Selecting a Monetary Indicator Evidence from the United States and Other Developed Countries

Transcription

1 Selecting Monetry Indictor Evidence from the United Sttes nd Other Developed Countries by MICHAEL W- KERAN I HERE IS long trdition of comrdeship between centrl bnkers of different countries. This is due not only to the similrity of professionl bckgrounds of centrl bnkers but, perhps more importntly, to the similrity of monetry tools nd technicl problems in the conduct of monetry policy. Most centrl bnks hve common set of monetry tools: 1 (1) the discount rte, or the price t which it lons reserves to the bnking system; () open mrket opertions nd window guidnce, or the quntity of direct reserves it provides; nd (3) reserve requirements, or the mount of reserves tht the bnking system is required to hold s idle blnces nd therefore cnnot use for lons nd investments. Which of these tools wifi be dominnt depends upon the institutionl nd finncil conditions of ech country. In the United Sttes, with its well-developed short-term finncil mrkets, the primry monetry tool is Federl Reserve open mrket opertions. In Germny, where the short-term money mrket is not An erlier version of this pper ws presented on June 1, 197, in Seoul, Kore, on the occsion of the twentieth nniversry of the Bnk of Kore. The uthor gives specil thnks to Professors Krl Brunner nd Alln Meltzer, nd to his collegues Leonll Andersen, Christopher Bbb nd Keith Crison, for helpful comments. These monetry tools represent indirect controls of the centrl bnk on the bnking system, becuse they constrin only the totl blnce sheet of the bnking system, nd the bnks re free to djust the individul components of their portfolio. The centrl bnk my lso hve monetry tools which directly ffect specific sectors of the bnking system s blnce sheet. Interest rte ceilings on time deposits constrin segment of the bnking system s libilities; quntittive limits on the mount of business lons restrict component of bnks ssets. The discussion with respect to the indictor question pplies to both direct nd indirect centrl bnk tools. well developed nd wheie lrge reserve m)ections come from blnce of pyments surpluses the pri m-iry monetry tool is chnges in reserve require ments (Mrndestreservepolitrk) In Jpn, where com mercil bnks re in lrge nd eontmuous debt to the centrl bnk the prmmry monetry tool is rtioning centrl bnk credit through the discount window (Mdoguchi Shzdo) In Kore the pnmry tools re reserve requirements nd rtioning t the discount window (C/in gu Kyu ~rn) Once monetry policy is determined nd the monetry tools ctivted, the next question centrl bnkers fce is re net monetry influences on the economy moving in line with policy? In world of uncertinty, this question cn only be nswered in the context of properly specified indictor of monetry influence on the economy. This rticle will (1) briefly discuss the need for n indictor nd the method of testing lterntive indictors; () develop the criteri of good indictor; (3) present sttisticl evidence regrding which indictor hs given the most consistently correct informtion for vrious periods of Americn history nd for recent experience of other developed countries; nd (4) consider the generl fctors which would mke one indictor superior to nother An indictor is defined here to be some redily observble economic time series which cn be used to scle monetry or fiscl influences on economic ctivity. If the indictor shows n increse, we wnt to be ble to sy with some confidence tht monetry or fiscl influences re esier or tighter, depending on wht sign the indictor is postulted to hve. Pge 8

2 FEDERAL RESERVE SANK OF ST LOUtS SEPTEMBER 197 In world of perfect knowledge of the finncil nd economic interreltionships in n economy, one would not need n indictor of monetry influence. Given prticulr monetry policy gol, such s to restrict totl demnd, the polieymker could directly link the mnipultion of his monetry tool (open mrket opertions, reserve requirements, or the discount rte) to desired chnge in totl demnd. This would be possible becuse, with perfect knowledge of the reltionships in the economy, the policymker would know the exct linkge between his mnipultion of the monetry tool nd its consequence with respect to totl demnd. Unfortuntely, we do not hve perfect knowledge bout the links between monetry tools nd finncil mrkets or between finncil mrkets nd rel mrkets. We know reltively little bout the trnsmission mechnism between centrl bnk ctions nd the finl effect on the economy. This uncertinty is not only due to lck of sttisticl dt, since it exists in ll countries irrespective of whether they hve strong or wek sttisticl gthering services. An exmple my help illustrte the problem of uncertinty in the implementtion of monetry policy. Suppose the United Sttes wishes to follow restrictive monetry policy. To do this the Federl Reserve my rise the discount rte, rise reserve requirements, or sell Government securities on the open mrket. However, ny of these movements in the monetry tools my not by themselves led to tight monetry influences on the economy. A rise in the discount rte my not rise the reltive price of centrl bnk credit if, becuse of n increse in the demnd for credit, money mrket interest rtes rise by s much s, or more thn, the rise in the discount rte. An increse in reserve requirements designed to impound reserves my be offset by n increse in Federl Reserve flot, becuse of rise in bnk trnsctions. The reserves lost by the bnking system through Federl Reserve selling of Government securities my be neutrlized by gold inflow. Some of these neutrlizing influences cn be ccounted for nd offset by the centrl bnk. However, given the current stte of knowledge bout economic reltionships, mny other fctors which could neutrlize Federl Reserve ctions re not known. The centrl bnk needs summry indictor of net monetry influences on the economy s check ginst whether the mnipultion of its monetry tools is chieving the previously estblished gols. By observing the movement of the indictor, the centrl bnk should be ble to determine whether monetry influences re expnsionry, contrctionry, or neutrl. If the indictor shows monetry influences re expnsionry, nd policy clls for contrction, then the monetry tools cn be mnipulted in more eontrctioniy wy. If the monetry indictor is moving in the snie direction s tht clled for by policy, then the monetry tools need not be mnipulted s vigorously s in the previous cse. inetnoci c-f 4 cst.tn.g I%O JUIIWOIOY 44 The indictor problem cn be considered either in the context of lrge structurl model or in the context of single eqution, reduced form pproch. The single eqution pproch wifi be used here The single eqution pproch to the indictor issue hs number of virtues. First, it includes most of the monetry nd fiscl vribles tht re components of the economic theories developed in most textbooks, nd which re used in the estimtions of most structurl econometric models. Generlly, it is these monetry nd fiscl vribles which re, within the frmework of these lrge models, the dominnt fctors influencing economic ctivity. Thus, if the monetry nd fiscl vribles re properly specified, the single eqution pproch will include the generlly recognized mjor fctors in economic stbiliztion. Second, there is considerble degree of uncertinty, given our lck of knowledge bout the economic world, s to the mjor chnnels by which these monetry nd fiscl vribles influence the economy. In consequence, it is useful reserch strtegy to consider these issues by employing the single eqution pproch where the trnsmission mechnism is not tm In the cse of lrge structurl model, theory is stted bout the interction of decision-mking units in the economy. Such theory would, nturlly, include informtion bout how monetry nd fiscl policy tools ffect economic ctivity. The monetry or fiscl indictor would be implicit in the hypothesized structure of the economy nd, by stndrd theoreticl nlysis, could be mde explicit. Different inchctors could be derived nlyticlly from lterntive theones bout the structure of the economy. If we re not certin which of the hypothesized economic structures js true then even if we hve the optiml indictor for ech structurl theory, we do not necessrily hve the tnse indictor of monetry or fiscl influences, For n exmple of nlyticlly deriving monetry indictors from number of structurl econometric models, see Richrd Zecher, An Evlution of Four Econometric Models of the Finncil Sector, Disserttion Series No. 1, Federl Reserve Bnk of Clevelnd Economic Ppers (Jnury 197). Pge 9

3 FEDERAL RESERVE BANK OF ST LOUIS SEPTEMBER 197 H specified. 3 Third, this pproch is consistent with wide rnge of theories (hypotheses) bout the structurl interreltions in the economy. The key to the single eqution pproch is the proper specifiction of the monetry nd fiscl vribles. On the fiscl side, there is generl consensus tht some mesure of chnges in government spending nd tx rtes trnsmits importnt fiscl influences. On the monetry side, there is controversy s to the pproprite mesure of monetry influences. Some economic theorists nd model-builders use vrious mrket interest rtes s mesure of monetry influences; others use vrious monetry ggregtes. To help resolve which clss of mesures provides the better indictor of monetry influences, sttisticl test is employed, For this test, representtive of ech clss of indictors is selected; long-term interest rte nd the nrrowly defined money stock. The test would llow us to ssert one of three propositions: (1) the money stock is superior to long-term interest rtes s n indictor; () long-term interest rtes re superior to the money stock s n indictor; or (3) neither the money stock nor interest rtes re clerly superior s n indictor. GritorG t Snorting i-sir I o.o/wn:to r There re no generlly ccepted criteri of good indictor with the single eqution pproch. Three criteri re suggested here \vhich re plusible, but not necessrily exhustive: (1) to be useful s guide to centrl bnk policy implementtion, n indictor should be responsive to the monetry tools of the centrl bnk; () in order to interpret movements in the indictor s expnsioniy or contrctionry, it should hve theoreticlly unmbiguous ssocition (or sign) with totl demnd; (3) to be of prcticl use to centrl bnkers, it should hve high degree of sttisticl ssocition (with the theoreticlly expected sign) with totl demnd. If the indictor chnges in vlue tody, we wnt to be ble to predict with some degree of confidence wht will hppen to totl demnd in the future. How do the money stock nd interest rtes compre with the criteri of good indictor? With respect to the first criterion, the centrl bnk s bility to sub- ~This, of course, would only give first pproximtion mesurement of impct, which could lter be reimed when we hve greter confidence in the structurl models. Indeed, the results of the single eqution estimtes could help guide structurl model-builders in the most fruitful direction. stntilly ffect interest rtes or the money stock is widely ccepted mong economists. This is bsed on the generl proposition tht becuse centrl bnk hs, in effect, unlimited finncil resources, it cn determine the vlue of ny finncil vrible, including interest rtes or the money stock (but not both simultneously). There hs been reltively limited mount of empiricl work directed to the question of responsiveness of monetry indictors to centrl bnk tools, but wht hs been done supports this generl proposition. 4 It is lso not hrd to find theoreticl justifiction for the role of both interest rtes nd the money stock s n importnt element in the trnsmission of centrl bnk ctions to the rest of the economy. Both the Keynesin Income-Expenditure Theory nd the Modern Quntity Theory of Money plce money nd interest rtes in strtegic roles. 5 These two theories differ substntilly with respect to how money nd interest rtes operte on the economy, but do not differ on the proposition tht both vribles re importnt. In the Keynesin theory, the money stock is positively ssocited nd interest rtes re negtively ssocited svith economic ctivity. In the Quntity theory, the money stock is lso positively ssocited with economic ctivity; however, the interest rte link to economic ctivity is mbiguous, becuse the link between money nd interest rtes is negtive in the short run but it could be positive in the long run. Both interest rtes nd the money stock pss the first two tests of good indictor, which leves the third criterion for differentiting between money nd interest rtes. Which of these two vribles hs been observed to hve the closest sttisticl ssocition (with the expected sign) with economic ctivity? 4 See A. Burger, An Explntion oft/se Money Supply Process, \Vdsworth Publishing Compny (forthcoming); Kern nd Bbb, An Explntion of Federl Reserve Behvior (1933-8), this Review (July 199); Alln Meltxer, Controlling Money, this Review (My 199); John Wood, A Model of Federl Reserve Behvior, Stff Economic Study No. 17, Bord of Governors of the Federl Reserve System; nd Zecher, An Evlution of Four Econometric Models of the Finncil Sector; C. Kufmn, Indictors of Monetry Policy, Ntionl Bnking Review, June 197, 5 Until recently, most econometric models long Keynesin lines hve ignored the explicit role of money. However, more recent wurk, specificlly the MIT-FRB model, hs included monetry ggregtes. Keynesin economic theory is comptible with either monetry or interest rte mesure of centrl bnk ctions. The quntity theory of money lso trets interest rtes s the strtegic price vrible which trnsmits monetry influences to the rest of the economy. See Milton Friedmn The Quntity Theory of Money A Resttement, in Studies in the Quntity Theory of Money, (Chicgo: University of Chicgo Press, 195), pp Pge 1

4 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, 197 St:ti.stion.i los-to of Altonzot.iv Indir trs A number of recent studies in this Review hve mesured the reltive impct of monetry nd fiscl influences on economic ctivity in the United Sttes nd in other developed countries. The single eqution tests were of the following generl form: where AY=o+cnL,M+ cx,af+e Y is mesure of economic ctivity (totl demnd) M is mesure of monetry influence F is mesure of fiscl influence A is qurterly chnge The symbol 5 stnds for the coefficient relting monetry influences to economic ctivity. The symbol is the coefficient relting fiscl influences to economic ctivity. The symbol represents the coefficient for the trend vlue of ll other influences on economic ctivity. The symbol e represents the error term or nontrend vlues of ll other influences on economic ctivity. These erlier studies found this single eqution pproch useful, s first pproximtion, in mesuring monetry nd fiscl influences on totl demnd. This sme single eqution pproch is used here to test lterntive monetry indictors. One difference from erlier studies is tht lterntive monetry indictors must be estimted in seprte equtions, becuse they re conceptully mesuring different spects of the sme pheuomnenon.~ All vribles re mesured s qurterly differences or chnges from one qurter to the next. The dt re drwn from fifty yers of Americn history (1919/Il to 199/IV), divided into totl nd five sub-periods, nd the postwr periods of five other developed countries: Cnd, Germny, Jpn, South Afric, nd the United Kingdom. 8 All equtions were estimted using the Almon distributed lg technique (see Appendix for discussion). The rtionle for this pproch to empiricl estimtion hs been discussed before nd will not be repeted here. The interested reder is referred to Andersen nd Jordn, Monetry nd Fiscl Actions: A Test of Their Reltive Importnce in Economic Stbiliztion, this Review (Novensber 198); DeLeeuw nd Klchbrenner, Comment, this Review (April 199), nd Kern, Monetry nd Fiscl Influences on Economic Activity The Historicl Evidence, this Review (November 199). 7 Fiscl indictors re included in the sttisticl estimtions, but re not considered explicitly in the text which is concerned only with monetry indictors. If the fiscl vribles hd not been included, the estimted coefficients of the monetry vribles could hve been bised or their sttisticl significnce over- or tinder-stted. 5 Detiled description of dt nd sources is given in the Appendix. For ech country, nd for ech period of Americn history, three tests were performed, nd the results re summrized in Tbles I, II, nd III of the Appendix. The first test consisted of regressing chnges in economic ctivity ginst chnges in Government spending, the Government tx rte, nd the money stock. Government expenditures nd tx rtes re the indi~ ctors of fiscl influence, nd the money stock is the indictor of monetry influence. The second test ws identicl to the first test, except tht chnges in longterm interest rtes were substituted for the money stock s the indictor of monetry influence. In the third test, the level of long-term interest rtes ws used for the monetry indictor.bo Severl interesting observtions could be mde on the bsis of these sttisticl results. However, just one question \vill be considered whether the money stock or interest rtes is more relible indictor of monetry influence. According to the discussion in the previous section, the monetry vrible which is most consistent in predicting future movements in economic ctivity is superior indictor. Predictble ssocition of one of number of independent vribles with respect to the dependent vribles is mesured by the t sttistic. A t sttistic of 1.9 or lrger for coefficient is considered sttisticlly significnt within the conventionl 95 per cent confidence intervls. A t sttistic of less thn 1.9 is not considered sttisticlly significnt. The higher the t sttistic, the greter confidence one hs tht the estimted coefficient is dr\vn from the sme universe s the true coefficient. To fcilitte comprisons of the t vlues for the monetry sum coefficients in the Appendix, they hve been grouped into Tble I. Of the eleven test periods six from the United Sttes nd five from other countries only in two Totl tx receipts re function of both the level of income nd the verge tx rte estblished by the Government. Only the tx rte cn be considered policy vrible, becuse chnges in tx receipts due to chnges in GNP re not directly controllble by Government ction. To tke ccount of this considertion, the tx vrible in this study is coinputed s n verge tx rte nn ll sources of income s follows Ats = where Tx is totl receipts nd 1 is nominl GNP. The chnge in the tx rte is scled by the level of (Y) to convert it into billions of dollrs equivlent. i ithere re two exceptions in the use of long-tenn rtes, Jpn nd South Afric. A short-term rte ws used for Jpn becuse the long-term rtes re subject to informl interest rte ceilings imposed by the government. A shortterm rte ws used for South Afric becuse no suitble long-term rte ws vilble. Pge 11

5 FEDERAL RESERVE BANK OF ST LOUIS SEPTEMBER, 197 ii)! t VALIJ $OP ALTERNA lye MONETARY INDICATORS Mo lory lnflvenc Moo ur*d by CI, ge Irs [eve Of Mon y Stock Interest Rote, Interest Stes Expec ccl rg P*srhve ~+ ~ Wegotite I Negohve 1 3 EstrrnGted Ssgrn United totes s Cctnedo Germ ny Jopon 145$ South Afr co tirul S ISt ICingdIem periods did chnges in interest rtes (column ) hvc sttisticlly significnt negtive vlue (U.S nd Cnd). In the other nine periods, the chnge in the interest rte coefficient ws significntly positive in one period (U.S ), nd sttisticlly insignificnt in the eight other periods. Clerly, chnges in interest rtes do not give systemtic or consistent indiction of monetry influences on economic ctivity nd thus re not relible indictor. The sttisticl test ws lso performed using levels of interest rtes nd first differences for the other vribles (column 3). These results re less stisfctory thn using chnges in interest rtes. Of the eleven test periods, only one hd sttisticlly significnt nd negtive coefficient. Tht result occurred for the United Sttes in Of the other ten cses considered, four re sttisticlly significnt but of the wrong sign (positive), nd six re sttisticlly insignificnt. The money stock, on the other hnd, hd positive reltionship with economic ctivity in ll eleven periods nd ws sttisticlly significnt in ll but one period, World \Vr II (U.S ). In spite of the wide diversity of institutions nd economic circumstnces represented in the different time periods nd different countries, chnges in the money stock hve lmost lwys led to predictble chnge in economic ctivity in the direction consistent with economic theory. With respect to the propositions considered on pge 1, the one which is most consistent with the evidence just presented is (1) the money stock is superior to long-term interest rtes s n indictor of monetry influence. It is possible tht different pir of monetry indictors would not hve supported the superiority of monetry ggregte over n interest rte mesure. However, such result is not likely, becuse most monetry ggregtes move in line with the money stock, nd most interest rtes move in line with the long-term bond rte. In this type of test, it is unnecessry for the mgnitudes of the movements to be similr. A second test of lterntive monetry indictors consists of looking t the verge qurter-by-qurter pttern of their ssocition with economic ctivity, in contrst with their totl (sum) ssocition with economic ctivity (Tble I). The chrts on the next pge present the results of such test for chnges in money nd chnges in interest rtes, where ech chrt cn be thought of s representing the pttern of sttisticlly estimted coefficients relting chnges in money (the solid line) nd chnges in interest rtes (the dotted line) to chnges in economic ctivity. Becuse the money stock nd interest rtes re mesured in different dimensions, the estimted coefficients hve been multiphed by the rtio of the stndrd devition of the independent nd dependent vribles, so tht the coefficients cn be compred directly. When the estimted coefficients re thus modified, they re referred to s Bet coefficients.~ The pttern of the Bet coefficients for the money vrible (AM) is very similr for ll periods nd countries represented. The coefficients hve consistently positive vlue through most of the tune periods, If there re ny negtive coefficients on the money vrible, they pper in the longest lg time period, usully in excess of t-4. The only exception to this stndrd pttern is the United Kingdom, where there is one virtully zero vlue of the AM coefficient in the middle of pttern of positive coefficients. The Bet coefficients for chnges in interest rtes (AR) lso hve degree of consistency. However, it is not the kind of consistency which increses poiicymkers confidence in interest rtes s n indictor. In ll but one cse, chnges in interest rtes show n initil positive ssocition with economic ctivity which only grdully diminishes nd becomes negtive ssocition fter three to five lgged qurters. The interest rte coefficient hs the theoreticlly expected negtive ssocition with economic ctivity consistently only in the cse of the United Sttes from lithe results for the Wr nd immedite Postwr periods for the United Sttes nd South Afric re omitted from the chrt, becuse of spce limittions. The pttern of the Bet coefficients for the omitted periods is quite similr to tht of the included periods. Pge 1

6 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER. 197 Bet Coefficients of Alterntive Monetry Indictors First Differences United Sttes - ~: ~sj~7~---- V U - / 4 - I I I I I I I I I I t+ t )tt3t4t t-17r$t9 +41 tlit +5 I I I I I AR _ A~ I tfit tltt -4-5 CANADA t+ 1 t t 3 t-4t5 Other eveloped Countnes GERMANY ll 9_ : ~ // 4 I I I I I ii t t +4 5 t4 +7 1$ 19 tic JAPAN I I I I I t 1 t +1 3 t4 S $4 t. UNWED kingdom _ ; HT11T1TH~~ ~IIiII t4lt tlttst tatsttttts Not Bet coeff crents re Ic c cog so the money tck ~&M~nd interest roes tar) h so ~ero ef i rots e ot imed o he products of the tog es son oefficsents or the e p two vorsoble nd the r so of the tondrd de to so o the sndepe d yors We tothe stndrd de solo ci hod pend t V sble AYI r~8gnptlog w to sole ted h bps the mrnscnum stndrd error of thee I mote od ed for de e sl freedom 15 Pge 13

7 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, to In the Cndin cse, interest rtes hve smll positive influence in the initil qurter, nd consistently negtive influences in ll subsequent qurters. It is not surprising tht these two cses re lso the only ones where the sum coefficients were sttisticlly significnt nd negtive, s described in Tble I, column. In the other six cses in the chrt, the pttern of the interest rte coefficients is such s to virtully wsh out ny consistent effect on economic ctivity. The erly positive influences re mtched by the lter negtive influences. This is lso consistent with Tble I, column, where the vlue of the t sttistic indicted tht these sme six cses hd sttisticlly insignificnt sum coefficients. The results presented in Tble I nd the chrt re highly consistent with ech other, nd provide strikingly strong cse tht monetry influences, mesured by chnges in the money stock, hve more predictble nd uniform pttern of effect on economic ctivity thn monetry influences, mesured by chnges in long-term interest rtes. Wht do these results imply for the monetry p 1 - icymker? If he desires to minimize his errors in predicting the effects of his ctions on the economy, he wifi use the money stock s n indictor of monetry influence. This selection is not dependent on his cceptnce of Quntity Theory view of the trnsmission mechnism. It is eqully consistent with Keynesin view of the trnsmission mechnism which lso postultes positive ssocition of money with economic ctivity. Rther, the selection is bsed on the empiricl observtion tht interest rtes hve proven to be misleding indictor in most periods, while the money stock hs proven to be n ccurte indictor in virtully ll periods. 1 Why is Mnes, Snperi.or to Interest Rtes s n Indictor? The empiricl results just discussed should not be interpreted s denying the centrl role of interest rtes in trnsmitting monetry influences to the rest of the economy. The lrge body of theoreticl literture on the prmount role of interest rtes is not in dispute. Most monetrists cknowledge the role of interest rtes in the trnsmission mechnism. 1 For nother study long similr lines see M. hmburger Indictors of Monetry Policy The Arguments sid the Evidence, The Americn Economic Review, My 197, nd M. Willms, An Evlution of Monetry Indictors in Germny, in K. Brunner, ed,, Proceedings of the Pint Europen Conference on Monetry Theory nd Monetry Policy, forthcoming. In world of perfect knowledge bout the finncil nd economic structure, both the money stock nd interest rtes would give identicl informtion bout monetry influences on the economy. 13 The indictor problem rises becuse there is ignornce t the empiricl level bout exct specifiction of the linkges of monetry nd other vribles in the economy nd the time lgs ssocited with them. The evidence which ws considered bove suggests tht the money stock hs n overwhelmingly more predictble ssocition with economic ctivity thn interest rtes. Knowledge is one of our scrcest resources, nd it pprently tkes less knowledge to properly evlute the impct of the money stock thn the impct of interest rtes. Conversely, to see the workings of interest rtes it tkes more knowledge of the workings of the economy thn we currently hve. There re number of possible resons for this stte of ffirs: 1. Difference between theoreticl nd ctul mesures. The rnge of interest rtes which re theoreticlly relevnt in indicting monetry influence on economic ctivity is much broder thn tht vilble in the published interest rte series. The trnsmission of monetry impulses to the rest of the economy opertes through chnging prices of wide rnge of ssets nd libilities, which is equivlent to chnges in their ssocited interest rtes. The vlue of finncil ssets reflected in the yield on ny one type of bond my be too nrrow to represent the wide spectrum of ssets nd libilities represented in the blnce sheets of households nd firms which trnsmit rnonetiy influences. The mesured money stock, on the other hnd, is much more complete enumertion of the liquidity position of ll households nd firms. Only commercil bnk demnd deposits nd currency issued by the centrl bnk nd Government cn perform the role of medium of exchnge. Even other finncil institutions must hold their working blnces s demnd deposits in commercil bnk. Therefore, the observed money stock series comes closer to theoreticl mesure thn the observed interest rte.. Difference between rel nd nominl vlues. It is generlly sserted tht it is chnges in rel interest rtes which ffect economic ctivity, but only chnges in nominl interest rtes re ctully mesured nd reported. The difference between rel nd nominl interest rtes is the result of the chnge in prices which is expected to occur between now nd the mturity of the finncil in- 13 See Krl J3runner nd Alln Meltzer The Nture of the Policy Problem in Trgets nd Indictors of Monetry Policy, (Sn Frncisco: Ghndler Publishing Go., 199). Pge 14

8 FEDERAL RESERVE BANK OF ST LOUIS SEPTEMBER, 197 strument. t Mesurement of these expected price chnges is both conceptully nd empiriclly difficult process, subject to mny errors. If nominl interest rtes re rising becuse of expected infltion in the future, the rel interest rte my ctully be unchnged or flling. Thus, to evlute monetry ctions in period of infltion or defltion by looking t nominl interest rtes my be misleding. This problem does not rise with mesurements of the money stock, becuse in its most generlly used form it is nominl vlues of money which influence nominl vlues of economic ctivity. 3. Confusion between supply nd demnd. Even if one could mesure rel interest rtes, the chnge in interest rtes my be due to chnge in the demnd for credit rther thn to chnge in the supply of credit, engineered by the centrl bnk. In period of economic expnsion, the demnd for credit increses, which pushes interest rtes up. In period of economic decline, there is typiclly reduction in the demnd for credit, which pushes interest rtes down. Such movements in interest rtes re not the result of centrl bnk ction but of feedbck from the rest of the economy. Yet, if interest rtes re used s n indictor of monetry influence, it would pper s if the centrl bnk hs tken countercyclicl ctions when, in fct, it my hve tken no ction t ll. This problem is not s serious when the money stock is used s n indictor. Most studies on the determinnts of the supply of money led to the conclusion tht centrl bnk opertions dominte the money stock nd tend to offset demnd-induced chnges in the money stock.m In other words, the behvior of the public, cting on the demnd side of the mrket, does not bis the money stock s n indictor of monetry influence s much s it does interest rtes. tt The difference between rel nd nominl interest rtes cn be presented s follows: r = r p, where r is the rel interest rte, r is the nominl interest rte, nd ~ is the rte of chnge in expected prices of goods nd services over the life of the finncil ssets. If price expecttions re formed very slowly, then the gp between rel nd nominl interest rtes will be smll. Until quite recently, this ws the generlly held position mong economists. However, Yohe nd Krosky (this Review December 199) hve developed new evidence which indictes tht price expecttions re formed quite rpidly, thereby creting substntil gp between rel nd nominl interest rtes even during reltively short periods of infltion nd defltion. 15 See John Wood, A Model of Federl Reserve Behvior, Stff Economic Studies No. 17, Bord of Governors, 198. Also, An Explntion of Federl Reserve Actions, this Review, July 199. Reply to Gomments on the St. Louis Position, August 199, nd Gomment, My Greter stbility in the demnd for money thn in the demnd for commodities. If our current stte of knowledge llows us to more ccurtely predict the demnd for money thn the demnd for goods nd services, then the money stock will he more closely relted to economic ctivity thn interest rtes in ny sttisticl nlysis. 1 This point cn be illustrted in stndrd Keynesin LM-IS frmework, s in Figure I. R Figure I The Demnd for Money (LMO) nd Commodities (IS-IS ) II I ~ ill I Y l~y 3 Y 1 I is 15 If the demnd for money is well specified, then the locus of points representing the LM curve cn be described by line (LM ). If the demnd for commodities, however, hs lrge rndom (stochstic) element, the IS curve cn be described only s bnd, the dimensions of which re IS IS. In this circumstnce, the link between ny interest rte R nd income would be represented by the gp Y - Y 1. On the other hnd, the reltionship between ny given money stock M (which is implied by given LM curve) nd income would be represented by the bnd - Y,. Becuse the spred between Y nd Y 1 is greter thn the spred between Y, nd Y,, the degree of sttisticl ssocition between chnges in R nd chnges in Y would be less thn between chnges in M nd chnges in Y t The rtionle for the greter stbility for the demnd for money thn the demod for commodities is presented by Willim Poole in Optiml Ghoice of Monetry Policy Instruments in Simple Stochstic Mcro-Model. Qurterly Journl of Economies, My 197. Y Pge 15

9 FEDERAL RESERVE BANK OF ST. LOUIS SEPTEMBER, Government controls. Governments historiclly hve imposed ceihngs on interest rtes. When such ceilings exist, interest rtes cnnot be used simultneously s n indictor of monetry influence on the economy. An indictor tht is not llowed to move with chnges in mrket forces cu give misleding nd wrong informtion. This point pplies only to the use of legl uthority to control interest rtes by flt. The use of stndrd monetry tools to control interest rtes does not, of course, weken its role s n indictor. Usully when n interest rte is used to mesure monetry influence, it is selected from mong those which re not under direct government constrints. For exmple, the corporte A bond rte, which is used in the sttisticl tests on the United Sttes, hs lwys been free of legl constrints. However, when the government controls one interest rte, like tht which bnks cn py on time deposits (Regultion Q), credit flows wy from bnks nd into other finncil mrkets in which the rtes re uncontrolled. These distortions in credit flo\vs could distort the interest rte quoted in those mrkets s n indictor of monetry influence. It is, of course, possible tht interest rte controls on time deposits could distort the money stock, especilly when money is defined to include time deposits. However, the money stock definition used here includes only demnd deposits nd currency, nd therefore the distorting effects of controls re pt to be minimized. The min point of this rticle is tht selection of n indictor of centrl bnk ctions need not be mde only on theoreticl grounds. If we re not certin of the theoreticl structure of the economy, the selection of the indictor cn lso be mde on empiricl grounds. We hve observed, in wide rnge of historicl nd institutionl contexts, tht the money stock is rchble nd predictble indictor of monetry influence, nd tht interest rtes re not. The resons for this difference in results stem lrgely from the fct tht it pprently tkes more knowledge bout the workings of the economic system to evlute the finpct of interest rtes thn to evlute the impct of the money stock. There re t lest five possible fctors responsible for this: (1) the reported interest rtes do not cover ll the finncil mrkets which trnsmit monetry influences to the rest of the economy; () the dt reported re of nominl interest rtes, while it is rel interest rtes which ffect economic ctivity; (3) it is difficult to distinguish chnges in interest rtes which re induced by demnd pressures of the public from those cused by centrl bnk ctions; (4) uncertinty bout the demnd for commodities reltive to the demnd for money increses the uncertinty of the reltion of interest rtes to economic ctivity; (5) Government interest rte ceilings in some mrkets induce rbitrge flows which distort the movements of interest rtes in other mrkets. We cn summrize these fctors by sying tht they represent the greter degree of knowledge we must hve bout the economic system to mke interest rtes successful indictor of monetry influence on economic ctivity. This does not imply tht the money stock is not subject to some of the sme uncertinties s those ttched to interest rtes. Rther, the sttisticl results suggest tht the uncertinties re less with the money stock thn with interest rtes. This rticle is vilble s Reprint No. 59 The sttisticl ppendix begins on the next pge. Pge 1

10 STATISTICAL APPENDIX The following tbles summrize the regression results which re the bsis for the ssertioos in the text. The only spect of these results \vhich is discussed in the text re the t vlues on the lterntive monetry vribles. There re other implictions which cn be drwn from these results. Specificlly, the fiscl vribles ply stronger role nd hve greter sttisticl significnce when n interest rte, rther thn the money stock, is used s the monetry vrible. This result is not surprising. Omitting money from the eqution llows the Government deficit to be finnced by increses in the money stock rther thn just through increses in debt sles to the public. Thus, following the nlysis of Fnd (this Review, Jnury 197), one would expect stronger mesured fiscl influence when interest rtes re the monetry vrible, nd wek fiscl influence when the money stock is the monetry vrible, This point nd others will be developed in future rticle. The Almon lg technique ws used to estimte ll equtions presented below. By constrining the distribution of coefficients to fit polynomil curve of n degree, it is designed to void the bis in estimting distributed lg coefficients which my rise from multicollinerity in the lg vlues of the independent vribles, The theoreticl justifiction for this procedure is tht the Almon constrined estimte is superior to the unconstrined estimte, becuse it will crete distribution of coefficients which more closely pproximtes the distribution derived from smple of infinite size. In order to minimize the severity of the Almon constrint, the mximum degree of the polynomil ws used in ech cse. The mximum degree is equl to the number of lgs Tubic I UNITED STATES Monetry nd Fiscl Influences on Economic Activity iuorter 3 y Frst Diftc-. ericp; Bill or of DoIicrbl, cy: M. AE.-.. Cnstnrt Moro~or Fiscl Pui,od tr.s Term ir IIL- r.cn InFli.encec - DV? CI,, _ i. ~Sun11 IS. ~ 5 ~1 I: I il.: V 18) 4.8) ( 43) (.1) 1.84 II ! ) 1.) I 11) 3 3.3) 19 Ill 199. II 1939 I 4 1 1/ ) (.48) I -9/I I 49) L47 HI 939-IV 194 I ) IL.9) ( 57) 11.4) / IV 195 F / ) (.) 13.33) 3.5 I 1953 III 199 t l5.3~) ( 1.51 (.15) 1.1 V Clu. Cr,. -. Atx CorsFcn Mon ~ory Fscl R - Pcrioo Lciq~ Te m Ir. lucric.o Ir f.t.r.ncc:. D-W (11 CI. (I S..m) 5i..m) (Si.m~ II I ( ).45) 1.8 II 1919 II 399 I I I.) ( ) J9 III ~99.II , ) 1 313) (1.53) I 14) 1.83 III IV 194 t / ) (1.71 (3.73; ) 1.7r I 947.IV 195 I ~ I3.9~) (.tc) (.13).9) 3 3~ II 99 F (?5~ 1311 (.381 (.47) L95 p H. i:. Oi.~.. i.,,,~.c... II. i ~ ii.-.. I I ri,:. ii,...,r..,.i i.. I,: I i \I.: :. :1 1:1... ~.,..::I:.., 1 i:...1 I i... :i, :I.~. t 1N l Li.,...~.:,;,..,.. 1.;.i.. ni:nh,. r:... I:. n. In: h I I.,~ I Pge 17

11 ~, 3- CI El;> - - P V to it ~o I - CC -h 51 -.tncd 3: -i - ~ ~ I 3 n Z,.5 ~ ~ ft ~ ~ g. - ~ ~ ~ -nt,,, tt~ rn~ ~ 8 (F: ~ -,, i~dz,.,, -nn -C ~ cc3 CC CC CO CC 3 OBt,n~ Cu-t3SQ ~ ~,~ - ~ Ez~~ fl o~o~o~s~ ~ tno- Ct ~&E ~~s~ft - 3,.,,-,,o- - ~ ~~t -~<z oq~ Oz. <- ozo - n3~t - ~ ~o~- t ~- CC ot~~ R~,- 3~- 3--~P5 tz~_ ~ - <~z - CC<~ zo ~ ~ o9 ~$~ 51 it - i-tn CD C?I ;> C- it C it, - O -CO 3;> -ll -p -1-5 ~ + I -- ~ >5 + >5 - -~~ I, O) ~!. N E. - ~ CD - 3 O C, C - z O ~z ~ so 3 +~- it - ~ -z z 5 ~ 3 z q SQ ~ ~ ~ gi z (t - -. to 4 ~ - ~ -C -I P1 P1 I D r Ill to N I D N zz 11 In r I: If) In P1 -U -4 P1 P1 ( -3 o ~

12