Strategies and tactics for monetary control

Carnegie-Rochester Conference Series on Atblic policy 18 (1983) S-104 North-Holland

STRATEGIES

AND

TACTICS

FOR MONETARY

CONTROL*

Karl Brunner of Rochester

University

and Allan H. Meltzer Carnegie-Mellon

University

Nearly twenty years have passed since we undertook, at the request of a Congressional committee, to “appraise Federal Reserve policy in the light of the Congressionalmandate.” See Brunner and Meltzer (1964a, 1964b, 1964~). At the time, there was much less public and professional scrutiny of Federal Reserve decisions and actions. Minutes were not released; money targets were not announced; there were no regularly scheduled oversight hearings; and the weekly and monthly data on the money stock were not the subject of widespread speculation and comment. Market participants, and many economists, used the level or direction of change in short-term market interest rates and free reserves total reserves minus required reserves and member-bank borrowing -- as the principal measures of monetary “tightness” or “ease.” Much of the professional criticism was limited to comments about excessive or insufficient “tightness” or “ease” during particular peri0ds.l Our study focussed on policy arrangements, and procedures for implementing policy decisions, not the specific actions taken in particular circumstances. A principal finding was the absence in Federal Reserve analysis of a systematic connection between policy decisions and money, prices, and output. At meetings of the Federal Open Market Committee, and in our interviews with members of that committee, there was much discussion about prices, balance of payments, output, and unemployment. There was little evidence of a systematic framework linking these measures with the actions taken. In place of a systematic framework, we heard and read rules of thumb that, typically, identified specific levels of free reserves or the “color, tone, and feel” of the money market with broad, imprecise goals for the economy. *This paper developed from the invited comments given in April, 1981 by Karl Bnrnner on the two research volumes prepared mainly by the staff of the Board of Governors. See Board of Governors (1981). Ray Lombra and Thomas Mayer made helpful comments on an earlier draft. 1~ notable exception is Friedman and Schwartz (1963). Other early studies of Federal Reserve Procedures by Meigs (1962) and Dewald (1963) appeared at about this time.

0 167-2231/83/$03.00 @ 1983 ElsevierSciencePublishersB.V. (North-Holland)

One of our main criticisms of the Federal Reserve, as an institution entrusted with the conduct of monetary policy, was that it had not produced a firm foundation for its actions. Absence of a foundation was revealed by the neglect of research on either strategic or tactical issues and, also, by the often vague and imprecise statements of the connections between Federal Reserve actions and their effects. We noted that serious consideration had been given to these issues in the twenties by Riefler, Burgess and Strong, but in the forties and fifties, very little analysis of strategic or tactical issues appeared or was carried out. We found that Federal Reserve policy was pro-cyclical, on average, and concluded that pro-cyclicality resulted from the importance given to free reserves in Federal Reserve operations and in judgments about the direction or thrust of monetary policy.2 The Federal Reserve was not immune to the criticisms. Research on monetary policy increased. Federal Reserve procedures and interpretations of policy changed; short-term interest rates and the demand for money replaced free reserves and the free reserve doctrine in the Board’s discussions. The Board’s staff developed a general equilibrium model of the economy. The model represents an effort by the staff to impose consistency in the relation of policy action to forecasts of the path of economic activity and prices.3 In the seventies, recognition of the desirability of monetary control increased, and much greater attention was given, at least verbally, to monetary aggregates. The role assigned to interest rates became more ambiguous. Interest rates were referred to less often as the “target” of monetary policy and more often as the means of influencing monetary aggregates. In October 6, 1979, a further evolution occurred. Greater emphasis was to be given to monetary aggregates. By controlling nonborrowed reserves total reserves minus member-bank borrowing - the Federal Reserve claimed that it would improve control of the growth rates of the principal monetary aggregates. The new procedures continued or revived old themes, however. Required reserves were a fixed proportion of deposits two weeks earlier, under the Federal Reserve rule known as lagged reserve accounting, so banks’ demand for reserves to meet requirements was satisfied by changes in borrowing and free reserves. The Federal Reserve revived major pieces of the free reserve doctrine (see Meltzer (198 1)). For example, Chairman Volcker (1980) again refers to increases in member-bank borrowing as evidence of monetary re-

2We discussed some of these issues in terms of targets terms “targets” and “indicators” have acquired several meanings, discussion. 3A recent paper by Lombra and Moran (1980) and Pierce suggest that, at times, the policymakers refused to accept the

60

and indicators for monetary policy. The so we limit usage of these terms in this comments constraint.

on the paper by Jordan

and

striction and uses arguments based on the free reserve doctrine to explain his reasoning. Similar statements are made by other spokesmen. After a year of experience, the Federal Reserve conducted a study of the new procedures. TWO volumes, containing papers assessing performance in the year following the October 1979 announcement, give evidence of the technical ability and professional standard of many staff members. The principal shortcomings of the volume are related to the general framework used for analysis and policy evaluation. The framework, a standard IS-LM model, augmented by a Phillips curve, reflects developments in economics during the fifties and sixties but largely ignores later developments. Notable examples are the central ideas emphasized in the method known as rational expectations and illustrated by the time inconsistency problem. Also, the volumes usually separate tactical procedures for short-term control of the monetary aggregates (or interest rates) from the influence of tactical decisions on the growth of various aggregates, the persistence of “errors,” and the resulting pro-cyclicality of monetary growth. This paper discusses several of the contributions in the two volumes. The following section gives a summary of the main findings in several of the principal papers. Next, we consider the work on tactics, particularly the paper by David Lindsey and others that evaluates the experience in the first year using revised procedures to control monetary aggregates. We find that the main conclusion of the Lindsey paper is based on an incorrect analysis. Strategic issues concern the control of inflation, methods of reducing the cost of disinflation, and the use of stable versus flexible rules for monetary policy. Several of these issues are raised in the papers by Richard Davis, Dana Johnson, and Jared Enzler, and in Stephen Axilrod’s introduction. We comment on a number of these issues before considering an issue that receives scarcely any attention either in the two volumes or in other Federal Reserve studies in monetary economics - the case for, or against, the activist strategy. Our preliminary examination finds no evidence that Federal Reserve operations reduce the variability of output.

A SELECTIVE

SUMMARY

AND

CRITIQUE

The announcement on October 6, 1979 expressed a determination to improve control of money growth by giving greater attention to monetary aggregates. The announcement was followed immediately by a substantial increase in the level and variability of interest rates on a wide range of assets. In March 1980, controls were placed on borrowing and lending as part of a

61

revised policy proposed to reduce inflation. The new policy was followed by a 9% decline in the level of output and a decline in interest rates for all maturities. Growth of narrowly defined money, MlB (annual rate), fell from 7% in the first quarter to -3% in the second. Controls on borrowing and lending were removed in July. During the third quarter, money growth soared to 14.6%, and interest rates rose to new levels. The staff of the Board of Governors, with some assistance from the Federal Reserve Bank of New York, analyzed the causes and consequences of the variability of money and its relation to the change in procedures. The papers in the two volumes can be classified into three groups. One group addresses strategic issues. Included here are papers by Richard Davis, Jared Enzler, and by Enzler and Lewis Johnson. A second group, including papers by David Lindsey and others, Peter Tinsley and others, David Pierce, Fred Levin and Paul Meek, and Peter Keir, considers issues that are mainly tactical. A third group, consisting of four papers, considers the effects of the new procedures on the economy and on the foreign exchange market. This group consists of two papers considering international effects by Margaret Greene and by Edwin Truman and others, a paper by Lawrence Slifman and Edward McKelvey discussing the effects on economic activity, and a paper by Dana Johnson and others concerned with interest rate variability. An introduction by Stephen Axilrod summarizes principal findings and comments on both the strategic and tactical issues. Space does not permit consideration of all of the papers or the issues they raise. We have chosen to discuss some of the strategic and tactical issues at greater length, and within these groups to select papers, or sections of papers, that raise central issues. We ignore, for example, most of the discussion of shifts in the demand function for money. The volumes add little to the existing literature, and as in most previous discussions of the “shifts,” authors do not separate shifts induced by regulation of interest rates and reserve requirements from other short-term changes in the demand for money. The four papers considering economic effects suffered from an obvious handicap. There was too little experience to evaluate and too much difficulty separating the effects of the change in monetary control procedures from other major changes during the year from random variation. There is useful information, however. One example is the paper by Dana Johnson and others, “Interest Rate Variability Under the New Operating Procedures and the Initial Response in Financial Markets.” The paper presents evidence suggesting that the variance of interest rates increased substantially in 1980 compared to earlier periods. Johnson and his colleagues find little evidence that an increase in liquidity

62

premiums raised long-term rates relative to short-term rates. Instead, the data suggest increases in rates across the maturity structure. This finding of correlation across maturities is difficult to reconcile with a pattern of transitory shocks affecting interest rates, but it is consistent with a pattern of perceived permanent shocks.4 Why would the market perceive the change in control procedures as a sequence of permanent shocks to interest rates? The Federal Reserve study offers two, noncompetitive explanations, one by Axilrod, the other by Tinsley et al. Both explanations use the augmented IS-LM framework and try, unsuccessfully we believe, to reconcile the data provided by Johnson with the maintained hypothesis. Axilrod argues that the change in procedures fostered substantial confusion among market participants and, thus, increased the variability of interest rates.5 We have difficulty reconciling this argument with the data. Uncertainty about the new procedures has a transitory effect that vanishes as the credibility of the anti-inflation policy becomes apparent. The effect of uncertainty about procedures should be concentrated on short-term rates, and the variability should decline once the market learns that the new procedures improve the Federal Reserve’s control. Neither the level nor the variability of interest rates declined in the following years. The observed pattern is consistent with a pattern of perceived permanent shocks and increased monetary variability, not the temporary or transitory shocks required for Axilrod’s hypothesis. An interesting study by Tinsley and others analyzes the response to deviations of money growth from the announced target path. A rapid return to the preannounced path increases the variability of interest rates but lowers the variability of money growth. A slow return permits greater variability of money growth, measured as deviations from the announced path or variance of errors, but market interest rates vary less. The authors define the “reentry problem” as the choice of the speed of return to the announced path for money growth. The Federal Reserve is permitted to trade-off the variance in measured money growth against the variance of market interest rates. Tinsley’s analysis applies only to the variability of short-term rates. It cannot explain the increased variability of long-term rates or the association between variability of money growth, short- and long-term interest rates. Further, the proper response to a transitory change in money is to do nothing. The proper response to a permanent or persistent change, if it can be identified, 4Brunner,

Cukierman,

and Meltzer (1980) analyze perceived permanentand transitory shocks interest rates. Cukierman (1981) extends the analysis of interest

to income,prices,unemployment,and rates. ‘The

point is developed

in Axilrod’s

summary.

63

is to remove the change rapidly enough to prevent the change from being incorporated in the public’s expectations and in interest rates. Tinsley’s analysis does not distinguish between anticipated and unanticipated changes in money. Hence, the study cannot distinguish between unanticipated changes in money growth that are expected to persist and, therefore, to change market interest rates permanently and changes in money growth that are expected to disappear. The neglect of anticipations is critical. The failure to introduce expectations is one example that illustrates our claim that the Federal Reserve study failed to incorporate developments in economic theory made during the past decade. Shocks to interest rates can be either real or monetary in origin. It is difficult to find a shifting pattern of permanent or persistent changes in real variables that is broadly consistent with observed changes in interest rates. Perceptions of growing real budget deficits cannot account for the observed oscillations. Positive and negative changes in the expected real deficit did not occur with a frequency or magnitude consistent with the changes in nominal interest rates. A monetary explanation fares better. Chart 1 suggests that there is a relatively strong correspondence between the annual rate of growth in total reserves and the level of interest rates. A similar relation holds if the annual rate of growth of the monetary base replaces total reserves. The pattern is suggestive of a strong association running from maintained rates of change in reserves (or base money) to perceived permanent changes in money and to premiums in market interest rates. The Federal Reserve typically explains the association shown in Chart 1 as the result of shifts in the demand for money. In the IS-LM framework, an increase in the demand for money raises interest rates. Prices fall; real balances increase; and interest rates return to their previous level. This explanation can, at most, account for temporary changes in short-term interest rates. We see no way to use the explanation to account for the persistence of high real interest rates on short- and long-term debt more than two years after the change in policy procedures. Further, as we show below, the interpretation does not follow from the Federal Reserve hypothesis. Their hypothesis, as currently formulated, does not permit anyone to identify all deviations of money with shifts in the demand for money. The change in policy procedures was introduced, in response to pressures from foreign governments, to increase control of money growth and to lower inflation. Successful implementation should have lowered the variability of money growth and strengthened the credibility of the frequent statements about determination to control inflation. The volumes give no evidence of a decline in the variability of money growth. We have computed the variance

64

CHART Bank

Left

Scale:

and Short-Term

Interest

Rates

Adjusted Bank Reserves Right Scale: Interest Rate on Three-Month Treasury Bills

-----

-

Reserves

1

Total

- 17.0%

7.25%-

-15.6%

1 ’ F U

2.25

1.00

-

-

,,lll,lll,lrmlllllrmrmIlll~llll~lrrl

~O.OO%

1982

1981 19.81 Reserve data are four-week moving averages.

65

of unanticipated changes in money growth for periods before and after October 1979. (See Table 1.) The variance of money growth more than doubled, following the change in policy procedures. With the benefit of more time, we know that the increased variability cannot be explained entirely by the credit control program of 1980. In 198 1 the variability of money growth remained above the levels preceding the October 1979 change in policy procedures, And shortand long-term interest rates, after adjusting for inflation, remained far above the levels that prevailed before the October 1979 announcement. Suppose the public perceived the October 1979 change as a hesitant and possibly reluctant abandonment of an inherited strategy and the adoption of a poorly designed tactical procedure to implement a strategy of monetary control. A series of conflicting statements and the absence of a clear procedure for implementing the new strategy, when combined with increased variability of interest rates and money, increases uncertainty and leads to excessive concentration on short-term changes in money. The public searches for clues to determine whether the new procedures are likely to produce the promised policy of disinflation. Persistent growth of money above target increases skepticism and raises interest rates; persistent growth of money consistent with the announced target reduces skepticism. This pattern continues. Whenever per&tent disinflation becomes more believable, risk premiums decline, and interest rates decline at all maturities. The increased variance of interest rates at all maturities was accompanied by a similar increase in the variance of the daily and monthly spot rate of exchange. The variance of forward rates appears to have been lower after the October change. The studies of exchange rates in the volume do not include an analysis of the time series structure of exchange rates on which to base a clear assessment, This is regrettable because the exchange market contains information about the public’s perceptions of shocks. Increased variance of spot rates and reduced variance of forward rates suggest that shocks are dominantly transitory, contrary to the hypothesis supported by the movements of interest rates and money. The study of effects on economic activity by Slifman and McKelvey suggests that the October change caused a change in inventory behavior. This study attributes the increased variability of interest rates to the change in procedures, but it does not relate the higher variance of interest rates to the choice of strategy and tactics. A principal finding of the paper is that the increased variability of interest rates raised the cost of holding inventories and lowered target levels of inventory. The evidence offered in support of this conclusion is recognized to be relatively weak and difficult to separate from other large changes that occurred during the year.

66

TACTICS The variance of money growth and interest rates around a chosen target path or level depends on the control procedures, or tactics, used by the Federal Reserve. If all shocks are known to be transitory, tactical problems are reduced to an issue of operating efficiency - whether greater operational efficiency in implementing a strategy is obtained by controlling the monetary base, reserves, interest rates, free reserves or another variable. When some shocks persist and others are transitory, the choice of tactics influences expectations about money growth, output, and inflation. Errors perceived as permanent (or persistent) induce changes in expected future values and, therefore, induce changes in borrowing and lending and in the demands for money and commodities.6 The choice of tactics raises or lowers the social cost of disinflation, under these conditions, by affecting the credibility of the public in Federal Reserve policy statements. Doubts about the Federal Reserve’s ability to control money growth are expressed frequently by Federal Reserve officials and others. Many of these statements assert that there is, at best, a loose connection between the money stock and the variables which the Federal Reserve can manipulate. In our studies for the Banking. Committee (1964) and later, we maintained that ineffectual control of money is not inherent in the money supply process; loose control of money results from the tactical decisions of the Federal Reserve. The Federal Reserve’s commitment to targets for free reserves (or nonborrowed reserves) or interest rates and the institutional practices7 it imposes or tolerates, we claim, increases the variability of money and reduces control. Analysis of these issues for the Shadow Open Market Committee by James Johannes and Robert Rasche (1979), and their forecasts, prepared for each meeting of the Committee, show that quarterly errors in forecasting the rate of change of the multiplier of the monetary base, generally, are less than 1% despite current institutional arrangements. These estimates are true forecasts, made in advance of the event. The implication of these forecasts is that, with strict control of the monetary base, money growth can be held within + 1% of the annual target within semiannual or quarterly periods. 6The importance of permanent and transitory shocks for the tactical problem should be clear from our discussion of the reentry problem analyzed in the paper by Tin&y and others. See also Meltzer (1980) who shows that when the frequency of persistent shocks to aggregate demand is large relative to shocks to the growth rate of output, there is a strong case for controlling monetary aggregates. When the principal persistent shocks affect the growth of output, the case for interest rate control is strongest. ‘IThese include interest rate ceilings, lag reserve requirements, an inflexible discount rate, a complex structure of reserve requirements, and the publication of seasonally adjusted numbers based on imprecise seasonal factors.

67

For many years, the Federal Reserve did not participate in public discussion of tactical issues. The volumes under review are a delayed break with this tradition. Several papers analyze tactical problems, and the issue involved therein is discussed in Stephen Axilrod’s introduction. In this section, we summarize several of the main findings in the principal paper on operating procedures by the Board staff. Next, we discuss some of the issues raised in that paper and in related papers on procedures. The volumes do not contain analyses of the interaction of operating procedures, money stock, interest rates, and the demand for money. To complete this section, we offer our own analysis of the Federal Reserve’s tactics. Lindsey and others studied the controllability of money using four distinct frameworks - (1) a “judgmental approach” developed at the Board; (2) the Johannes-Rasche procedure used by the Shadow Open Market Committee to forecast the multiplier of the monetary base; (3) the “monthly” econometric model developed by the Board’s staff; and (4) an alternative model developed at the Federal Reserve Bank of San Francisco. There are three versions of the judgmental procedure, reflecting three types of information. The staff chose four reserve measures - nonborrowed reserves, total reserves, the nonborrowed monetary base (the base minus borrowings from the Federal Reserve); and the monetary base - as control variables and made forecasts of multipliers for three definitions of money, M-IA, M-IB, and M2. We concentrate on M-W and reproduce the relevant parts of Tables 4 through 8 of the paper. Tables 4 through 6 report errors in monthly forecast at annual rates. Tables 7 and 8 report quarterly errors of forecast at annual rates. The errors in forecasting the rates of change of the multipliers are the errors in forecasting the growth rate of M-1B using a particular reserve aggregate from October 1979 to October 1980. Several findings from Table 4 suggest that the choice of tactics is not an irrelevant detail. Monthly absolute errors at annual rates range from 4.0%, using the Board’s monthly model and controlling the monetary base, to 31.4%, using ,the San Francisco model and nonborrowed reserves. Control of the monetary base produces the lowest average absolute error for each forecasting technique separately. Control of either the nonborrowed base or the monetary base produces the lowest mean errors for ail reserve measures. Mean absolute errors are largest for all forecasting techniques, using nonborrowed reserves. The authors argue that the estimates in Table 4 are misleading because they do not correct for the endogeneity of the various reserve measures.8 The

%-II

e mean error reported in Table 4 for the Johannes-Rasche procedure differs little from the error computed using the “true” (made-in-advance) forecasts at the semi-annual meetings of the Shadow Open Market Committee. These data are available for examination.

68

Lindsey’s Monthly

Average

Table 4

Error Statistics for Monthly Bate of Change of Money Judgmental and Econometric Forecasting Techniques October 1979 -October 1980 (in annual&d percent)

Multipliers,

NSA

M-U Mean error

Mean absolute error

RMS

2.7 3.0 1.8

12.7 9.9 8.6

14.9 12.0 11.0

4.8

20.7

26.2

-9.0

18.9

24.7

Model

13.6

31.4

36.0

Total Reserves Board Judgmental Initial Intermeeting Period r’ Adjusted Intermeeting Period L’ 2’

4.3 -3.4

8.9 6.5

10.3 8.2

-1.9

5.0

6.1

-3.0

14.6

16.6

-2.8

8.2

9.5

-5.6

16.0

20.4

0.2

1 .o

8.4 8.1

10.4 10.5

0.3

4.5

5.6

a.6

9.9

11.7

-35

6.1

8.2

2.4

9.4

11.2

Forecasting Techniques By Reserve Measure Nonborrowed Reserves Board Judgmental Initial Intermeeting Period L’ Adjusted Intermeeting Period -1121 Current Month 3_’ Johannes-Rasche Board Monthly San Francisco

Current

Model

Month

3_’

Johannes-Rasche Board Monthly San Francisco

Model Model

Nonborrowed Monetary Base Board Judgmental Initial Intermeeting Period U Adjusted Intermeeting Period L/ z1 Current

Month

L’

Johannes-Basche Board Monthly San Francisco

Model Model

69

a0*

Lindsey’s Monthly

Average

Table 4 continued

Error Statistics for Monthly Rate of Change of Money Judgmental and Econometric Forecasting Techniques October 1979 -October 1980 (in annualized percent)

Multipliers,

NSA

M- 1B Mean error Forecasting Techniques By Reserve Measure Total Monetary Base Board Judgmental Initial Intermeeting Period l’ Adjusted Intermeeting Period r’ 2’ Current

Month

3_’

Johannes-Rasche Board Monthly San Francisco

Model Model

Mean absolute error

RMS error

-I .3 -0.3

6.1 6.7

7.2 8.5

-0.6

4.1

4.7

a.5

8.0

9.0

-1.9

4.0

5.0

-2.4

6.6

8.6

‘From October 10, 1979 to February 6, 1980, projection errors of old M-1 are reported and M-U. AB the percent errors are annualized by 12 and include the October 29-November period.

for M-IA 19, 1980

lEnor of initial multiplier forecast adjusted for intermeeting changes in targeted reserve path. 3From October 1979 to January 1980, projection errors of old M-l are reported for M-L4 and M-U and projection errors of old M-2 are reported for M-2. n.a. -- not available

70

point is that when errors in money growth are observed during the period, the path of the reserve aggregate changes, Such changes in the reserve aggregate say nonborrowed reserves - are not considered when computing summary statistics, so the observed monthly changes (and variances) of the money stock and the associated reserve multiplier are overstated. The staff used two procedures to “correct” for the endogeneity of the reserve aggregates. The results are shown in Table 5 and 7 and Tables 6 and 8 of the paper by Lindsey and others. Table 5 compares the predictive performances of Board and the San Francisco models. This Table and Table 7 correct the observed multiplier error for changes in the path of reserves made within the period. The absolute errors using nonborrowed reserves fall substantially below the values in Table 4, but they ‘remain above the errors made on the assumption that the nonborrowed base is controlled. The results using the San Francisco model fall within a relatively narrow range when allowance is made for the fact that all errors are annualized. Tables 6 and 8 compare the same two models using an alternative procedure. The nonborrowed base again produces the lowest average absolute error and the lowest root mean square error, but the superiority of the nonborrowed base compared to nonborrowed reserves is less than before. With but one exception, all the mean absolute errors now fall within the range 4.0% to 8.4% at annual rates, or, as monthly averages, l/3% to l/2%. This is substantially better than the control actually achieved. The quarterly data, in Table 8, reduce the mean quarterly average error (at annual rates) to the range 0.6% to 1.8%, using one of several different procedures and any of the four control variables. Three main conclusions can be drawn from the study: (1) The results confirm that monetary control is technically feasible. Acceptable control can be achieved within a oneto twoquarter period. The results from several procedures and all models - Johannes-Rasche, the Board, and the San Francisco - deny the repeated assertion that the unstable demand for money destroys the controllability of money growth. (2) Memos attached to the tables and statements in the text confirm that lagged reserve requirements reduce control. “The system of lagged reserve accounting makes the monthly average required reserve ratio by type of deposit quite unstable...and quite unpredictable as well.” (Board of Governors (1981), II, p. 52 of the paper by Lindsey and others.) A change to contemporary reserve accounting would improve monetary control.

71

Month

Path z’

3_1

Total Reserves

Base 3_i

a1

Reserves

Total

Nonborrowed Monetary Base Board Monthly Model Given Exogenous Nonborrowed

Total Reserves Board Monthly Model Given Exogenous San Francisco Model Given Exogenous

Econometric Procedures by Reserve Measure Nonborrowed Reserves Board Monthly Model Clen Actual Nonborrowed Reserves San Francisco Model Given Actual Nonborrowed Reserves

Current

FOMC Interim Money Stock Targets IntermeetIng Period Path 11

4.4

0.1

-1.7

0.5

(continued)

3.9

8.3

13.2

4.8

-2.1

4.2

-1.0

5.6

10.6

18.6

5.6

6.5

5.0

a.4

0.8

4.6

0.0

-2.1

-1.0

4.9

5.5

6.8

13.2

4.0

7.5

9.0

18.6

4.9

8.2

5.0

6.0

9.8

7.6 4.2

eN0r

erroI

9.8

RMS

Mean absolute Man erro*

WlB

Mean absolute error RMS error

Table 6 PROCEDURE

Lindsey’s SECOND

M-1B

1.6

4.6

Table 5

PROCEDURE

-0.9

WOI

Mean

FIRST

Lindsey’s

Erro.r Statistics for the Monthly Rate of Growth of the Monetary Aggregates, NSA Actual Versus Targeted and Actual Versus Predicted from Econometric Models October 1979 -October 1980 (in annualized percent)

rate.

Table 5

5.3

7.9

2.8

Mean error

and M-1B.

All the percent

-0.3

-2.5

0.1

errors

are annualized

6.4

8.4

4.0

8.9

10.4

4.9

RMS error

funds

by 12 and include for M-IA and M-1B and errom for old M-2 are reported for new M-2. of this reserve aggregate given the judgmental prediction of the federal

for M-L4

8.1

9.9

3.2

eN0r

Mean absolute error

RMS

PROCEDURE

Mean absolute error

SECOND

Table 6

M-1B

I

Lindsey’s

M-1B

PROCEDURE

are reported

FIRST

Lindsey’s

1979 to January 1980, projection errors for old M-l are reported level of each reserve measure is equal to the model’s prediction

na. -- not available

2From October 3The exogenous

for old M-l

0.4

Base &’

Total

0.9

-3.7

Base &/

MtWl error

Total Base 3_’

Nonborrowed

1From October 10, 1979 to February 6, 1980, errors the October 29-November 19,198O intermeeting period.

Total Monetary Base Board Monthly Model Given Exogenous San Francisco Model Given Exogenous

San Francisco Model Given Exogenous

Table 5 and Table 6 continued

Error Statistics for the Monthly Rate of Growth of the Monetary Aggregates, NSA Actual Versus Targeted and Actual Versus Predicted from Econometric Models

Lindsey’s

Current

~te-fon-on;;~~~

Month

Path L’ 3_’

1,

21

0.2

Base *’

0.6

0.8

0.7

2.2

3.1

1.2

0.1

0.9

(continued)

1.1

1.0

2.8

4.4

1.6

1.2

1.1

2.4

WIOf

errOr

1.8

RMS

absolute

Mean

M-1B

0.8

-2.3

Table 7

PROCEDURE

-0.8

-0.7

Reserve-s *’

Total

a.3 a.4

Base 4_1

Reserves *’

Total

Nonborrowed Monetary Base Board Monthly Model Given Exogenous Nonborrowed San Francisco Model Given Exogenous I&borrowed

Total Reserves Board Monthly Model Given Exogenous San Francisco Model G&n Exogenous

Econometric Procedures by Reserve Measure L1 Nonbolrowed Rezsxvee Board Monthly Model Given Actual Nonborrowed Reserves San Francisco Model Given Actual Nonborrowed Reserves

FOMC

mor

MW

FIRST

Lindsey’s

-0.1

a.2

0.7

-2.3

-0.1

-0.8

-0.4

-0.3

eIIOI

Mean

Errox Statistics fox the Quarterly Rate of Growth of the Monetary Aggreg&es, NSA Actual Versus Targeted and Actual Versus Predicted from Econometric Models October 1979 -September 1980 (in annualized percent)

SECOND

Table 8

1.0

0.6

1.8

3.1

1.0

0.8

0.9

1.8

error

absolute

Mall

M-1B

PROCEDURE

Lindsey’s

1.3

0.6

2.4

4.4

1.3

1.0

1.1

2.4

elT0r

RMS

Monetary Base Board Monthly Model Given Exogenous San Francisco Model Given Exogenous of three

4The exogenous level is equal to the model’s n.a. - not available.

prediction

periods,

1.6

1.2

ammalized

by a factor

2.1

1.5

error

RMS

errors in each calendar

by a factor

of 4.

intermeeting for new M-2.

1.6

0.0

of four adjoining

1.7

absolute error

Mean

M-1B

of the federal funds rate.

annualiied prediction

quarter,

Table 8 PROCEDURE

-0.7

of 4. (Averages

I

SECOND

Lindsey’s

for M-1,4 and M-1B and errors for old M-2 are reported

absolute error

Mean

M-1B

of this reserve aggregate given the judgmental

percent

Table 7

PROCEDURE

are reported

intermeeting

errors for old M-l

adjoining

averages of monthly

1980, projection

as three-month

1979 to January

as averages

0.4

Total Base *’

calculated

-1.2

en01

Mean

Total Base g1

errors calculated

October

3Quarterly

lFrorn

l”Quarterly” errors give very simii results.)

Total

I

FIRST

Lindsey’s

Lindsey’s Table 7 and Table 8 continued Error Statistics for the Quarterly Rate of Growth of the Monetary Aggregates, NSA Actual Versus Targeted and Actual Versus Predicted from Econometric Moiiels

periods

2.2

2.0

RMS error

(3) The study recognizes that the traditional operation of the discount window lowers the degree of control. A similar conclusion is reached by Keir. (See Board of Governors (198 l), I, p. 8 of the paper by Peter Keir.) The period considered is relatively short, and no comparison is made with earlier periods when different methods of control were used. A longer period also would have permitted a comparison with the longer series of forecast errors available for the Johannes-Rasche model. Nevertheless, despite substantial shocks and large uncertainty during the period, the data show that considerable improvement in monetary control can be achieved. More than a year later, none of the institutional changes to improve control has been made. A Critique of the Study The study makes a strong claim that the results in (their) Table 4 are misleading. The claim is that under a nonborrowed reserve or interest rate control procedure, the base is an endogenous variable determined jointly with the money stock so that the reported errors for the multipliers of the base (and total reserves) are biased. As already noted, the authors claim that the results in Table 4 overstate the errors made in estimating the multiplier for unborrowed reserves. Expressed concern for the endogeneity problem leads the authors to minimize the relevance of the findings reported in Table 4 and to emphasize the findings in Table 6, particularly. 9 The authors’ argument permits them to dismiss the Johannes-Rasche results as irrelevant. There are three parts to the authors’ argument. First is the endogeneity of the base, total reserves, and other monetary aggregates under current (and past) operating procedures. There is no dispute about endogeneity under the Federal Reserve’s procedure. Endogeneity does not imply the operation of an important bias lowering the forecasting error associated with the monetary base multiplier. Second, the study adjusts the multipliers to correct for changes in nonborrowed reserves made in response to unanticipated deviations of monetary growth from target. This adjustment is reported in Table 5 of Lindsey’s paper. We accept this adjustment as relevant and correct, but we do not accept the assertion that the Johannes-Rasche procedure is biased and should be dismissed. The Federal Reserve study does not support this claim analytically. Third, Lindsey’s study makes a further adjustment when simulating the two Federal Reserve models. This adjustment, reported in Table 6, appears to us mistaken.

gAxilmd’s summary ass&s three times that the connection between the money stock and the monetary base is “looser” or suffers “greater slippage” than the relation with nonborrowed reserves. We find no support in their Tables 4, 5, or 6 for this claim. The relevant results are in the tables reproduced for our paper.

76

The authors’ procedure is not fully described in the text, but we believe we have a correct understanding.10 Consider the linear system

Ay1+uy2+bxl +cx*+u=o, where Y 1 is a vector of endogenous variables, y2 is the endogenous monetary base,

~1

is the exogenous federal funds rate, x2 is a vector of all remaining

exogenous variables, and u is a vector of disturbances. A and C are matrices of appropriate form; a and b are scalars. The first step of the procedure inserts observed values of all exogenous variables, ~1 and ~2, into the linear system and derives the expectation of y2, the monetary base, conditional

on xl and x2. The expected value of the mone-

tary base obtained in this step is unavoidably exogenous with respect to the disturbance terms. The second step of the procedure, to remove endogenous components from the errors made when controlling the monetary base, uses the expected values of the base (~2) associated with an observed xl as an (exogenous) input. To compute errors for the money stock (or monetary growth), the realizations (observed values) of the disturbance vector, U, are used also. These disturbances were obtained from a model estimated on the assumption that xl is exogenous. With the base (32) fixed at its expected value, y2 cannot reflect the effect of the random disturbances. The authors adjust xl (the federal funds rate) to satisfy the equation.

This step makes xl depend on the random disturbances; xl is

no longer exogenous, as assumed in the underlying model. The authors’ case for dismissing the results shown in (their) Table 4 depends critically on the allegations that endogenous influences produce major distortions in variables treated as exogenous in the monetary control process. Since the authors’ procedure is open to the same,criticism that they level at the Johannes-Rasche procedure, they must dismiss the endogeneity introduced into x 1 by their procedure as a matter of little importance. They must claim that the disturbance vector, u, is independent of the choice of exogenous variables. The sequence of “true” forecast errors using the Johannes-Rasche procedure provides relevant evidence. The continued reliability of the procedure during a period with different policy regimes supports the claim of robustness with respect to these changes in regime. 1 1 The same claim cannot be made for 10~ letter from Kenneth that was used. We accept, for current ’ lln Lindsey

(1981)

Kopecky purposes,

of the Board’s staff helped us to understand that the linear system is correct.

the claims about endogeneity

is given.

77

are repeated,

the procedure

but no new evidence

or analysis

the Federal Reserve model simulated by Lindsey. The Federal Reserve study suggests that several of the equations in the model changed or become less reliable. Comments on the borrowing, exchange rate, and interest rate equations indicate that these relations shifted in unforeseen ways after October 1979. Shifts in these equations suggest that the Lucas criticism of econometric models is applicable to the Board’s model. Shifts of the models’ equations invalidate the procedure used by Lindsey. The errors reported by Lindsey and others exaggerate the control prpblem. Their tables report mean monthly errors and root mean square errors as annualized values. The monthly values of mean absolute errors are multiples of the mean errors. This distorts the information and suggests that monetary control faces a staggering problem. The corresponding quarterly errors are much closer to the mean monthly error than to the mean absolute monthly errors. This suggests low serial correlation of monthly errors. The proper method of presentation depends on the underlying stochastic process. Annualization is an assertion, possibly unintended, about the persistence of errors. It suggests that the observed monthly errors are a permanent part of the process and that errors are serially correlated. Since the relation between monthly and quarterly errors shows little evidence of serial correlation, errors appear to be transitory during the sample period and tend to cancel within a few months. 12 Axilrod offers an alternative justification for the use of reserves as a control variable. He claims that, by controlling reserves, the Federal Reserve can offset variations in the public’s holdings of currency more effectively than by controlling the base. Governor Wallich has frequently made a similar claim. A comparison of the variance structures of the multipliers for nonborrowed reserves (or total reserves) and the monetary base establishes that a positive covariance between changes in the ratio of currency to demand (or checkable) deposits and the ratio of time to demand deposits is a necessary condition for Axilrod’s statement. Negative covariances are necessary for the superiority of the base. Further, the truth of Axilrod’s claim is difficult to reconcile with the differences in forecast errors for nonborrowed reserves, total reserves, and the monetary base using the Johannes-Rasche procedure. Fed Tactics and the Demand for Money The research volumes contain no discussion of the framework used by the Board or its staff, and there is no attempt to relate the choice of tactics to 1’Lindsey (1981) adjustment for the discussion staff studies.

again reports annualized errors, without of this issue at the meetings in January

78

comment. and April

His discussion makes no 1981 called to discuss the

the theory of money. In earlier work, Brunner and Meltzer (1964b), we traced the development of the free reserve doctrine from the notions advanced by Riefler, Burgess, and Strong in the 1920s. The emphasis, then, was on the role of borrowing. Borrowing was taken as exogenous. Banks were said to be reluctant to borrow, so increases in borrowing raised interest rates. Later, this thesis was extended to include excess reserves, as negative borrowing, and excess reserves minus borrowing became known as free reserves. The Federal Reserve operated on free reserves to affect interest rates and economic activity. Sometime in the sixties the free reserve doctrine was replaced by a version of the IS-LM model.13 Emphasis shifted from free reserves and borrowing to the demand for money, but the tactics shifted much less than the verbiage. The Federal Reserve continued to set targets for short-term interest rates and to focus principally on interest rates and money market conditions. Pressure from Congress and the academic community eventually forced the Federal Reserve to give more attention to the money stock, Ml, and other monetary aggregates in discussions of monetary policy, but these pressures had little influence on operating procedures until October 1979. Moreover, as recently as 198 1, the Board’s monthly model continued to treat the federal funds rate as a policy variable controlled by the Federal Reserve. Other interest rates are seen as determined by the federal funds rate - which is to say, by the Federal Reserve with limited influence from market forces. See Anderson and Rasche (1982). Despite increased reliance on econometric models by the staff, the open market committee continued to use a vague and imprecise framework for tactical decisions. The committee chose to avoid the conflicts that would arise if there were a serious effort to form a common view of the monetary process. See Lombra and Moran ( 1980).14 Modelling the process used by the Federal Reserve gives an impression of greater precision and more agreement than the facts warrant. Nevertheless, we use a model of the process to capture the main ideas of the action. l5 Officials and staff usually explain sustained, large positive or negative deviations of money from its path as an unanticipated shift in the demand for money. Suppose money demand is a distributed lag function of the federal funds rate, ffr, with ffrt-i used to represent all relevant past values of the funds 13Elements of the free reserve doctrine after October 1979. See Meltzer (1981).

coexisted

in the sixties

and returned

in a modified

form

14Lombra and Moran (1980) show that the discussion was highly ambiguous, and the views of the process were not consistent. The use of interest rates as an instrument for monetary control shifted easily, given this ambiguity, into a process of controlling interest rates. 15Sweral comments at the conference suggested that similar models have been used by others to represent the Federal Reserve’s approach. David Lindsey supplied references to several versions of the model, using lagged and contemporaneous reserve accounting, exogenous base, exogenous total reserves, etc. Early examples are Brainerd (1964), de L.eeuw (l%S), and Dewald (1966).

79

rate. Y is the projected value of aggregate income, and M is the stock of money. M, = M,-,

+ M,

= hVfrt+

Y, u2, $

Equation (1) is a relation between the federal funds rate and the change in the money stock, The Federal Reserve’s tactical procedure distinguishes between the long-run target path, M*(t), and the short-run target, M**(t). Short-run targets are set to reduce deviations between actual M, and M*(t). M**(t) - M,-l = AM**(t) is the short-run target for money growth. Until October 6, 1979, the Federal Reserve chose the federal funds rate that it believed was consistent with AM** given its forecast of Y, the relevant past history, and the expected value of q> t’ l6 An obvious problem with the procedure is the absence of a constraint linking AM**(t) to M*(t). The choice of an appropriate constraint requires either knowledge of the structure of 242, t or a procedure that minimizes the cost of ignorance. The reason is that persistent or permanent deviations can be mistakenly perceived as transitory, or vice-versa, and demand shocks can be mistakenly perceived as supply shocks, or vice-versa. Suppose that a persistent positive shock to aggregate demand occurs but is perceived to be transitory. With ffrt fixed, AM, exceeds AM**(t). Since the shock is perceived to be transitory, positive deviation of

ffrt+l

= ffrt,

and there is a second

Mt from M*(t) and M**(t). A series of positive shocks

affects anticipations, and therefore affects borrowing, the demand for money, the rate of intermediation, and the money stock. Unless one translates the accumulated deviation of M from M ** into a setting for the federal funds rate, any change in the funds rate can be either too large or too small to minimize the cost of returning to the desired path. The October 1979 change can be viewed as an attempt to let interest rates change with transitory changes in borrowing and the demand for money. Nonborrowed reserves, NBR, were to be used to keep Mt closer to the chosen short- or long-run path. Let BR, RR, and ER represent borrowed required reserves, and excess reserves, respectively, where

NBRtBR=RRtER.

reserves,

(2)

16 According to Lombra and Moran (1980) and other observers, the open market committee often changed the staff estimates of the value of w*(t) which they believed to be consistent withffrj. This can be interpreted as a different estimate of Euj or Y.

80

Free reserves F = ER - BR, so NBR=RR+F

(3)

Under current policy arrangements, required reserves depend on past deposits. Using two weeks as a period to reflect the current rule,

where

Pf-1 =PU(X,+qt-lL

The ratio, ~1,has the same determinants as the money multiplier associated with the monetary base. We summarize the systematic components of the multiplier in X and the stochastic components in u 1. The staff study and other recent work suggest two alternative hypotheses about free reserves or borrowing. The alternatives are written as (5a) and (5b).17

F, = d.ffrt, disct, h, ~3).

(5a)

ffrt =f(F,, disct, ~3).

(5b)

Both relations assume a negative association between the funds rate and free reserves. One, (5b), is the old Riefler, Burgess, Strong “reluctance theory” of the relation between borrowing and interest rates; (Sa) reverses causality and emphasizes portfolio adjustments in response to relative prices. The discount rate, disc, has a positive coefficient, and the administration of the discount window, h, supplements the discount rate as a control instrument for borrowing in equation (5a). Equations (1) (3), (4) with either (5a) or (5b) provide two hypotheses about the structure of the monetary process and the relation of shifts in the demand for money to interest rates and the money stock. The portfolio adjustment hypothesis uses (5a); this hypothesis reduces to a system of two equations, (6a) and (1). l’For convenience, we use u3 in both equations. Formally, this is incorrect because the stochastic processes are not identical. The difference is not simply a formality. Statements made by the Federal Reserve in 19378 to justify increases in reserve requirements are a logical implication of (Sb) but cannot be obtained from (5-a). The statements claimed that excess reserves were a residual that could be removed without further consequences for money and credit. Also, the Federal Reserve claimed that, in the absence of borrowed reserves, central bank policy had no effect on money.

81

(64

NBR, -gcffrrt, disct, h, ~3~) = i.dXt-1,

(1)

ult-1)

Mt = XVfrrt, ffrt-i,

Wfrt..l

,ffrt-i.

K u2t-1).

Y. ~2t) .

The “reluctance theory” reduces to three equations. Equation (1) is the same, but (6a) is replaced by (5b) and (6b).

1.

(6b)

A principal difference between the two hypotheses is that the reluctance theory makes Ft depend only on u 1t- 1 and uzt- 1; under the alternative hypothesis, F, depends also on ~43~.l8 A linear approximation

to the solution

can be written in the general form expressed by (7a) and (7b). a(LjMt = /3(L)NBRt + Y(L)ZQ + 6(L)Xt

.

(74

U’b) Money and the federal-- funds rate have the same autoregressive structure, and the parameters 0, 6 and p, 6 are related by the structural equations. In (7a) ut is the vector 1~1~~1, ~2~~1, 9, in Ub)

ut = hilt-l,

~3~1,

uzt-bJ ujtl.

and X, is a vector of remaining variables; All lag coefficients

of a(L) and p(L) are

negative. Officials and staff explain most surprises in money growth and deviations from the target path as the result of unanticipated changes in money demand. Such statements acknowledge that restrictions on interest rates continue to have a decisive influence on money growth. Inspection of (7a) and (7b) .shows, however, that there is no reliable basis for the identification of errors in money growth with shifts in the demand for money, as is commonly done. 19 The shock vectors u and u contain ~3, the stochastic element in F or ffr from 18We note that the staff report, undertaken to learn about the Federal cedures, never investigates these implications and never discusses the policymukers’ the reluctance theory. lgSee the discussion of Figure 1 above,

82

Reserve’s control prohypothesis based on

(5a) or (5b). Even with P and M fixed in t- 1, surprises reflect the combined effects of ~2 and ~3 under the portfolio adjustment hypothesis (7a). The size of u3 depends on the Federal Reserve’s ability to specify an operationally useful hypothesis about banks’ use of discount facilities. The staff study suggests that this problem has been addressed but has not been solved; the borrowing equation is not very reliable and appears to have become less reliable following the October 1979 change. The equation is replaced, at times, by extrapolation of past values. The worsened performance of the borrowing equation suggests that the shift from interest rate control to control of free reserves changed the variancecovariance structure of at least some of the equations in the staff model. The tactical procedure for controlling money has a major difficulty that has not been overcome. The procedure depends on two relations - the demand for money and the borrowing relation - that, according to the staff and the policymakers, are highly volatile and unreliable. Axilrod (“Overview”, p. 32) attributes the “looseness” of monetary control to the borrowing relation; Lindsey ( 198 1, p. 18) locates the problem in the demand-for-money function. Both errors contribute to the deviations of M from its targets, M* or M**, according to the Board members’ policy models. Equation (8) is a simpler version of (7a).

M, = al NBR, + a2 qt + a3 qt +x,

.

In advance of period t, policymakers have to choose the best estimates of u2 and u3 to determine the value of NBR, that achieves MF. The “true” values of u2 and u3, Eu2 and Eu3 are not known. No one knows the extent to which past shocks are permanent or transitory or the market in which shocks originate. The staff or the policymakers choose expected values, E,*uZt and EFz+ conditional on available information at the time decisions are made. The size of the difference between the true values and the perceived, expected values reflects the reliability of the underlying relations. The true conditional variance of deviations from target depends on the accuracy of the perceptions given the information available at 1, as shown by (9). (9) The conditional variance recognizes that the values of the stochastic terms in t-l are given and known. The unconditional variance includes the latter terms

83

and, consequently, is larger.. It would be interesting to compare the predictive quality of this framework to the forecasting procedure for money multipliers developed by Johannes and Rasche. Axilrod finds the accommodative element arising from the borrowing relation to be advantageous. The procedure accommodates shocks to demand by allowing M to deviate from its path. This alleged advantage has obvious drawbacks. Permanent shocks to the demand for money drive money from its path and force the Federal Reserve either to respond or change M*. There is latent skepticism and a high degree of uncertainty about which choice the Federal Reserve will make at any time. Further, accommodation is not limited to unanticipated changes in the demand for money. The Federal Reserve also accommodates past shocks in the multiplier, /.L Accommodation of borrowing raises an important question. Where does the money come from? Neither the “reluctance” view nor the portfolio view of the Federal Reserve answers the question. Once NBR is fixed, equations (7a) and (7b) determine money and the federal funds rate, given X, and the shocks. The money stock adjusts to every shock to the demand for money within the unit period. The financial system plays no role. The money stock appears mysteriously without involving any behavior or action by banks. Intermediation and the credit markets have no influence on money or interest rates. There is one key interest rate, ffr. Other rates, as in the Federal Reserve monthly model, depend mainly on ffr.20 To answer the question, the Federal Reserve must distinguish between money and credit and treat the banks’ demand for earning assets as associated with -- but not the mirror image of -- the banks’ supply of deposits. Once this is done, all shocks to money demand do not affect the stock of money, and all financial shocks are not shocks to the demand for money. See Brunner and Meltzer (1968). The separation of money and credit (or more serious attention to intermediation) would alter the Federal Reserve’s beliefs about the influence of monetary impulses and random shocks and about their ability to control money by controlling interest rates and unborrowed reserves. Further, explicit recognition of intermediation would permit identification of the distinct influences of money market and credit market disturbances on monetary control.

*%he San Francisco model breaks with this tradition, but the available version of the model does not identify separately one equation for earning assets and one for money. The inconsistency arising from nonidentification of the two equations is discussed by Anderson and Rasche (1982).

84

STRATEGY Two issues involved in the choice of a monetary strategy receive attention. One is an analysis of the choice between interest rates and money. The other is the choice between activist and nonactivist strategies. In this section, we discuss the papers by Enzler, Enzler and Johnson, and Davis. Then we provide evidence about their claims and more generally about the achievements of activist policymaking. Monetary or Interest Rate Strategy In a deterministic world, invariant under the choice of strategy, there is no basis for choosing between an interest rate strategy and a monetary strategy. l&zler and Johnson subject an IS-LM model to stochastic shocks. They augment the model by introducing a Phillips curve and a policy feedback relation. The only policy rule considered is an activist policy that is not unlike Federal Reserve policy. The rule requires the Federal Reserve to adjust the short-term interest rate in response to deviations of the money stock from its target and deviations of money growth from its target. Enzler and Johnson use the model to answer two questions:21 (1) Do “plausible values” of the parameters produce cycles when the Federal Reserve seeks to achieve monetary targets? (2) Can an alternative policy be formulated “which would eliminate or greatly attenuate the cycles” for sets of parameter values that produce cycles under monetary targeting of the type described? The answer to both questions is an obvious yes. We do not require lengthy analysis to know that it is possible to select parameter values, label them “plausible,” and show that particular feedback rules produce cycles for the selected values. It is not difficult to change the conclusion. By replacing the distributed lag of short-term real rates with the expected short-term real rate in the aggregate-demand function, one can obtain at least one complex root with modulus exceeding unity (under an interest rate control strategy). This result is obtained if the price level adjusts slowly. Further, most economists are not surprised to learn that one can find a policy rule that improves performance relative to the arbitrary policy which they choose as a bench mark. In the Enzler and Johnson paper, the alternative rule sets “final targets” for output growth and inflation. Enzler and Johnson conclude (p. 21): “The scope of the experiments performed was broad enough to give some reason to believe that these results would be qualitatively representative of a wide range of models... . The results targeting.

21The authors give the reason for the emphasis on the instability One of the Governors requested that the study be done.

8.5

induced

by money

stock

would almost certainly not be representative, however, of currently fashionable models in which expectations variables are specified to be identical with the model’s outcomes for those variables.” Many economists will resist the characterization of rational expectations as a current fashion without relevance to the issues addressed in the paper. The problem with an interest rate strategy, analyzed by Sargent and Wallace (1975) is independent of the model used by Sargent and Wallace. The problem can be avoided by imposing arbitrary assumptions about information or by denying that people learn, but at this stage, the problem should not be neglected or dismissed. Yet, none of the papers comparing strategies recognizes this fundamental, problem brought out forcefully in models using rational expectations. Enzler, in a separate paper, uses several models to investigate three issues by simulating responses. First, Enzler studies whether a less variable path for the policy instruments would have altered the outcomes achieved in 1980. Second, he compares different speeds of return to the target path. Third, he compares the response of variables of major interest to four types of disturbances, using both monetary control and interest rate control. Each of the shocks (to consumption, wages, prices, and demand deposits) is assumed to be permanent. The results are similar to those obtained in earlier experiments with similar models. The best strategy depends on the type of shock. Enzler (p. 31) suggests that a good policy rule “would take into account both the relative likelihood of the various kinds of shocks in the period before the nature of the shock could be determined, and the nature and magnitude of the shock once the determination could be made.” This statement begs the key question about choosing strategies. Where do we get the information on which to base inferences before the nature of the shock is known? Enzler’s conclusion lends no support to the case for policy activism or for any particular strategy. Enzler recognizes, but dismisses, the problem of structural invariance under alternative monetary rules. He questions the relevance of the “Lucas effect” on the grounds “that it is still a matter of great dispute.” This statement must puzzle anyone who believes that there is a more than modest accumulation of evidence showing that people learn and adapt.

Activist or Non-Activist Strategy Despite lapses with respect to rational expectations, the technical quality of several of the studies is high. The Davis paper is an exception. Davis addresses the issue of activist policy, a central issue for many years in pro-

86

fessional discussion of monetary policy. It is regrettable that the staff volume fails to contribute a serious analysis of this basic problem. Much of Davis’s paper is a discussion of what he calls “practical monetarism.” There are neither references to papers developing the analysis that Davis discusses nor references to Friedman, Schwartz, Fellner, Lucas, Brunner and Meltzer, the Shadow Open Market Committee, or other “practical” monetarists. There are, however, some creative interpretations of the case against policy activism. Davis never refers to the central argument against policy activism made in Friedman (1953) and recently presented in detail by Brunner (1980). These arguments are independent of any specific hypothesis about money - monetarist or nonmonetarist. Instead, Davis claims that monetarists make the following arguments against activism: (1) dominance of the monetary impulse in the determination of aggregate demand; (2) stability of the demand for money, or a vertical LM curve; (3) internal stability of the economic system; and (4) long and variable lags in the response to monetary policy. The length and variability of lags has been used as part of the rationale for a nonactivist strategy.22 The other arguments have been mentioned by “monetarists” in unrelated contexts, or are attributed to “monetarists,” but they have never been used as a main argument against policy activism. Some have not been used at all. For example, from our very first work on the demand for money, in the early sixties, we have denied that the demand for money is independent of market interest rates. Some combination of Davis’s arguments can be used to establish a sufficient condition for monetary control, but they cannot exclude an activist monetary policy. Nor do they offer a necessary condition for nonactivist policies. Davis’s discussion of the case for activist policymaking is, again, impressionistic and casual. He considers the “information extraction” approach used in Kareken, Muench and Wallace (1973) and later in B. Friedman (1976). These papers show that monetary targetting (or intermediate targetting) is inefficient in a specific technical sense. Their results depend on a specific assumption about information. Policymakers are assumed to know the deterministic and stochastic structure of the economy. Davis notes the difficulties in recognizing and correctly interpreting the shocks that occur, but he does not relate the interpretation problem to the general problem of choosing policies in an uncertain world subject to permanent and transitory shocks affecting suply and demand on many markets. 22Pischer and Cooper (1973) paper by Fischer and Cooper, however.

demonstrate

tbat thii argument

87

fails. Davis does not refer

to the

Davis argues that a nonactivist strategy prevents the Federal Reserve from responding to short-run problems of the financial markets. He fails to consider the possibility that “credit crunches” are a result of the sudden shifts in policy characteristic of activist policy; thus the occurrence of “credit crunches” is not an argument for activist policy. A main issue here is whether the Federal Reserve’s activism increases or reduces variability. Davis provides no evidence. His discussion also fails to distinguish between the role of a lender of last resort for the financial system and daily or weekly policy operations. expect the monetary authority to function as lender of last “Monetarists” resort; at least, we do. The problem of defining money and the role of financial innovation are, as usual, mentioned as reasons for activism or against nonactivist policy rules. The alternative of removing regulations that encourage innovation and distort the monetary aggregates is not mentioned. There are no quantitative statements about the magnitude of the’ problem or of the comparative magnitude of “noise” induced by past activism. There is no mention of the effect of inflation on financial innovation in a system with noninterest-bearing required reserves. Does the Federal Reserve Raise or Lower the Variability of GNP?23 Variability of monetary velocity‘ and shifts in the demand for money are common reasons for policy activism. Davis and Axilrod repeat the familiar argument that, in principle, the Federal Reserve can offset changes in the demand for money or velocity. The problem is that the Federal Reserve usually does not know whether the shifts are permanent or transitory until sometime after they occur. Also, the Federal Reserve cannot locate the source of the shock quickly. The case for activism requires the Federal Reserve to reduce variability. The argument against activism rests on the claim that the Federal Reserve does not have sufficient information to reduce variability. See Friedman (1953), Brunner (1980). Lack of information can prevent activist policy from offsetting shocks to the demand for money even if the Federal Reserve uses the best available model of velocity (or the demand for money).24 Consider the case in which velocity follows a random walk around a constant trend. Knowledge of the systematic process provides no basis for activist policy; neither does perfect knowledge of the stochastic process, It is obvious that, in this case, activist policies to offset changes in the demand for money or velocity may introduce

2%

e are indebted to Angelo Mascaro and Vicki Farrell for their mating the models from which these results are obtained. 24~n, ail shocks to the demand for money are shocks to velocity,

88

help in formulating but the converse

and estiis not true.

more variability into GNP than they remove. To measure the variability in GNP introduced used the formulas

Atn GNP=ARM*+A!Zn

Vy+p+

by monetary policy, we

v

where * indicates systematic components and /.L and v are the random components in AQn M and ARn Vi, respectively. The variance of A!?n GNP is 02(AQn GNP)=02(AQn M*)+02(AQn

VT)+2o(Mn

M*, A!& VT)

(10)

+ u2p + u2v + 20(/J, v) M is defined as the sum of currency and checkable deposits, and VI is the ratio GNP/M. To estimate the systematic and random components, we used ARIMA models for !Zn M and Iln VI. The systematic component is the optimal forecast of a univariate stochastic process, and the random component is the current innovation. Our procedure separates the systematic and random components and permits us to study, separately, the Federal Reserve’s performance with respect to the systematic components and with respect to the innovations. The models were estimated using nonseasonally adjusted data from the cycle peak in third quarter 1953 to the cycle peak in first quarter 1980 and for two subperiods. To control for seasonal variation, we estimated the ARIMA seasonal factor using first differences of corresponding quarters of adjacent years, Mt - M,-4, Mt+l - M,-3, etc. Then, an ARIMA model separated the seasonally adjusted data into systematic and random were relatively simple moving averages or second order. The systematic components Avy. We computed the variances and the

components. Most of the processes autoregressive processes of first or are the expected values, &I4* and covariances of the systematic com-

ponents and of the innovations. The first two columns of Table 1 compare computed following equation (lo), to the variance requiring constant monetary growth. The latter is the variance of Alln VI. The difference between the

the variance of APn GNP, of A!2n GNP under a rule measured provisionally, as two columns is a measure

of the variance of Alln GNP resulting from policy activism. Our procedure partitions the difference into a systematic and random component.

89

For the systematic component, denoted S in Table 1, the difference between activist policy and constant money growth is u2 APn M* + 2a(AKn M*, Ann VT).

This sum can be positive or negative. Constant money growth removes this contribution to the total systematic variance of GNP growth. The remaining systematic variance is measured (provisionally) by the variance u 2 Alln VT. The first line in Table 1 shows that, for the period 1953-80, the systematic component of the variance, 0.90, exceeds 02APn VT. The difference between the two, 0.20, is the (positive) contribution of activist policy to the variance of GNP growth. The second line of Table 1, denoted I, compares the random components. If u2/.l + u2v + 20(/J, v) < u2v ) activist policy has reduced random fluctuations. Line 2 shows that u2v is the larger of the two, so it is clear that the covariance of the random components is negative and larger than in absolute value ~2. Line 3 is the variance of GNP grtwth, the sum of the systematic and random components:Column (1) shows the measured variance of GNP growth; column (2) is a (provisional) measure of the variance of GNP growth that would have resulted from a policy of constant money growth. The difference between the two, .06, is smaller than the difference between the systematic components but remains positive. The computations show a consistent pattern. For the two subsamples, and for the period as a whole, Federal Reserve activism increased the variance of the systematic component relative to the variance that would be achieved with constant monetary growth. In fact, the systematic covariance, u(AQn M*, ARn VT), is positive for 1969-80 and, by a lesser amount, for the period 1953-80 as a whole. Activist monetary policy produced twice as much variance in the systematic component of Alln GNP during 1969-80 as would have occurred if money growth had been constant. The Federal Reserve misjudged the oil shocks and other events of this period. The error added to the risks borne by consumers and producers and lowered welfare. The difference between the contributions to u2AQn GNP from the innovations suggests that the covariance, u(~, v), is small but negative. This is consistent with the hypothesis that the innovations are statistically independent.

90

Table 1 Actual

Period

Variances

of

A

GNP and Hypothetical Variances Money Growth in Percent

(2)

(1) Actual

Constant

Ml’

M growth

Vl

(3) Actual B, m

VI

with Constant

(4) constant B growth

(5) Variance of !&I v.

1953-80 s I

0.90+ 1.02

0.70 1.16

l.OO++ 1.03

0.78 1.13

0.56 1.28

1.92

1.86

2.03

1.91

1.84

0.12 0.94

0.56 1.00

0.55++ l.Ol++

0.54 0.99

0.49 1.13

1.66

1.56

1.56

1.53

1.62

3.75+ 2.16

1.83 2.09

2.33++ 1.91

2.27 2.05

0.90 1.24

5.91

3.92

4.30

4.32

2.14

1953-69 s I

1969-80 S I

S = systematic component I = innovation + q positive covariance of &n

M* and &n

<

++ = positive covariance of &n B and hPn m All data computed from cyclical peak to peak. Quarterly according to NBER chronology.

91

peaks occur

in 1953-3,1969-4,1980-l

No particular significance should be read into the positive or negative differences for I. The same appears to be true for the combined results, obtained by summing the S and I results for the period as a whole and for the first subsample. The results for 1969-80 suggest, however, that activist monetary policy increased the variance of AQn GNP. The findings reported in the first two columns of Table 1 are subject to three criticisms. First, the Federal Reserve does not fully control M, so it cannot reduce the variance of AQn M or the covariance of the systematic components of AQn M and AQn Vl to zero. Some of the systematic variance is unavoidable. Second, the results are biased against constant monetary growth. The reason is that constant growth of money or the monetary base avoids large changes in the rate of inflation. Price levels would continue to fluctuate, but the range within which such fluctuations occur would be constrained. Third, the variance of AQn Vl would change if the Federal Reserve shifted to a nonactivist policy. We defer consideration of the third criticism until our response to the first criticism is complete. The Federal Reserve can control its own balance sheet and, therefore, can hold the quarterly growth of the monetary base to its path. Strict control of the base (B) requires several changes in institutional practices recommended by many economists, including several of the authors of papers in the staff report. There is little reason to doubt that if the practices are changed appropriately, the (quarterly) variance of AQn B can be reduced approximately to zero. The covariance of Qn B with other variables then falls toward zero also. Columns (3) and (4) report the test of activist policy using AQnM=AQnm+AQnB=AQnm*+AQnB*+e+/3 in place of AQn M. Now, m is the money multiplier and e and /3 are innovations in the ARIMA processes for m and B, respectively. Asterisks again denote the systematic components. There are now three variances and three covariances for each period or subperiod reported in column (3): a2(AQn GNP) = a2(AQn B*) + 02(AQn m*) + 02(AQn VT) + 2[o(AQn B*, AQn m*) + u(AQn VT, AQn B*) + .u(AQn m*, AQn V*)l + a20 + u2e + u2v 1 + 2[o(P, E) + u(P, VI + NE, VII

92

(11)

For the systematic component in column (4) we set a2(AQn B*) + 2[o(AQn B*, Ann WI*) + a(AQn B*, AQn VT)] = 0,

(124

and for the random components, we set u2p + 2[43,

E) + a(& v)] = 0.

The result for the systematic component during the full period, 195380, favors constant rate of base growth. The main conclusion to be drawn from columns (3) and (4) is that there is again no evidence to support the claim that activist policies reduce the variability of nominal GNP. There is no evidence in the comparisons that activist monetary policy has contributed to economic stability. Constant growth of the base can be set to achieve the growth of money that is consistent with the GNP growth path. As an alternative, base growth can be set on a path related to GNP growth directly. Column (5) shows the variance of ARn GNP computed for the alternative procedure. VO is the ratio of GNP/B. Comparison of column (5) with the other columns suggests that this procedure was superior to activist policy during. the period and performed at least as well as the other monetary rules, The absence of any positive covariance gives the procedure involving VO a clear advantage for the period 1969-80.25 A rule requiring constant growth of the base would have eliminated 3/4 of the systematic variance in ARn GNP for this period. A striking result of the comparison of activist and nonactivist policies is our finding that, in every two-way comparison of systematic components, constant money growth lowers the variance of A!Zn GNP. Results for the innovations are generally closer together, as expected from the nature of these random components. Taken together these results do not provide a strong case for activist policies. Further consideration suggests two reasons why the case for activist policies is likely to be weaker than Table 1 suggests. First, the ARIMA models overstate the amount of information about the various processes available when decisions are made. The models separate random and systematic components, including seasonal factors, with greater accuracy than is usually available. One reason is that the models have a larger sample on which to base information 25Witb a currency drain, a run on the banks, or in other financial crises, we believe the monetary authority should function as lender of last resort. We would expect the authority to abandon constant base growth during such crises.

93

about the changing nature of shocks, seasonal adjustments, and structure because they include information beyond the period in which the shocks occur. Second, some of the variance of velocity reported in Table 1 arises because variable monetary growth induces changes in market interest rate and expected inflation. Variation in these variables induces changes in the demand for money and in velocity. As previously noted, the variance of AQn VI would change if the Federal Reserve adopted a policy of constant money growth. To investigate the effect of variable money growth on the variability of velocity, we separate AQn VT (or AQn Vc) into three terms. As is well-known from any standard macroeconomic theory, velocity depends on real and monetary factors. The variance of the systematic component, u2(AQn VT), is, therefore, the sum of (1) the contribution of real shocks, (2) the contribution of AQn M*, and (3) the contribution of the covariance of monetary and real shocks. Constant money growth sets the second term to zero and makes real and monetary shocks independent,. so the third term is also zero. The total variance o2(AQn VT), and the systematic component of a2(AQn GNP), is equal to the contribution of real factors. An activist policy can lower u2 AQn V; relative to a rule requiring constant money growth if, and only if, the third term is sufficiently negative. 26 We have not presented information that permits an unambiguous conclusion to be drawn. However, some of the data reported in Table 1 are suggestive. We know from Table 1 that the (total) covariance between AQn VT and AQn M* is less than one-half of the systematic variance of the monetary aggregate. This was shown by comparing columns (1) and (2) or columns (3) and (4) or by the comparison of column (5) with columns (1) and (3). A claim that the covariance of real and monetary shocks is sufficiently negative to reduce u2 AQn Vi implies that the Federal Reserve offsets real effects on velocity to a greater extent than it offsets monetary effects that occur independently of monetary policy. This seems unlikely and has no obvious rationale. The monetary effects include shifts in the money multiplier, m, and shifts in the demand for nominal money balances. The Federal Reserve usually explains the variability of money growth as evidence of their efforts to offset shifts in the demand for money. Further, the available evidence does not show the required pattern. During the period in which the (real) 1973-4 oil shock has a large effect on the systematic variance of AQn GNP, the covariance of AQn M* 26 Column that the comparison of constant monetary

(4) introduces a more complicated problem. Equations (ll), (12a) and (12b) show involves additional terms. Nothing of substance is added by a discussion of the effect growth on the value of each of the systematic terms.

94

and ARn VT is positive. During this period, our computations

show the largest

gain from constant monetary growth or constant base growth. The Shifting Demand for Money Advocates of “flexibility” and discretionary activism often argue that the demand for money is “unstable” or “shifting.” Evidence to support this position usually depends on residuals from estimated demand equations for money. Most of the estimates are for single equations; thus they do not separate the effects of regulation on the stock of a particular monetary aggregate from changes in demand. Changes induced by regulation of interest rates and inflation are often treated as random shocks to the demand for money. This section presents results of a study of the time series structure of two measures of velocity, VO and Vl, during the three most recent decades. Equations were estimated as first differences of seasonally adjusted quarterly data of the logarithms of velocity. A first-order moving average of APn V was used for all estimates.27 All shocks to the demand for money are shocks to velocity, but the converse is not true. Table 2 shows the computed trends in velocity for the past thirty years and for each decade separately. For VO the trend is approximately 0.6% per quarter, and for T/l, 0.76% during the thirty-year period. There is no significant difference between the trend rates of changes of I/O estimated for the fifties and the seventies, and there is no evidence of a higher trend in recent years. If velocity is (approximately) a random walk, innovations in the growth of velocity are permanent shifts in level. There is no way to predict these shifts, and unless all changes are known to be either permanent or transitory, no one can expect to offset changes in V without increasing the variance of ARn GNP. The Vo process shows evidence of becoming a random walk in the sixties and remaining a random walk in the seventies. The coefficient, fi, describing the first-order process for VI is not significantly different from zero in the sixties, but the random walk in VI did not continue through the seventies according to these estimates. The standard deviations of innovations in Qn Vo and Qn VI computed for the seventies are lower than the corresponding estimates for the fifties. This finding is a clear contradiction to the many statements about shifts in velocity or the demand for money by Federal Reserve spokesmen and others. In fact, a puzzling feature of these data is the absence of a large change in trend or 27Estimates for Y2, using M2, were computed. These show similar :! I Yo and Vl estimates were generally not significant after lag 1.

95

patterns.

Serial correlations

Table 2 Tie

Series Analysis

of Velocity*

Base Velocity, Period

YO

Trend

P

%

1951-2 1981.3

to

0.0061 (5.87)

-0.212 (2.35)

0.010

1951-2 1961-1

to

0.0073 (3.25)

-0.501 (3.45)

0.012

1961-1 1971-1

to

0.0055 (5.30)

-0.085 (0.5 1)

0.007

1971-1 1981-3

to

0.0059 (4.91)

0.098 (0.63)

0.010

Monetary

Velocity,

Y,

1951-2 1981-3

to

0.0076 (7.11)

-0.206 (2.25)

0.010

1951-2 1961-1

to

0.0067 (2.64)

-0.526 (3.80)

0.013

1961.1 1971-l

to

0.0073 (6.47)

0.196 (1.11)

0.007

1971-l 1981-3

to

0.0087 (20.23)

0.635 (5.25)

0.009

* &n

Yt = Trend

t - statistics

+ Er - /%Zzql

in parentheses

96

variability under conditions of relatively high and variable inflation. We conclude that these data show no significant evidence of a change in the trend growth of velocity.28

CONCLUSION: PROCYCLICAL MONEY

GROWTH

The two volumes analyzing Federal Reserve operations are comparable to a conference volume. They contain several papers that are informative, some that will continue to be discussed, and some that do not succeed. Our long experience with conference volumes suggests to us that this is not an unusual mix. M’ore impressive is the fact that most of the papers were written within a short period by the staff of a single agency. A main purpose of the studies was to evaluate the first year of experience with the October 1979 change in operating procedure. One year is a short period; not much can be learned about the effects on GNP, prices or other broad aggregates, and not much is said about the quality of the new procedures relative to other (for example) past procedure>. Adding to the problem of evaluating the change in procedure after so short a period is the experience itself. Money growth was more variable than usual. The annualized percentage growth rates of money (Ml& in the four quarters beginning with fourth quarter of 1979 are, respectively, 4.7, 7.0, -3.0, and 14.6. The stock of money declined during the recession and rose during the expansion, influenced no doubt by the credit control program in the spring of the year and its elimination in the summer. Nominal GNP growth for the same four quarters is 8.8, 12.6, - 1.l, and 11.8. Money growth shows the familiar pro-cyclical pattern -- the same pattern we and others have discussed many times. The imposition and elimination of credit controls probably explains part of the pro-cyclicality of money and GNP, but the studies do not claim that the problems are specific to the year. And they are not. There is no evidence that pro-cyclicality has been eliminated. Continued pro-cyclicality is not easy to reconcile with the greater importance allegedly given to control of monetary aggregates. Neither is the fact that the Federal Reserve was no more successful in reaching its targets for M-1B in 1980 and 198 1 than in prior years. In both years, the Federal Reserve missed the announced MlB target by more than two percentage points.

28There is some evidence that the fiist+xder process affecting The coefficient 0 changes from negative to positive, as shown in Table 2.

97

VI

changed

in the seventies.

Several of the studies suggest that institutional changes would reduce errors and improve control of money. Elimination of seasonal adjustment, a return to contemporaneous reserve accounting, simplification of the reserve requirement structure, and improvements in borrowing arrangements are mentioned. There is a strong case for these changes as steps to improve monetary control. But the proposed changes, desirable as they are, are unlikely to eliminate pro-cyclicality or have much influence on the Federal Reserve’s ability to hit their announced annual targets. In the language of our paper, the proposed changes are tactical improvements, not a strategic change in procedures. A review of the tactical procedures shows that the federal funds rate continues to have an important role. Until October 1979, the Federal Reserve set targets for the funds rate and allowed reserves and money to adjust to demand. In October 1979, the Federal Reserve reversed the process. Targets for nonborrowed reserves are chosen to achieve targets for total reserves and money. The estimation of borrowed reserves and money depends on the funds rate, so the errors made in predicting interest rates, borrowing, and the demand for money continue to move money and reserves far from their target paths at times. The volumes do not include an analytic restatement of the roles of interest rates and reserves in the monetary control process. Our formalization of the links shows that errors in forecasting the demand for money and errors in the borrowing equation introduce random elements into the equations determining reserves and money. Statements about the cause of deviations by staff and policymakers, and parts of the work done by current and former staff members, are consistent with our analyses. These statements and papers attribute most deviations from monetary targets to shifts in the demand for money or shifts in bank borrowing from the Federal Reserve. If all the errors in the borrowing and the demand-for-money equations are transitory random shocks, the use of interest rate targets puts random fluctuations into the stock of money but has no lasting effect on money. An improbably long series of transitory shocks is required to induce the observed pattern of persistent, pro-cyclical changes in the demand for money. See Friedman (1959). To be consistent with observed changes in money and velocity, the sequence of serially correlated, transitory shocks must repeat with a frequency equal to the frequency of a standard business cycle. Our studies of velocity suggest that quarterly changes in velocity are well-described by a moving average of random shocks. Velocity has a stochastic trend, given by the permanent component, and a transitory random component. The best forecast of velocity, conditional on all available information, is that velocity follows an augmented random walk. Additional evidence, for a longer period, is Gould et al. (1978).

98

Permanent or persistent changes in velocity cannot be identified at the time they occur. Neither the Federal Reserve nor anyone else can forecast the interest rate that holds money growth on a stable path unless they can separate persistent and transitory shocks. An activist policy in a world of unidentifiable permanent and transitory disturbances commits us to a risky strategy. It is likely to introduce more variance than it removes. Computations of the systematic and total variance of GNP suggest that activist monetary policy does introduce more variance than it removes. The difference between a policy of constant monetary growth and an activist policy is, typically, not large, but further analysis suggests that the computed differences are underestimated. The main reason is that a policy of constant monetary growth, pursued on a sustained basis, removes one cause of changes in velocity - shifts in the demand for money in response to perceived shifts in the growth rate of money. The analysis of velocity helps to explain why Federal Reserve policy is a main cause of pro-cyclical money growth. The Federal Reserve sets targets or bands for interest rates. Unless the Federal Reserve correctly identifies permanent and transitory changes when they occur, the market interest rate they choose will, at times, be above or below the equilibrium rate consistent with full employment. Suppose the public finances a persistent increase in aggregate spending by supplying assets to banks and by reducing cash balances. The Federal Reserve prevents the induced rise in interest rates by increasing bank reserves. The stock of money rises, lowering the increase in measured velocity and increasing average cash balances. The initial increase in aggregate spending is followed by additional increases, some induced by the desired increase in spending, some by the unanticipated increase in 44. The Federal Reserve eventually offsets the bulge in money. Interest rates rise, but the Federal Reserve cannot know where to set the interest rate because it cannot identify the persistent and transitory components of the change in aggregate demand. A similar analysis applies to movements of money, spending, and interest rates during recessions. Our analysis does not imply that there is a single, unchanging constant rate of money growth. Persistent changes in the growth of productivity may change the growth of output and, therefore, change the rate of money growth consistent with a maintained average rate of inflation. Sustained changes in the rate of growth of output cannot be separated promptly from the many sustained and transitory changes to the level of output, so they do not provide a rationale for activist policy. The two volumes do not offer a clear, coherent statement of the case for activist policy, for interest rate targets, or for the strategy currently used to

99

conduct monetary policy. We know of no such evidence elsewhere that supports continued use of activist policies. Perhaps we have overlooked relevant evidence. We look forward to seeing the empirical basis of the case for activist policy set forth with as much care as is spent on the tactical issues in the volumes we have discussed here.

100

References Anderson, R.C. and Rasche, R.H. (1982) What Do Money Market Models Tell Us About How to Implement Monetary Policy? Washington: American Enterprise Institute (forthcoming). Board of Governors, Federal Reserve System (1981) New Monetary Control Procedures, Washington, D.C. Brainerd, W.C. (1964)

Brunner, K. (1980)

Volumes

I and

II.

Financial Intermediaries and the Theory of Monetary Control, Yale Economic Essays, 4: 43 l-82.

The Control of Monetary Aggregates, Controlling Monetary Aggregates 111, Proceedings of a Conference, Federal Reserve Bank of Boston. l-65.

(1980)

, Cukierman, A., and Meltzer, A.H. Stagflation, Persistent Unemployment and the Permanence of Economic Shocks, Journal of Monetary Economics, 6: 467-92.

(1964a)

and Meltzer, A.H. General Features of the Federal Reserve’s Approach to Policy. Washington: House Committee on Banking and Currency.

(1964b)

(1964~)

and The Federal Reserve’s Attachment to the Free Reserve Concept. Washington: House Committee on Banking and Currency. and An Alternative Approach to the Monetary Mechanism. Washington: House Committee on Banking and Currency.

Brunner, K. and Meltzer, A.H. Liquidity Traps for Money, Bank Credit and Interest Rates, (1968) Journal of Political Economy, 76: l-37.

101

Cukierman, A. (1981)

de Leeuw, F. (1965)

Dewald, W.G. (1963)

(1966)

Interest Rates during the Cycle, Inventories and Monetary Policy - a Theoretical Analysis. Carnegie-Rochester Conference Series on Public Policy, 15: eds. K. Brunner and A.H. Meltzer. Amsterdam : North Holland.

A Model of Financial Behavior, ed. J. Duesenberry. The Brookings Quarterly Econometric Model of the United States. Chicago: Rand McNally.

Free Reserves, Total Reserves, and Monetary Control. Journal of Political Economy, 71: 141-53.

Money Supply versus Interest Rates, National Banking Review, 4: 509-22.

Fischer, S. and Cooper, P.J. Stabilization (1973) 81: 847-77. Friedman, B. (1977)

Friedman, M. (1953)

(1959)

(1963)

Policy and Lags, Journal of Political Economy,

The Inefficiency of Short-Run Monetary Targets for Monetary Policy, Brookings Papers on Economic Activity, 2: 293-335.

The Effects of Full Employment Policy on Economic Stability: A Formal Analysis, Essays in Positive Economics, 117-32. Chicago: University of Chicago Press.

The Demand for Money: Some Theoretical and Empirical Results, Journal of Political Economy, 67: 327-5 1. and Schwartz, A.J. A Monet&y History of the United States, 1867-1960. Princeton: Princeton University Press for the NBER.

102

Gould, J.P., Miller, M.H., Nelson, C.R., and Upton, C.W. (1978) The Stochastic Properties of Velocity and the Quantity Theory of Money, Journal of Monetary Economics, 4: 229-48. Johannes, J. and Rasche, R.H. (1979) Predicting the Money Multiplier, nomics, 5: 301-26.

Journal of Monetary

Eco-

Kareken, J., Muench, T., and Wallace, N. (1973) Optimal Open Market Strategy: The Use of Information Variables, American Economic Review, 63: 156-72. Lindsey, D. (198 1)

Non-borrowed Reserve Targetting and Monetary Control, Board of Governors, The Federal Reserve. (unpublished).

Lombra, R. and Moran, M. (1980) Policy Advice and Policymaking at the Federal Reserve, Carnegie-Rochester Conference Series on Public Policy, 13 : eds. K. Brunner and A.H. Meltzer. Amsterdam: North Holland. Meltzer, A.H. (1980)

(1981)

Central Banking: Some First Principles, Annual Monetary Review, 2: London: Centre for Banking and International Finance, City University.

Consequences of the Federal Reserve’s Re-attachment to Free Reserves, (unpublished). Carnegie-Mellon University.

Sargent, T.R. and Wallace, N. (1975) Rational Expectations, the Optimal Monetary Instrument and the Optimal Money Supply Rule, Journal of Political Economy, 83: 241-55. Volcker, P.A. (1980)

Recent Developments in Monetary Bulletin (Dec.), 943-52.

103

Policy, Federal Reserve