53

Daniel L. Thornton

Daniel L. Thornton is a research officer at the Federal ReserveBank of St. Louis. Rosemarie V. Mueller provided research assis-tance.

The Effect of Monetary Policy onShort-Term Interest Rates

HE “liquidity effect” plays a central role inKeynesian theory of the transmission of monetarypolicy. It is based on the notion that the demandfor money is negatively related to the nominalinterest rate.1 Other things the same, an exogenousincrease in the money stock depresses nominaland real interest rates, stimulating aggregatedemand.

Even though theorists acquiesce to the liquidityeffect as a theoietical proposition, it is often chal-lenged on efficacy grounds. It is argued that

changes in the money stock do not leave all otherthings unchanged. Monetarists, such as Friedman(1968) assert that the liquidity effect is, at best, onlytemporary; the ultimate effect of more rapidmoney growth is higher inflation (or, mom-c impor-tantly, expectations of higher- inflation) and, conse-quently, higher nominal interest rates. New classi-cal economists argue that the real interest rate isdetermined by basic tastes and technology con-siderations, which are slow to change.’ If increases

in the money supply primarily affect the market’sexpectations of inflation, nominal interest rateswill rise immediately.

Estimates of money demand equations, espe-cially short-run equations, indicate that moneydemand is very interest inelastic, suggesting that

there is a strong liquidity effect.3 Most other em-pirical work, however, has estimated the total ef-fect of changes in monetary policy on interestrates. A wide i-ange of methodologies have pro-duced diverse and sometimes conflicting results.This article is an attempt to consolidate the evi-dence on the responsiveness of interest t-ates tomonetary changes. Various methods for estimatingthe relationship between interest rates and mone-tary impulses are reviewed and then applied to acommon data set. Also, the analysis implicitlyincorporates the possibility that the money stock

is endogenous in the sense that the money multi-plier depends on the interest i-ate.4

‘Until fairly recently, most forms of money were non-interest-bearing. Consequently, the opportunity cost of holding moneywas represented by the nominal interest rate. A largeportion ofMl now is held in the form of interest-bearing NOW accounts.The opportunity cost of this component ot Ml is the spreadbetween market rates and the rate paid on these deposits.

‘Recently, Niehans (1987) has argued convincingly that thedescription of the rational expectations school as “new classi-cal economics” is a misnomer. He argues that its emphasis oncontinuous market-clearing constitutes afundamental breakfrom both classical and neoclassical economics.

‘Manyeconomists, for example Carr and Darby (1981), believethe liquidity effect implied by these equations to be implausiblylarge.

4The interest sensitivity ot the multiplier is shown in models ofthe money supply process. For example in Thornton (1982),the behavioral equations are assumed to be linear; thus, al-though the multipliers are not functions of the interest rate perse, they arefunctions of the interest elasticities of these behav-ioral equations.

54

THE LIQUIDiTY EFFECT

The liquidity effect is defined as the interestresponsiveness of the demand for money in a sim-ple model of liquidity preference where the mone

stock is assumed to be controlled directly andexogenously by the monetary authority.’ For’ ex-ample, consider the following specification of thedemand for- nominal money

(1) M” = Lii, Py), L <0, L,,, L,>0,

where M, i, y and P denote the nominal money

stock, the nominal interest rate, real income andthe price level, respectively. Ifthe money stock istaken as exogenous, M’ = M, the market equilib-rium condition is

(2) M = Lii, Py)

Hence, the liquidity effect is defined as

(3) di = (i/L,)dM.

While the theoretical relevance of the liquidityeffect is acknowledged, analysts genem-ally amguethat it may be partially or totally offset quickly byother effects, both direct and indirect, of moneystock changes. To see this, assume that the pricelevel is positively’ related to the money stock andreal output is negatively related to the interest

rate. That is,

p = P(M), P > (I

and

y = yti), y’ <0.

Substituting the above expressions into equation2, the effect of an exogenous change in the moneystock on interest rates is

(4) di = (i — L~P’yldM/(L,+L,Py’I.

This measure m-efiects not only the interest sensi-

tivity of the demand for money, I,, hut the directeffect of money stock changes on the price level,L,,P’y, and the indirect effect of interest iates on

income, L,Py’ -

The effect of an exogenous change in money on

interest r-ates given by equation 4 is strictly smallerthan the liquidity effect of equation 3 because ofthe income and price level effects. According to

the Keynesian transmission mechanism, the lower’nominal and, at this point real interest rate, stimu-lates aggregate demand and, hence, real income.The rise in m-eal income increases the demand formoney, causing interest i-ates to mise; this mitigatesthe initial liquidity effect. Equation 4 also incorpo-rates the direct price level or the ‘‘Keynes effect’’ -

An increase in the nominal money stock causesthe price level to rise, which in turn causes thereal money stock to decline, resulting in an in—ct-ease in interest rates.°

If money stock changes affect output or pricessufficiently m-apidly, then the income and pm-ice

level effects will oftiet, at least in pam-t, the declinein interest t-ates associated with the liquidity ef-fect. Moreover-, it may be difficult to find a statisti-

cally significant negative relationship betweenchanges in the money stock and changes in theintem’est rate if the data am-c averaged over a long

period.’ Indeed, if financial market pamticipantsanticipate the rise in income or the price level,these effects will he meflected in market interestm’ates immediately; thus the observed change ininterest rates associated with a money stockchange might be small even over short timepenods.

The “Fisher Effect”

In addition to the income and price level effectsincorporated in equation 4, them-c is also the possi-bility of the “Fisher effect.” Fisher (1930) am-guedthat, in the absence of differences in holding costs,the ieal, risk-adjusted return on assets should bethe same regardless of the units in which the as-sets are expressed. Consequently, the i-eturn on

physical assets should be the same as the returnon credit contracts denominated in fixed units ofnominal money. This implies that the interest rate

on dollar-denominated contracts will reflect the

‘Because the liquidity effect usually is discussed in modelswhere the money stock is assumed to be controlled by themonetary authority, it has become synonymous with the inter-est responsiveness of money demand. In amodel where themoney stock is endogenous, it may be more appropriate tothink of the liquidity effect in termsof the impact of an exoge-nous change in monetary policy on interest rates. This wouldreflect not only the slope of the money demand function, butthe slope of the money supply function as well.

‘For notational convenience, equation 1 is written without im-posing the usual assumption that LC) is linear homogenousofdegree one in P.

‘This maybeone reason why Peek (1982) and Wilcox (19S3a),Makin (1983) and Hoffman and Schlagenhaut (1985) obtaineddifferent results using similar dataand methodologies. All usedthe biannual Livingston survey dataon inflation expectations;however. Makin, Hoffman and Schlagenhauf interpolated thedata and estimated a quarterly model, while Peek and Wilcoxused biannual data.

55

market’s expectation of inflation over the duration

of the contract. Hence, if an increase in moneygrowth pmoduces expectations of more rapid in-flation, the nominal interest rate will rise.’ Theexistence of a contemporaneous price expectationeffect mitigates and possibly eliminates the liquid-ity effect on the nominal interest m-atesY

The Effect of an Endogenoras MoneySupply

Until now, the money supply has been assumedto be controlled exogenously by the Federal Re-

serve. In the modern financial system however,the total money stock is determined not only bythe policy actions of the Federal Reserve, but by

the portfolio decisions of depository institutionsand the public. That is, the money supply is com-posed of both “inside” and “outside” money. Gen-

erally, there is no sense in which one can measurethe effect of a change in the stock of endogenous,inside money on interest rates.” Instead, the effect

of monetary changes on the interest rate is mea-sured in terms of changes in outside money.

For’ example, assume that the money supply isendogenous in that the usual money multiplier is

a function of the interest rate. That is, let themoney supply be expressed as

(5) M’ = mliiH, m’ > 0,

where H denotes the stock of “high-powered,”outside money and mU) denotes the usual moneymultiplier. Setting IS) equal to Ii) results in theequilibrium condition

16) mtiiH = Lii, Plmii)H)yti) I.

Consequently, the effect of an exogenous change

in the stock of high-powered money on the inter-est rate is given by

(7) di =

(1— L,yP’)mdH/(L, + L,It + (L~yP’— 1)m’H).

The responsiveness of interest rates measured by17) is strictly smaller than that given by (4) for anidentical exogenous change in the money supply,that is, mdli = dM.

The Role of Monetarv PolicyObjectives

There is an exception where it would be appro-priate to measure the effect of monetary changeson interest rates in tei-ms of the total money stockdespite the presence of inside money. This occurswhen the monetamy authority is targeting the totalmoney supply and when it is forecasting andquickly offsetting the effect of other factors on thesupply of money.” For example, suppose that theFederal Reserve is targeting the total money sup-ply but controls only H directly. Ifm were to rise,

say due to a decrease in the public’s desire to holdcurrency relative to checkable deposits, the Fedwould attempt to offset the effect of the rise in themoney stock by reducing H. If the Fed anticipated

the rise in m and changed H by the appropriateamount immediately, there would be no change in

the money supply or interest rates associated withthe change in H. Estimates of the responsivenessof interest rates to changes in H would be biaseddownward, IL on the other hand, the Fed does notrespond instantaneously, interest rates would benegatively associated with changes in H. tn con-trast, assume that there is an exogenous increasein the demand for money. If the Fed responds

‘Thereader should note that there is a somewhat subtle differ-ence between equating the liquidity effect to shiffs in the stockof money and shifts in the growth rate of money. The problemhere is that the Fisher effect, which relates the level of nominalinterest rates to the rate of inflation, is fundamentallydynamic.The bridge that links these concepts can be found in the mone-tary growth models where, in long-run equilibrium, both themonetary growth rate and the nominal interest rate are con-stant. An exogenous increase in the growth rate of moneyproduces a liquidity effect and potentially a Fisher effect. Thisdifference is also rellected in empirical work. For example,compare the approach of Gibson (1970b) with that of Caganand Gandolti (1969).

‘Theoutcome depends on a number of factors, including thehomogeneity of the demand for real money with respect to theprice level. If there is no money illusion, thenominal interestrate must rise point for point with the expected rate of inflation.Consequently, it the inflation consequences of an increase inthe growth rate of the money stockare fully anticipated, thenominal rate must risewith the acceleration in money growth.

“See Patinkin (1965), pp. 297—301,tor a good discussion of thispoint. Of course, this does not apply to exogenous shifts in thestock of inside money, such as agold discovery under agoldstandard.

“SeeThornton (1984) for a discussion of this point in termsofthe issue of debt monetization. Also, see Mishkin (1982) for adiscussion of the effects ofthis form of money stock endoge-neity or estimates of the market’s response to changes in themoney stock.

Also, Mishkin (1981) and Robinson (1988) use M2 to measurethe responsivenessof interest rates to changes in the moneysupply. This is odd since changes in M2 are much more likelyto be related to factors other than policy changes.

56

instantly to offset the effect of this increase on themorley stock, inter-est m’ates will rise while the

money r’emains unchanged and the stock of high-

powered money is reduced. If the Fed does notrespond instantaneously, both interest r’ates and

the money stock will initially rise, then interestmates will continue to rise as the money stock falls.The point hem’e is that whether the total moneystock or’ the stock of high-powered money shouldhe used depends on whether the Fed is trying tocontt-ol the money stock and on how r-apidly it isresponding to other- factors that influence money.

This observation has implications for empim-icalwork. Lf the Fed is attempting to control the totalmoney stock and if the Fed moves reasonablyquickly to offset the effect of other factors, measur-ing the responsiveness of interest m-ates in tetms ofthe total money supply would be appropr’iate even

if day-to-day or week-to-week shocks were notoffset instantaneously.

To deter-mine whether the estimated respon-siveness of interest m-ates is sensitive to the mone-tary vamiahle used, alter-native measures of the

monetary impulse ar-c used. This is necessarybecause the Fed often relies on multiple objectivesand is not explicit about them.2 Of course, if m’ is

small, the choice of a monetary variable will herelatively unimportant.

Polic Belated Endogeneily

The endogeneity of the money stock discussedabove is based upon the economic response ofdepository institutions and the public to changesin nominal inter-est i-ates. Another- monetarv—policvrelated view holds that the money supply is en—dogenous whenever the Fed is using shor-t-ter-minterest m-ates as an intermediate policy target. In

this instance, the Fed merely adjusts the moneystock to shifts in the demand for or the supply ofmoney over which it has no control. In the case ofexogenous shifts in the money supply function,the Fed neutralizes the effect of such shifts onnominal interest through appropriate open mar-ket operations.” As a result, both the nominalmoney stock and the interest rate are unchanged.In the case of shifts in the demand for- money, theFed uses open market oper-ations to accommodatechanges in the demand for money. The interestrate m-emains unchanged, hut the money stockchanges.

This type of endogeneity cr-eates severe prob-lems for isolating the t-esponsiveness of interest

rates to monetary changes because only the mat--ket equilibi-ium values of the intem-est rate are ob-served. Since the interest rate is unchanged, de-

spite changes in the money stock, the responsive-ness of interest r’ates to changes in the moneystock appear-s to he nil.’~Ifthe Fed offsets only

part of a demand shift, however, money stock andinterest rate changes will be positively correlated.

Ifonly par-t of the exogenous supply shifts areoffset, money and intei-est rates will be negatively

correlated, Consequently, statistical analysis mayshow a positive, negative or no statistically signifi-cant relationship between interest rates andmoney gr-owth, despite the fact that it is preciseh’

because of the liquidity effect that compensatoryopen market operations are under-taken.

If the Fed reacts instantaneously to these

shocks, evidence of the effect of changes in themoney stock on interest rates can he obtained¼Jthpr’ecise knowledge of the Fed’s interest i-ate

tamget. Llnfortunately, such information is gener-ally unavailable.” Alter-natively, a time interval

short enough to isolate the response of the market

“For example during most of the 1 960s and the early I970s, thepolicy directives of the Federal Open MarketCommittee to theTrading Desk were stated in terms such as “maintain theexisting degree of credit restraint.” Even when the Fed wastargeting the monetaryaggregates in the late 1970s and early1980s, the policy directives often were stated in terms ot multi-ple monetary aggregates and in loose terms, such as “runsomewhat above the upper limit ot the target range,” More-over, the money growth obiectives frequently were conditionalon movements of other variables such as the federal fundsrate.

“The Fed’s reaction to offset asupply-side shift is referred to as“defensive open market operations.” Stabilizing the normalinterest rate will be effective only if the change in the moneystockdoes not give rise to inflationaryor deflationary expecta-tions. Proponents of this view would argue this will not happenbecause the Fed is merely accommodating shifts in the de-mand for money.

“In termsof a more formal model, let H’ be thestock of high-poweredmoney required to hit some target interest rate 1’, i.e.,V = L(i’,Py)/m(i). From this, dH/dPy = LPy/m(i). The changein the equilibrium interest rate associated with ashift in thedemand for money is given by di/dPy — [LPy/(L, — m’H) I+ [m(i)/(L, .~ m’H)l(dH/dPy). Substituting in for dH/dPy, yieldsdi/dPy = 0,

“At times, the Fed’s announced ranges for the federal fundsratewere fairly narrow, It is difficult to use these ranges to modelthis relationship, however, because the relationship betweenthefederal funds mate and theT-bill rate, which is usually usedto estimate the responsiveness of interest rates to monetarychanges, is itself not very stable.

57

to the Fed’s actions could be used. In the absenceof such detailed information or such a rich dataset, it is important to measure the effect of mone-tary changes on interest rates during periods in

which the Fed was attempting to exert greatercontrol over the money supply.”

A REVIEW OF METHODOLOGIES

One method of estimating the responsiveness ofinterest rates to changes in the money stock, used

by Cagan and Gandolfi (1969) and more recentlyby Melvin (1983) and Bi-own and Santoni (1983), isto regress the change in the nominal interest rate

lAi) on a distributed lag of unanticipated changesin the nominal money stock, SM”. That is, the

equation

K18)Ai,”a,+ I~3iAM~,+e~

i=0

is estimated. The random error, e, is assumed tobe identically and independently distributed witha mean of zero and a constant variance, a’. that is,e is iid(Q, a2). This equation is estimated with ordi-nary least squares IOLS).

A second approach used by Peek (1982), Wilcox

l1983a), Mehra (19853, Hoffman and Schiagenhauf1985) and Peek and Wilcox (1987) employs an IS-

LM, aggregate demand/aggregate supply model.”

In this modeL commodity demand is a function ofthe real interest tate and money demand is a func-tion of the nominal interest rate. While specificmodels differ, the following specification encom-passes the essential features. The IS curve is given

by

(9)y;’= a, — a,r; + a,Z, + v,,

and the LM curve by

(10) (M,—P,) = b,, + b,y~— b,i, + h,X, + v,.

[Unless otherwise stated, all variables are in loga-rithms.I y~denotesthe deviation of real GNP fromits “natural i-ate” (or full employment level), and P

and m denote the price level and real interest r’ate,respectively. Z, and X, ar-c vectors of vamiables that

influence the demand for commodities andmoney, respectively, and v,, and v,, are stochasticdisturbances such that v is iid(0, a~l,v,, is iidlo, o’/,(and Ely,, v,,) = 0 for all t. The model is closed by

the Phillips curve

Ill) P = P~+ cy~,

whete the superscr-ipt “e” denotes the expectationbased on information known before period t.Equations 9, 10 and 11 are solved for the r-eal inter-est rate. The result is substituted into the Fishetequation,

(121 i, = r, + ir~,

where ‘mm denotes the rate of change in the pricelevel, to yield a quasi-reduced form equation forthe nominal interest rate

(131 i, = A, + A,Z,a, + A,X,b, — A,(M, — P1

+ A,ir~’+ u,.

The responsiveness of the interest mate to realmoney stock changes, A, = [(c+b,)a + bJ’ >0,captures not only the “liquidity effect” (hJ, hut

also the net effect of all other factor’s that influencethe equilibrium interest rate.

While equations 13 and 8 appear- quite different,they are both i-educed-form equations. The funda-niental differences are that equation 13 is stated inlevel rather- than first-difference fotm and that itexplicitly includes factor-s. in addition to themoney stock, that could affect nominal interestrates. The absence of these factors from equation 8could be justified by arguing that it is a final-formequation, not simply a reduced-form equation. Onthe other’ hand, estimates of the mesponse of inter-est rates based on equation 8 could be biased ifvariation in other factors that affect intei-est ratesis not controlled for.”

Another difference is that equation 8 incorpo-rates a distributed lag of unanticipated money,while equation 13 uses only the contemporaneous

“It should be noted that Mishkin’s (1981, 1982) approach ofusing unanticipated money does not circumvent this problem.In this instance, unexpected changes in the money stock dueto demand and supply shocks are different, so that the coeffi-cient on unexpected money wilt be different depending onwhether the shock emanates from the demand or supply side.Moreover, the effect of an unexpected change in the moneysupply will be different from the effect of ashock to the moneysupply.

“Also, because equation 13 is a quasi-reduced form, the vari-ables Z,, IC, P~,M, or ,mf may be correlated with theerror term.Consequently, OLS estimates of these equations may beinconsistent. Of course, the same would be trueof equation 8 ifthe money stock is endogenous. This observation is the basisfor Mehra’s (1985) work.

“Actually, this approach was used earlier by Sargent (1969,1972).

58

level of actual money. The structure of equations9—fl can be modified, however, to replace themonetary variable by its unexpected component; adistributed lag of unanticipated money also canbe included by appealing to “price-stickiness” orBlinder and Fisher’s (19811 inventory adjustment.”

A third methodolo~’has roots in the rational

expectations/efficient market literature.” Mishkin(1981, 1982) and, more r-ecently, Hardouvelis (19861and Robinson (1988) estimate the equation

(14) i, — i~= a,, + a,l, + aJM, — M~)

+ a,(y — y~+ a~n~— i~ +;.

I denotes the set of information that market par-

ticipants have available to them at the beginning ofthe period, while ; denotes the error ter-m. Mish-kin characterizes equation 14 as the “m-ational ex-pectations analog of the typical money demand

relationship found in the literature.”

Mishkin derives equation 14 by using the ef-ficient market/rational expectations model to ar-gue that

i, — i~= 3W, — W~jj3+ w~,

Furthermore, equations 8, 13 and 14 are alter-na-tive representations for the nominal interest rate.Thus, they can be compared directly using stand-ard nested and/or nonnested test procedures.

EMPIRICAL ESTIMATES OF THELIQUIDIYV EFFECT

The empirical estimates presented hem’e coverthe period from 1958.08 to 1987.06. Prior studieshave generally used quarterly data when estimat-ing equations 13 and 14 and monthly data whenestimating equation 8. This study uses monthlyobservations for all specifications. The month pe-riod is short enough that the liquidity effect is lesslikely to be weakened by subsequent income, pricelevel or- inflation-expectations effects. On the otherhand, many of the variables that might reasonablyenter equations like 13 are unavailable on amonthly basis, so that the estimates are subject toa potential omitted-variables bias.

The variables used are

y = the real value of the industrial production

index,

TBR = the three-month Treasury bill rate,

where ~N, is a vector of variables that reflect the‘information relevant to the determination of

short-term interest rates” and w, denotes the errorterm.” He then solves a monetary equilibriumcondition for the interest rate in terms of all theother variables that enter-the money demandfunction, that is, variables which appear- as argu-ments in equation 1. He includes these variables inW,, arguing that they are part of the relevant infor-mation set. Of course, any right-hand-side vai-iablein equation 13 could be considered an element ofW, simply by broadening the theoretical frame-work, Consequently, equation 14 differs from theother specifications pmimarily in its explicit andcomplete reliance on the efficient markets/rationalexpectations paradigm.

P = the CPI,

M = the Ml definition of the money stock,

MB = the Federal Reserve Bank of St. Louis ad-justed monetary base,

= the Federal Reserve Bank of St. Louis ad-justed nonborrowed reserves.

Two measures of unanticipated changes in themoney supply are used here. The fir-st is thechange in the growth rate of money. Cagan andGandolfl use changes in the growth i-ate of moneyto proxy such changes, arguing that the mnarketshould respond only to unanticipated changes inthe money stock,” Today, the unanticipated change

“For example, see Makin (1983) and Hoffman and Schla-genhauf (1985).

“Dwyer (1981) has an alternative rational expectations frame-work where, because the same factors affect both the ex-pected inflation rate and thereal interest rate, they give rise toa set of cross-equation restrictions that can be tested,

“Mishkin (1982), p.66.

“Mishkin (1982), p. 64.

“Cagan and Gandolfi (1969) p. 279, state “It is hard to deter-mine to what extent monetarychanges at anyparticular timeare anticipated, but presumably asteady growth rate willsooneror later come to be reflected in a corresponding rise in

prices (allowing for the growth rate of real income). Conse-quently, changes in the monetary growth rate will tend to pro-duce, every time they occur, a response in interest rates.Gibson (1970a) uses a similar equation based on an analo-gous argument; however, Gibson (1970b) regresses firstdifferencesof the interest rate on first differences of the moneystock,

and

NBR

59

in the growth rate of money typically would beobtained by subti-acting expected money growth,estimated using some time-series method, fi-omactual money growth. Nevertheless, because Ca-gan and Gandolfi’s pm’ocedure has been utilized byall who have estimated equation 8, their measure

of unanticipated money is used to see if theresults are sensitive to the form of the unantici-pated monetajy var able.

Additionally, unanticipated money is measuredby )AM-AM°t,where AM’ is a time-series represen-tation of past AM. In this instance, the expectedvalues of M, y and P are obtained by regressingeach on a six-month distributed lag of itself andthe other variables, including changes in the ‘Urea-sury bill rate.”

This study uses three monetary policy variables:

Ml, the adjusted monetary base (MB), and nonbor-rowed reserves (NBR). The monetary base is used

often as a measure of exogenous monetary policy.NBR is used because some would argue that it is abetter- measure of the exogenous monetary im-pulse than MB because depositomy institutions’borrowings fr-om the Federal Reserve are related to

the interest rate. Also, the Fed used a NBR-oper-ating procedure to control the money stockfl-our October 1979 to October 1982, Since the Fedwas primarily targeting Ml growth during thisperiod, however. unanticipated Ml growth may hea better measure of the exogenous monetary im-pulse during this period.

Alternative nieasures of the monetary impulseare used to see whether’ estimates of the respon-siveness of interest rates to monetary impulses aredependent on the variable used.

Initially, the equation6

15) ATBR, = a, + ~ a,zXTBB1 , + 13MV~i=1

+ ,iPVr + 8yV~’+ E,

is estimated, The unanticipated monetary variable,MV’, is alter-nately proxied by AM1, AMB, ANBR,(AM1-AMP), IAMB—AMW) and (ANBR—ANBR’),”

The unanticipated price (WI and income (yV’)variables are alternatively measured by iSP and Ayor lAp — iSP’) and (iSy— Sy”!” This specification,and other-s which follow, include a finite distrib-

uted lag of the dependent variable to capture anyeffect of past information,”

OLS estimates of equation 15 for the period1959,08—1987.06 and two subperiods, 1959.08—1973.09 and 1973,10—1987,06, are presented intables 1—3, The split was made at 1973 .09 because

II) it marks the well-known break in the deniandfor money, (2) it roughly coincides with the demise

of the Bretton Woods agreement and (3) it alsoroughly coincides with the beginning of an era inwhich the Federal Reserve claimed to pay increas-ing attention to the growth rate of the monetaryaggregates .“ The equation is estimated with andwithout PV” andyV” to determine how sensitivethe r-esults are to these variables.

The results indicate considerable variability in

the statistical significance of the effect of the mon-etary variables on interest rates, both across timeand across monetary variables. During the entireperiod, ther-e is a small but statistically significantnegative effect for three of the unanticipated mon-etary variables. The largest statistically significantnegative effect is obtained when AMI is used, butthere is a statistically significant negative response

of interest rates when the unanticipated growth ofnonborrowed reserves is used, whether it is mea-sured by ANBR or (ANBR — ANBR’9.

The results in tables 2 and 3 indicate that the

responsiveness of interest rates to monetary im-pulses is sensitive to the sample period. Whenpre-1974 data are used (table 2) the effect is statis-tically significant only when the unanticipatedchange in the growth mate of nonborrowed re-

“This is similar to the multivariate time-series approachof Mish-kin (1981) except that a distributed lag of the ~TBR is includedin all regressions. It is important to include all relevant variablesthat affect interest rates. Wickens (1982) has argued that ifthey are not included, the expectations cannot be efficient,

Also, there wassome experimentation with alternative laglengths. The lags used here appeared to work well and pro-duced white noise residuals.

“When (AM — AM’) is used, AM denotes the annualizedfirstdifference of the log of the variable. AM, however, is the firstdifference of the annualized growth rate of the variable. Thesame is true for all other variables.

monetaryvariable, then AP and A~areused as the corre-sponding unanticipated price and income variables.

“The coefficients on the lagged dependent variable are notreported. In nearly every instance, they were jointly significantat the 5 percent level.

“Hafer and Hem (1982)date the break in money demand at1973.04, while Lin and Oh (1984) date it at 1972,02. TheUnited States formally broke from the Bretton Woods accord inlate 1971.

The Federal Reserve Open Market Committee stated adesireto place increasedemphasis on the growth of certain monetaryaggregates at its January 15,1970 meeting; Congress passedResolution 133 requiring the Board of Governors to set long-run ranges for the aggregates on March 24, 1975,

“The unanticipated monetary, price and income variables arematched in the regressions. That is, it AM1 is used as the

60

serves is measured by (ANBR — ANBR’) and whenPV’ and yV” are omitted. Even in this case, how-ever, the strength of the effect is small,

In contrast, there is a statistically significant

negative effect during the latter’ period (table 3)when AM1 or- NBR, in either- form, is the monetaryvariable. These results are interesting because theysuggest that the response of interest rates isstronger during the latter period, when the Fedclaims to have paid more attention to monetaryaggr-egates arid when Melvin (1983) reports thatthe effect vanishes. Finally, the coefficient for un-

anticipated base growth measured by (AMB —

AMBI, is significantly positive during this period,

Both quantitatively and qualitatively, the resultsare similar whethem’ the unanticipated price orincomne variables are included. Accounting for thepossible effect of unanticipated inflation or in-come growth does not appear to be important inmeasuring the effect of trnanticipated monetary

growth on interest rates,” The effects of unantici-pated inflation and income growth are highly sig-nificant for the entire period, but they are muchless so during the individual suhpem-iods.’°

“This result is not too surprising in the case where the unantici-pated variables aremeasured by the difference between actualand expected. It is usually assumed, either explicitly or implic-itly, that in the case where the expectation-generating equa-tions are jointly estimated with the “structural” equation, theunanticipated components are mutually orthogonal. (Estimatesindicate that this condition is reasonably satisfied for the speci-fications used here). When these variables are measured inthis way, the regressors of equation 15 are nearly mutuallyorthogonal. Consequently, the parameter estimates of onearenot likery to be affected by the absence of the others,

“This could be a manifestation of the heteroskedasticity in thedata, In general, heteroskedasticity may causethe reportedstandard errors of the parameters of OLS to be biased, andthey can be either too large or too small.

61

:2 K”’KK\K’1’17; K~V/~KKK~ ~K, ,~2’ K\’,’ 42 ~ ~ ~: ~

K / <, / / ,, 7/

“ / “ K KK K

KKJ K> >KK~K>~K /\ / 2 / N ~K K\K< K

K ~~‘K ,~, KV*ISflt(W~ Pr K4

-. K ~, St a ~flK ‘K 264**a K~S~ 1~8~KK

K, ~ tS* a’S/ 0’S $~

~ K> K.K~ 50’‘K ~ ~

K lIt SOt~ =‘~~ ~ KSt

/ K2

K / K KKK

/ / ~‘KK, ii : ‘oa K,, ~

2 KK> K” ®D>. ‘K / 8~ sS4>~’

~~2’Y’ ,T~ ~<c~y~*F~1~~ ‘~ ~ K

7/ / ~KKK~ ‘~K> 4* .7 ‘K / > r..,K ,,‘/ , 2 2”A >~“ K’s’ / >7 >> ‘2’,’ A K

,\K~f/’~’K,’’’>JK~Kc2/2’’>\,\ 72as //:KV’2/~K’Kø/\’K*$*/K\KK2’, / >2 V // >4 / V

)AKØ~KSS:y~2

L~ ~K~/> ~t’ /K,

2‘K -88k “~ / , t28~ / V K

K ‘KKK2> V/f “ ‘K/a K — >2 —~ ,,K~ ,V •~$42~

/ N ~ > K,,< i$0’ “K’ ~ / ~“ / <K ‘K 2’ ~ V”4~>

V V / >2K V $~KV ‘V / K>, “K K

/ / / // K / ~ K~tfl ~ KK K

KKK/K/:KK >AIK, ‘K’//:2 ,2,,, , ~‘K”~ >~/ K >2 ~K2 K’ ‘K~’ KK K?:KK K~K>’K 1.t.’,~’r-K’ K ~ AKK;/ ‘ K’ ‘K~

KK> K \/ / K ‘K ‘K K

K KKK “: ,~WffiS*s 0~te8flan s~*y~othesó~b~S ~hG~t4$SlSt~ K K

K> ~ 2”KK ‘2 KKK,KKKK1,, K KK 2 K KK K K ‘ K>2 K KK

‘KK KKSóS$ K<sS~e< 3~,~>K K \K K

K KK/ K K K K KK K K KK< K> K K KK 2’KK K K

Because the results could be specific to the form

of equation 15, the equation

6 36(16) ATBB9 = a, + Z cçATBB. + ~ ~MV~K, + 6,,

i=0

was estimated using the same data for the sameperiods.’ These r-esults, repor’ted in tables 4—6, arestrikingly different from those in tables 1—3. For’the entire period ltable 4) there is no statisticallysignificant, negative response of interest rates,even initially, when AM1 or AMB is used, More-over’, the sum of the coefficients is significantlypositive for both monetary variables. These resultsare consistent with those meported by Cagan andGandolfi (1969), Brown and Santoni (1983) andMelvin (1983). when ANBR is used, however’, there

is a significant initial negative m’esponse of interestmates for the entire period, and the sum of the

coefficients is negative and significant.

The results using the unanticipated monetaryvariable measured by (AMV—AMV’( are consider-

ably different from those using AMy.” For both Mitand MB, few coefficients ar-c significant and mostof these ar-c positive. Also, while the sums of the

coefficients are positive, they are not statisticallysignificant. When NBR is used, the initial coef-ficient is negative and significant, hut the sum ofthe coefficients is positive and not significant.

Most of the results for the pre-1974 per-iod table5) are qualitatively the same as those for the entireperiod. One exception is for ANBR—ANBR”l, whenthe initial coefficient is negative hut not significant

“Cagan and Gandolti (1969) used 38 lags, Melvin (1983) used36 and Brown and Santoni (1983) used 24. Because of thelong lags involved, it was necessaryto delete the first threeyears from the entire estimation period and from the first sub-period when (AMV — AMy’) is used as the monetaryvariable.

“OLS estimates of the standard errors of the coefficients arebiased downward when unanticipated monetaryvariables aremeasured by (AMy—AMy’). Consequently, the reportedt-ratios overstate the significance of the effect of unanticipatedmonetary impulses. See Pagan (1984) p,234.

62

The results for- the post-1973 period (table 6) ar-cdifferent when NBR is used, The initial negativeresponse of interest rates is larger during the post-1973 period and is statistically significant regard-less of how unanticipated nonborrowed reservesare measured. The sums ofthe coefficients, how-ever, are not significantly different fromn zero.Thus. while the magnitude of the negative effect islarger’ during this period, it is not permanent. Theresults for the Ml and MB measures are similar tothose ofthe entire period.

Tests ofAlternati’ve Spec~ficalions

Tables 1—6 show that the results are sensitive tothe specification of the monetary variable and to

the sample period. Consequently, it is important

to test which monetary variable, if any, best cx-plains changes in the interest rate, To this end, thespecifications with alternative monetary variablesare tested against one another using the Davidsonand MacKinnon (1981) J-test. In order for the testto favor specification A over specification B con-clusively, the infor-mation in B must not be signifi-cant when specification A is the null hypothesisand the information in specification A must be sig-nificant when B is the null,

Table 7 presents the test r-esults which, thoughlargely inconclusive, favor Ml and NBR when un-expected money is specified in AMy form. Thisis due solely to the post-1973 per-iod, however-.When the monetary variables are specified in(AMy—AMy’) fom’m, the results tend to favor- NBR.t’

and the sum ofthe coefficients is positive and sig-nificant.

“Although not reported here, the results of the J-test applied tothe specification given by equation16 were also inconclusive.

A K, ‘K K K K’ / > K ~> ~ 7” ~2 ,,K”K A,, K 7KK ,‘ A A K ~7 ‘>7 A

“K ~ KKK ,, K,K’ ~ ,, ~7/ K /2 / ~ / K ~ 7K /7 K

~ /,I,K9:’K’;:K;~/5, I ~~~,‘:t1K’ ~ K> /K>KK>/,KKVK,7K ‘K’~~::: ~ ‘K K ,,K~ ~ ~K , ~Le4*~:t4*t* ktc~fl~&~tflfl;tT~~~i’~>2 ~ ~ 4 $e>*~ ~ ~ ~.. >s , K

1 ‘‘,,;K, , ~.7 ,>‘K’ ,‘:>7~’ ~ K ,>,,~‘ ~ 2”,”~’~/VK,’2’7 ,, >2KKKK>2’/>> , ,7 ,KKK/> ~K K ~ ~ K ‘K ~ ‘K ~ “ K, ,‘ ~ ‘, .. 4~:’ ‘ ‘K~ ~ “> ‘K K’, KK ‘K’~ A /K”,,/’ ~ t,AK ~ “V K> ~‘K, K’ ,,

‘K ‘

4/ ~1,,

5>/K,

K

K’

K

:2

K

‘K

K~K

K

:

:~KK

K

K

~‘,

~

T

1’

K

21

A

‘2~K5><K

;~ ‘s/K/K/ <

25>’ ‘2’”

‘K >2 KK ,,

K,

V/’ KKKK,K7’AK’ K>~

~

~>7/~’,

K’ VK ‘V OK

~

‘KK 1/,K>K

‘K ‘K

K’K

K ~ K

‘KK

KK/KKK>2KK>

K,

K//K

K1~,KK,

7K > , K~

K ‘:K

A’,

‘C K’KKK

K V

~/ ‘‘K

K,’K4a

~

>7 ~>7*A

S~J

‘K~

IKK2’,,K’K4K

~Ø>K >~

;‘2?”$

K ~K~~Y5~

~7

Jjj2 iag~j>g,ntt’~t’i 8~1net ~i ~ : >‘~1~S~<~:‘~~*‘ ,_>~ ,Krtn eK’ ~ >~“/‘ øOgo~S2 ~ ,, ‘K /

?2 ~ K/K77K/>2/,/;K,72’5’K’KKK’K:KK7K:,K/,~~/~///,KK:,,K ,:/KK22’4/2 : ~ ,K “K K ,, ‘KKK VA / ~ ‘KK’ ,

2>K ‘K ~ K” KK

t, K2 , p,2/ K’~ >7 <‘7/’ /7 K , /7,’K 7 ‘K>,$, K K> ‘ 4”e”4~ 2K ,,,, >~~K~’ ‘ / ~ K’‘K’’~ ‘~“ .7’ ‘2 “K‘K’2’,2t~nzra,nt,n~ ~aEtfatp*~t~ / ‘K*’,W ~

~ eo4a~po~14$~-4 ,1~9sa& /‘~ó’”~ ,spoK r •~ K,

~ >~‘>7<’2, A ‘> A K>, , :~‘,‘1’~,~’~’,KK K K”KA’K//K2’’/’ /‘K>22 /K/’’5, ,7/7’(, , //K,/K~ ,“ ‘K ,>~‘ A’>” K<K K A ,K’,A ~, A K’ ‘K// V7

/ A K “K‘K”K ‘K ~ A ‘, ;~ >K / /7, ~2’,’, L~ittn ~t+t4t~~i~a ‘*~E~ ~Y7 I(K; jK

~ e$tdl4’, \K ¶ KQ’ ~ ‘~-* —‘ K ~

,K 2,7>27 /KK7 ~

KK, K / 2i/’,K>,KV KK / , :‘K’ V 2 >7

r ~,, 7 K c >< ‘K’~ ‘‘K>K’ “‘K “ K “K ‘2K ,~

~ a4~ai;iii,i&td2~IK ~ 7/K ~

~~ocoKDfl~ysoo ~Krø2~ 4’ dS~00s~a~4~ °‘“ ‘K

I >2K’, >>K >~ K’2K,’> //K/ K >>,$,75>7/~7KK~KK>7~>K:1~’K7K K,K’/~ KK< K’7VKKK’KK ,~2‘,K K~ ,,K K ~ ~ ,, K K, K, ~ ,~K

‘ /7’>,, /,‘K ,~ ~>K /KK ~ K ó4

‘~ OS~K 3~L*t~IduISKT~~ KNR~l~K~8ZiIoe.7K..:l%4 ‘KLw*4 ° ~ ~ o~~~a~,ogq d o’*=a d4-~ K’’->n ‘ K

,,K,K> ~ /K’KK1K ‘2’, ~ 7 ;K~K~’>/’KKK~

K K K7

K ~ K, K> K>~’ K K ‘K ~ K,’ KKK V

5 ~ K:/ K K~,KKK.7 K K’,/ 2>2K V ,~ Kv’,/2 Kfl

~K,KKcK~K 27K >7 >~ ~ ~ ~5>~K ‘5>>’K,’ K’~ >~‘‘ K/K,,KKKAK,K ‘K”5>’V~U “~‘ 5> ~‘ ~ ~2‘‘ K ~ “ <KK,~,K2’’

4 “ 7/ A, 2K ‘4 ~ >,‘K K/ ~ ,K~ ‘ V ~ K7

2/ 7 VA K

~ ~ i~~w~so~&w.t,szgankgs K

j~ ‘ /~‘*~t~’~>& b*_tt?a~oo_k~d / ~

/‘ ‘K K/ ~ ~V K ,,V /,>K,,

1*r ‘Pa? ~ s>nKDwst4’e ,,, >t~ 4-~ ‘—p~Ø’*~a’ ~

K/K’K/KK/,/,’K>75>>, ‘‘ s~,’K,’K ‘~ 7’> A,’>, ,2KKK,K,K/2 /,KK>’’/>N2K,7,i7’2/14>>

>7 K ,,7” 7 ‘K,, / 4’ K’ / ‘K K,’KK/~’2” 7 7/ /

k K~ ~

‘,2

’2K,

1/’>>7 /~ / 2’ K V~ >‘, “,,i KI*C~St L:~t4d~tt~ fl’~*i4**:ffrCfl

/ ,~$s ttK 0/ 0~4~

t~a ~K,’ K “

/ /K’KK/ KK>’KKK/, K~’7> K”>’,> V // /,>,‘ /K/’,’,,\/KK K’K7N/’,, ,‘/‘~,>1~K> K / AK K KK/ K K> ‘K / / ‘41 ,..7 K~ ~K 7/ AK

>~‘‘KK >1 K’ ‘K ~ 2K’ K K>~’ K>2

~Ji4EEi i/~hs /~IKlK~Mt4K K~gj/~~~K/

*~a”°r~’S,e nO’~ar at des a’~P_/K/7~

CS

,K,>K,KK~\,KKKK///<>2KK~>2KK/~KKK\KK/K/’K/K’K/,’2’K~K

/ ~, ,,,~ K /KK/ ‘K>> /KVK ‘KA K 2, /

-? ut aIstahbadSsiatb ~‘ In / rn

<<*cK

/>,C5>/

‘K

K K”

: K

K,,

~ CKK

~

~K2’

~K

1K

~:>K

~

‘~K>

~2’

~

SS

eL,2

>1~

&K

~r$,,Kt,’

1S

~I ~

‘1

1”2’~:

I~nj ~ ~ g~f”’,d K,K>/?,//’, ~,,‘

7 2’ K 2, ~‘ C “,> “27

K ‘K,> AK K /7/ ~ ~

~ E ~ /// 7 2 “ ‘K ,~“ / A/K>> > “~K K>K,~ ~ “K ‘ / ‘7 ‘K’ ‘K V <,/>‘ // /,/‘K,

K — / >7/ “K / K’K>K’ / K’ / >2>2 7,, 2>>N K ~K 9~’ V K’~5 ?,r/

/ K, ~ / >K>K’ ‘K ‘A’ 5’’ “K ‘2’> >2 ~ / /K/’K”

7 ‘V”’’’’K ‘‘‘‘K’ ,5>’~KK>K ‘KK /,‘“‘KN, ‘A~’~”’~75>K>/’‘K ‘K/K V K ~‘/,/‘K~~ ‘‘‘ ~ K’ ,,7~,,A, V~/ ,, K 7,,’, > K’/ ,, K7

7~/Ky/

7~K / ‘, ‘ A, ‘‘V>2’ “, 4>7~’K,K>’,K4’ “ a~”

‘KrV

~5>~

,

‘K~,

C‘

~‘~•:g

!V

~‘

eK

,5>‘5>2’

A/~‘

2’

/~

K~

—

/ ~“~flK>’K>, ~ ~ ‘

~

2’ s,,, ,~ /4’, 1 :>2~ VK ,, V, ~ K> K>~ ‘K>,,’’ V “K / K>’ / ‘ “2’”K’ ‘~t’ “>2,V’,,

.7’,’ ,7 /‘7K>’K>K>, , 7 7, 7 V ‘ V ‘>—‘>,‘ , ‘K 2’j/ / K K S’dsw~’*

,

‘2’, >7K> V / K’ / V ‘ ~>K /7 / V‘77’~ ‘K K,,7/,K, ‘7K>,,.7 ‘>>‘‘‘V/1t, ‘K’ ‘K>CK>’~””KK”K K>s’n ~ fl~ E4E;~4:

,>c>”2”1

’,>K>,~’K>~ K’,,’//K2’/V, “K 7K~ K>>~>K,,’K>,7K~,\KK>>/>,>7,KKtKK>V VK K 5> K>>>, / / V ‘ ,

‘K K 5> ‘2” ‘ , 75>2

r;:f :04 a~: 4g~ojS:tes1tK K,, ‘K~K>K’’,’ “>5>”2’AK,

72’2’ ~,,‘‘K,~”>>’”’’2’> ‘2’5>~~’>>~’*5>’i’1>V

/7K\// ,K K> ~,‘‘‘K’’/ KV ‘‘‘K> ‘ ,5>~’ ~ ‘2’ K~ ‘K K/ ‘K V ‘ / V, /7’ “77 ‘~

K~C ,~c’~”.~2’~è” ~2’~”’4,4 ‘5Ø~1~K>V’K~~ /< ~>>K’<”K} ¶11P’ ~ ~

/ / V 7’ K K / K / / K > K’ K/K

/ K 77 ,,,,K’\/K>K,,V,, ‘K’’ ‘ ‘‘> K’ /K7,~/ ‘ /

/ fl’s”>’ ‘ ~ à*” a~d4r ~ K> K >~*/ ~ p>~a”C ~

~ :~V ~t°~ ¶7/~ ~ 4’ V

7/ ‘ K / K’,, / “K ~,

K> K,j~ ‘2~5>~ ,#/ ~ >~‘ ‘ ‘K’*t,2’ ~ a~’~

“>~ S ~g K~~geg,~g p

z>Ø~>’

‘V

‘K’K> K’

I K

~p

~ V

~ ‘>7

4~K~

Vg,/

~

~K V

K

1

~/

‘

>

,

~

>

K

KK

K

K>

~

/g2’

‘K’s — K/ ‘~ i,K>’~~/~2‘~~- ~ ~5> ~ ‘2’ ~ ‘ ,~ ‘ ~ ‘,K’/K>t2 2’ ‘ K

K ~‘ ‘ ~K~//K, K,K> K’2’7>,’/ KK2’’/K ‘ ~,/“’‘,K>K 5>~’K>’/2’7’2’ K/K,’ ~ ~ I a~ KKK/ Ka K, / 7 /K, 7 7 4’ ,~K~’ 7 47 K 7 , K ~ 7 / >>7 K K V

K K Va / >~* SKI K ~KC VK K K K /‘K V K / , K> K>

2’ ‘KKKK 7~2K> K K K K 5> K> K K> K K

2’ , K K>~ K> K K K K//~//KK~/o a~~’ j’” a~Ko *00,0 ~ar2’a ,>Ka’0~K K

I K S~’~fl K75~$r2 0~o ‘K&artffl~t

t K / K V K K >‘ K K K

K KK ~ a ~ae ~ K ap a’ K

~ K

K> ‘K ‘~‘ -KaK5 “‘ ~—oos o~a’,— >—*rqK o~aso~

‘K K~/2’2’~>~ K~~4t /KKKKK>~\/KK\2’K ~t Y~tk~KK:K:K

KK K K K K> K SK> K’K K> K ““> K ~K KK ~ K~ K K

K ~ a~s~s°8’~,,KKKKK KKKK ‘K’KK>KKKKK> KKKKKK’K K>, KKKK2’K K K OK> K K

K K KKK / KK ‘KK 2’ KK K K’K KK ‘ 7> KK 2’ K KK K K

K K ~ K

K ~ KT KKt~’1 ~IKK1’?~ ro,,o>,~! ~KaK!KK5~1 K2’ K KK K

5> KKKKKK\KKC KKK KKK 2’ >>K KKKK K\K\ KKKKK K K

K ,,

K 2’ ‘ K~ K K t~oo~0J~Q’p’a doaK—Ofl’a 5OO~~~O 0 K

KK ~ K~K~ KK> 2K /KK K K

K KVK K KKK KK K / K KK

K KKK KK KK KKKKK KKK K~KK K K K K K K KK K K KKKK K

K K K K K K K K K K K K

K KK>KC K K ,K ,K,2’ ~ KKKKK:K7CKK CK K KKKK KK2’K>K’, KCKK2’KKKK> ‘K, K ‘K’K KK KK>K>KKK K~K~KKKKK K

K K K ‘K / ‘‘K ,> KK KKK

K KK~ /KKK K K. ~ ‘KK :: :‘ K’2’

K 2’ ~ K K, ‘K ‘K1

K,, /Ks KK

K K> 2’ salgLs,lskbLnLssbs ks is! > Kit 2J~2’s ~ °~~‘ — K,” “

KK>K “ K‘K ,‘ ‘S K>’ K>K> ‘K K” ,,K/’, ‘> ‘5> KK>// ,‘CK>”K 7> , >2’~K2’2’K’>41> 7’> ,2 ‘‘K5>’7,’’”5>K K,,~ >2’, K K’

K,, K> ~ ~ /7 K> N KK7

K>’ K ‘K >~ ‘ 2’ ‘‘K ,V ~

‘K , K K> ~ ‘K , K> , K>, ‘ K~, C>2~ /7, ‘K VA/K AK ~‘ K<~

‘K 7 K S 141 I~S1I1I~~S*4~EIK>SWCJ,j~UI ~ t~’~ 3K ~ P ~c~~—n,do~ara’—’a— ~t~,Ce~’dl>r~c’,tC °‘>~‘t>K>’ flK>rV

K , 7’ KK>,’’~’’ ‘ ,K 2’/K/’ K>K,K, ‘,1K,/ ,K ~K/>2 K>’,,, K, ‘V ,‘>K>/7’’’> ‘ KKK >~1KKAK>’K~’K/

~K, ‘ 7’ 7 7/ / ‘K K ‘ 7 ‘2K ,,, K ,, / <V “

K ‘ K/’t ‘7 K,,,,, , , 7> 7’ ‘KK/ K, K> ~ A/,’

K7&j , ‘K ‘ >2~K>7,7 7,K ~, K’ K ~ , 2’

‘AK a ~ c~~Ifl*gt~~~~ifliSn~ 2~~ flK K> 7’ ‘ K’~~,’V K’~92’~’ ~~ ~ ‘~ ~ ~ K> ‘W’~’ ‘~‘ ‘Y ,~‘> ~ “ ~5 ‘ ~

K a’KØ’ 0’~>Ø ~o~r2d~StOSO4 ~ ~ “*-

>2’ KK>,K>K ‘‘K”’’> ,~KK’’’,,’V~/’ , K,, >7 ‘ ,V,,,/ > K ‘ ‘/~,K /7K>, /, ‘K‘K> ‘2” K K> ,,K’,/,, ~ K .7K ‘ 1, 7’> >K’/K> 7’>,” ,K,,K,, 7, K , ,A>K>’’ ‘ ,,‘‘K’,’>’K / ~ ‘ 7 7 7 7K , K 2 V K 7 ‘K ‘V ‘ ~ 7

:~>. ~:flrj~’~j1.I~tizs tfl~7fl44vK’KJ ‘K ‘~a

~K ‘K ~ “~‘ ,, a d a Ø~” ,,~,, ‘2’s ,>~*>ra~a~’4~,,a ‘ate 4$ K ~ ,

A ~ ~ K~,>22’:,:KK//>~5>/1> P>>: ~/‘7 ‘K> K~ ‘ , K A ~ ~ ‘K 7 / V >K>~ KKK K

K K K ‘K ~K ‘ K>~ ‘‘K ~ K ‘a”>24~’ K>f~,,’ ‘‘K ~ ‘ ‘‘5 >2 K’K”’ > ~ S‘ ‘ 7 ~ ~ aI~1I aRE*t~d~~5~ tat ~n

>7 ~t C ~C ~,d’C >0>5>0 ~ S~0~K ~ *‘ ~V $>2’2’~, ;K/KK>5>aKCK:K>k>7cK>,,~~K>KK/K2 ,‘

2’’Kl’’~

7,K,,lK

5>lll’,”

5>2’’~K>>

2’K’K’K>~

2’lK>>,’~l

K ‘K 7 >7K’ K K’K ‘K ‘ K’K’

> K 7’ 7 :2’ // , ‘K,’ 7 K K ,, 7 ~2’, K, /

2’ ‘1c’~,,~:~> : t I au re~st~çgjg)~ z~1’i~*~7/, K>” S~Sc t~ OS’dS”d~S O~,d 5 a’d08 o’r~D~. ‘t-’ a ~“

2”’ K,/K>’K>K>7K> >>/K>’>,,71’~2C,4C, K2’2’KK K’,7>’K,2’1, ‘K~ ,K2’>2’K;5>>K7~/K ‘‘K>>2”’CY”5>1Y2”>‘75>2,>~5>2’/7 1 ~ , 7# ~//4> 7 K’ , V

7K >‘‘ K K

/ K>, ~‘ ~K K’ , ‘K ,‘ ~ V 7 ‘K’KK >2’, ! K ‘K’ >‘K :~~ ~,1’K f2’KK’2 ‘~1>”,,~/~‘K ~J “~‘~j; :,~ !~fr’! :~,:> /~J~~7/ >742’2’~ ~

A/ ‘K K KK V /2’ /7/ >7 ‘K’ ~ ‘K

K ~, :~,~ ~7 ‘~‘15>~i ~i~:~ 1’/>CC>/~K;2C<K2’,<KKKK>>21:’C~2’:C>C2’K1K2’7K/ ~,>

/ V I : z41 en ‘KSf_it~w*~*r.K””

7’2’a’~ •~s—a—s>—n 4 i> S a’~~’4>~—t4’a2’ t~,z~, —4~— S

/ ‘“7~>2K’ 2’,’”~V’’72’7, V’>’’K,~,14K>’K~’5>’2’,K K/727>K >KK/C,2//’K>~,CK K> 2”2,’2’V 7 >>K / ‘K 7’ / / “ ,> K:,

‘K K> “~ *‘ ,7 ,/ ‘K K “7 7 ‘K V K7~Kr ~ t~2’— ea a —a ~a’ d~o~Øa ‘OKWK ,~4’g’~ola’

>77>/K, ‘K>>>,~~/,/~<7> 5>K ‘“KA,’C,K*’ >~C2’ , 222>2C7,,K>’~ K>’>, KK>/’K ,‘i>V’C,’ ‘K”KK>>’7K

/ 1 ~,1~t I ~~2’~’> “ /5>~ / K> 5>2> ,,~

7 K,~ (attfl ~rs°11~ Aj’K~,~ j ;~‘jg ~ as>‘a ~ ~, ~ ‘o—,’c~ ‘ a ‘a 2’~t—w~2*02’ 00

K>> K’, ‘C”~>. 2””7’”7’K’>7’ ,‘,‘i~”’, 2,”, ‘,KI,K’’7 $,K//K7/”72’4CK/>’>~,AK>K/~’7>~/V4’K’’K

K’K / K 2’ 7 K> / K 5> / I ,42’VKK>7/ 7 /

K, K,’>> >‘> , ~,‘ 2’ ‘‘‘‘ ‘‘‘4’ /KKKK ‘ 7K>~> ‘‘K> A ,

1. ~ *,‘,r,r~$ Jt~r4flU/ ,*K>K,>7,,CVbKK

C *,1KK>2’,2>K,KK2’2’/,/,K’K,>>KK>,,i~>2C/~~KI2’CK ,,,“1,, ,77K>7/7>//’7,7K>K/ 7 K K ,A/ 7 >7

V”K K/’K’K’ ‘ ‘K ‘

7 / K / ~‘ :~“ ~>2’ !~‘1’7t~~:.K 7K “~ II” ‘K V’ 7> ,V K

‘ K,,7

‘“K

‘K

V

K~

1’K:~vK>K>~(K’

I’>

¶f:~ 7

ti’>>~

>E~It

~LV

I~1

1~4’.

K>KK 7>

“2

K>41’

~

P~’,~,

‘>~,

v~;~,~5>’I

IK

K , KV’K ‘,:~c ,ti’ K / A ,,7 K> ‘~ “ C” Ky’> ,, K>, >)~,,

flK>’ K,K ‘ “>> ‘‘K ~ C’ 1~,,’> V’

77P “C~~ / K>7/ K ,‘K ~KVK_ ‘K>> f~KK

66

As a further test of the robustness of the resultsto the model specification, equations of the gen-eral form of equation 13 are estimated. This speci-fication has been estimated in such diverse waysand with such a wide array of regressors that anexhaustive evaluation is difficult. Instead, the ap-proach here relies on the fact that this specifica-tion differs from the others primarily in that it hasbeen estimated in level, rather’ than first-difference, form.” Some studies include measuresof expected and unexpected inflation and unan-ticipated money growth; other-s include expectedinflation, some measure of income growth, and a

measure of the change in the growth rate ofmoney. In the fommer studies, inflation expecta-tions are generated as they are in the rational ex-

pectations models; in the latter, they are usuallyderived from survey data. Furthermore, Mehra1985) and Wilcox 1983 a,b( measure the change in

the money supply by the annualized growth rateof money over a shorter period relative to itsgrowth rate over a longer period.

Consequently, two equations are estimated tocapture the essence, if not the exact form, of varia-

(171 TBR, = a, + £ a>TBR,. + j3L1Q,

and

i=l

+ ~tAt~ + BAy+ Ki~+ ~,

6(18)TBR, = a,+ £ aJBR.+ 13(AMV-’AMV”(,

i=l

+ j.rAAP—AP’L + S(Ay—Avi, + AP7 + e,,

LIQ is the negative of the difference between theannualized growth mate of Ml during the last threemonths and its annualized gr-owth r’ate over thepmior’ 12 months, A~is the change in the growthrate of the price level and Ay is the change in levelof the industrial production index. Because equa-tion 18 includes AP~the estimated standar-d errors

of AP~fromthe usual two-step estimator of equa-tion 18 are biased, Consequently, equation ISisestimated using a full-information, maximum-likelihood (FIML( method used by Mishkin 1981,1982(.”

“One exception to this is Peek who, although he specified theequation in level form, appears to have estimated it in first-difference form, See Peek (1982) p. 986.

“Lquation 18 is estimated simultaneously with the equationsthat generate the expected rates of monetary growth, inflationand real output growth, imposing the implied cross-equation

restrictions. Also, because equation 18 includes a distributedlagof the level of TBR, the equations used to generate theseexpectationsare modified to include the level of interest rates,

Estimates of Equation 13 tions of this specification. These equations ar-c

6

67

‘2’7 C’’~,’,K>5>.’5>”~/ 2’K>7>’V~’. 77>/K>K,,’K2’7’/5>>7’/’K>K>,’ K>K’~’> K,C’C/ >//K,>K/ V,’7C<>,V,/ K ~/“,‘ ;7’/ ~>, 2,, ,),‘2,’ ‘a’,~>’K ~ TsbJe~,’ I ~‘, 1>1<’ K K K ‘>K K

~ / ,,Z,1

‘ 1K>>’K~2/’K2’ 2K ‘KK>KK

1W’ t~ffl~Ø4~j/Vj4Q; K ~AP, K>~S~2’7K2 I~K k>,SEt K

V I “‘ 2’ 2’ K K> K

K ~ K ®~K~/ a, ‘øS “ / K~ >&*/ , 2’K / / >2054 /K fl*’ ‘K V K ,~Ø), K

K /~,K> 4$ >1 K K KK K

K K A, ~ ,,4~ CV)> “ K1K>K’K>K K K

2, K~’ >$~> / ~ V’/rK~4$K> ‘~ ~K a~r GTe> $~4 2’ K

K> ‘ , 4?) fO2’2~2” -‘ VK K K K

K~ ‘K ,, 2,41’ / , “ 12’’> ‘1’9~‘1%tØwi~~ ,‘ K K ~, K K

‘‘ >70U8 ~ K, / 4M~~975 S48>>~~~“5> C’ ‘K> 40*5>’~ S) ~ ‘K> / K

7’

a’ / >1w,/ 2’,is e~ K ~ a5>’C’ K,> ~‘ S V >$2” S ~‘ ‘ fl’ I

>V /,)hJBR2’ // “ ~“ K K V ‘4’ ~I” ~ e4 S4 K“2’ ,J$3t K,, 40 V~ ‘~‘~‘/ / ‘“>7’ 5>

/ / ‘KKK “V K ,K2’V19#~*”Vjfl~IO$> V ~ VVK, / ‘K/ J%.aK>’- 5>,>.~, ~ ‘ ‘~K V ‘K’,K 7

/ “2,1*4’ ,‘ 5sf> /A “K ,~s “ 7>,, / K> V 4*42” 2’ // 40444’ ‘>4 ‘/V “2’> “ /5>”

,‘ 9’,> ‘7>’ “5>1’ K’ I “a2” 4t / 4Ør ,, , >g~ C”,K K

7 / ‘K // >> 48 ~‘e>Qe’~ ~ ‘K/K > K

K Os a’ V ~ espK>, “44(1 “8047”V/, ~o> 1”I7’>~12”1,>7 “ C”>2>~K> ‘2,’~ \K>4~KKK

K7K>,K//,K’KK// ‘‘K -~>- - ‘‘‘2K> 2”’ K ‘<K,’-’K> ,,K,/, K>”>,’/ KK> K K

K I K ace ffi~>d I~4f~Jt$S*YàfWbIØ~ abIcAihesAd b be’ flosS*> )he~slAan*1eSa* K K>

K dP*4~~t/ ~K> ‘K>> ‘> ‘7’ , C ‘K N K> K>

K K ~ ~fl~lAM K K K K> K>

K> K>KK>KK, ‘KK>>KKKK/K>’2~/ >1K K>K> KKK> KK K>’K “K K K ~K K K>’K’KK KKK>< K K K K

KK>’K K K K K K K> K

Table S presents estimates of equation 17,~0Theresults indicate that interest rates show no statisti-cally significant negative response; however, thecoefficient for- NBR for the pre-1974 period isnearly significant at the 5 per-cent level, The signifi-cant positive relation between LIQ and the level ofthe Treasury bill rate during the entire period,when either-MI or- MB is the monetary vam-iable, isattributable solely to the post-1973 period.

The magnitude of the coefficients on ~P and ~yand, in the case of Ay its statistical significance,

depends on the period. The positive coefficient onP is statistically significant regardless of the sam-

ple period; howevem’, the estimated magnitude ofthe coefficient is sensitive to the sample period.

Table 9 presents estimates of equation IS, Unan-ticipated inflation is significant in all three periodsonly when NBR is the monetary variable. Bothunanticipated income and inflaUon are significantduring the post-1973 period for all monetary vari-ables, Surpm-isingly, anticipated inflation is signifi-

“Some econometric issues should be addressed becauseequations are estimated in both level and first-difference form,The issues center around whether the variables on both theleft- and right-hand sides of the equations are stationary. If theright-hand-side variables are non-stationary, then the reportedstandard errors from fhe level equation will be incorrect even ifthe left-hand-sidevariable is stationary. Onthe other hand, ifboth the left- and right-hand-side variables are stationary, thereportedstandard errors from the first-difference specificationwill be inconsistent because the error term from this equationwill be serially correlated. Most tests of macroeconomic time-series variables, like the ones used here, suggesf that they arenot stationary in the levels, e.g., Nelson and Plosser (1982);however, these tests are not powerful against the alternative

hypothesis that the data aregenerated by a stationary ARprocess wifh close to a unit root, In this instance, estimates ofthe level equation would be appropriate, though the samplesizenecessary for appropriate inferencesmight be large.Because fhe objective is to see whether the results aresensi-tive to the specification of the equation, we are agnostic aboutwhether the level or first-difference specification is best.”

Because of the lags involved in the construction of LIQ, it wasnecessaryto shorten the estimation periodfor the first twoperiods. They begin at 1960.05, rather than 1958.07.

2 ‘K //K’K>/ ~K> ‘ ~“$ N ~ K77K>~ K>’ ‘K>~K V K’ ‘ 2’K’/,/KKK’K KK>K>/KK

K ,K /K~KK>’7 >K,,, KK 2 \~,, K K” /K KK/KK >KK’K’ ‘ K~/ , ‘K>/ ,,,K> ~,‘ K7KK”,7 >>41~’ , ~>~>7~’>, IK’ ,K/K>V,’J,/ ,,5>,~N>>2K>~>

7\.7\\/ >7 K

Ktabls,e, KKK> “ K K K>K K>, K’> K> K>

K>> K> “ K I ~KKK 2,t,,K , , -._ K> 3.. ,K” ~, K t~i~>2, K K

K’ K CORS$$ K,7 jj - K> K>~ K>’ K> K

‘>~ tS*-’tThs ,. >1 .

K 508 ~ ~ ‘nor> Kr

“2’ K ) K (5$K> K K> ~

K> N Ka _e* ‘ K ~ ~ K$~,‘K/K> 1K KK’>K ~‘KK2’ ‘ KKK 2’ >~l>MKK K>’’>l K~ 7’’ K >“~ “V* ‘K K

‘K Kt~Øa >2’> K> ‘44’)> t5t~ “>oogt , K>

K K ,~‘ K>K>K, Wil~/, “ Pt) K

K> ~ K 2’K.KK~> K K K K> t$toaK~~Ktl*q92’ K

K> MI ~ KK7>2’tØ K K> 7’ 4>>,~ >000’ K’K,5~t’K> K>

K ‘ K ‘K44’SS)> K>, SSK ,2’> ‘~ 7 1$) K> ‘K K~2’5~~KK K

,M$ / K~8~’ V 1>V ~>2*’ 4> > V I K

K ,, 7>,,~ V “K K> K> 54 > K>

“K >>,,~ ,J~ft2’/K >K 144 , 7> K A ~ V V 5>”, IS‘‘K K’’ ‘/1’ ‘CK ‘2’, ,K ~K K 7>”/ ‘ / K K ‘ V~ AK’ ,‘K (‘K >K K/ 2K J8)>~>~>>‘ K> ‘K K>,2’ ~ _,,‘,~ _‘K -, ~ ‘ 2/” ‘K’K=_~ K >tK” K

2’ /7 7 ~ — / V ~ / ~K> K

CCI / ‘ 1” />C’Q~/’’ >,V’ ‘ >2K//4w 2,’ K K

K V~K ‘7 / /‘ V ‘ / ~(6>~ ,K, f‘(‘I/, ‘,9’~t’>2’Ir5>’,’a’I277/>V’~~S,2’>C’~I7’>~;>4rnrc>I2’I;>’ 75>~34’K’K ‘>‘>>‘>~ >~>~~~K,I’K>

‘7> / *00, >2 ,, 48)> / ‘KK, K // V / K K’” ‘K >2’ ‘ ~~KKKKK

~ /K>~~$> 03 / >1> “>2’ ~> ,‘GIV , (400 K K K

,,JV5>K’C’K,>2\4ØK KC’,K>>CIIV\I/K2KIKK>>I4’ />K>/,>7qs>2KKKK

~4~~jK>’K tKKt~/ Vi Y “7> 7KAKKK K

I’ [I K K>~K K’ K ~ Wpo’ ~ ~ 1K KK K K K K

7K’>, V ~ K “ K> K K

cant only during the prc-1974 period, and then is attempting to control money. Since the Fed wasonly when Ml or’ MB rs used. attempting to control Ml through a nonborrowed-

- resemvcs operating procedure from October 1979With respect to the responsrveness of interest to October 1982. mnore precise estimates of the

rates to monetary changes, tim results are consist- --. responsiveness of uiterest rates should be ob

ent with those i-epor ted in tables 1—b. A significant ‘ ‘ , .. -- tamed dunng thms period. the limited number of

negative effect is obtained during all three periods- . monthly observations prcvcnts using specrfica-only when NBB is th monetamv vanable, More-

- ‘ , trons with a large number of parameter’s, howcver,ovem-, the effect ms larger dunng the post-1973 pe- -

• - ‘ ‘ the number of observations can be expanded byr’iod, when a significant negative effect is also ob-

- ‘ ‘ employing weekly data. The weekly time periodtamed with Ml as the monetary vanablc, Hence,

has thc added advantage that the responsivenessthe results am similar whether the int rest rate is -

- . - ‘ . of interest r-atcs to monetary changes is cven lussspecified in Icvel orflist-diffcience form. likely to be contarmnated by income and inflation- expectations effects,

Unfortunately, using wer-kly data precludes theThe iesponsiveness of interest iates should be income and pnce variables. - Previous results,

greatest during periods when the Fedumal Reserve however, indicate that a statistically significant

‘Cunningham (1987)and Cunningham and Hardouvetis (1987) report no drrect evidence consistent with a strong response ofalso useweekly data and proxy changes in prices by the BLS interest rates.22-commodity spot price index and income by unemploymentclaims, They acknowledge the weakness of these proxies and

69

effect is just as likely to show up in relatively sim-ple and parsimonious specifications like equation15. Also, the results indicate that the significance

of the effect is relatively unaffected by the form ofthe unanticipated monetary variable, Conse-quently, specifications like equations 15 and 16without the price and income var’iablesl can be

used to estimate the responsiveness of interestrates to changes in the money stock with weekly

data.

Estimates of equation 15 using monthly data forthe period fi-om 1979.10 to 1982,09 are presentedin table 10. They are similar to those for the post-1973 period. When ~M1 is the unanticipated mon-etary variable, the coefficient is negative and sig-nificant at the 5 percent level if unanticipatedoutput and inflation are included, and marginally

insignificant if they are not. For MB, the coefficient

is positive and statistically significant only ifLXMB — ~MB”) is used and the other variables areexcluded, When NBR is used, howevem, the coef-ficient is negative and highly significant tegardlessof whether-the other- variables are included. Fur-them-more, the estimated coefficients are larger-than those obtained for the entire post-1973 pe-riod, and the adjusted R’ is about twice that of theother monetary aggregates. These results are inkeeping with the nonbor-rowed-reserves operatingpiocedure used during the period. Nevertheless>the coefficients are small, indicating that a 1 per—cent increase in the gr-owth i-ate of nonborr-owedreserves results in an about four- to six basis pointsdecline in the monthly Treasury bill rate,~s

Table 11 presents results using weekly data.’~

‘8See Thornton (1988) for adiscussion of the borrowed-reservesoperating procedure.

3iAn equation similar to 16 was also estimated using weeklydata. The results are not qualitatively different from thosereported in table 11.

70

There is no statistically significant response of

equation 15 without the pr-ice and income vari-ables, r’egardless of the monetary variable used.

The results suggest that intem’est rates do not re-spond over a period as short as a week, but dorespond over a period as long as a month,~”

One possible reason for the disparity betweenthe weekly and monthly results is that the data areaverages of daily figures and the aver-aging processmight mask the response of interest rates whenweekly data are used.’ Consequently, the equa-tions using weekly data were re-estimated withthe change in the Treasury bill rate measured bythe difference in the Treasury bill i-ate on consecu-tive Wednesdays. Though not reported here, theresults are qualitatively the same as those shownin table 11-Consequently, the insignificant re-

sponse of interest rates is not due to averaging.

SUMMARY AND CONCLUSION’S

This article estimates the responsiveness of

interest rates to monetary changes using alterna-tive specifications that have been used in the liter—ature and alternative monetary variables, Theequations are estimated over the same time peri-ods using the same data. Several interestingresults emerge from this study.

First, estimates of the iesponse of interest ratesare relatively insensitive to the specification em-ployed; they are, however’, sensitive to the mone-tary variable used. A significant negative responseof interest mates is most likely obtained if nonbor-i-owed r-eserves is used as the monetary variable.

Second, a negative and statistically significant

relationship between Ml om’ nonborrowed reservesand interest mates is mor-e likely to be obtainedduring periods when the Fed was placing greateremphasis on monetary aggregates. The most con-sistent and statistically significant negative effect isobtained using nonbom’i’owed reserves, a monetaryvariable that is likely to reflect the independentactions of the Federal Reserve. Nevertheless, thefact that them-c is a significant effect using nonbor-

40Hardouvelis (1987) estimates an equationsimilar to equation16 using quarterly data for the period 1979,04 to 1982.03 andreports avery large negative and statistically significant effectof unanticipated money on the three-month Treasury bill rate.He finds no significant effect for the ii quarters prior to1979,04 or during the 12 quarters after 1982,03. He interpretsthis as evidence of a strong liquidity effect during the periodwhen the Fed was targeting the money stock. Since he doesnot adjust for the credit controls during the first and secondquarters of 1980, however, his atypically large interest rateresponse may be due to unusual movements in money andinterest rates during these quarters. For example, the money

stockdecreased at a5.9 percent annual rate during the firstquarter of 1980, while the three-month Treasury bill rate in-creased by 316 basis points (measured as Hardouvelis doesfrom the last month of the quarter). The money supply in-creased at a 21 percent rate during the second quarter of 1980andthe Treasury bill rate declined by 813 basis points.

41The monthly dataused here are also averages of daily figures.Mishkin (1982) argues that misleading results about marketefficiency can be obtained using averaged data, and reportsthat he obtained substantially worse fits when he estimated hisequations using quarterly averaged data,

71

rowed reserves regardless of whether the Fed is ________ - “Interest Rates and Monetary Policy,” Journal ofconcerned about the mnoney stock or intem’est rates Political Economy (May/June 1970b), pp. 431—55.is anomolous. Hafer, R. W., and Scott E. Hem, “The Shift in Money Demand:

What Really Happened?”, this Review (February 1982), pp.Third, estimates of the responsiveness of inter- 11—16,

est rates are sensitive to the time period chosen. Hamlen, Susan S., William A. Hamlen, Jr., and George M.Generally, there is no statistically significant re- Kasper. “The Distributed Lag Effects of Monetary Policy onsponse of interest rates from 1958.08 to 1973.09 Interest Rates Using Harmonic Transformations with a Pre-test,” Southern Economic Journal (April 1988), pp. 1002—11.regardless of the monetary variable used. In con- Hardouvelis, Gikas A. “Monetary Policy and Short-Term Inter-trast, a statistically significant negative effect is est Rates: New Evidence on the Liquidity Effect.” First Boston

obtained using both Ml or nonborrowed reserves Working Paper 86—28(1986).after 1973.09. ________ - “Monetary Policy and Short-Term Interest Rates:

New Evidence on the Liquidity Effect,” Economics LettersFourth, the results are sensitive to the periodic- (1987) pp. 63—66.

ity of the data. In particular, in the specifications Hoffman, Dennis L., and DonE. Schlagenhauf. “Real Interestestimated oyei the period from October 1979 to Rates, Anticipated Inflation, and Unanticipated Money: A

October 1982, there is a significant negative effect Multi-Country Study,” Review ofEconomics and Statistics(May 1985), pp. 284—96.when monthly nonborrowed reserves are used,

Lmn, Kuan-Pin and John S. Oh. “Stability of the U.S. Short-Runbut not when weekly nonborrowed reserves are Money Demand Function, 1959—81,”Journalof Financeused. (December 1984), pp. 1383—96.

Finally, the evidence shows that even when Makin, John H. “Real Interest, Money Surprises, AnticipatedInflation and Fiscal Deficits,” Reviewof Economics and Statis-

there isa significant negative response of interest tics (August 1983), pp. 374—84.rates, the measured response is small. Mehra, Yash, “Inflationary Expectations, Money Growth and

the Vanishing Liquidity Effect of Money on Interest: A FurtherREFERENCES Investigation.” Federal Reserve Bankof Richmond Economic

Review (March-April 1985), pp. 23—25.Blinder, Alan S., and Stanley Fischer, “Inventories, Rational

Expectations and the Business Cycle,” Journal of Monetary Melvin, Michael, “TheVanishing Liquidity Effect of Money onEconomics (November 1981), pp. 277—304. Interest: Analysis andImplications for Policy,” Economic

Inquiry (April1983), pp. 188—202,Brown, W. W., andG. J. Santoni, “Monetary Growth and the

Timing of Interest Rate Movements,” this Review (August’ Mishkin, Frederic S. “Monetary Policy and Long-Term InterestSeptember 1983), pp. 16—25. Rates: An Efficient Markets Approach,” Journalof Monetary

Economics (January 1981), pp. 29—55.Cagan, Phillip. The Channels of Monetary Effects on Interest

Rates (NBER, 1972). . “Monetary Policy and Short-Term Interest Rates:An Efficient Markets-Rational Expectations Approach.” Jour-

Cagan, Phillip and Arthur Gandolfi. “The Lag in Monetary nalof Finance (March 1982), pp. 63—72,Policy as Implied by the Time Pattern of Monetary Effects onInterest Rates,” American Economic Review (May 1969), pp. Nelson, Charles R., and Charles I. Plosser. “Trends and277—84. Random Walks in Macroeconomic Time Series: Some Evi-denceand Implications,” Journalof Monetary Economics

Carr, Jack, and Michael R. Darby. “The Role of Money Supply (September 1982), pp. 139—62.Shocks in the Short-Run Demand for Money,” Journal ofMonetary Economics (September 1981), pp. 183—99. Niehans, Jurg. “Classical MonetaryTheory, New and Old,”

Journal ofMoney, Credit and Banking, (November 1987), pp.Cunningham, Thomas J. “Observations About Observations 40924

About Liquidity Effects.” The Difficulties of Exploring aSimpleIdea,” Federal Reserve Bank Atlanta Economic Review (Sep- Pagan, Adrian, “Econometric Issues in the Analysis of Regres-tember/October 1987), pp. 14—22. sions with Generated Regressors,” International Economic

Review (February 1984), pp. 221—47.________ and GikasA. Hardouvelis, “Temporal Aggregation,

Monetary Policy, and Interest Rates,” Federal Reserve Bank Patmnkin, Don. Money, Interest, and Prices, 2nd ed., (Harperof Atlanta Working Paper 87—04 (May 1987). and Row, 1965).

Davidson, Russell, and James G. MacKinnon. “Several Tests Peek, Joe, “Interest Rates, IncomeTaxes, and Anticipatedfor Model Specification in the Presence of Alternative Hypoth- Inflation,” AmericanEconomic Review (December 1982), pp.eses,” Econometrica (May 1981), pp. 781—93. 980—91.

Dwyer, Gerald P. “Are Expectations of Inflation Ratiqnal or Is Peek, Joe and JamesA. Wilcox, “Monetary Policy RegimesVariation of the Expected Real Interest Rate Unpredictable?” andThe Reduced Form for InterestRates,” Journal ofMoney,Journalof Monefary Economics (July 1981), pp. 59—84. Credit and Banking (August 1987), pp. 273—91,

Fisher, Irving. The Theory of interest (NewYork: Macmillan, Robinson, Kenneth J. “The Effect of Monetary Policy on Long-1930). Term Interest Rates: Further Evidence from an Efficient-

Friedman, Milton. “TheRole of Monetary Policy,” American Markets Approach,” Federal Reserve Bank of Dallas Eco-nomic Review (March 1988), pp. 10—16.EconomicReview (March 1968), pp. 1—17.

Gibson, William E. “The Lag in the Effect of Monetary Policy Sargent, Thomas J. “Commodity Price Expectations andTheon Income and Interest Rates,” Quarterly Journal ofEco- Interest Rate,” The Quarterly Journal of Economics (Februarynomics (May 1970a), pp. 288—300. 1969), pp. 12740,

72

“Anticipated Inflation and The Nominal Rate of Wickens, MR. “The Efficient Estimation of EconometricInterest,” The Quarterly Journal of Economics (May 1972), pp. Models with Rational Expectations”, Review of Economic212—25. Studies (1982), pp. 55—67.

Thornton, Daniel L. “Simple Analytics of the Money SupplyProcess and Monetary Control,” this Review (October 1982), Wilcox, James A. “Why Real Interest Rates Were So Low inpp. 22—39. the 1970s,” American Economic Review (March 1983a), pp.

“Monetizing theDebt,” this Review (December1984), pp. 30—43.

________ “The Borrowed-Reserves Operating Procedure: , “The Missing Fisher Effect on Nominal InterestTheory and Evidence,” this Review (January/February 1988), Rates in the 1 950s,” Review of Economics and Statisticspp. 30—54. (1983b), pp. 644—47,