previous four years—the impulse statistic reflects an easy-to-calculate measure of a variable that could be accurately controlled by any central bank that chose to do so.

Given the velocity growth feature of the proposed measure, its magnitude at a given time will reflect the implied medium-term growth rate of nominal GDP. The inflationary implications are readily obtained by comparing the monetary impulse with the trend growth rate of real GDP for the economy in question. Time series of the impulse measure have thus been plotted with low-inflation bands, centered around a rate of 2 percent, reflecting an appropriate average rate of annual real GDP growth plus 1-3 percentage points. These plots have then been used for a brief discussion of the 1965–93 inflationary experiences for each of the major industrial economies, and the paper has argued that the impulse measure provides a useful characterization of monetary policy behavior over those years. As the velocity adjustment term is designed to incorporate effects of technological and regulatory change in the payments and financial industries, the impulse measure should also be useful for analyzing policy stances in the future.

Appendix Description of Data

All data on nominal and real GDP are seasonally adjusted quarterly series. These series, as well as the data on consumer prices, the yield curve, and broad monetary aggregates, are obtained from national sources compiled by IMF staff. For the monetary base, various sources were utilized. These are described for each of the seven major industrial countries in the following paragraphs.

United States

The series came from the “adjusted monetary base” series that was developed by the Federal Reserve Bank of St. Louis and is available on the WEFA database. Monthly (seasonally adjusted) values were averaged to obtain the quarterly series.


These figures were taken from McCallum (1993). The basic underlying data were provided by Bank of Japan end-of-month series on “cash currency issued” and “deposits from deposit-money banks.” The latter series was adjusted for changes in reserve requirements, as explained in McCallum (1993). Then quarterly averages were calculated, added to the cash series, and the sum was seasonally adjusted to provide an adjusted monetary base series extending from 1963:1 to 1992:4, The extension through 1994:1 was based on Bank of Japan data available from Nikkei Services.


For 1960 to October 1968, the reserve series used was the sum of “total reserve required” and “excess reserves” published in the Bundesbank’s Monthly Report, Table II.A.5. For November 1968 to February 1978, the reserve series used was “actual reserves,” from Table IV.3 (a) and then Table IV.2 (a) of the Monthly Report, Beginning in March 1978, vault cash in banks began to count as reserves, so the reserve series became “actual reserves” plus “deductible cash balances” from Table IV.2 (a) and then Table V.2 of the Monthly Report. Beginning in August 1990, the reserve and vault cash data cover the entire deutsche mark currency area. In order to make the reserve data correspond to the data on currency in circulation (which cover unified Germany beginning in July 1990) the reserve and vault cash figures for July 1990 were scaled up by the respective series’ average ratios of unified to west-only figures for the next three months.

The resulting (monthly) reserve series was adjusted for reserve requirements changes, converted to a quarterly average series, seasonally adjusted, and added to quarterly averages of the monthly seasonally adjusted “currency in circulation” series.29 The latter series was provided by the Bundesbank and is reported in the Bundesbank’s Statistical Supplement to the Monthly Report, Table 1.2.


For 1960–77, data on notes and currency in circulation and bank reserves (quarterly averages), constructed by Sylvie LeCarpentier for a doctoral thesis, were provided by the Bank of France. For 1978-93, data on reserves and on coins and currency in circulation were obtained from the IMF’s international Financial Statistics.30 Data on vault cash (end of quarter) were obtained for 1990-93 from various issues of the following Bank of France publications: Statistiques Monétaires et Financières Trimestrielles (Tables 06 and 1.8); and Bulletin de la Banque de France, Supplément Stattstiques (Table

Basically, the monetary base series is the sum of these components, but the reserve series had to be adjusted for frequent major changes in reserve requirements. This adjustment procedure was necessarily more complex than in the case of Japan and Germany because reserve requirements were nonexistent before 1967 and became extremely low in 1992—possibly low enough to be nonbinding. The adjustment was not calculated as the product of the unadjusted reserve figure and rr0/rr (as is mentioned in the text and as was done for Japan and Germany); instead, the multiplicative adjustment factor for France was rr*/max(rr, rr*), where rr is the current required ratio for sight deposits and rr* is the ratio that would be held voluntarily in the absence of requirements. This rr* value was estimated to be 3 percent prior to 1967 and then to decline geometrically toward a value of 1.76 percent in 1994:1 (the value of each quarter is 0.9951 times the value of the preceding quarter).

The quarterly series for the adjusted monetary base is then equal to adjusted reserves plus the sum of notes and coins in circulation and those held by banks. After 1990:3, however, “reserves” includes currency in banks, and so the base is simply notes and coins outside banks plus adjusted reserves. The quarterly values of the monetary base were seasonally adjusted; annual values are averages of quarterly figures.


The basic sources were two monthly series, not seasonally adjusted but adjusted for changes in reserve requirements, provided by the Bank of Italy. The first series, from February 1962 to September 1990, presents end-of-month reserve holdings. Beginning in October 1990, reserve requirements have been specified instead in terms of average monthly reserve holdings, and the second series, from October 1990 to May 1994, is accordingly on a monthly average basis.31 The monthly values were averaged to yield quarterly figures, which were seasonally adjusted.

United Kingdom

For the United Kingdom, a quarterly series labeled “MO, break-adjusted, sa,” beginning in 1969:3, was obtained from the Central Statistical Office database. For quarters before 1969:3, the monetary base series developed by Capie and Webber (1985) was utilized. It was seasonally adjusted by means of the (multiplicative) ratio-to-moving-average routine of the Micro TSP software package and spliced on to the later series using 1969:3 as the splice date.


A monthly adjusted monetary base series, seasonally adjusted, was obtained from the CANSIMS database, based on Bank of Canada data. Because reserve requirements are being phased out, the adjustment for reserve requirement changes was intended to determine what the monetary base would have been had reserve requirements been zero over the entire period. The resulting adjustment amounts to the noninclusion of required reserves. However, this method may be somewhat problematic, as quantities desired for transaction purposes will exceed quantities required when requirements are small.


    CapieForrest and AlanWebberA Monetary History of the United Kingdom 1870-1982 (London and Winchester, Massachusetts: Allen and Unwin1985).

    EstrellaArturo and Gikas A.Hardouvelis“The Term Structure as a Predictor of Real Activity,”Journal of FinanceVol. 46 (June1991) pp. 55576.

    FischerStanley“International Macroeconomic Policy Coordination” in International Economic Cooperation a National Bureau of Economic Research Conference Reported. by MartinFeldstein (Chicago: University of Chicago Press1988).

    FriedmanMilton and Anna J.SchwartzA Monetary History of the United States 1867-1960 (Princeton, New Jersey: Princeton University Press1963).

    HargravesMonica and Garry J.Schinasi“Monetary Policy Financial Liberalization and Asset Price Inflation” Annex I World Economic Outlook World Economic and Financial Surveys (Washington: International Monetary FundMay1993).

    International Monetary Fund“Price Stability” Box 2 World Economic Outlook World Economic and Financial Surveys (Washington: International Monetary FundMay1993).

    McCallumBennett T.“Specification and Analysis of a Monetary Policy Rule for Japan,”Monetary and Economic Studies Bank of JapanVol. 11 (November1993) pp. 145.

Bennett T. McCallum is the H.J. Heinz Professor of Economics at Carnegie-Mellon University. Work on this paper was begun when Professor McCallum was a visiting scholar in the Research Department. The authors are indebted to Peter B. Clark and David T. Coe for advice and encouragement, to Robert P. Flood for comments on an earlier draft, and to Toh Kuan, Sheila Bassell, and Sungcha Cha for expert assistance.

Whether the aggregate spending measure should pertain to asset exchanges as well as production (or consumption) Hows is an issue, raised by Hargraves and Schinasi (1993), that will be reserved for a future study.

That actual central banks, at least in the major industrial countries, do not use the base as a control instrument is well known and will be discussed below.

Other approaches could be explored, such as extracting the “permanent” component of the velocity series by means of time-series decomposition methods.

Actual inflation targets and monetary policy objectives vary among the major industrial countries, but annual inflation of 2 percent may be considered roughly consistent with price stability. See International Monetary Fund (1993).

The estimates of long-run real growth rates are not intended to be refined measures. They are simply averages of realized rates over 1961-92, rounded to the nearest ½ of 1 percent (annual basis). The values obtained in this manner are 3 percent, except for Japan (4 percent for 1972-92), Canada (4 percent), and the United Kingdom (2 percent). It would be straightforward to adopt more sophisticated estimates of potential output growth.

As a sum of liability items on the central bank’s own balance sheet, the base could be monitored daily—or even more frequently—and adjusted by open market operations whenever observations depart from desired values.

Much has been made recently or the “stability” of M2 velocity in the United States over the past 30 years. But this phenomenon—more accurately described as the absence of an upward or downward drift—most emphatically did not prevail prior to 1960; see Friedman and Schwartz (1963, p. 640). As reasons for the change in behavior are not known, explicit or implicit predictions of future drift-free behavior should be regarded with some skepticism.

Recent papers on the information content of the term structure spread and the spread between short-term rates on private and government securities have explored their predictive value for aggregate variables. One of the explanations offered for the information content of these spreads is that they reflect the current stance of monetary policy. Some of these studies stress, however, that it is the real effects or monetary policy, not the effects on prices, that are relevant. See, for example. Estrella and Hardouvelis (1991). To the extent that the information content does derive from moneiary policy, however, the choice between using such interest rate spreads or the monetary impulse measure is in part an empirical one, and some assessments are made in the discussion on alternative monetary indicators.

Data underlying the measures are available from the authors.

The importance of adjusting for reserve requirements can be illustrated by the case of France in 1975. The International Financial Statistics’ “reserve money” series fell from FF 152.3 billion at the end of 1974 to FF 119.4 billion at the end of 1975, which might appear to be a contraction of epic proportions. Over the same time, however, the ratio of required reserves on sight deposits fell from 17 percent to 2 percent. With initial required reserves in excess of FF 50 billion, roughly FF 44 billion were freed by the reduction in the required ratio, and the overall effect was not strongly contractionary.

In the case of the United Kingdom, the series is not adjusted, but reserve requirements are low enough to be nonbinding.

In Germany, reserve requirements on demand deposits varied according to bank size until March 1977, after which they varied on a “progressive scale” based on the first DM 10 million of a bank’s deposits, the next DM 10-100 million, and so forth.

For Japan, the reference value of rr chosen is the 2.5 percent rate that prevailed without change front April 198I until October 1991. For Germany, the reference value is the 12.1 percent rate that prevailed from February 1987 to February 1992.

Seasonal adjustment has also been required for the Italian series.

Inflation did indeed fall beginning in 1990, but not all the way to 2 percent.

In addition, during much of 1992 and 1993, mortgage refinancing activity contributed greatly to faster Ml and reserve growth in the United States because some mortgage funds were mandated to hold demand deposits for a period following the prepayment of mortgages.

The 1986 episode is mentioned by Fischer (1988, p. 33).

Between 1986 and their peak in 1989, stock prices in Japan more than quadrupled, while goods price inflation as reflected in the consumer price index or GDP deflator remained below 2 percent annually. It is noteworthy in Chart 3 that the velocity adjustment term in 1987-90 is relatively large and negative, possibly reflecting in part the shift in the pattern of transactions during this period away from the flow of goods and services and toward assets whose sales are not included in GDP, For a discussion of the role of monetary policy in the asset price inflation in Japan—and, to a lesser extent, in the United States and United Kingdom—see Hargraves and Schinasi (1993).

The sharp decline in the IM measure in 1986 is an artifact of the reserve adjustment procedure used in this paper. In May 1986, reserve requirements on lime and saving deposits were reduced substantially, leading to a considerable reduction in reserve demand. However, reserve requirements on demand deposits were not changed at the time, so the reserve adjustment procedure described above does not capture this change.

The 0.12 magnitude exceeds the fraction of west German GDP represented by east Germany but is smaller than the analogous fraction in terms of domestic demand.

The 1990 values of the GDP and base variables are averages of quarterly values. Two quarters of the year passed before unification; therefore, even if the full impact occurred in 1990, the appropriate adjustment to the annual average growth rates is to assign half to 1990 and half to 1991.

During 1979-82, the monetary aggregate then being targeted, sterling M3, grew very rapidly—well above its target range—thereby inappropriately signaling that monetary policy was loose. The instability of M3 velocity at that lime was recognized and attributed to financial innovation and the removal in 1980 of the “corset” that had suppressed M3 growth.

The severe choppiness of the unsmoothed quarterly IMI measures that have not been reported comes primarily from the DLB series, rather than from DLVBAR (which is already smoothed), and presumably reflects the fact that the various central banks have not regarded stability of base growth rates as a desideratum. An exception to this statement is Germany, where central bank money was a target variable for a number of years.

The IMIAVG measure is equivalent to one based on growth rates calculated for each quarter relative to the same quarter one year earlier.

The adjustment made for German unification in the quarterly data was to subtract 7 percent from the growth rate of the monetary base in 1990:3, 3 percent in 1990:4, and 2 percent in 1991:1, which amounts to the adjustment of magnitude 0.12 made to the annual data. The caveats noted above in the discussion of the adjustment to the annual series apply here as well.

The sharp increase in inflation in 1980-81 is not matched by a correspondingly large rise in the impulse measure in prior yearn. However, as the other panels in Chart 11 show, alternative measures of monetary conditions also do not seem to capture this episode.

The choice of a broad aggregate differs across countries so as to conform somewhat with national preferences. The aggregates used are as noted in the charts.

Because of data limitations, one difference between the calculated base scries for Germany and those for other countries is that currency held by hanks is not counted in “currency in circulation.”

The quarterly series for coins is an end-of-period series. For 1978 to 1990:3, a quarterly series on reserves and currency was obtained by averaging monthly data. For 1990:4 to 1993, end-of-quarter values for currency and reserves were used because vault cash—which began to count toward reserve requirements in 1990: 4-is published as an end-of-quarter series.

The Bank of Italy’s Statistical Bulletin reports growth rates of its adjusted monetary base scries: underlying level data were obtained from the Bank of Italy.

    Other Resources Citing This Publication