How Should Monetary Policy Be Conducted in an Era of Price Stability?

Transcription

1 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? Lars E.O. Svensson For several decades, high inflaion has been he main hrea o moneary sabiliy. During he 1990s, low and sable inflaion has been achieved in many counries. This may be he sar of a new era of price sabiliy. This paper discusses some issues relaed o how moneary policy should be conduced in such an era. Before we can alk abou an era of price sabiliy, we need o be clear on he definiion of price sabiliy. In Secion 2, I discuss wo definiions, low (and sable) inflaion, which implies base drif in he price level, and price-level sabiliy, which does no imply such a base drif. Mos curren discussion and moneary-policy formulaions involve he firs definiion, low inflaion. Price-level sabiliy has he advanage of reducing long-erm price-level uncerainy, which should be beneficial for long-erm planning and invesmen decisions. According o convenional wisdom, however, i has he drawback of increasing shor-erm variabiliy of inflaion and oupu, and is, herefore, no o be recommended. I argue ha his convenional wisdom is ill founded, and ha he relaive advanages of he wo kinds of price sabiliy are a raher open issue warraning furher sudy. Insead, in a decade or so, when cenral banks (hopefully) maser mainaining low and sable inflaion, he ime may be ripe for seriously considering price-level sabiliy as a goal for moneary policy. Accumulaed experience and research may hen allow a more reliable evaluaion of he relaive advanages of he wo regimes. The res of he paper is mainly 1

2 2 Lars E. O. Svensson concerned wih low inflaion, alhough some of he discussion also applies o price-level sabiliy. The res of Secion 2 discusses he choice of suiable objecives for moneary policy. This boils down o specifying a loss funcion corresponding o flexible inflaion argeing. Tha is, sabilizing inflaion around an inflaion arge, bu also puing some weigh on sabilizing he real economy by sabilizing he real oupu around poenial oupu. I also briefly discuss he choice of a suiable index and arge level for inflaion. An era of price sabiliy will, obviously, only arise if cenral banks succeed in mainaining price sabiliy. In Secion 3, I discuss how cenral banks can bes mainain price sabiliy. I discuss hree alernaive ways: (1) a commimen o a simple insrumen rule, a rule for seing he insrumen, like he Taylor (1993) Rule for he federal funds rae or he Melzer (1997) and McCallum (1988) rules for he moneary base; (2) forecas argeing, seing he insrumen such ha he cenral bank s condiional forecass for inflaion and he oupu gap approach he inflaion arge and zero, respecively; and (3) moneary argeing, where he insrumen is se such as o achieve a arge growh rae for a broad moneary aggregae, like M2 or M3. Alhough simple insrumen rules prominenly figure in curren research and policy discussions, and he research on simple insrumen rules has conribued imporan insighs, I find ha a commimen o hese rules is neiher a desirable nor a pracical way of mainaining price sabiliy. Furhermore, in line wih mos previous research, I find ha moneary argeing is an inferior way of mainaining price sabiliy. Insead, forecas argeing, which is, indeed, already praciced by successful cenral banks, seems o be he bes way of mainaining price sabiliy. 1 An era of price sabiliy will no only bring low and sable inflaion bu also low and sable inflaion expecaions, ha is, credibiliy for a low-inflaion policy. Cenral bankers ofen seem obsessed wih credibiliy. In Secion 4, I argue ha here are good reasons for his and discuss he benefis of credibiliy (in he sense of privae inflaion expecaions anchored on he inflaion arge). I show ha credibiliy makes he rade-off beween inflaion variabiliy, oupu-gap variabiliy and insrumen variabiliy more favorable, and ha i pus he economy

3 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 3 increasingly on auomaic pilo, such ha less conrol and acivism need o be exercised by he cenral bank. These circumsances make i easier for he cenral bank o mee is arges. Furhermore, credibiliy helps avoid deflaion and a liquidiy rap, he opic of he nex secion. Curren inflaion is lower han in many decades, which, ogeher wih he siuaion in Japan, has brough he poenial hrea of susained deflaion and a liquidiy rap ino focus. In Secion 5, I discuss he hrea of deflaion and a liquidiy rap. More precisely, I specify wha a liquidiy rap is and discuss how i is avoided, and how o ge ou of i if already rapped. Aliquidiy rap is a siuaion wih zero nominal ineres raes and persisen deflaion and deflaionary expecaions. Since money and nominal bonds are hen perfec subsiues, moneary policy becomes ineffecive and has no effecs on nominal and real prices and quaniies. I argue ha ransparen inflaion argeing of an announced moderaely posiive inflaion arge is he bes way of avoiding a liquidiy rap. I also argue ha cenral banks should make advance coningency plans for emergency measures o be used if a series of unforunae shocks were o push he economy close o or ino a liquidiy rap. These measures include increasingly aggressive and unorhodox open-marke operaions, as well as preparaions for coordinaed fiscal and moneary expansions. These measures also serve a role in escaping from a liquidiy rap if already rapped. Since moneary policy may be ineffecive on is own, fiscal policy, boh wih regard o a fiscal expansion and o nominal public deb managemen, is likely o have an imporan role in escaping from a liquidiy rap. Secion 6 presens some conclusions. Appendices A and B presen some echnical deails on inflaion variabiliy under inflaion argeing and price-level argeing (relaed o Secion 2.1) and on moneary and fiscal policy in a liquidiy rap (relaed o Secion 5). Defining price sabiliy 2 Before we can alk abou an era of price sabiliy, we need o be clear on how o define price sabiliy. In his secion, I discuss his defini-

4 4 Lars E. O. Svensson ion, in paricular he disincion beween price-level sabiliy and low inflaion. I also discuss how o formulae he appropriae loss funcion for a policy aiming a price sabiliy, including any concern for sabiliy of he real economy, as well as he choice of price index and arge level. Price-level sabiliy vs. low inflaion Wha is he appropriae definiion of price sabiliy? The mos obvious meaning of price sabiliy would seem o be a sable price level, price-level sabiliy. Neverheless, in mos curren discussions and formulaions of moneary policy, price sabiliy insead means a siuaion wih low and sable inflaion, low inflaion (including zero inflaion). The former definiion implies ha he price level is saionary (or a leas rend-saionary, saionary around a deerminisic rend). The laer definiion implies base drif in he price level, so ha he price level will include a uni roo and be non-(rend-)saionary. Indeed, he price-level variance increases wihou bound wih he forecas horizon. Thus, referring o low inflaion as price sabiliy is, indeed, somehing of a misnomer. Le me refer o a moneary-policy regime as price-level argeing or inflaion argeing, depending on wheher he goal is a sable price level or a low and sable inflaion rae. Noe ha if argumens in favor of a small posiive inflaion rae are acceped, an upward-sloping price-level arge pah may be preferable o a consan price-level arge, and sill achieve he desired rend-saionariy. In he real world, here are currenly an increasing number of moneary-policy regimes wih explici or implici inflaion argeing, bu no regimes wih explici or implici price-level argeing. Whereas he Gold Sandard may be inerpreed as implying implici price-level argeing, so far he only regime in hisory wih explici price-level argeing occurred in Sweden during he 1930s (see Fisher (1934) and Berg and Jonung (1999); his regime was quie successful in avoiding deflaion). Even if here are no curren examples of price-level argeing

5 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 5 regimes, price-level argeing has been subjec o an increasing ineres in he moneary policy lieraure. 3 A frequen resul, which has emerged as he convenional wisdom, is ha he choice beween price-level argeing and inflaion argeing involves a rade-off beween low-frequency price-level variabiliy on one hand and high-frequency inflaion and oupu variabiliy on he oher. Thus, price-level argeing has he advanage of a reduced long-erm variabiliy of he price level, which should be beneficial for long-erm nominal conracs and ineremporal decisions, bu would come a he cos of increased shor-erm variabiliy of inflaion and oupu. The inuiion is sraighforward: In order o sabilize he price level under price-level argeing, higher-han-average inflaion mus be succeeded by lower-han-average inflaion. This would seem o resul in higher inflaion variabiliy han under inflaion argeing, since base drif is acceped in he laer case and higher-han-average inflaion need only be succeeded by average inflaion. Via nominal rigidiies, he higher inflaion variabiliy would hen seem o resul in higher oupu variabiliy. 4 However, his inuiion may be misleadingly simple. In more realisic models of inflaion argeing and price-level argeing wih more complicaed dynamics, he relaive variabiliy of inflaion in he wo regimes becomes an open issue. As shown in Appendix A (which reproduces some resuls in he appendix of he working-paper version of Svensson [1999g]), his is he case if here is serial correlaion in he deviaion beween he arge variable and he arge level. For insance, if he price level displays mean reversion oward he price-level arge under price-level argeing and inflaion displays mean reversion oward he inflaion arge under inflaion argeing. Svensson (1999g) and Vesin (1999) give examples where he absence of a commimen mechanism implies ha he rade-off beween inflaion variabiliy and oupu variabiliy becomes more favorable under price-level argeing han under inflaion argeing. 5 For some empirical macro models (boh small and large), reacion funcions wih responses of he insrumen o price-level deviaions from a price-level arge lead o as good or beer overall performance (in erms of inflaion and oupu variances) han wih responses o inflaion deviaion from inflaion arges. 6 Ineresingly, a price-level arge may have special advanages relaive o an inflaion arge in avoiding persisen deflaion, because

6 6 Lars E. O. Svensson an unanicipaed deflaion ha makes he price level fall below he price-level arge will, if he price-level arge is credible, resul in increased inflaion expecaions. These increased inflaion expecaions will, by hemselves, reduce he real ineres rae and sabilize he economy, even if he nominal ineres rae remains unchanged. 7 I believe hese resuls show ha he relaive properies of price-level argeing and inflaion argeing are far from seled. In paricular, he poenial benefis from reduced long-erm price-level variabiliy and uncerainy are no ye well undersood. Sill, I believe ha low and sable inflaion may be a sufficienly ambiious underaking for cenral banks a presen. However, once cenral banks have masered inflaion argeing, in perhaps anoher decade, i may be ime o increase he ambiions and consider price-level argeing. By hen, research and experience may provide beer guidance abou which regime is preferable. 8 The res of he paper will refer o low inflaion, which allows base drif in he price level, raher han price-level sabiliy, which does no allow such base drif. Relucanly, I will occasionally refer o low inflaion as price sabiliy, even wihou using quoaion marks. Some of he discussion below is applicable o boh price-level sabiliy and low inflaion, hough. Specifying he loss funcion Wha is he appropriae loss funcion for a cenral bank aiming a low inflaion? As subsaniaed below, here is considerable agreemen among academics and cenral bankers ha he appropriae loss funcion boh involves sabilizing inflaion around an inflaion arge and sabilizing he real economy, represened by he oupu gap. This can be represened by a quadraic loss funcion, (2.1) [ π π *) 2 λ * 2] 1 L = ( + ( y y ), 2 where he subscrip refers o he period, π is an index of inflaion, π * is an inflaion arge, y is (log) oupu, y * is poenial oupu (so ha y y * is he oupu gap), andλ>0is he relaive weigh on ou-

7 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 7 pu-gap sabilizaion. 9 As in Svensson (1999d and 1999f), I find i pracical o refer o his loss funcion wihλ>0(or concern abou sabiliy of he real economy in general) as flexible inflaion argeing, and λ=0 as sric inflaion argeing. 10 Because curren inflaion and oupu are, in pracice, predeermined, curren moneary-policy acions can only affec fuure inflaion and oupu. Furhermore, inflaion, oupu, and, in paricular, poenial oupu are observed (or esimaed) wih measuremen errors. Then, he objecives of moneary policy need o be expressed as an ineremporal loss funcion of expeced discouned fuure losses, (2.2) E δ τ L+ τ, τ=0 where E denoes expecaions condiional on he cenral bank s in- δ 0< δ< 1) is a discoun facor. formaion in period and ( ) Whereas, here may previously have been some conroversy abou wheher inflaion argeing involves concern abou real variabiliy, represened by oupu-gap variabiliy and corresponding o he second erm in (2.1), here is now considerable agreemen in he lieraure ha his is, indeed, he case: Inflaion-argeing cenral banks are no wha King (1997) referred o as inflaion nuers. 11 As shown by Ball (1999a) and Svensson (1997a), concern abou oupu-gap sabiliy ranslaes ino a more gradualis policy (a leas in sandard simple models of he ransmission mechanism). Thus, if inflaion moves away from he inflaion arge, i is more gradually brough back o arge. Equivalenly, inflaion-argeing cenral banks lenghen heir horizon and aim a meeing he inflaion arge furher in he fuure. In conras, sric inflaion argeing would involve meeing he inflaion arge a he shores possible horizon and, hereby, generae considerable oupu-gap volailiy. 12 As furher discussed in Svensson (1999d), concerns abou oupu-gap sabiliy, simple forms of model uncerainy, and ineres rae smoohing all have similar effecs under inflaion argeing, namely a more gradualis policy. Sveriges Riksbank has expressed similar views. 13 The Chancellor s remi o Bank of England

8 8 Lars E. O. Svensson menions undesired volailiy of oupu. 14 The minues from Bank of England s Moneary Policy Commiee are also explici abou sabilizing he oupu gap. 15 Several conribuions and discussions by cenral bankers and academics in Lowe express similar views. Ball (1999b) and Svensson (1998) give examples of he gradualis approach of he Reserve Bank of New Zealand. Indeed, a quoe from he ECB (ECB Monhly Bullein, January 1999, p. 47) also gives some suppor for an inerpreaion wihλ>0(as well as some weigh on minimizing ineres rae variabiliy):... a medium-erm orienaion of moneary policy is imporan in order o permi a gradualis and measured response [o some hreas o price sabiliy]. Such a cenral bank response will no inroduce unnecessary and possibly self-susaining uncerainy ino shor-erm ineres raes or he real economy... Thus, i seems non-conroversial ha real-world inflaion argeing is acually flexible inflaion argeing wih some concern for variabiliy in he real economy, corresponding o λ>0 in (2.1). Furhermore, as discussed in some deail in Svensson (1999c), here is general agreemen ha he oupu arge under inflaion arge should equal poenial oupu, so as o avoid any average inflaion bias. 16 Wha index and which level? Which price index would be mos appropriae? Sabilizing he CPI, inerpreed as a cos-of-living index for he average consumer, should simplify he consumers economic calculaions and decisions. The CPI also has he advanage of being easily undersood, frequenly published, published by auhoriies separae from cenral banks, and very rarely revised. Ineres-relaed coss cause well-known problems wih he CPI, hough: An ineres-rae increase o lower inflaion has he perverse shor-erm effec of increasing inflaion. This presens a pedagogical problem in he cenral bank s communicaion wih he general public. To avoid his problem, Bank of England and Reserve Bank of New Zealand have inflaion arges defined in erms of CPIX (RPIX in Briain), he CPI less ineres-relaed coss. 17 The Eurosysem has

9 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 9 defined price sabiliy in erms of he HICP, which also excludes ineres coss. Furhermore, changes in indirec axes and subsidies can have considerable shor-run effecs on he CPI. Differen measures of underlying inflaion, core inflaion, ry o eliminae such effecs. Eliminaing componens over which moneary policy has lile or no influence serves o avoid misleading impressions of he degree of conrol. The disadvanage of subracing componens from he CPI is ha he remaining index becomes more remoe from wha maers o consumers and less ransparen o he general public. I may also be difficul o compue in a well-defined and ransparen way. Opinions generally differ on wha componens o deduc from he CPI. My own view is ha deducing ineres-relaed coss and using CPIX, ogeher wih ransparen explanaion of index movemens caused by changes in indirec axes and subsidies, is an appropriae compromise. Criics of inflaion argeing and proponens of moneary argeing someimes criicize inflaion argeing for being concerned abou shor-run inflaion and emporary changes in inflaion, and argue ha moneary argeing would indirecly imply argeing permanen componens of inflaion only. 18 I seems ha his issue is much beer confroned by explicily selecing he appropriae inflaion index o be argeed. Indeed, as already noed above, differen measures of underlying inflaion (as suggesed by he erm iself) are explicily designed o exclude emporary disurbances and focus on persisen componens of inflaion. Furhermore, he pracice of measuring inflaion as welve-monh moving-average inflaion raes is, by iself, a way of averaging ou some of he wider and more ransiory movemens in inflaion. Finally, forecass of inflaion, in pracice, focus more on persisen changes in inflaion and less on ransien ones, since he former are, by naure, easier o predic. 19 Wha level of he inflaion arge is appropriae? Alhough zero inflaion would seem o be a naural focal poin, all counries wih inflaion arges have seleced posiive inflaion arges. The inflaion arges (poin arges or midpoins of he arge range) vary beween 1.5 percen (per year) in New Zealand; 2 percen in Canada, Sweden, and Finland (before Finland joined he EMU); and 2.5 percen in he Unied Kingdom and Ausralia (he Reserve Bank of Ausralia has a

10 10 Lars E. O. Svensson arge range of 2 o 3 percen for average inflaion over an unspecified business cycle). The Bundesbank had a 2 percen inflaion arge for many years (called unavoidable inflaion, price norm, or mediumerm price assumpion ). During 1997 and 1998, i was lowered o 1.5 o 2 percen (which could perhaps be ranslaed ino a poin inflaion arge of 1.75 percen). The Eurosysem has announced annual increases in he HICP below 2 percen as is definiion of price sabiliy, which has been inerpreed as he ranges 0 o 2 percen or 1 o 2 percen; he Eurosysem appeared o have used a poin inflaion arge of 1.5 percen in consrucing is reference value for money growh. 20 Tha he inflaion arge exceeds zero can be moivaed by measuremen bias, non-negaive nominal ineres raes and possible downward nominal price and wage rigidiies. 21 Two percen is he borderline in Akerlof, Dickens, and Perry (1996), who sudy he effecs of he downward rigidiy of nominal wages. One percen is he borderline in Orphanides and Wieland (1998), who examine he consequences of non-negaive nominal ineres raes. These sudies indicae ha inflaion arges below hose borderlines risk reducing average oupu or increasing average unemploymen. 22, 23 Alogeher, I believe ha i may be more imporan o announce an explici inflaion arge (a poin arge or a range) han wheher he arge (he midpoin of he range) is 1.5, 2, or 2.5 percen. Asymmeric inflaion arge implies ha inflaion below he arge is considered o be equally bad as he same disance above he arge (which is he case if inflaion argeing is represened by a symmeric loss funcion like (2.1)). An asymmeric inflaion arge may induce an upward would, from his poin of view, be beer han jus a range. As argued in Secion 4, here should be subsanial benefis from inflaion expecaions sabilizing a he inflaion arge. A poin inflaion arge hen gives a well-defined focal poin for inflaion expecaions ha ener ino wage seing and oher conracs. Wih a range of 0 o 2 percen, for insance, here is a large difference beween inflaion expecaions sabilizing a 0 percen or 2 percen. Neverheless, a symmeric range would, in urn, be beer han an asymmeric formulaion like he Eurosysem s below 2 percen. As furher discussed in Secion 5, hese aspecs may be paricularly imporan in order o avoid a liquidiy rap wih persisen deflaion a possibiliy ha seems less remoe

11 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 11 given recen developmens in Japan. A symmeric and small posiive inflaion arge would seem o be he bes defense agains a liquidiy rap wih persisen deflaion and deflaionary expecaions. Mainaining price sabiliy 24 An era of price sabiliy will obviously no maerialize unless cenral banks succeed in mainaining price sabiliy. The cenral bank hen needs o minimize expeced fuure discouned losses, (2.2), consising of a weighed sum of squared deviaions of inflaion from he inflaion variables and squared deviaions of oupu from he oupu gap, (2.1). Then, inflaion and he oupu gap are he cenral bank s arge variables, ha is, he variables enering he cenral bank s loss funcion. The cenral bank s conrol variable, is insrumen, is, in pracice, a shor ineres rae like he federal funds rae in he Unied Saes. 25 The cenral bank s ask is hen o se is insrumen so as o bes minimize he ineremporal loss funcion. To achieve his, he cenral bank mus have a view of he ransmission mechanism of moneary policy, ha is, how he insrumen and he curren sae of he economy affec he fuure pah of he arge variables. In his secion, I will discuss he cenral bank s framework for policy decisions, he principles for he seing of is insrumen. I will discuss hree alernaive decision frameworks or principles for seing he insrumen, namely commimen o a simple insrumen rule in subsecion 3.1, forecas argeing in subsecion 3.2, and moneary argeing in subsecion 3.3. Commimen o a simple insrumen rule A large par of he moneary-policy lieraure uses he concep of rules in he narrow sense of a prescribed reacion funcion for moneary policy. Insead, as in Svensson (1999c), I find i helpful o use he concep of moneary policy rules in a wider sense, namely as a prescribed guide for moneary policy. This allows insrumen rules, prescribed reacion funcions, as well as argeing rules, prescribed loss funcions or prescribed condiions for he arge variables (or forecass of he arge variables).

12 12 Lars E. O. Svensson Thus, in his subsecion, I am concerned wih insrumen rules, ha is, a rule prescribing he cenral bank s insrumen as a paricular reacion funcion, ha is, a paricular funcion of variables observable by he cenral bank. Suppose we make he unrealisic assumpion ha he cenral bank can commi, once and for all, o a paricular reacion funcion for all fuure periods. Furhermore, assume ha he cenral bank knows he ransmission mechanism and he curren sae of he economy. Under hese assumpions i is, in principle, possible o find he opimal reacion funcion ha minimizes he cenral bank s ineremporal loss funcion, (2.2). The opimal reacion funcion under commimen is a very complex consrucion and would normally be a funcion of all he relevan variables ha describe he sae of he economy (as well as lags of hese variables). 28 This is likely o be oo complex o be verifiable. For his and oher reasons discussed below, a commimen o he opimal reacion funcion is no realisic. Consider insead he class of simple reacion funcions, meaning reacion funcions wih few argumens. A ypical simple reacion funcion is he much-discussed Taylor (1993) rule, i = r + π * + 15.( π π * ) + 05.( y y * ), Where i is he federal funds rae and r is he average real ineres rae. 29 Under he assumpions saed above i is, in principle, possible o find he opimal simple reacion funcion (under commimen), he reacion funcion in a paricular class of simple reacion funcions minimizing he ineremporal loss funcion. Asufficienly simple reacion funcion is easily verified. In principle, we can, hus, conceive of a commimen o a simple reacion funcion, a commimen o a simple insrumen rule. We realize ha, under a commimen o a simple insrumen rule, he cenral bank need no longer be forward looking. I need only be forward-looking once and for all, when deciding o which simple reacion funcion i will commi. Afer ha, i can jus se he insrumen mechanically according o he simple rule. 30

13 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 13 Mos of he curren and previous discussion of moneary-policy rules is in erms of commimen o alernaive insrumen rules (see, for insance, McCallum (1997) and he conribuions in Bryan, Hooper and Mann and Taylor [1999b]). The research on insrumen rules has conribued many imporan insighs. 31 Neverheless, I do no believe ha a commimen o an insrumen rule is eiher a pracical or desirable way of mainaining price sabiliy, for several reasons. Firs, here are considerable pracical difficulies in deciding, once and for all, which insrumen rule o follow. In general, Currie and Levine (1992) have shown ha he opimal simple reacion funcion does no only depend on he model and he loss funcion bu also on he sochasic properies of he shocks and he iniial sae of he economy. 32 This is a considerable problem, because he model and he sochasic properies of he shocks are no exacly known (o say he leas). Sill, i may be possible o find a simple reacion funcion ha works olerably well in differen models. This idea is promoed and examined in several papers by McCallum and recenly resaed in McCallum (1997). Resuls of Levin, Williams, and Wieland (1999) for a se of models of he U.S. economy indicae ha a simple reacion funcion may be relaively robus in his sense. Sill, in a class of reacion funcions as resriced as Taylor-ype reacion funcions wih ineres-rae smoohing (where he federal funds rae depends on inflaion only, he oupu gap and he lagged federal funds rae), here is considerable variaion in he suggesed magniudes for he hree coefficiens, as is apparen from he papers published in Taylor (1999b). Second, a commimen o an insrumen rule does no leave any room for judgmenal adjusmens and exra-model informaion. In pracice, moneary policy canno (a leas no ye) rely on models only. As furher discussed in Svensson (1999h), he use of judgmenal adjusmens and exra-model informaion is boh desirable in principle and unavoidable in pracice. Furhermore, a commimen leaves no room for revisions of he insrumen rule when new informaion and research resuls in revisions of he model. For boh hese reasons, a commimen o an insrumen rule would be inefficien.

14 14 Lars E. O. Svensson Third, alhough a commimen o a simple insrumen rule seems echnically feasible, such a commimen is unheard of in he hisory of moneary policy, mos likely for obvious reasons. I would involve commiing he decision-making body of he cenral bank o reacing in a prescribed way o prescribed informaion. Moneary policy could be delegaed o he saff, or even o a compuer, and i would be compleely saic and mechanical and no forward looking. Such a degradaion of he decision-making process would naurally be srongly resised by any cenral bank and, I believe, argumens abou is inefficiency would also easily convince legislaors o rejec i. In pracice, here is, herefore, no commimen mechanism ha commis he decision-making body o reacing in a prescribed way o prescribed informaion. In real-world moneary policy, decision-making under considerable discreion is more or less unavoidable. As Blinder (1998, p. 49) pus i, Rarely does sociey solve a ime-consisency problem by rigid precommimen...enlighened discreion is he rule. 33 Insead, a bes, he commimen is provided by he objecive of moneary policy. 34 Thus, we are unlikely o ever see a commimen o a mechanical simple insrumen rule for moneary policy. Because any given insrumen rule is likely o be more or less inefficien in cerain siuaions, here would be frequen incenives o deviae, ofen for very good reasons, due o new, unforeseen informaion (a crash of he sock marke, a crisis in Asia, he floaing of he Brazilian real, ec.) and corresponding sound judgmenal adjusmens. Therefore, a simple insrumen rule is no incenive compaible, and, in he absence of commimen, frequen deviaions would occur. 35 Thus, alhough alernaive insrumen rules can serve as informaive guidelines (as emphasized in Taylor [1993]), 36 and decisions ex pos may someimes be similar o hose prescribed by he simple insrumen rules, a rigid commimen o a simple insrumen rule is no a realisic subsiue for a forward-looking decision framework. Indeed, insead of making a forward-looking decision once and for all only, a he ime of a commimen o a simple insrumen rule, a cenral bank aiming a price sabiliy needs o be coninuously forward looking and have a regular cycle of decision making. To quoe Greenspan (1994, p. 244),

15 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 15 Implici in any moneary policy acion or inacion, is an expecaion of how he fuure will unfold, ha is, a forecas. The belief ha some formal se of rules for policy implemenaion can effecively eliminae ha problem is, in my judgemen, an illusion. There is no way o avoid making a forecas, explicily or implicily." 37 Therefore, I now urn o a pracical and realisic, and, indeed, already praciced, way of mainaining price sabiliy, namely by way of forecas argeing. 38 Forecas argeing Moneary policy affecs he economy wih considerable lags. Normally, curren inflaion and oupu are, o a large exen, deermined by previous decisions of firms and households. Normally, curren moneary policy acions can only affec he fuure levels of inflaion and he oupu gap, in pracice, wih subsanial lags and wih he oal effecs spread ou over several quarers. This makes forecass of he arge variables crucial in moneary policy. Le us preliminarily make he assumpion ha he ransmission mechanism is approximaely linear, in he sense ha he fuure arge variables depend linearly on he curren sae of he economy and he insrumen. Furhermore, make he preliminary assumpion ha any uncerainy abou he ransmission mechanism and he sae of he economy shows up as addiive uncerainy abou fuure arge variables, in he sense ha he degree of uncerainy abou fuure arge variables only depends on he horizon bu no on he curren sae of he economy and he insrumen seing. I is hen a sandard resul in opimal-conrol heory ha so-called cerainy equivalence applies, and ha opimal policy need only focus on condiional mean forecass of he fuure arge variables, forecass condiional on he cenral bank s curren informaion and a paricular fuure pah for he insrumen. 39 Because his means reaing he forecass as arge variables, he procedure can be called forecas argeing. This procedure involves making condiional forecass of inflaion

16 16 Lars E. O. Svensson and he oupu gap, condiional on differen pahs of he ineres rae (he cenral bank s insrumen), using all relevan informaion abou he curren and he fuure sae of he economy and he ransmission mechanism. 40 Then, he ineres rae pah is chosen for which he corresponding condiional forecass minimize he ineremporal loss funcion, (2.2), which, in pracice, means ha he inflaion forecas reurns o he inflaion arge and ha he corresponding condiional oupu-gap forecas reurns o zero, a an appropriae pace. If he inflaion forecas is oo high relaive o he inflaion arge a he relevan horizon (bu he oupu-gap forecas is accepable), he ineres rae pah needs o be raised. If he condiional inflaion forecas is oo low, he ineres rae pah needs o be lowered. The chosen ineres rae pah is hen he basis for he curren ineres seing. 41 In regular decision cycles, he procedure is hen repeaed. If no new significan informaion has arrived, he forecass and he ineres rae pah are he same and ineres rae seing follows he same ineres rae pah. If new significan informaion has arrived, he forecass and he ineres rae pah are updaed. This is he procedure recommended by Blinder [14] and referred o as dynamic programming and proper dynamic opimizaion. Compared wih many oher ineremporal decision problems ha households, firms, and invesors solve one way or anoher (usually wihou he assisance of a sizeable saff of Ph.D.s in economics), his paricular decision problem is, in principle, no overly complicaed or difficul. 42 Forecas argeing requires ha he cenral bank have a view of wha he policy mulipliers are ha is, how ineres rae adjusmens affec he condiional inflaion and oupu gap forecass. Bu i does no imply ha forecass mus be exclusively model based. Insead, i allows for exra-model informaion and judgmenal adjusmens, as well as very parial informaion abou he curren sae of he economy. I basically allows for any informaion ha is relevan for he inflaion and oupu gap forecass. Under he above assumpions of a quadraic loss funcion and an essenially linear ransmission mechanism, ogeher wih addiive uncerainy, he cerainy-equivalence resul implies ha he mean forecass are he relevan arge variables, regardless of he degree of

17 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 17 uncerainy. When he uncerainy abou he ransmission mechanism is non-addiive, ha is, here is uncerainy abou he policy mulipliers, or if he ransmission mechanism is characerized by significan nonlineariies, cerainy-equivalence no longer applies, and he mean forecass of he arge variables are no sufficien. Insead, he balance of risks and, indeed, he whole probabiliy disribuion of he arge variables maer. Forecas argeing can hen be generalized from mean forecas argeing o disribuion forecas argeing. Disribuion forecas argeing. Disribuion forecas argeing consiss of consrucing condiional probabiliy disribuions of he arge variables insead of mean forecas only. Thus, for a given ineres rae pah, he cenral bank consrucs he join condiional densiy funcion of he random pah of inflaion and he oupu gap, condiional upon all informaion available in period and a given ineres-rae pah. Then, his condiional probabiliy disribuion is used o evaluae he loss funcion (2.2) wih (2.1). This can eiher be done numerically or informally by he decision-making body of he bank. In he laer case, he decision-making body is presened wih he probabiliy disribuions for a few alernaive ineres-rae pahs and hen decides which pah and disribuion provides he bes compromise. Disribuion forecas argeing is already praciced o some exen. Inflaion-argeing cenral banks have moved beyond mean forecas argeing by considering he balance of risks. Furhermore, Bank of England and Sveriges Riksbank have developed mehods for consrucing confidence inervals for he forecass published in heir Inflaion Repors (see Blix and Sellin [1998] and Brion, Fisher, and Whiley [1998]). 43 Bank of England presens fan chars for boh inflaion and oupu, and Sveriges Riksbank gives confidence inervals for is inflaion forecass. 44, 45 Furhermore, scruiny of he moivaions for ineres-rae changes (including he minues from Bank of England s Moneary Policy Commiee and he Riksbank s Execuive Board) indicae ha boh banks occasionally depar from cerainy-equivalence and ake properies of he whole disribuion ino accoun in heir decisions for insance, when he risk is unbalanced and downside risk differ from upside risk.

18 18 Lars E. O. Svensson Transparency. The above procedure may seem o involve an excessive amoun of discreion. Wha is here o ensure ha he forecass are unbiased and ha he ineres-rae decisions aken follow from proper forecas argeing? 46 This is where ransparency eners in a crucial way, and where some inflaion-argeing cenral banks (in paricular Reserve Bank of New Zealand, Bank of England, and Sveriges Riksbank) have broken very new ground. Transparen inflaion argeing means being explici abou he numerical inflaion arge (and increasingly explici abou he weigh on oupu-gap sabilizaion, as discussed in Secion 2.2 above). As argued in more deail in Svensson (1997a and 1999c), ransparen inflaion argeing can be inerpreed as a argeing rule, a commimen o do whaever i akes o minimize he loss funcion. Jus having an explici inflaion arge goes a long way oward such a commimen. Bu he hree cenral banks menioned have gone furher, by regularly publishing heir inflaion forecass and he reasoning and main informaion behind hese, and by using hese forecass o moivae heir policy decisions. No only should his be he bes way of moivaing policy decisions, his ransparency also opens up he banks analysis and reasoning o ouside scruiny (especially when minues from policy meeings are also published), maximizes he possibiliy for ousiders o spo biases and deviaions, and provides he bes incenive in he hisory of moneary 47, 48 policy for he cenral banks o do heir job well. Moneary argeing and he role of money Recen ineres in moneary argeing has been simulaed by he view ha moneary argeing is he reason behind Bundesbank s ousanding record on inflaion conrol and he possibiliy ha he Eurosysem would choose moneary argeing as is moneary policy sraegy. However, wih regard o wheher moneary argeing lies behind Bundesbank s success, as discussed in, for insance, Svensson (1999c), a number of sudies of Bundesbank s moneary policy have come o he unanimous conclusion ha, in he frequen conflics beween sabilizing inflaion around he inflaion arge and sabilizing money-growh around he money-growh arge, Bundesbank has consisenly given prioriy o he inflaion arge and disregarded he moneary arge. 49 Thus, Bundesbank has acually been a moneary

19 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 19 argeer in words only and an inflaion argeer in deeds. Furhermore, he Eurosysem has srongly rejeced moneary argeing as a suiable sraegy on he grounds ha he relaion beween prices and money may no be sufficienly sable and ha he moneary aggregaes wih he bes sabiliy properies may no be sufficienly conrollable (see Issing). On he oher hand, he Eurosysem has assigned a prominen role o is money-growh indicaor, he deviaion of M3 growh from a reference value. Allan Melzer has, in a series of papers wih Karl Brunner and in Monearism: The Issues and he Oucome (1998), for insance, consisenly argued for a prominen role for money as an indicaor arge or insrumen. 50 Some 25 years ago, several auhors concluded ha inermediae-variable argeing in general (and moneary argeing in paricular) is inferior in mos circumsances. 51 When would moneary argeing, meaning money-growh argeing, be opimal? This requires ha he ransmission mechanism is recursive in a paricular way. The insrumen and he sae of he economy mus affec he arge variables exclusively by firs affecing money growh and hen by money growh affecing he arge variables. 52 Suppose inflaion is he only arge variable (ha is, assume sric inflaion argeing). Then, schemaically, we need o have insrumen sae of he economy money growh inflaion. Money growh would hen be he only deerminan of inflaion. If so, money-growh argeing ha is, sabilizing money around a money-growh arge corresponding o he inflaion arge would be equivalen o sabilizing inflaion around he inflaion arge. 53 Clearly, such recursiveness of he ransmission mechanism, ha money growh would be he sole deerminan of inflaion, is an exreme and unrealisic case. In he real world, and in reasonable models, here are several channels of ransmission from he insrumen o inflaion. The ransmission mechanism is oo complex for inermediae variables in he above sense o exis. Tha is, he ransmission mechanism is no recursive in he above sense. 54

20 20 Lars E. O. Svensson Therefore, inermediae-variable argeing in general, and moneary argeing in paricular, is no a good moneary policy sraegy. However, here is one excepion o he general nonexisence of suiable inermediae-arge variables. As discussed in Svensson (1997a and 1999c), one se of inermediae-arge variables always exiss namely condiional forecass. We can always wrie insrumen sae of he economy condiional forecas of arge variable arge variable, where he arge variable differs from he condiional forecas by an error ha is uncorrelaed wih he insrumen and he informaion abou he sae of he economy. Formally, condiional forecass of arge variables can be seen as inermediae arge variables. As Mervyn King (1994) saed early in he hisory of inflaion argeing, inflaion argeing can be inerpreed as having inflaion forecass as inermediae arges. Thus, money-growh argeing is opimal (for sric inflaion argeing) only if money growh is he sole predicor of fuure inflaion. Since money growh is no he sole predicor of inflaion (nor even he main predicor a horizons relevan for moneary policy), 55 how inefficien would money-growh argeing be? Rudebusch and Svensson (1999) examine his issue wih U.S. daa in an empirical model of U.S. inflaion, oupu, and money. In he model, inflaion is deermined by a convenional acceleraionis Phillips curve, where inflaion is deermined by lagged inflaion and he oupu gap. The oupu gap is deermined by a convenional aggregae demand equaion, where i is deermined by he lagged oupu gap and he real federal funds rae. Money is deermined by a convenional error-correcion moneydemand equaion, where he change in (log) real M2 balances adjuss o he lagged deviaion of (log) real balances from a convenional long-run money-demand equaion. These equaions all fi he daa quie well. 56 Thus, he ransmission mechanism in his model is he convenional ineres rae channel: inflaion expecaions are sicky, so he nominal federal funds rae deermines he real federal funds rae, which affecs

21 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 21 he oupu gap one quarer ahead which, in urn, affecs inflaion wo quarers ahead. M2 has no direc role in he ransmission mechanism and is deermined separaely from he money-demand funcion. Since money demand is demand for real money, nominal money and nominal prices are highly correlaed in he long run, as in he daa and any reasonable moneary model. Sill, his long-run correlaion is irrelevan a he horizon relevan for moneary policy. The esimaed money-demand funcion is well behaved and money is quie conrollable. Neverheless, he resuls unambiguously show ha alhough moneary argeing can, no surprisingly, achieve he same average inflaion as inflaion argeing, i would be quie inefficien for he Unied Saes, in he sense of causing much higher variabiliy of inflaion and he oupu gap han inflaion argeing (boh variances roughly doubles). Furhermore, seing money-demand shocks equal o zero, and hus assuming a compleely sable money demand, only marginally reduces he inefficiency of moneary argeing. Thus, couner o convenional wisdom, he resuls indicae ha he reason why moneary argeing is inefficien is no he insabiliy of money demand. Insead, regardless of he sabiliy of money demand, since money growh is no a predicor of inflaion and he oupu gap, sabilizing money growh does no mean sabilizing inflaion and he oupu gap. The dynamics of money demand is such ha he reacion funcion for he federal funds rae resuling from sabilizing money growh is quie unsuiable for sabilizing inflaion and he oupu gap, also if here are no shocks o money demand. In conras o he model in Rudebusch and Svensson (Working Paper 1999) described above, he so-called P* model (see Hallman, Porer, and Small [1991] and Töder and Reimers [1994]) assigns a direc role o moneary aggregaes in deermining inflaion. Inflaion is hen no deermined by he lagged oupu gap bu by he lagged price gap, he gap beween he price level and he long-run equilibrium price level ha would resul wih he curren money sock if oupu were a is poenial level and velociy were a is long-run equilibrium price level. The price gap is equal o he negaive of he real money gap, he difference beween curren real balances and long-run equilibrium

22 22 Lars E. O. Svensson real balances. The P* model is ypically seen among proponens for moneary argeing as providing a heoreical raionale for focusing policy deliberaions on he behavior of moneary aggregaes. 57 Neverheless, Svensson (1999b) shows ha alhough he P* model gives a prominen role o moneary aggregaes in he form of he real money gap, i does no provide a raionale for money-growh argeing. For he euro area, Gerlach and Svensson (Working paper 1999)provide a preliminary sudy of he relaionship beween inflaion, oupu, money, and ineres raes using reconsruced hisorical daa. The P* model is shown o have subsanial empirical suppor. The real money gap is shown o have subsanial predicive power for four-quarer inflaion 4 and 8 quarers ahead, wih some addiional informaion in he oupu gap. Gerlach and Svensson also consider a Eurosysemype money-growh indicaor, four-quarer growh of M3 relaive a Eurosysem-syle reference value. They find ha he money-growh indicaor has lile or no marginal predicive power for fuure inflaion. Thus, hey find lile empirical suppor for he prominen role ha he Eurosysem has assigned o is money-growh indicaor. In he end, a raional reamen of indicaors suggess ha he weighs on any given indicaor depend exclusively on is power in predicing fuure inflaion and he oupu gap. I seems ha moneary aggregaes should have no special role beyond ha, and ha any weigh on moneary aggregaes should exclusively depend on heir predicive performance for fuure inflaion and he oupu gap. In he P* model and (according o Gerlach and Svensson [Working paper 1999]) in he euro area, inflaion seems o be affeced via a srong direc real-money channel. Melzer (March 1999) and Nelson (Working paper 1999) repor an empirical direc real-money channel affecing demand, separae from he effec of a shor real ineres rae (alhough Nelson inerpres his as a proxy for he effec of a long ineres rae). Woodford (1999d) finds ha any direc real-money effec on consumpion and demand is likely o be very small. In he model in Rudebusch and Svensson (Working Paper 1999), here is no direc real-money channel o inflaion and oupu. Even if here are srong such real-money effecs on demand and/or inflaion, his is no an argumen for moneary argeing (or an argumen for a Eurosysem-syle money-growh indicaor).

23 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 23 A mos, i is an argumen for including real money aggregaes among he many indicaors affecing he forecass of he arge variables, inflaion, and he oupu gap. Benefis of credibiliy An era of low and sable inflaion also creaes low and sable inflaion expecaions, eiher by forward-looking observers who judge ha he cenral banks goals are now o mainain price sabiliy and ha he environmen is unlikely o preven hem from achieving heir goals, or by more backward-looking observers using hisory o assess fuure inflaion oucomes. Expecaions of low and sable inflaion can be inerpreed as good credibiliy for he low-inflaion regime. Le me now discuss he role of credibiliy and he poenial benefis of credibiliy in a moneary policy aimed a low inflaion. Defining and measuring credibiliy Blinder s (1999) favorie definiion of credibiliy involves words maching deeds : A cenral bank is credible if people believe i will do wha i says. Wih an announced policy goal, credibiliy hen boils down o privae expecaions being consisen wih he goal. For a cenral bank wih an explici inflaion arge, i is hen naural o define credibiliy as privae inflaion expecaions coinciding wih he inflaion arge, and o le deviaions of privae inflaion expecaions from he inflaion arge indicae a lack of credibiliy (boh when inflaion expecaions are above and below he arge). Because inflaion expecaions can be measured or esimaed, for insance, from surveys or from nominal and real yield curves, his allows he explici measuremen of he degree of credibiliy. Inflaion-argeing cenral banks regularly include measures of privae inflaion expecaions in heir inflaion repors. 58 Char 1 shows survey daa of inflaion expecaions for he Swedish CPI of invesors on he Swedish bond marke. 59 For each quarer, he hin line shows mean inflaion expecaions for he nex wo years. The semi-hick line shows he same for he nex five years. The hick line shows he implied expecaions for years hree o five. The inflaion

24 24 Lars E. O. Svensson Percen/year 7 Char 1 Inflaion Expecaions, Sweden 1991:1-1999:2 Percen/year 7 6 For nex 3-5 years (implied) 6 Inflaion expecaions For nex 5 years For nex 2 years Survey dae arge of 2 percen per year, wih a olerance inerval of plus/minus 1 percenage poin, was announced by he Riksbank in January 1993, o be effecive from 1995 onward (shown as horizonal solid and dashed lines). We see ha credibiliy was low during he firs few years of he inflaion-argeing regime, wih expecaions for years hree o five far above he upper bound of he olerance inerval. From 1997, inflaion expecaions have been well inside he olerance inerval, and from 1998, five-year expecaions and hree- o five-year expecaions have been close o he 2 percen inflaion arge, indicaing ha he credibiliy is now good. Thus, one of he many benefis associaed wih an announced explici inflaion arge is ha he degree of credibiliy is easily measured. The absence of an explici arge, as in he case of he Fed, or a somewhaambiguous definiion of he arge, as in he case of he Eurosysem, 60 makes he credibiliy less well-defined and less observable.

25 How Should Moneary Policy Be Conduced in an Era of Price Sabiliy? 25 Benefis of credibiliy Cenral bankers ofen appear o be obsessed wih he noion of credibiliy, for which I believe here o be a good reason. Credibiliy, indeed, brings considerable benefis o a moneary policy aimed a low inflaion. So, wha are hese benefis? In a convenional model of inflaion deerminaion, inflaion is deermined by inflaion expecaions and coss, he laer, in urn, depending on (among oher variables) he oupu gap or unemploymen. For given inflaion expecaions, moneary policy hen mainly affecs inflaion via is effecs on real aciviy. 61 Now, wih an explici inflaion arge, credibiliy means ha inflaion expecaions some wo years ahead and more are sable and close o he inflaion arge. Compared o a siuaion wih inflaion scares (Goodfriend [1995]) and flucuaing inflaion expecaions, his eliminaes an imporan source of disurbances o inflaion. Furhermore, credibiliy inroduces considerable mean reversion of inflaion oward he inflaion arge. Consequenly, here is less need for moneary policy o affec real aciviy in order o keep inflaion close o he arge. As a resul, i is easier for he cenral bank o fulfill he inflaion arge, and a more favorable rade-off beween inflaion variabiliy and oupu gap variabiliy arises. 62 Furhermore, in his convenional model of he ransmission mechanism, he moneary-policy insrumen, a shor nominal ineres rae, affecs real aciviy according o he following sequence. Because of sicky inflaion expecaions, he shor nominal rae affecs he shor real ineres rae rae. Expecaions of fuure shor nominal raes and fuure inflaion affec longer real raes. These, in urn, affec real aciviy. Now, if inflaion expecaions are sable around he inflaion arge, an imporan source of disurbances in he above sequence is removed. Furhermore, wih sable inflaion expecaions, he impac of he shor nominal rae on he shor real rae is more direc and sable. For insance, for a shock increasing inflaion, he required increase in he shor nominal rae o achieve a given increase in he shor real rae is