Implementing Inflation Targeting in Brazil

Transcription

1 Implemening Inflaion Targeing in Brazil Joel Bogdanski Alexandre Anonio Tombini Sérgio Ribeiro da Cosa Werlang Absrac Brazil has pu in place an inflaion-argeing framework for moneary policy in mid-1999, less han six monhs afer moving o a floaing exchange rae sysem. This paper presens he macroeconomic background ha has led o he shif in moneary policy regime, and describes he general insiuional arrangemens and operaional framework ha has been adoped. The paper also discusses he basic modeling approach ha has aided he decisionmaking process in he iniial phase of inflaion argeing in Brazil. We describe he family of small-scale macroeconomic models ha has been used for informing and disciplining discussions abou moneary policy wihin he Cenral Bank. These models conain few equaions and few variables, bu carry a considerable heoreical conen and provide a sylized represenaion of he moneary policy ransmission mechanism. They are easily undersood, and especially suiable for simulaion of a wide range of issues. We conclude wih he main lessons ha may be drawn from he iniial Brazilian experience wih inflaion argeing. Depuy Head, Research Deparmen, Cenral Bank of Brazil. Head, Research Deparmen, Cenral Bank of Brazil. Vice Governor, Cenral Bank of Brazil.

2 Implemening Inflaion Targeing in Brazil 1. Inroducion Brazil has recenly pu in place a new framework for moneary policy. Afer moving o a floaing exchange rae sysem, he governmen defined inflaion arges for he coming years and assigned o he Cenral Bank he responsibiliy and he operaional independence o conduc moneary policy in order o mee he inflaion objecive. Inflaion argeing requires ha moneary auhoriies adop a forward-looking aiude and ake preempive acion, given he lags beween policy decisions and heir effec on oupu and prices. In Alan Greenspan s words, Implici in any moneary policy acion or inacion is an expecaion of how he fuure will unfold, ha is, a forecas. Indeed, we believe ha wha inflaion-argeing cenral banks acually do is inflaion forecas argeing. Raher han reacing o presen facs, moneary policymakers make decisions based on condiional forecass of fuure inflaion, condiional on alernaive ineres rae pahs and on he bes esimae of he curren sae of he economy and he probable fuure developmen of exogenous variables. I is crucial o develop a basic modeling framework o allow policymakers o exercise heir judgemenal analysis in a srucured and quanified way. Economic models are jus anoher ool available o guide policy decisions under uncerainy abou he sae of he economy and he size and naure of he shocks ha consanly hi i. Noneheless, simple models can help clarify economic problems by focusing on a small number of facors hough essenial for heir undersanding. In his paper, we presen he basic modeling approach ha has aided he moneary policy decision-making in he iniial phase of he inflaion-argeing regime in Brazil. We hank oher members of he Cenral Bank s research eam, in paricular Fábio Araújo, Paulo Springer de Freias, and Marcelo Kfoury Muinhos. Needless o say, he views expressed here and any errors are our own responsibiliy. 2

3 Secion 2 describes he general insiuional arrangemens and operaional framework adoped for inflaion argeing. Secion 3 describes he family of small-scale macroeconomic models ha has been used for informing and disciplining discussions abou moneary policy wihin he Cenral Bank. These models conain few equaions and few variables, bu carry a considerable heoreical conen and provide a sylized represenaion of he moneary policy ransmission mechanism. They are easily undersood, and specially suiable for simulaion of a wide range of issues. Secion 4 deals wih he basic requiremens for running simulaions. Secion 5 presens brief conclusions. 2. The Brazilian IT framework 2.1 In search of a nominal anchor for moneary policy The sabilizaion process in Brazil, iniiaed in mid-1994, successfully brough annual inflaion down o one-digi figures in less han hree years. This process included a wide program of economic reforms. For insance, he size of he public secor was subsanially reduced in his period hrough privaizaion of sae companies operaing in secors like elecommunicaions, chemisry, railroads, banking and mining. Likewise, he rade liberalizaion was deepened hrough reducion of impor ariffs and eliminaion of non-ariff barriers. The financial sysem was submied o a full-fledged resrucuring wih unsound insiuions liquidaed, merged or resrucured and prudenial regulaion updaed. The main source of inflaionary ineria, he auomaic indexaion of prices, wages and oher conracs, was subsanially reduced. Annual oupu growh averaged 3.4% in real erms in , even hough unemploymen sared o rise in However, despie is relaive success, he sabilizaion process involved a gradualis approach owards many srucural economic problems ha remained unsolved. A muchneeded definiive fiscal adjusmen was coninually posponed because, in par, he governmen coaliion was no sufficienly convinced of is urgency. So, Brazil remained 3

4 vulnerable o a confidence crisis, which became a realiy when he inernaional financial urmoil culminaed wih he Russian moraorium in Augus The confidence crisis generaed a large capial fligh from emerging markes. Brazil raised shor-erm ineres raes and announced a srong ighening of he fiscal regime. A he same ime, he governmen negoiaed a prevenive financial suppor package wih he IMF, oaling US$ 41.5 billion. The governmen was iniially successful in implemening he fiscal package, bu marke confidence coninued o erode up o January 1999, also reflecing concerns over he newly eleced governors commimens o adjusing heir saes public finances. Following srong pressures on foreign exchange reserves, he Cenral Bank was forced o abandon he crawling peg o he dollar 1. Afer a brief aemp o conduc a conrolled devaluaion, he real was forced o floa on January 15. As a consequence of his abrup change in regime, mos of he Cenral Bank s Board of Direcors was replaced. Due o Brazilian peculiariies, he new Board ook office only in he beginning of March. 2 In he absence of a well-defined guidance for moneary policy, he exchange rae averaged R$1.52/US$1 in January and R$1.91/US$1 in February, compared wih R$1.21/US$1 prior o he change in regime. Inflaion rose sharply: he wholesale price index increased 7.0% in February, while he consumer price index rose 1.4%. This led privae analyss o foresee a huge deerioraion of all macroeconomic fundamenals. The new Board ook office on March 4 h and immediaely worked on wo main frons. The firs was o calm down he nervous financial markes. The expecaion ha an inflaion hike could bring he real raes of reurn on public deb insrumens o he negaive range was he firs o be aacked. 1 The official exchange rae policy a ha ime consised of an inended nominal devaluaion of 7.5% p.a., while annual inflaion was near 2%. 2 In Brazil, he Federal Senae mus formally approve he nominees o he Cenral Bank s Board. The process consiss of wo seps. The firs is a preliminary open hearing in he Commiee of Economic Affairs. The second is a discussing and voing session, where he 81 Senaors decide by simple majoriy wheher o approve or rejec he nominee. Alhough he Governmen coaliion has a large majoriy in he Senae, he process is ime consuming. 4

5 The Moneary Policy Commiee (Copom), whose members are he Governor and Depuy Governors, decided o raise he basic shor-erm ineres rae (Selic) from 39% p.a. o 45% p.a., aking ino accoun ha he fuure conracs for he nex mauriy were already rading a 43.5%. An imporan novely was inroduced, he bias on he ineres raes, which delegaed o he Cenral Bank s Governor he power o change ineres raes during he period beween wo ordinary Commiee meeings (hey used o be 5 weeks, prior o he adopion of he inflaion argeing framework). The Copom esablished a downward bias, meaning ha ineres raes could be lowered a any ime before he nex scheduled meeing 3. For he firs ime, also, he Commiee released a brief explanaion righ afer he meeing (he minues used o be released only afer 3 monhs). Is iniial words were mainaining price sabiliy is he primary objecive of he Cenral Bank. And i wen on o say ha: (1) in a floaing exchange rae regime, susained fiscal auseriy ogeher wih a compaible moneary auseriy suppor price sabiliy; (2) as fiscal policy is given in he shor run, he conrol over inflaionary pressures should be exered by he ineres rae; (3) observed inflaion is due o he currency depreciaion, and markes expec a furher rise in he price level his monh; (4) he basic ineres rae should be sufficienly high o offse exchange-based inflaionary pressures; and (5) so, we decided o raise he basic ineres rae o 45% p.a., bu wih a downward bias, for if he exchange rae reurns o more realisic levels, keeping he nominal ineres rae ha high would be unjusified. Indeed, he bias was used wice before he nex meeing: he ineres rae was reduced firs o 42% and hen o 39.5%, following a reversal of he exchange rae overshooing and a reducion boh in observed and expeced inflaion raes. The second fron was he iniiaive o propose he adopion of inflaion argeing as he new moneary policy regime. Alhough i is clear from he Copom s press release ha IT was already in he minds of he Board s members, here was a lo of work o do in he insiuional area. For example, he Cenral Bank has never been graned formal insrumen independence o conduc moneary policy. Moreover, even a he Bank, very 3 On he oher hand, under a downward bias, if he Governor needs o raise ineres raes, an exraordinary Copom meeing mus be called o ake he decision. 5

6 few saff members knew wha an IT framework was abou. The echnical skills needed o develop adequae inflaion-forecasing models were scaered unevenly hroughou he Bank s deparmens. In paricular, here was no Research Deparmen: each deparmen used o make is own research effors, usually o solve immediae demands and no devoed o hink coherenly abou he fuure. Once hese problems were deeced, heir soluion was sraighforward. The new floaing exchange rae clearly required a new nominal anchor for economic policy. Moneary policy, along wih srenghened fiscal adjusmen and a firm wage policy in he public secor, would be insrumenal in prevening he recurrence of an inflaionary spiral and ensuring a rapid deceleraion of he rae of inflaion. Inflaion argeing was he mos suied framework o achieve economic sabilizaion under a flexible exchange rae regime, wih he arge iself playing he role of he nominal anchor. Wih sound argumens, i was no difficul o convince he Presiden, he Finance Miniser, and heir senior economic advisors ha IT could work well in Brazil. The IMF saff was mos recepive o he proposed new framework for moneary policy, and showed ineres in organizing an inernaional seminar on IT, where he discussions could benefi from he experience of a number of cenral banks and academics. Wihin he Cenral Bank, a Research Deparmen was creaed by he end of March. Iniially, hree research areas were opened: IT, financial risk and pricing and microeconomics of banking. The IT group, consising of 14 researchers, sared o sudy he lieraure. Two key books were very useful: Bernanke e al. (1998), Inflaion Targeing: Lessons from he Inernaional Experience, Princeon Universiy Press, and Taylor, John B. (ed.) (1999), Moneary Policy Rules, Universiy of Chicago Press. Some oher aricles and books enered a minimum ki, a mandaory reading for anyone in he group. (i) General readings and cases: King, Mervyn (1997), The Inflaion Targe Five Years On; Massad, Carlos (1998), La Poliica Monearia en Chile; Masson, P.R. e al (1997), The Scope for Inflaion Targeing in Developing Counries; Taylor, John (1999), A Hisorical Analysis of Moneary Policy Rules. (ii) Opimizaion models: Backus, David e al (1986) The Consisency of Opimal Policy in Sochasic Raional Expecaions 6

7 Models ; Currie, David e al (1993), Rules, Repuaion and Macroeconomic Policy Coordinaion; Svensson, Lars (1998), Open-Economy Inflaion Targeing. (iii) Applied work: Taylor, John (1993) Discreion versus Policy Rules in Pracice; Taylor, John (1994), The Inflaion/Oupu Variabiliy Trade-off Revisied. Afer careful planning, he IT group sared o work in he design of he insiuional framework and he modeling of he ransmission mechanism of moneary policy. Brazil has grealy benefied from he discussions and consulaions held during he Seminar on Inflaion Targeing, joinly organized by he Cenral Bank of Brazil and he IMF s Moneary and Exchange Affairs Deparmen, which ook place in Rio de Janeiro (May 3-5, 1999) 4. A general consensus ha emerged during ha meeing can be summarized as follows: Low and sable inflaion was singled ou as he primary long-run objecive of moneary policy, and inflaion argeing was regarded as an effecive framework for guiding moneary policy. In paricular, inflaion argeing was seen as providing a nominal anchor boh for moneary policy and inflaion expecaions, making his anchor idenical o he long-run objecive of moneary policy; providing more ransparency and accounabiliy o he design and implemenaion of moneary policy; faciliaing is communicaion, undersanding, and assessmen; and providing effecive policy guidance by focusing policymakers aenion on he long-run consequences of shor-erm policy acions. (Brazil Seleced Issues and Saisical Appendix Inernaional Moneary Fund, July 16 h, 1999). 4 The objecive of he seminar was o review he experience of a number of developed and emerging economies in implemening inflaion argeing, and o provide an opporuniy for Brazilian economiss and policymakers o discuss plans o implemen a similar framework in Brazil. Expers from Ausralia, Canada, Chile, Israel, Mexico, New Zealand, Sweden, Unied Kingdom, and Unied Saes made presenaions on heir counry s experiences. Researchers from he Cenral Bank of Brazil presened heir iniial work on inflaion argeing, eniled Issues in he Adopion of an IT Framework in Brazil, May

8 2.1 General seing On July 1 s, 1999, Brazil formally adoped inflaion argeing as he moneary policy framework. The Presiden of Brazil issued Decree No of June 21, 1999, whose key poins are lised bellow. The inflaion arges will be esablished on he basis of variaions of a widely known price index; The inflaion arges as well as he olerance inervals will be se by he Naional Moneary Council on he basis of a proposal by he Finance Miniser; Inflaion arges for he years 1999, 2000, and 2001 will be se no laer han June 30, 1999; for he year 2002 and subsequen years arges will be se no laer han June 30, wo years in advance; The Cenral Bank is given he responsibiliy o implemen he policies necessary o achieve he arges; The price index ha would be adoped for he purposes of he inflaion argeing framework will be chosen by he Naional Moneary Council on he basis of a proposal by he Finance Miniser; The arges will be considered o have been me whenever he observed accumulaed inflaion during he period January-December of each year (measured on he basis of variaions in he price index adoped for hese purposes) falls wihin he olerance inervals; In case he arges are breached, he Cenral Bank s Governor will need o issue an open leer addressed o he Finance Miniser explaining he causes of he breach, he measures o be adoped o ensure ha inflaion reurns o he oleraed levels, and he period of ime ha will be needed for hese measures o have an effec; and The Cenral Bank will issue a quarerly inflaion repor ha will provide informaion on he performance of he inflaion argeing framework, he resuls of he moneary policy acions, and he perspecives regarding inflaion. On June 30 h, 1999, he Naional Moneary Council (CMN) issued a Resoluion 5 on he definiion of he price index as well as on he inflaion arges. The Broad Consumer Price Index (IPCA) repored by he Naional Bureau of Geography and Saisics (IBGE) was chosen for he purpose of gauging inflaion arges. The arges were se a 5 Resoluion No

9 8% for 1999, 6% for 2000 and 4% for 2001 accumulaed annual variaions by he year-end. Tolerance inervals of ±2% for each year were also defined. The seleced price index IPCA covers a sample of families wih personal income beween 1 and 40 minimum wages and has a broad geographical basis. I includes 9 meropolian areas (São Paulo, Rio de Janeiro, Belo Horizone, Poro Alegre, Recife, Belém, Foraleza, Salvador and Curiiba) as well as Goiânia and he Federal Disric. The raionale behind he adopion of decreasing arges beween 1999 and 2001 had o do wih he naure of he recen inflaion in Brazil. I is imporan o disinguish beween an inflaionary process and a emporary inflaion rise due o a shock. In he firs case, here is a coninuous acceleraion in he price level. In he second, here may be only a once-and-for-all change in he price level, wih no furher upward pressure. The Brazilian case belongs o he second caegory: he currency devaluaion ha sared in mid-january 1999 was a shock ha forced a realignmen of relaive prices. Before i occurred, Brazil was experiencing price sabiliy: average CPI inflaion was 1.7% 6 in As here were no indicaions of he presence of an inflaionary process in Brazil, a gradualis disinflaion sraegy was no recommendable. The CPI inflaion rae should reurn o is 1998 level as soon as he relaive prices realignmen is finished. Thus, i was no only possible bu also desirable for he governmen o se a decreasing inflaion arge pah. An imporan issue o discuss is he choice of he full inflaion rae as reference for he arge, and no some core inflaion measure. Perhaps, he bes echnical procedure would have been o purge some iems from he full index, exemping i from emporary and once-and-for-all shocks. Neverheless, adoping a headline index was essenial for credibiliy reasons, a leas in he beginning of IT implemenaion. Unforunaely, Brazilian sociey has experienced several price index manipulaions in a no so disan 6 In he case of CPI measured only in São Paulo ciy by IPC/FIPE, inflaion was negaive (-1.8%) in

10 pas, and so would be suspicious abou any change relaed o suppressing iems from he arge index. Anoher relaed issue is he absence of escape clauses in he insiuional arrangemen. In he case he arges are breached, he Cenral Bank Governor will be required o issue an open leer addressed o he Miniser of Finance explaining he underlying causes, he measures o be adoped o ensure convergence o he arges and he ime period required for hese measures o have an effec. The combinaion of he use of headline inflaion and he absence of escape clauses jusifies he adopion of he relaively wide 2-percenage poin olerance inerval around he cenral arge, and cerainly makes he announced arges much igher han hey may iniially appear. I is imporan o emphasize ha moneary policy decisions should be aken on he basis of he wides informaion se available. Therefore, a mix of models should be under consideraion when looking for an adequae reacion funcion, and producing inflaion forecass and heir probabiliy disribuions. I should also include privae secor percepions abou he expeced pah of economic variables, exra-model informaion, leading indicaors and any oher judgmenal knowledge ha helps predic inflaion. A final issue is he ransparency of he IT framework. As par of he iniial seing, an effecive communicaion process was esablished so ha he populaion will be able o undersand and monior he decisions of he Cenral Bank and o know he reasons why forecased and accumulaed inflaion may be deviaing from arge. The Copom mees a regular monhly inervals and decisions are aken by majoriy voe. The decisions are announced immediaely afer he meeing ends, someimes followed by a press release explaining briefly he reasons why he decisions were aken. In he second half of 1999, he Copom minues were published 2 weeks afer he meeings. In he beginning of 2000, his inerval has been reduced o only one week. 10

11 Finally, here is a quarerly inflaion repor discussing he main issues relaed o he performance of he inflaion-argeing regime 7. I includes deailed explanaions of he resuls delivered by pas decisions and a prospecive analysis for fuure inflaion, wih special emphasis on he assumpions made in he forecasing process ha generaed he moneary insrumen decisions. Minues of he previous Moneary Policy Commiee meeings are republished in he repor. 2.2 Operaional framework The Research Deparmen of he Cenral Bank of Brazil has developed a se of ools o suppor he moneary policy decision process. Among hem are small-scale srucural models of he ransmission mechanism of moneary policy o prices, complemened wih shor-erm inflaion forecasing models, leading inflaion indicaors and measures of core inflaion. There are also surveys of marke expecaions of inflaion, growh, and oher relevan economic variables, colleced by he Economic Deparmen. In order o build small srucural macroeconomic models, he Cenral Bank sudied he various channels of ransmission of moneary policy. These include he ineres rae (a policy insrumen), he exchange rae, aggregae demand, asse prices, expecaions, credi and money aggregaes, wages, and wealh. Given he macroeconomic characerisics of he Brazilian economy, he main conclusions are he following: (i) ineres rae affecs consumer durables and invesmen in a period beween 3 o 6 monhs. Moreover, he oupu gap akes addiional 3 monhs o have a significan impac on inflaion. In sum, he moneary policy ransmission hrough he aggregae demand channel akes beween 6 o 9 monhs o fully operae; and (ii) hrough a direc channel: changes in nominal ineres rae affecs conemporaneously nominal exchange rae, and he laer affecs, also conemporaneously, he inflaion rae hrough "impored" inflaion; (iii) given hisorical low leverage of he Brazilian corporae secor along wih he very sric credi and moneary policies implemened wih he Real plan, he credi 7 Copom has already issue hree Inflaion Repors June 30 h, Sepember 30 h, and December 30 h. They are available a he Cenral Bank homepage hp:// 11

12 mechanism has no operaed and is imporance in erms of channeling ineres rae impacs on inflaion has been negligible. 8 The srucural models are complemened by a se of shor-erm models. These complemenary models include Vecor Auoregressive (VAR) models and Auoregressive Moving Average (ARMA) ime-series models and serve hree basic purposes: (i) providing an alernaive shor-erm forecas for he inflaion rae and, herefore, permiing a consisency check wih he forecass resuling from he srucural models; (ii) permiing he use of he inflaion forecas resuling from hese models for he purposes of esimaing (wih he srucural model) he ex ane real ineres rae (which is an explanaory variable in he aggregae demand equaion in some of he esimaed srucural models), as well as in he forward-looking ineres rae rule (which is one of he equaions in he srucural models); and (iii) allowing o simulae shocks o specific componens of he IPCA, like for insance, changes in prices se by he public secor. 3 The basic modeling approach Benefiing from inernaional advice, gahered during he Seminar on Inflaion Targeing and on consulaions wih he Bank of England, he Cenral Bank has esimaed/calibraed a group of srucural models wih he main objecive of idenifying and simulaing he mechanism of moneary policy ransmission in Brazil, including he main channels of ransmission as well as he lags involved. A simple srucural model wih he following basic equaions can summarize his family of models 9 : 8 Wih recen measures aimed a reducing he difference beween borrowing and lending raes (bank spreads), he credi channel will cerainly become imporan for he operaion of moneary policy under he IT framework. See sudy published by he Research Deparmen of Cenral Bank of Brazil Juros e Spread Bancário no Brasil, Ocober 1999 (available a hp:// 9 The same modeling approach is widely used for policy analysis. Ball (1997) presens an illusraive woequaion model; Ball (1998) and Baini and Haldane (1999) show compac exensions o small, open economies, highlighing he main ransmission channels a work. 12

13 (i) (ii) (iii) (iv) an IS ype equaion expressing he oupu gap as a funcion of is own lags, real ineres rae (ex ane or ex pos), and real exchange rae 10 ; a Phillips curve expressing he rae of inflaion as a funcion of is own lags and leads, he oupu gap, and he nominal exchange rae (and imposing he long-erm neuraliy condiion); an uncovered ineres pariy condiion relaing he differenial beween exernal and domesic ineres raes wih he expeced rae of devaluaion of he domesic currency (he Real), and he risk premium; and an ineres rae rule, alernaively fixed rules on nominal or real ineres raes, Taylor-ype rules (wih weighs for conemporaneous deviaions in inflaion and oupu), forward-looking rules (wih weighs for deviaions of expeced inflaion from he arge), and opimal deerminisic and sochasic rules. This family of models allows several reduced-form specificaions, depending on which issues he Copom wans o discuss in deail. An example may clarify he modeling approach used in Brazil. Suppose ha he governmen is fully commied o a fiscal adjusmen, so ha he arges for he primary surplus of he consolidaed public secor will be observed. In his case, he fiscal policy will produce imporan effecs on aggregae demand, which should be explicily aken ino consideraion. One possible way o incorporae his informaion ino he model is o include a fiscal variable direcly in he IS equaion. In his specificaion, wo variables represen policy insrumens: he ineres rae and he primary fiscal surplus. The firs is a Cenral Bank insrumen and he second is a Treasury insrumen. The diagram in Figure 1 summarizes hese assumpions, showing he basic relaionships involved. 10 Due o he relevan sample period used in he process of esimaion, which involved a managed exchange rae regime, and due o he small relaive weigh of ne expors (when compared o he oher componens of aggregae demand), he real exchange rae was no found saisically significan. 13

14 Variables ha influence counry risk Risk Premium Toal Primary Defici/GDP UIP Exchange Rae IS Aggregae Demand IS Phillips curve UIP Inflaion Ineres Rae Rules/ Exogenous Pah Ineres Rae Figure 1: Model diagram. IS curve The sandard specificaion of an IS curve could be, in a quarerly frequency: (I) h = 0 + β1h 1 + β2h 2 + β3r 1 β + ε where: h log of oupu gap r log of real ineres rae [log(1+r)] ε h demand shock. h The addiion of a fiscal variable can be accomplished by including he erm pr log of (1+) public secor borrowing requiremens (PSBR), primary concep, as a percenage of GDP: (II) h = 0 + β1h 1 + β2h 2 + β3r 1 + pr 1 β + ε hf ε hf demand shock. 14

15 Firs of all, he oupu gap hisorical series should be consruced. I is usual o sar wih a measure of poenial oupu. Several echniques are commonly used o calculae poenial oupu: 1) he exracion of a linear ime rend from hisorical GDP daa; 2) smoohing ou he GDP series hrough filers like Hodrick-Presco; 3) Kalman filers; and 4) esimaes of producion funcions. In he Brazilian case, he linear rend and HP filer were preferred since boh produced similar resuls. The oupu gap was hen obained by he difference beween acual and poenial GDP, allowing a direc esimae of he fiscal IS curve. Long run calibraion of he IS curve In order o esimae he demand side of he reduced-form srucural model, a quarerly daa sample beween 1992:I and 1999:III was seleced. Hence, he esimaion resuls were heavily influenced by pos-real Plan daa (1994:III o 1998:IV). I should be noed ha he Real Plan was implemened along wih a managed exchange-rae regime ha was very insrumenal in reducing inflaion and keeping i low. However, domesic ineres raes were basically se wih he objecive of mainaining he managed exchange-rae pariy. In his old regime, he exchange rae had he role of a nominal anchor o sabilize inflaion, while moneary policy was conduced o aain a balance of paymens posiion compaible wih he desired pariy. In sum, wihou judging he success of his old regime in erms of inflaion sabilizaion and is susainabiliy over ime, i is reasonable o conclude ha equilibrium real ineres rae were necessarily high. This is no necessarily he case in an environmen of high inernaional liquidiy. Neverheless, beween he end of 1994 and he beginning of 1999 emerging economies faced several episodes of worsening in he exernal financial condiions. Under he floaing exchange rae regime (in place since January, 1999) and he inflaion-argeing framework (as of July, 1999), i is reasonable o sae ha equilibrium real ineres raes should differ subsanially from wha hey were in he previous regime. The ransiion effecs due o he new equilibrium level of real ineres raes called for a long-erm calibraion of he demand side reduced-form model. 15

16 The calibraion is sraighforward. In he long-run seady sae, he raio of governmen deb o GDP should remain consan, along wih a balanced budge (zero primary fiscal surplus) and zero oupu gap. This implies ha he long-erm equilibrium real ineres rae mus equal he poenial GDP growh rae. In he fiscal IS curve specificaion, his is equivalen o = β 0 r β 3. So, he long-run calibraion can be done by esimaing he fiscal IS curve wih he addiional resricion on he pair (β0,β3), whose raio mus equal he long-erm equilibrium real ineres rae. Phillips curve The supply side of he economy is usually modeled wih a Phillips curve specificaion, direcly relaing price inflaion o some measure of real disequilibrium (ypically he oupu gap), inflaion expecaions, and real exchange rae changes. Three varians are presened below. The coefficiens á on he righ side of all equaions, excep for he oupu gap one, are consrained o sum o uniy o ensure he long-run vericaliy of he Phillips curve, i.e. ha inflaion is neural wih respec o real oupu in he long run. Backward-looking specificaion (III) π + ε b b b b F = α1 π 1 + α 2π 2 + α3 h 1 + α 4 ( p + e ) Forward-looking specificaion (IV) π + ε f f f f F = α1 π 1 + α2 E ( π + 1) + α3 h 1 + α 4 ( p + e ) Combined (average) specificaion b f (V) π = f b f b f b f b ( α1 + α1 ) α π α2 ( α3 + α 3 ) ( α4 + α ) ( π + 1) + 2 π h E where: π log of price inflaion h log of oupu gap p F e log of foreign producer price index log of exchange rae firs-difference operaor ( p E( ) expecaion operaor, condiional on informaion available a ime ε b,ε f, ε n supply shock. F + e ) + ε n 16

17 The backward-looking specificaion can be moivaed by he assumpion of adapive inflaion expecaions. I is simple o esimae and, wih only wo lags, i is able o reproduce fairly well he rich inflaion dynamics of pas daa. However, i is vulnerable o he Lucas criique. Is predicive power should be weak due o he recen changes in moneary policy and exchange-rae regimes, which probably have alered he formaion of inflaion expecaions and he shor-run inflaion/oupu radeoff. The forward-looking specificaion is an aemp o overcome he parameer insabiliy commonly found afer srucural breaks. I is also moivaed by he naural assumpion ha, as he inflaion argeing regime gains credibiliy, expecaions end o converge o he argeed value. However, i raises difficul esimaion issues abou he appropriae measures of expecaions, especially when reliable survey daa are no available. Differen assumpions abou he expecaions mechanism were esed, bu in general he esimaions led o a weighed average of pas and fuure inflaion, wih a leas 60% on he forward-looking componen. Neiher he research saff nor he Copom members were comforable wih hese resuls for wo reasons. Firs, hey implied a degree of credibiliy ha was no expeced o be achieved so early and hey did no mach he curren surveys of marke expecaions. Second, hey generae an inflaion/oupu dynamics wih almos no ineria and consequenly a fas adjusmen of boh real and nominal variables, which is no believed o yield a reasonable represenaion of realiy. A soluion o balance ou he forward and backward-looking varians was o combine hem. The average of he previous wo specificaions of he Phillips curve (ogeher wih he oher equaions in he complee model) exhibis he desired dynamic properies of he economy, wih inflaion persisence due o sluggish adjusmen forced by he backward-looking erms, while keeping a forward-looking componen hough o be increasingly imporan in he ransiion period afer he changes in moneary policy and exchange-rae regimes. For he purpose of running simulaions o invesigae he implicaions for inflaion and oupu of differen moneary policy rules, i is easy o experimen wih alernaive assumpions abou he expecaions formaion mechanism. For example, expecaions 17

18 can be aken exogenously from a marke survey, ogeher wih an addiional hypohesis abou how hey reac o new informaion. Or expecaions can be calculaed recursively in order o be model-consisen. The passhrough The passhrough of exchange rae changes o domesic inflaion is a key issue in he Phillips curve se-up. Several linear and non-linear specificaions for he passhrough coefficiens have been esed, reducing o four he alernaives implemened in he preferred simulaion ool. The firs one is a sandard consan coefficien; simply esimaed from a suiable sample of pas daa. The second one is a quadraic ransfer from exchange rae variaions o inflaion. The hird one is a level-dependen coefficien. I is esimaed under he assumpion ha he passhrough depends also on he level of he (log) nominal exchange rae. The las one is a quadraic funcion of he nominal exchange rae level, moivaed by a simple parial equilibrium model in which exchange-rae devaluaions shif he supply curve of compeiive producers of radable goods 11. All non-linear varians inend o capure more precisely he effecs of a emporary exchange rae overshooing 12. For he small number of observaions available in a quarerly frequency, however, heir resuls were very close o he linear varian and consisen wih inernaional evidence ha he passhrough coefficien is inversely proporional o he degree of real exchange rae appreciaion a he momen prior o he devaluaion. The equaions below summarize he four alernaive specificaions. α 4 = consan α α α = 4 ( + F α 41 α 42 ( p + 1 e 1)) 4 = ( α + α e ) 1 = α where: p F e E 41 E E α + α log of foreign producer price index. log of exchange rae. exchange rae (R$/US$). 11 See Goldfajn and Werlang (1999), Appendix. 12 Dornbusch (1976) presens he overshooing resul in a differen conex. 18

19 Exchange Rae - Uncovered Ineres Pariy The nominal exchange rae is deermined by he uncovered ineres pariy condiion, which relaes expeced changes in he exchange rae beween wo counries o heir ineres raes differenial and a risk premium: (VI) E e +1 e = i i F x where: e log of exchange rae i log of domesic ineres rae i F x log of foreign ineres rae log of risk premium Taking he firs difference E F e+ 1 E 1e e = i i x and assuming for simpliciy ha he expecaion change follows a whie noise process 13 E e + 1 E 1 e = η, i is possible o specify he exchange rae dynamics as: (VII) e = i + x i + η. F There are wo exogenous variables in his equaion: he foreign ineres rae and he risk premium. Given he relaive sabiliy of foreign ineres raes, reasonably accurae projecions can be obained from conracs raded in inernaional fuures markes. However, he risk premium which can be measured by he spread over Treasury bonds of Brazilian sovereign deb has presened high volailiy in he las years. The risk premium is usually associaed o macroeconomic fundamenals and a number of oher subjecive facors ha are no easily anicipaed. Hence, wo alernaive approaches have been considered. The firs is o gaher he opinions of Copom members abou he fuure evoluion of he counry s risk premium, condiional on he overall scenario and based on anecdoal evidence, ranslaing i o an exogenous expeced pah ha will be used in simulaions. The second approach is o make assumpions linking he risk premium behavior o he main objecive facors hough o influence i, leing i be endogenously deermined by he model. 13 This is equivalen o a random walk wih moneary surprise, where a surprise is characerized by changes in ineres rae differenials or in risk percepion. 19

20 An assumpion consisen wih he fiscal IS curve is ha he risk premium will respond o he fiscal sance, wih any perceived improvemen in he consolidaed public secor fiscal posiion reducing he premium accordingly. Addiionally, a number of oher facors may have a considerable influence on expecaions and hence, on he risk premium. A lis of hese facors would ypically include inernaional liquidiy condiions and ineres raes, foreign capial markes performance, commodiies prices, curren accoun balance perspecives, and counry raing. The link wih he UIP condiion is hrough he following equaion: (VIII) n X = γ 1 X 1 + γ 2 PR 3 + γ j Z j, j j= 3 where: X risk premium in basis poins (SoT) PR PSBR, primary concep, as a percenage of GDP Z variables ha influence counry risk. Moneary policy rules The primary insrumen of moneary policy is he shor-erm ineres rae se by he Cenral Bank. To run a simulaion in any of he model varians 14, i is necessary o choose a moneary policy rule. The rules can be divided in hree basic families: fully exogenous ineres rae pahs, linear combinaion of sysem variables and opimal response funcions. Fully exogenous ineres rae pah This family of rules provides a direc way o inpu any ineres rae pah in he model. This is useful o analyze he consequences of an expeced ineres rae rajecory, such as ha implied by financial marke insrumens or he implici pah considered in he governmen budge. A paricular rule of his family can be very helpful for insiuional communicaion. The quarerly Inflaion Repor radiionally presens inflaion and oupu growh forecass consruced under he assumpion ha he shor-erm ineres rae will remain consan a 14 See Blanchard and Kahn (1980) for references on solving forward a sysem of difference equaions under raional expecaions. 20

21 he curren level along he projecion period. This projecion is made clear by means of an inflaion fan char, which shows he probabiliy disribuion around he cenral forecas for each quarer. By visual inspecion, i is possible o infer wheher moneary policy should be alered and in which direcion 15. Linear combinaion of sysem variables The ineres rae rule in his family is a linear funcion of some sysem variables. For example, moneary policy can reac conemporaneously o oupu gap and deviaions of inflaion from arge, as shown in he equaion below. When λ=1, his is equivalen o a sandard Taylor rule, while when λ (0,1) his is a Taylor rule wih ineres rae smoohing. The α s can be se arbirarily or using specific opimizaion procedures available in he simulaion ool. * (IX) i 1 λ ) i + λ( α ( π π ) + α h + ) = ( α 3 where: π log of inflaion. π* log of inflaion arge. h log of oupu gap. i log of ineres rae. This ype of rule is insrumenal o analyze he sysem behavior under he choice of a paricular se of α s. However, he se of α s obained by opimizaion procedures can frequenly be very uninuiive 16. Opimal Trajecories An opimal rule can be found using wo basic opimizaion mehods available for simulaion. The firs one is a deerminisic opimizaion made considering he expecaion for he sysem variables equivalen o he model own realizaions. The second uses sochasic simulaion. In he deerminisic case, he objecive funcion is given by equaion (X), and in he sochasic case i is given by equaion (XI). Noice 15 Brion, Fischer and Whiley (1998) explain how o inerpre inflaion forecass presened as fan chars. Haldane (1997) discusses how he inroducion of a parially subjecive probabiliy disribuion may help clarify he policy maker s assessmen of he curren economic sance. 16 See Svensson (1997). 21

22 ha due o he cerainy equivalence principle, for linear consrains, equaions (X) and (XI) are equivalen. N L 1 + r + r 2 + r 3 + r r= 1 * (X) = [ λ ( E( π ) π ) + λ [ E( h )] + λ ( i ) ] N L 1 + r + r 2 + r 3 + r r= 1 * (XI) = [ λ E[( π π ) ] + λ E( h ) + λ ( i ) ] These wo opimizaion mehods may be used wih a fully arbirary ineres rae rajecory 17 or wih a rajecory given as funcion of inflaion, oupu gap and he lagged ineres rae. The deerminisic case is useful for simulaing alernaive scenarios during a forecas eam meeing since his procedure is fas. On he oher hand, he sochasic case is more accurae and gives he confidence inervals for he implied ineres rae rajecory, hough i requires subsanial compuaion ime, making his procedure unfeasible during a Copom meeing. Basic srucure Combining equaions (II), (V), (VII), and (VIII), along wih a choice of formaion mechanism of inflaion expecaions, passhrough specificaion, and a moneary policy rule, he basic framework for simulaion and forecasing is deermined, which is consisen wih he relaionships shown in Diagram Requiremens for simulaion The Copom members exchange views wih he saff and choose relevan possible shocks. These shocks are hen sylized and inroduced ino he srucural models. I is imporan o noe ha given he simplified naure of he macro models, he saff is required o carefully idenify he form, he inensiy, as well as he iming of he inervenions. 22

23 The inroducion of shocks in he simulaion process involves a previous work on how much he economic agens have already anicipaed hem. This is paricularly rue for nominal variables. The Copom members have o carefully assess he se of shocks proposed in he simulaion exercises. Once he Copom defines he relevan shocks and he saff prepares hem for inroducion in he macro models, some addiional definiions are required for simulaion purposes. For insance, i is necessary o make he following choices: he ineres rae rule (a fixed nominal rae or a Taylor-ype rule or a rule based on he deviaions of expeced inflaion from arge or a predeermined rajecory for nominal or real raes); and (ii) a mechanism for he formaion of inflaion expecaions. Given hese definiions, he following resuls can be obained: (i) inflaion forecass (cenral pah and confidence inervals around he median) wih definiions of a measure of dispersion (variance) and of risks (asymmeries); (ii) forecass for oupu; (iii) he rajecory for ineres raes (boh nominal and real) resuling from he various reacion funcions; and (iv) dynamic simulaions of exogenous shocks. Simulaions permi he visualizaion of he ransmission mechanism of moneary policy implici in hese simplified models, wih he ineres rae affecing he nominal exchange rae conemporaneously and he oupu gap wih a lag; he nominal exchange rae affecing he impored inflaion and, hus he inflaion rae conemporaneously; and he oupu gap affecing he inflaion rae wih a lag. The simulaion of he srucural models is based on he selecion of a core scenario ha involves he mos likely hypohesis and a se of alernaive scenarios represening he perceived risks of deparure from he basic hypohesis. A careful assessmen of he various hypoheses is a necessary condiion for balanced decisions on he insrumen of moneary policy. Naurally, he resuls from he simulaion exercises are combined wih oher elemens in making policy decisions. In paricular, forecass canno be limied o hose produced by 17 In boh cases he opimizaion is consrained o he model in use. 23

24 models. Alernaive sources such as marke surveys and forecass, and informaion on inflaion expecaions embodied in financial insrumens need also be considered in he decision process. 5. Concluding remarks Brazil has implemened is inflaion-argeing framework in a very shor ime. Even hough he arge for 1999 has been me, i is oo early o discuss is success, bu here are some crucial poins ha deserve menioning. Firs, IT involves several elemens ha mus be properly addressed: a well defined quaniaive arge for he inflaion rae in he medium erm; an insiuional commimen o his arge as he overriding objecive of moneary policy; increased ransparency of he moneary policy sraegy hrough communicaion wih he public and he markes abou he plans of moneary auhoriies; and increased cenral bank s accounabiliy for achieving is inflaion arges. Second, he cenral bank saff engaged in moneary policy advice should focus iniially in he following issues: ge a clear view of he moneary policy ransmission mechanism and decide which channels are bes suied o be explored in order o mee he inflaion arge; develop simple and small srucural models of hese channels of ransmission, being able o undersand and explain he behavior key macro variables; use a preferred model o discipline he discussion wih he policymakers; monior incoming daa and use judgmenal analysis in he firs forecasing quarers. Communicaion effors in an IT framework are vial. I is fundamenal o explain clearly o he public a large, o financial marke paricipans, and o he poliicians he goals and limiaions of moneary policy (wha he cenral bank can do in he long run is o conrol inflaion; wha i canno do is o raise economic growh hrough moneary expansion); he numerical values of he inflaion arges and on which grounds hey were seleced; and how he arges are o be achieved, given curren economic sance and expeced fuure developmens. 24

25 In Brazil, he moneary auhoriies chose he full disclosure sraegy, in he same line as he Bank of England. The publicaion of Inflaion Repors is an inegral par of he communicaion effors, allowing he general public o undersand and assess he qualiy of he moneary policy decisions, in a coninuous process ha ulimaely leads o earning credibiliy and permis achieving he inflaion arges wih lesser coss. 25

26 References Backus, D.K., and J. Drifill Inflaion and Repuaion. American Economic Review, 75, no. 3 (June). Ball, L Efficien Rules for Moneary Policy. NBER Working Paper No Ball, L Policy Rules for Open Economies, NBER Working Paper No Baini, N., and A.G. Haldane Forward-Looking Rules for Moneary Policy. Bank of England Working Paper no. 91. Blanchard, O.J., and C.M. Kahn The Soluion of Linear Difference Models under Raional Expecaions. Economerica, 48, no. 5 (July). Brion, E., Fischer, P. and Whiley, J The Inflaion Repor Projecions: Undersanding he Fan Char. Bank of England Quarerly Bullein, no. 38. Cenral Bank of Brazil, Inflaion Repor. June, Sepember, and December issues. Dornbusch, R Expecaions and Exchange Rae Dynamics. Journal of Poliical Economy, 84, no. 6 (December). Goldfajn, I. and Werlang, S.R.C The Pass-hrough from Depreciaion o Inflaion: A Panel Sudy, PUC-Rio Working Paper no.423. Haldane, A.G Some Issues in Inflaion Targeing. Bank of England Working Paper no. 74. Svensson, L.E.O Inflaion Forecas Targeing: Implemening and Monioring Inflaion Targes. European Economic Review, 41, no. 6 (June). 26