Monetary Policy Rules and Inflation Targets in Emerging Economies: Evidence for Mexico and Israel

Transcription

1 Moneary Policy Rules and Inflaion Targes in Emerging Economies: Evidence for Mexico and Israel Rebeca I. Muñoz Torres Universiy of Leiceser January 2005 Absrac This sudy analyses he main deerminans of moneary policy in wo emerging economies: Mexico and Israel. For ha purpose we esimae reacion funcions for each counry of he form inroduced by Taylor (1993) and exended by Clarida e al. (1998) in erms of forward-looking rules. Due o he share feaure of hese counries facing currency and financial crises his sudy consider he alernaive o incorporae a financial variable such as he level of foreign reserves in he reacion funcion of he cenral bank. The main resuls sugges ha when seing moneary policy, cenral banks in hese counries look beyond jus inflaion, aking ino accoun oher variables. Moreover, movemens in he exchange rae seems o play an imporan role especially in he case of Israel where here is no a clear commimen o price sabiliy. This involves he inroducion of inflaion argeing principles along wih policies based on fixed or managed exchanges raes. Keywords: Ineres rae rules, Inflaion argeing, exchange raes, GMM Classificaion JEL: E52, E58. I hank Panicos Demeriades, Anhony Garra, Sourafel Girma and seminar paricipans a he Universiy of Leiceser and Universiy of Wesminser for heir valuable commens and discussions. Financially suppor from he Naional Council of Science and Technology (CONACYT) in Mexico is graefully acknowledged. Any remaining errors are mine. Address: KE Building, Deparmen of Economics, Universiy of Leiceser, Leiceser, LE1 7RH, UK. E- mail: rim1@le.ac.uk 1

2 1. Inroducion High levels of inflaion characerised many emerging economies a he beginning of he nineies. As a resul differen mechanisms o aim a reduce inflaion were se in place. In his respec, a variey of moneary policy frameworks have been suggesed. Three opions are he mos sudied ones; he exchange rae regime, he moneary regime and he inflaion-argeing regime (see Mishkin, 1997). Inflaion argeing (IT) however has become increasingly popular in emerging economies where is implemenaion has enhanced moneary policy credibiliy. There is a greaer weigh on inflaion conrol and more willingness o vary ineres raes according o inflaion expecaions (see Schaecher e al., 2000). Counries such as Brazil in 1999, Chile in 1990, he Czech Republic in 1997, Israel in 1991, Mexico in 1994, and Souh Africa in 2000 have adoped inflaion argeing. The success of hese emerging economies in reducing inflaion along wih IT principles, has generaed an increased ineresed in feedback rules for hese counries (see Mohany and Klau, 2004). These rules link shor-erm ineres raes conrolled by he cenral bank o he rae of inflaion and/or is deviaion from he argeed rae. Several sudies focus mainly on eiher individual counries or regional experiences (see Amao and Gerlach, 2002; Corbo, 2002; Bar-Or and Leiderman, 2000). Noneheless, lile aenion has been paid o he comparabiliy of emerging counries wih similar experiences in heir ineres rae seing behaviour and wheher consisency is achieved or no by incorporaing addiional variables o he radiional cenral bank reacion funcion. Corbo (2002) finds ha, in seing heir ineres raes emerging economies end o look beyond inflaion and focus on oher objecives as well. Moreover here is no consensus regarding he imporance of IT and is main requiremens a he ime his framework has been adoped in hese counries. Differen crieria are observed as regards he main prerequisies ha should be considered a he ime IT is adoped (see Agenor, 2000; Schaecher e al., 2000; Masson, e a., 1997). The presen sudy analyses he process hrough which ineres raes have been deermined in wo emerging economies Israel and Mexico hrough he implemenaion of similar procedures o achieve lower and more sable levels of inflaion. The main resuls sugges ha when seing moneary policy, cenral banks in 2

3 hese counries look beyond jus inflaion, aking ino accoun oher variables such as foreign reserves. Specially, movemens in he exchange rae seem o play an imporan role especially in he case of Israel where here is no a clear commimen o price sabiliy. This involves he inroducion of inflaion argeing principles along wih policies based on fixed or managed exchanges raes. The res of he chaper is organised as follows. Secion 2 examines how moneary policy has been conduced in he wo emerging economies. Secion 3 inroduces he idea of moneary policy rules, specifies a baseline ineres rae rule and describes he esimaion procedure. Then, secion 4 presens he empirical resuls and he final secion concludes. 2. The pracice of Moneary Policy in Emerging Economies We can use he erm moneary policy sysem, o denoe some coheren framework for making moneary policy decisions and for explaining hem o he public. In recen years, several emerging economies have been forced o adop more flexible exchange raes as a consequence of serious financial or currency crises. Therefore, hey have had o find a differen nominal anchor o guide domesic moneary policy over he medium and long erm. As a resul, some of hese counries have inroduced inflaion arges and since hen hey have followed successful sraegies for disinflaion during he ransiion owards a fully-fledged IT regime. 1 Bearing his in mind his secion reviews he progress made by wo emerging economies, Israel and Mexico, which have followed similar procedures o achieve moderae levels of inflaion and where he adopion of IT principles seem o be he mos imporan aspec. Israel is a good example of how IT has been adoped during a process of disinflaion along wih a crawling exchange rae band. On he oher hand, Mexico has been considered as one of he few cases of a Lain American counry moving owards he floaing exchange rae regime in conjuncion wih an implici inflaion arge sysem. The sance regarding he exchange rae policy emerges as he mos imporance difference. This issue has been considered essenial no only in he case of emerging economies bu also as one of he main requiremens o successfully implemen IT. 1 This seems no o be he case for indusrialised counries where disinflaion had largely been compleed by he ime of he IT adopion. 3

4 I is imporan o noe ha alhough more emerging economies have recenly inroduced IT principles (see Fig. 1) heir shor experience consiues a resricion o carry ou reliable economeric analysis. Noneheless, he experience of Israel and Mexico may be considered as pioneering cases for emerging economies in heir effor o reduce high levels of inflaion. 2 Furhermore, he differen posure concerning he exchange rae policy in hese wo counries consiues an ineresing issue o be analysed. The following secions briefly inroduce how moneary policy has been recenly conduced in he wo seleced counries in heir effor o reduce inflaion. Figure 1. Inflaion rae a adopion of IT in Emerging Economies Chile Inflaion rae (%) Israel Mexico Poland Czech Republic Souh Africa Korea Brazil 0 Dec-88 May-90 Sep-91 Jan-93 Jun-94 Oc-95 Mar-97 Jul-98 Dec-99 Apr-01 Beginning of ransiion o IT 2.1. Mexico Afer nearly a decade of sluggish economic aciviy and high inflaion, he Mexican economic auhoriies liberalised he rade secor in 1985, adoped an economic sabilisaion programme, known as he Paco a he end of 1987, and gradually inroduced marke-oriened insiuions. A he beginning of 1988, he nominal exchange rae was fixed and became he main anchor of he ani-inflaionary policy. 2 Chile is considered as he firs emerging economy ha announced inflaion arges in Chile s experience however has been he mos horoughly sudied and is implemenaion relies more from he pracice of developed counries. 4

5 Neverheless, he vulnerabiliies accumulaed during his process years of large capial inflows and financial liberalisaion plus he negaive exernal and domesic shocks faced by he economy during 1994, led o he currency and financial crisis of December The cenral bank, under severe pressure in he foreign exchange rae marke, decided o le he peso floa (Gil-Diaz and Carsens, 1996). Several measures were implemened during To conain he inflaionary effecs of he devaluaion, a igh moneary policy was adoped. To resore he credibiliy of he Bank of Mexico, he auhoriies reieraed publicly ha he primary objecive of is moneary policy was o sop he inflaionary effecs of he peso depreciaion and o rapidly bring down inflaion o moderae levels. The bank also commied o increased ransparency of is own acions by making available more imely informaion on is balance shees as well as oher moneary and fiscal indicaors. Following he measures aken by he governmen, he peso became more sable, ineres raes began o fall, economic aciviy recovered rapidly and inflaion came down o manageable levels. Inflaion declined from 51.7 percen in 1995 o 12.3% in 1999 and he cenral bank announced, for he firs ime, a 10% inflaion arge for he year 2000 before he Minisry of Finance submied o he Congress he economic programme for he year (see Fig. 2). This move conribued o he increase of accounabiliy of he cenral bank regarding is inflaion objecives. In addiion, he Bank of Mexico sared publishing inflaion repors in Afer he implemenaion of hese measures i can be said ha Mexico has moved o a fully-fledged IT regime (see Corbo and Schmid- Hebbel, 2001; Schmid-Hebbel and Werner, 2002) Israel Inflaionary developmens in Israel are closely linked o exchange rae policy and demand pressure. A he beginning of he 1980s, he rae of inflaion rose persisenly and a increasing rae, reaching hree-digi levels. This coninued unil mid-1985, when he Economic Sabilizaion Programme (ESP) was inroduced. The principal role of moneary policy under he ESP was o defend he exchange rae policy (Haldane, 1995). The exchange rae agains he dollar was se as he nominal anchor for prices. Neverheless, he Israeli currency (shekel) was devaluaed several imes beween 1986 and In March 1990, he widh of he band was increased o ± 5 5

6 percen and in December 1991 he horizonal band was replaced by a crawling-band regime. The crawling-band regime was inended o reduce uncerainy wih regard o he developmen of he real exchange rae, in order boh o benefi he business secor and o reduce speculaive capial movemens. Since hen, he widh of he exchange rae band has been increased on several occasions and he focus of moneary policy has shifed oward inflaion argeing (Leiderman and Bufman, 2000). 3 An explici medium-erm orienaion for moneary policy was esablished in 1996 when he governmen announced a long-erm objecive o bring inflaion down o he average rae of he Organizaion for Economic Cooperaion and Developmen (OECD) counries by In 1999, he Minisry of Finance announced wo-year arges of 3-4 percen for 2000 and The Bank of Israel only recenly sared publishing semi-annual inflaion repors (March 1998), bu i refrains from making explici quaniaive inflaion forecasing. Overall, Israel s experience wih inflaion argeing can be broadly considered a success. Wih he excepion of he large overshooing of he inflaion arge in 1994, since 1992 he inflaion arges have eiher been me or have been exceeded by less han 1 percen. Furhermore, he shekel has been consisenly hiing he lower (sronger) end of is crawling band. In shor, Israeli inflaion argeing provides a imporan example of how moneary policy can be made more flexible wih no apparen loss of credibiliy and effeciveness. In sum, he conduc of moneary policy in hese wo counries suggess ha inflaion argeing can become a successful medium-erm sraegy for moneary policy in emerging economies (see Fig. 2). A iniial sages while credibiliy was being buil and inflaion was being gradually reduced, here was a ransiion period where he inflaion arge was no explici and was usually saed as achieving a gradual reducion of inflaion owards indusrial counries levels. Once enough progress was made in reducing inflaion, hese counries sared o use explici arges publicly announced a he end of he previous year. They adoped IT as a means of balancing he uncerainies of heir exernal economic environmen, paricularly he behaviour of he exchange rae, wih he need o anchor he public s inflaion expecaions. Israel oped for a crawling exchange rae band which has been gradually widened and 3 The mid-poin rae is adjused on a daily basis according o a presen gradien, deermined by he gap beween Israel s inflaion arge and expeced inflaion in is rading parners. 6

7 Mexico was forced o adop a floaing exchange rae as a resul of a severe pressure on is currency. Figure 2. Inflaion Raes in Israel and Mexico ( ) Widening Ex-Rae Band Inflaion rae ( % ) 12 8 Fixed Ex-Rae Regime 'ESP' Crawling-band regime Fully-fledged IT Israel 4 0 'Israeli Currency Devaluaion' (Shekel) Year Fixed Exchange Rae Regime 'Paco' Inflaion Rae (%) 'Tequila crisis' Flexible Exchange Rae Regime Mexico Fully-fledged IT Year The role of he exchange rae under IT has been a debaable issue. A counry ha chooses a fixed exchange rae sysem subordinaes is moneary policy o he exchange rae objecive and consequenly moneary policy jus suppors ha rae. Under a floaing exchange rae, on he oher hand, moneary policy is no consrained by any rule. In oher words, moneary policy serves as nominal anchor of he 7

8 economy. In heory an exchange rae arge could coexis wih an inflaion arge so long as he moneary auhoriies make clear ha he laer objecive has prioriy if a conflic arises (Masson, Savasano and Sharma, 1998). In his respec, he analysis of ineres rae rules o evaluae he main deerminans of moneary policy in hese wo counries seems o be an appropriae approach o evaluae no only he role of he exchange rae bu also addiional variables ha may influence he sance of he moneary policy seing ineres raes. The following secion inroduces and discusses in more deail his procedure. 3. Moneary policy rules: Saisical framework and daa Commonly, moneary policy analysis has been approached in wo ways. The firs one requires an undersanding of he mechanisms hrough which moneary policy affecs he economy. In oher words, moneary auhoriies mus have a precise assessmen of he iming and effec of heir policies on he economy. This ransmission mechanism includes ineres rae, exchange rae and oher asse price effecs, and he so-called credi channel. The second approach, on he oher hand, evaluaes he condiions under which moneary auhoriies reac o change heir moneary policy. This idea can be summarized in erms of a policy reacion funcion ha indicaes how policy will reac o he shocks and coningencies ha will ineviably hi he economy (see Cecchei, 1998). The success of many counries in reducing inflaion, ogeher wih he adopion of formal inflaion arges by a growing number of cenral banks, has generaed an increased ineres in his kind of feedback rules for IT. Policy rules are seen as a guide according o which discreion should be used. Amao and Gerlach (2002), for example, evaluae IT in emerging and ransiion economies finding ha an imporan disinguishing feaure of inflaion argeing is ha i leads o a more sysemaic ineres rae response by he cenral bank o inflaion. To analyse he main deerminans of moneary policy, his sudy focuses on he laer approach Baseline specificaion Two ypes of rules have been he mos sudied ones in recen empirical work: (1) The inflaion-forecas-based rule, and he (2) Taylor rule, which is by far he mos popular and he one his sudy is based on. Rules based on inflaion forecass make he change 8

9 in he policy insrumen a funcion of he deviaion of a condiional forecas of inflaion in some fuure period from he arge rae of inflaion, as follows: r αr + γ [ E ( π ) ] (1) = 1 + k π The condiional inflaion forecas serves as a feedback variable, and he inflaion arge dicaes he necessary degree of insrumen adjusmen. This rule allows policymakers o adjus he horizon of he inflaion forecas, depending on he lengh of he ransmission lag for moneary policy and i uses all relevan-informaion for he predicion of inflaion 4. Neverheless, here are poenial difficulies such as he risk of indeerminacy and he inabiliy o disinguish beween demand and inflaion shocks. Taylor (1993), on he oher hand, proposed a feedback policy of he following form: 5 r = r + α ( π π) + (1 α)( y y) 0< α < 1 (2) where r is he long run equilibrium real ineres rae, π is he arge inflaion rae and y is he poenial oupu. This rule indicaes ha when he economy is in equilibrium, ha is, when he inflaion rae is equal o is arge rae and oupu is equal o poenial, he real ineres rae is also in equilibrium. Taylor s rule says ha he cenral bank should raise he ineres rae when inflaion and oupu rise above heir arge levels, wih he inflaion response being somewha greaer han he oupu response. Conversely, when inflaion or oupu falls bellow is arge, he cenral bank should reduce ineres raes. The heoreical foundaions of he rule proposed by Taylor has been formalised by several auhors. Svensson (1997), Clarida e al. (1999) and Woodford (2001), for insance, have shown ha approximaions o his rule can be obained from an opimizaion process where he cenral bank minimises a quadraic loss funcion in inflaion deviaions from is arge and he oupu gap, subjec o a sandard New Keynesian macroeconomic framework. In paricular, his sudy follows he work 4 The inflaion-forecas-based rule is differen from he Svensson s inflaion-argeing rule. Svensson argues ha he cenral bank should use an inflaion forecas as an explici inermediae policy arge. Moreover, Svensson s approach requires ha he cenral bank solve an opimal-conrol problem. 5 For a more deailed analysis of he hisorical evoluion of moneary policy rules and heir implicaions, see Taylor (1999). 9

10 done by Clarida e al. (1999, 1998) where he auhors posulae a forward-looking version of he rule proposed by Taylor. Specifically, [ E( π Ω ) π )] + [ E( y Ω ) y ] r = r + + n γ + k + k β (3) where r denoes he long run equilibrium nominal ineres rae 6, π indicaes he + n percen change in he price level beween periods and +n (expressed in annual raes) and y is real oupu. The variable he arge for inflaion and r is he ineres rae se by he cenral bank; π is y is given by poenial oupu, defined as he level ha would arise if wages and prices were perfecly flexible. In addiion, E is he expecaion operaor and Ω is he informaion available o he cenral bank a he ime i ses ineres raes. 7 The policy rule given by eq. (3) has some applicaions in boh heoreical and empirical grounds. Empirically, his reacion funcion offers a reasonable good descripion of he way major cenral banks have performed recenly. Clarida e al. (1998), for insance, esimae various reacions funcion o evaluae moneary policy in Europe. They applied a forward-looking version of he Taylor rule, which provides evidence o suppor ha heir baseline forward-looking specificaion works quie well agains various alernaives. Clarida, e al. (2000), esimae a forward-looking moneary policy reacion funcion for he poswar US economy and consider how his rule evolved over ime. In paricular, ineres rae policy in he Volcker-Greenspan period appears o have been much more sensiive o changes in expeced inflaion han in he pre-volcker period. Few sudies can be found in he case of emerging economies. Mohany and Klau (2004) for example, evaluaes cenral banks ineres rae seing behaviour in a group of emerging economies. They found ha in mos of hese counries he ineres rae responds srongly o he exchange rae and in some cases he response was higher han o changes in he inflaion rae or he oupu gap. Anoher sudy by Corbo (2000) which esimaes reacion funcions in erms of forward-looking rules for six Lain American counries showed ha jus Chile uses 6 By consrucion i is he desired nominal rae when boh inflaion and oupu are a heir arge levels. 7 I is highly possible ha when chosen he arge ineres rae, he cenral bank may no have direc informaion abou he curren values of eiher oupu or he price level. The specificaion proposed by Clarida e al. (1998) allows for his possibiliy. 10

11 moneary policy wih a clear commimen o achieve he arge inflaion. 8 Theoreically, policy implicaions given by eq. (3) for he cyclical behaviour of he economy will depend on he sign and magniude of he slope coefficiens, β and γ. To illusrae he basic inuiion, consider he implied arge for he real ineres rae, given by: rr, [ E ( π Ω ) π )] + [ E ( y Ω ) y ] rr = rr + ( β 1) γ + + (4), k k k where rr r E π Ω ) and (, k rr r π is he long run equilibrium real rae. I is assumed ha he real rae is saionary and is deermined by non-moneary facors in he long run. As a resul, r r is a consan and is independen of moneary policy 9. As eq. (4) makes clear he sign of he response of he real rae arge o changes in expeced inflaion and he oupu gap depends on wheher β is greaer or less han one and on he sign of γ, respecively. To he exen ha lower real raes simulae economic aciviy and inflaion, ineres rae rules characerized by β>1 will end o be sabilizing while hose wih β 1 are likely o be desabilizing or, a bes, accommodaive of shocks o he economy 10. Thus, he coefficien β urns ou o be a key parameer a he momen of assessing cenral bank s response. If a moneary policy rule like eq. (4) wih β > 1 offers a good approximaion o he process hrough which ineres raes are deermined, hen moneary policy works as an auomaic sabiliser of inflaion around is arge. In oher words, moneary policy plays he role of being a nominal anchor for he economy. In he same sense when a rule like eq. (4) wih γ>0 offers a good approximaion o he process hrough which ineres raes are deermined, moneary policy is said o be an auomaic sabilizer of oupu around is poenial level. In addiion, his specificaion allows he conrol of inflaion and he sabilisaion of oupu o be idenified as independen objecives 11 and considers all he informaion available a he ime he cenral bank se ineres raes. 8 The res of he counries are: Colombia, Cosa Rica, El Salvador, Nicaragua and Peru. 9 If he analysis is carried ou among differen sub periods here exiss he possibiliy of having shifs over ime in he long-run equilibrium rae. Therefore, he real ineres rae is consan only wihin he sub period esimaed. 10 In his case alhough he cenral bank raises he nominal ineres rae in response o a rise in inflaion, i does no increase i sufficienly o keep he real rae from declining. 11 In he radiional Taylor rule model, for example, he cenral bank reacs o lag inflaion as opposed o expeced inflaion. Then i is no clear if he cenral bank can respond independenly o inflaion and he oupu gap. 11

12 Ineres rae smoohing The policy reacion funcion given by eq. (3) assumes an immediae adjusmen of he acual ineres rae o is arge level. In oher words, i does no allow for any uncerainy in policy acions oher han ha associaed wih incorrec forecass of he economy. I also assumes ha he moneary auhoriies have perfec conrol over ineres raes. Cenral banks, however, adjus ineres raes more cauiously owards is desired level han sandard models predic (Goodfriend, 1991). To allow for his possibiliy, a parial adjusmen equaion is considered. 12 In paricular, Clarida e al. (2000,1998), proposed he following relaionship for he acual ineres rae, r : r = ( 1 ρ ) r + ρ( L) + υ (5) r 1 where he parameer ρ (0,1) reflecs he degree of lagged dependence on he ineres rae or he degree of ineres rae smoohing. The specificaion also includes an exogenous random shock o he ineres rae, υ ha i is assumed o be iid (0,1). The ineres rae arge r is given by eq. (3). Paricularly, each period he cenral bank adjuss he ineres rae o eliminae a fracion (1-ρ) of he gap beween is curren arge level and is pas value. In addiion, o obain an esimable equaion he auhors define: α r βπ and x y y. Then combining he parial adjusmen eq. (5) wih he arge model (3) yields he following policy reacion funcion: r = 1 ) { α + β E[ π + n Ω] + γ E[ x Ω]} + ρ r 1 ( ρ + υ (6) Finally, assuming ha expecaions are raional expeced values are replaced by realised values in order o obain he following equaion: r = 1 ) α + (1 ρ ) β π + n + (1 ρ ) γ x + ρ r 1 ( ρ + ε (7) 12 Oher explanaions for slow adjusmen of ineres raes include fear of disruping financial markes and disagreemen among policy-makers. 12

13 where 1 ρ) { β ( π + n E[ π + n Ω]) + γ ( x E[ x Ω])} ε ( + υ is a linear combinaion of he forecas errors of inflaion and oupu and he rue disurbance υ Esimaion echnique The esimaion of eq. (6) is no sraighforward as some of he variables considered in he righ hand side of he equaion may be endogenous. Moreover misspecificaion errors may be expeced. These difficulies invalidae convenional leas squares sandard errors since he cov (π,ε ) 0 and/or cov(x, ε ) 0. To obain unbiased and consisen esimaors he model is esimaed using he General Mehod of Momens (GMM). 13 In his respec, le U symbolize a vecor of insrumens known when r is se. Suiable insrumens include any lagged variables ha help forecas inflaion and oupu, as well as any conemporaneous variables ha are uncorrelaed wih he curren ineres rae shock υ. These sraegy implies a se of orhogonaliy condiions, which provide he basis for he esimaion of he parameer vecor (α,β,γ,ρ) ha are given by E[ε U ] = 0. Combining his condiion wih eq. (7) gives us he explici se of resricions used by GMM: E[ i (1 ρ ) α (1 ρ) β ( π ) (1 ρ) γ ( x ) ρ i 1 U ] = 0 (8) + n I should be emphasized ha GMM naurally provides a suiable economeric framework for esing he validiy of he model (see Hansen and Wes, 2002; Jagannahan e al., 2002). When he number of insrumens is larger han he number of parameers o be esimaed, he model is said o be over idenified. In his case, Hansen s J-es is used o es if he over idenifying resricions hold. Addiional variables So far, he discussion has no considered he inclusion of addiional variables o he baseline equaion. In pracice, he oupu gap and inflaion are no he only relevan objecives from a pracical poin of view. Ball (1999) and Svensson (2000) for insance argue ha moneary policy rules should ake ino accoun oher variables 13 For a deailed descripion of his mehod see Greene (2000) and Enders (1995). 13

14 such as foreign ineres raes or he exchange rae which migh reflec uncerainies abou fuure inflaion faser han expeced inflaion or he expeced oupu gap 14. In addiion, some cenral banks in emerging economies, such as he cenral bank of Chile, Czech Republic and Israel have a formal objecive wih regard o currency sabiliy. I is herefore useful o examine he consequences for moneary policy of inroducing addiional variables in he policy makers objecive funcion. To his end a number of simple alernaives o he baseline model are considered. Le z denoe a variable oher han inflaion or oupu ha may poenially influence rae seing (independenly of is use for forecasing). For each alernaive specificaion, he following equaion is uilised: r = 1 ) { α + β [ π + n π + n ] + γ [ x+ k ] + ϕ [ z+ j z+ j ]} + ρ r 1 ( ρ + ε (9) The alernaive model is hen esimaed in he same way as in he baseline model. This specificaion allows us o evaluae wheher he direc effec of z on policy is quaniaively imporan. Addiional variables commonly consis of changes in he nominal exchange rae, money, foreign ineres raes or any oher variable, which may have influence in he process hrough which ineres raes are deermined. The variables included in his sudy are changes in nominal exchange raes and money supply, foreign reserves, a lending boom indicaor and he curren accoun. 15 The change in prices over he previous year is also considered in order o ge a direc es of he forward-looking specificaion versus he backward-looking one. We examine each of hese variables below. I is plausible o expec ha a variable ha would be imporan in he conex of small open economies would be he exchange rae. 16 Recen work by Ball (1999) and Svensson (2002) on small open economies models has examined he role of he 14 Clarida, Gali and Gerler (2001) however presen a small, open-economy model in which he effec of he exchange rae on oupu and inflaion is capured in he parameers of he policy rule. The difference is in he quaniaive response (magniude of he parameers β and γ) of ineres rae o expeced inflaion and he oupu gap. The inuiion behind his argumen is ha if movemens in he exchange rae will have an effec in inflaion and/or oupu, hen hese effecs should be capured by changes in he expeced deviaion of inflaion from is arge and in he expeced oupu gap. 15 Choosing he variables ha could have played a role when seing moneary policy, we draw on he work of Kaminsky and Reinhar (1996, 1997) on he economic indicaors for banking and currency crises. 16 When he pass-hrough of he exchange rae ino prices is high, he exchange rae is likely o assume special imporance for moneary policy. 14

15 exchange rae in moneary policy rules explicily. Their resuls highligh he idea ha a depreciaion of he currency calls for an increase in he ineres rae 17. On he oher hand, Taylor (2001) argued ha alhough he exchange rae is imporan, paricularly in emerging economies, any reacion of he ineres rae o his variable would be indirec, hrough is effec on inflaion and he oupu gap. Addiionally, cenral banks may reac o changes in he exchange rae o mainain financial sabiliy raher han price sabiliy. In his respec, Calvo and Reinhar (2002) for insance, aribue such fear of floaing behaviour o he high-risk premium hey have o pay because of heir low insiuional and policy credibiliy. Cenral banks in emerging economies are also ypically concerned wih financial sabiliy. Thus, hree more variables are considered in he augmened ineres rae rule as poenial indicaors for exernal crisis. The firs one is foreign reserves. Many emerging economies affeced by currency crises in he 1990s los a large par of heir foreign reserves rying o defend he value of heir currencies and evenually, hey had o abandon he fixed exchange rae (Kaminsky e al., 1997). Due o his reserve rundown associaed wih he financial crisis, counries have been accumulaing reserves even hough hey have adoped more flexible exchange rae regimes. This srenghening of reserves may be seen as a mehod of building up credibiliy and herefore may be relaed o he process of having more sabiliy. Hence, incorporaing foreign reserves in he reacion funcion of he cenral bank migh resul in a monioring variable for emerging marke economies regarding financial sabiliy. The second variable is consruced as he raio of he logarihm of he claims on privae secor o Gross Domesic Produc (GDP). This lending boom variable indicaor has been mainly used o explain many banking crises, including Chile s in 1982 and Mexico s in During a boom, credi o he privae secor increases rapidly making he banking secor more vulnerable (see Gourinchas e al. 2001). Moreover, i can be argued ha he higher he raio of his variable, he higher he poenial of economic growh. Therefore more pressure on fuure inflaion and consequenly he need of raising ineres raes. The hird variable is he curren accoun defici. The curren accoun is ypically used as one of he main leading indicaors for fuure behaviour of an economy and is imporance has been mainly highlighed in emerging 17 The significan increase in he ineres raes in Mexico afer he currency depreciaion a he end of 1994 migh be an example of such a reacion. 15

16 economies. Corbo (2002) for insance, finds ha cenral banks in Lain America show an imporance preference o sabilise curren accoun deficis in he balance of paymens even hough his is no announced explicily. I is worh menioning ha he addiion of hese variables consiues mainly a shorcu o avoid a very complex model in which oupu and inflaion may be affeced by hese facors. Table A1 in he appendix summarises various sudies on moneary policy rules, which have exended he radiional version of he Taylor rule. These sudies consider no only he formulaion of forward-looking rules bu also he inclusion of addiional variables. 3.3 Daa The daa are quarerly ime series spanning he period 1982Q1-2001Q4 in he case of Mexico and he period 1988Q1-2001Q4 in he case of Israel. Daa is mainly aken from he Inernaional Financial Saisics (IFS) CD-Rom regarding oupu, inflaion, ineres raes and foreign reserves and from he respecive cenral bank s websies. 4. Empirical resuls The saring equaion of he empirical analysis is: r = 1 ) α + (1 ρ ) β π + n + (1 ρ ) γ x + ρ r 1 ( ρ + ε (10) where β is he inflaion coefficien expeced o be greaer han one and γ he oupu gap coefficien expeced o be greaer han zero. The oupu gap is calculaed using he Hodrick-Presco filer and he Core Price Index (CPI) o calculae he rae of inflaion. 18 Three differen specificaions are considered. The firs one uses acual values insead of expeced of forecas ones (Baseline model). The main reason is he lack of reliable daa for he enire sample period in he absence of official and 18 Following he lieraure, he mos common mehod o consruc an oupu gap series is based on he Hodrick-Presco filer. In his respec, poenial oupu is deermined as he oupu level ha simulaneously minimises a weighed average of he gap beween acual and poenial oupu and he rae of change of he oupu rend. In addiion o his mehod a linear and a quadraic rend are used o obain alernaive measures of he oupu gap. The series are consruced as he deviaion of he naural logarihm of he real GDP from is poenial value using he alernaive mehods. Alernaive measures of inflaion, on he oher hand, are also aken ino accoun. These include a measure of he core inflaion, which excludes he relaively more volaile componens of he CPI and he GDP deflaor. The unemploymen rae is also considered when available. 16

17 commonly acceped forecass. In he second specificaion acual values of he explanaory variables are replaced by heir deviaions from rend values (Gap model). This is mainly because of he shor experience of hese counries wih a fully-fledged IT and herefore wih he lack of sufficien daa for he announced inflaion arges. The las specificaion includes daa on inflaion expecaions available in each counry (Forward-looking model). As i will be discuss laer on his implies however he esimaion of eq. (10) for a shorer sample period. The analysis is based on he primary assumpion ha wihin shor samples all variables are saionary (see ables A2 and A3 in he appendix). 4.1 The case of Mexico The Mexican experience is ineresing for wo reasons. Firs, i represens an emerging economy under a floaing exchange rae regime ha has been able o reduce inflaion o one-digi levels. Second, he Mexican experience provides an ineresing case of sudy for oher developing counries considering he possibiliy of moving owards he floaing exchange rae regime in conjuncion IT principles. In oher words, i is paricularly ineresing because he adopion of inflaion argeing has been a useful mechanism for imposing discipline on moneary policy Baseline specificaion To accoun for imporan moneary policy reforms a dummy variable is aken ino consideraion. This dummy permis us o evaluae issues regarding srucural sabiliy wihou he need o break he sample up which would be problemaic given he shor sample period. 19 In he case of Mexico, he dae seleced considers he differen measures implemened by he moneary auhoriies as a resul of he currency and financial crisis a he end of These measures include cenral bank independence, he adopion of a floaing exchange rae, he commimen o price sabiliy and he 19 According o Clarida e al., 2000 he sample period mus conain sufficien variaion in inflaion and oupu and mus be sufficienly long in order o idenify he slope of he coefficiens in he policy reacion funcion. Chuecos (2003) and Corbo (2002), however esimae policy rules wihin shor samples arguing sufficien variabiliy in heir esimaions. 17

18 announcemen for firs ime of inflaion annual objecives. 20 In his respec, he dummy equals 0 for he pre-reform period; ha is, before 1994 and akes he value of one oherwise. The dummy variable is included in he esimaed equaion as a consan as well as ineracing wih each of he explanaory variables. In he former case, i.e. i = c1 + c2 π + n + c3 x + c4 dummy + ρ i 1 + ε he dummy urns ou o be saisically insignifican bu is inclusion improves he fi of he policy rule. Alernaively, when he dummy is ineracing wih oher erms, i.e. i = c1 + c2 π + n + c3 x + c4 dummy + c5 dummy π + n + c6 dummy x + ρ i 1 + ε jus he erm when he dummy variable is ineracing wih he inflaion erm becomes saisically significan. 21 This resul suggess no only he exisence of some kind of srucural break bu also provides informaion of how effecive has been he implemenaion of differen sraegies in reducing he levels of inflaion (β value). Thus, he following ineres rae rule is esimaed: 22 r = 1 ) α + (1 ρ) β π (1 ρ) γ x+ 2 + (1 ρ) δ dummy95 π ρ r 1 ( ρ + ε (11) The resuls of his baseline equaion are presened in he firs row of able As menioned before, β urns ou o be a key parameer a he momen of assessing cenral bank s response. In his case, he parameer β is posiive and saisically greaer han one. 24 This resul suggess ha moneary policy in Mexico, hrough is effec on ineres rae, has effecively sabilised he economy. The cenral bank does no accommodae inflaionary pressures. Concerning he smooher parameer ρ, is high value suggess no only considerable ineres rae ineria bu also ha he cenral bank is concerned abou smoohing adjusmens. In oher words, a value abou 0.80 implies ha he iniial adjusmen (same quarer) in he ineres rae is only 20%. 20 A dummy variable for 1988 o accoun for he oil-price shock is also considered. Neverheless jus in few cases his is significan and he resuls do no improve considerably. 21 Differen specificaions were carried ou. In almos all of hem he dummy boh as a consan and ineracing wih he oupu gap urns ou o be saisically insignifican. 22 Alhough he Bank of Mexico does no direcly deermine ineres raes, heir behaviour is seered by he sance of moneary policy. Several auhors assume shor ineres raes as he insrumen of moneary policy alhough hey embrace differen moneary policy regimes. This is he case for example of Aron and Muellbauer, 2002; Corbo, 2002; Merha, 1999 and Clarida, e al., Two ses of insrumens are considered. The firs one includes one o four-period lags of he explanaory variables, in oher words, of he inflaion rae, ineres rae and oupu gap. The second se of insrumens adds o he firs se, lagged informaion on he changes in he nominal exchange rae (depreciaion rae) and moneary aggregaes. The resuls sugges a beer fi of he policy rule when he second se of insrumens is used. 24 The Wald es is used o verify ha he value of he parameer is saisically greaer han one. 18

19 Parameer γ alhough has he expeced sign is saisically no differen from zero. 25 The value of he dummy variable urns ou o be posiive and saisically significan suggesing imporan changes in he srucure of he economy as a consequence of he currency crisis. Is posiive value also reinforces he role of moneary policy as a nominal anchor afer 1995 when he moneary auhoriies made explicily heir commimen o price sabiliy. Anoher way o evaluae he baseline moneary policy rule is o compare he esimaed ineres rae rule wih acual ineres raes. Tha is, o compare presen values wih hose ha would have been observed if during he sample period he ineres rae had been deermined by he esimaed policy rule. Figure 3 shows he acual and simulaed values for he ineres rae. Overall, he moneary policy rule does a good job in following he direcion of acual ineres raes. In general he fied values are above he curren ones. In he pos-reform period (afer 1995) his paern becomes clearer. These resuls can be inerpreed as suggesing a more expansionary moneary policy han wha i should have been if i had been deermined in accordance wih he proposed reacion funcion. Resuls from he baseline case sugges ha moneary policy in Mexico has been consisen wih he objecive of achieving lower levels of inflaion. The following secion, however, considers he inclusion of oher variables as addiional objecives in he cenral bank s reacion funcion which may describe beer how ineres raes are deermined in Mexico. Augmened moneary policy rules for Mexico An augmened moneary policy rule is used o es he role of macroeconomic variables oher han inflaion and oupu. For his purpose, he following equaion is esimaed: r = 1 ) α + (1 ρ) β π+ 3 + (1 ρ) γ x+ 2 + (1 ρ) δ d95 π+ 3 + (1 ρ) ϕ z + ρ r 1 ( ρ + ε (12) 25 Torres, A. (2002) and Marinez, Sanchez and Werner (2001) found similar resuls. Their sample period however is shorer saring in

20 where z represens any variable, oher han inflaion and oupu ha may influence he process hrough which ineres rae is deermined. Figure 3. Acual ineres rae vs. dynamic ineres rae rule Baseline specificaion Dynamic 'ineres rae rule' Acual ineres rae The backward-looking componen of moneary policy To check wheher ineres rae are deermined more in a forward-looking han backward-looking manner he augmened ineres rae rule is esimaed defining z as he observed lagged inflaion ( π 1 ). If ineres raes are deermined in a relaively backward-looking model he nominal ineres rae would be expeced o rise enough o increase he real ineres rae (φ>1) raher han when inflaion is generaed in a forward-looking form. The resuls presened in able 1 sugges ha here is a significan conribuion of he lagged value of inflaion in he deerminaion of ineres raes in Mexico alhough his is no good enough o sabilise he economy (φ<1). The same performance is observed wih he value of he parameer β. In oher words, β becomes significanly less han one suggesing ha he response from he auhoriies o inflaion shocks is no sufficien o sabilise he economy. Finally, he significan decrease in he value of parameer β once he lagged value of inflaion is included in he esimaion suggess eiher high persisence of inflaion or ha par of he informaion is already incorporaed in he explanaory variables. 20

21 Money and he exchange rae Following Taylor (2001), he ineres rae reacion funcion esimaed includes he change in he nominal exchange rae where an increase means depreciaion and vice versa. 26 Resuls repored in able 1 indicae ha parameer φ is posiive and saisically differen from zero. However, β is boh saisically no differen from zero and no longer greaer han one. Parameer γ remains insignifican. I seems ha par of he informaion enclosed in his variable is also incorporaed in he observed inflaion and oupu gap. This in an ineresing resul in he conex of a floaing exchange rae regime where he cenral bank seems o reac o effecively mainain price sabiliy raher han o mainain a specific level of he exchange rae. Base specificaion (3.85) Table 1. Moneary Policy Rule: Augmened esimaions 1 Mexico: α β pre-reform γ δ φ ρ J-es (5.28) 0.34 (2.26) (13.55) (0.78) Backwardlooking (2.09) (2.12) (0.30)) 0.76 ( 2.85) 0.84 (5.54 ) 0.67 (10.93) (0.72) Depreciaion ^ 8.43 rae (1.73) 0.50 ( 0.74) (0.45) 0.61 (1.04 ) 6.38 (2.06) 0.75 (8.35) (0.72) 2.73 Money growh ( 0.42) 1.83 (6.26) (0.61) 0.02 ( 0.06 ) 0.29 ( 2.70) 0.81 (13.41) 9.28 ( 0.81 ) 4.18 Foreign reserves ( 0.65 ) 1.41 (8.01) ( 0.42) 1.15 (2.14) 0.88 ( 3.71) 0.80 (21.06) Lending boom indicaor ( 1.15) 1.15 (5.87) 4.25 ( 0.08) 0.39 ( 1.22 ) (1.18) 0.78 (13.36 ) (0.84) Curren accoun ( 2.62) 1.65 (9.71) ( 0.82) 0.84 (3.49) ( 3.16) 0.63 (13.57 ) (0.92) 1 The numbers in parenhesis are -saisic values, excep in he J-es column where hey represen he p value o rejec he hypohesis ha over-idenifying resricions hold. Saisically significan a 95% confidence level ^ Saisically significan a 90% confidence level The money supply, alernaively, is added o he regressors of he baseline equaion. In his case z is defined as he annual variaion of he sum of money and quasi money 26 Taylor (2001) assumes ha cenral banks respond o he level raher han o he change of he nominal ineres rae. 21

22 (firs log difference). 27 The resuls show ha he parameer φ urns ou o be negaive. This is conrary o wha one would expec since higher money supply growh should moivae an increase in he ineres rae in order o preven inflaion from rising. Curren accoun, foreign reserves and lending boom indicaor Afer several aemps o reduce inflaion o sable raes, Mexico has evenually achieved his goal even experiencing profound economic crises. So, in paricular, his secion invesigaes if he cenral bank behaviour has changed and if i is valuable o consider prevenion variables in is policy reacion funcion. The resuls of he augmened moneary policy rule are presened in he second par of able 1. As in he baseline case, parameer β is saisically larger han one and parameer γ remains saisically no differen from zero. Regarding he lending boom indicaor, his seems no o conribue a he ime he cenral bank has o make a decision concerning is moneary policy posiion. Noneheless his is no he case for he change in foreign reserves and he curren accoun defici alhough boh coefficiens presen unexpeced signs Gap model An alernaive version of he baseline model is esimaed where acual values of he explanaory variables are replaced by heir deviaions from rend values (gap model). In his respec he Hodrick-Presco filer is used o measure he rend values. This is mainly because of he shor experienced wih a fully-fledged IT sysem and herefore wih he lack of sufficien daa for he announced inflaion arges. 28 In addiion, i is imporan o noe ha he inflaion arge is no considered o be consan over ime as in Clarida e al. (1998). This value could be ime-varian for a counry ha has an objecive o reduce inflaion, gradually, owards indusrial counry levels. Poor resuls however are obained from hese esimaions where parameers become saisically insignifican and wih he wrong signs. These resuls may be because of he high insabiliy observed in he main variables a he beginning of he 80s. To address his issue he gap model is reesimaed for a shorer period of ime ( ). The 27 Including he quarerly change in he esimaion of his variable does no aler he resuls. 28 The sample period under esimaion is longer han he one under his moneary policy framework. 22

23 resuls presened in able 2 in general suppor he previous findings. Parameer β is always greaer han one and saisically significan suggesing he commimen of he moneary auhoriies o sabilise he economy even before is ransiion o a fullyfledged IT. Two excepions are when he depreciaion rae and he lending boom indicaor are added suggesing ha hese variables conain informaion already given in he main variables. In all he augmened policy rules he signs of all esimaed parameers are he expeced ones and saisically significan. Parameer ρ decreases in his sample period suggesing ha approximaely half of he iniial adjusmen is achieved in one quarer. Parameer γ is saisically no differen from zero in all he cases. Table 2. Gap model : Baseline and augmened policy rule 1 Mexico: Baseline Backwardlooking Money growh (GR) α β γ φ 1 ρ J-es (14.64) (10.89) (10.01) (9.75) 1.91 (6.25) (8.38) 2.55 (0.42) ( 0.76) 8.16 ( 0.14) (6.36) (2.09) 0.25 (14.43) 0.34 (5.79) 0.52 (11.58) (0.89) 6.94 (0.95) 7.61 (0.94) Depreciaion rae (19.41) 0.29 (2.78) (0.68) 0.54 (13.76) 0.16 (4.44) 9.34 (0.85) Foreign asses (GR) (9.68) 2.71 (6.84) (0.61) (2.44) 0.53 (11.74) 7.63 (0.97) Lending boom indicaor (3.75) 0.53 (2.64) (0.29) 0.21 (3.19) 0.81 (33.55) 9.34 (0.85) Curren accoun (23.86) 2.49 (6.05) ( 0.96) (1.61) 0.66 (23.86) 9.17 (0.95) 1 The numbers in parenhesis are -saisic values, excep in he J-es column where hey represen he p value o rejec he hypohesis ha over-idenifying resricions hold. Saisically significan a 95% confidence level ^ Saisically significan a 90% confidence level The role of inflaion expecaions The exisence of reliable daa on marke-based inflaion expecaions is very useful in an inflaion-argeing conex. However, as Bernanke and Woodford (1997) indicae key limiaions of marke-based inflaion expecaions need o be considered. Specifically, measured inflaion expecaions are backward looking and reflec 23

24 adapive behaviour by he public. Therefore, a variable ha in principle should conain a srong forward-looking elemen is largely deermined by pas performance. Furhermore, in a regime where he auhoriies conduc inflaion forecas argeing based on marke-based inflaion expecaions as one of he key indicaors, and where here has been a gradual learning abou he ani inflaion process sance of he auhoriies, i is very difficul o deermine he properies of marke based inflaion expecaions. Taking ino accoun hese difficulies, a hird specificaion is esimaed using expeced raher han acual values for he inflaion rae. This policy rule however includes jus he sample period in which his daa is available ( ). The daa on inflaion expecaions is obained from he Cenral Bank of Mexico. Consequenly, he inflaion gap is calculaed as he difference beween inflaion expecaions a he end of he year and he value of he inflaion arge announced by he cenral bank. The resuls shown in able 3 are quie remarkable. All he coefficiens are saisically significance a he 5% level of significance and wih he expeced values. Parameer β coninues above uniy suggesing ha Bank of Mexico has been concerned wih inflaion sabilizaion. The coefficien of he oupu gap for his sample period ( ) urns ou o be saisically significan and wih he expeced value in all he esimaions. The moneary auhoriies seem o be concerned wih oupu flucuaions. In paricular, he resuls imply ha he auhoriies increased ineres raes when a rise in he gap beween real GDP and is rend is expeced. The smoohing parameer however decreases subsanially in his period. On average almos 90 percen of he adjusmen is achieved in he firs quarer. This may reflec he gains in erms of policy credibiliy which makes he moneary policy sance more effecive. Similar conclusions can be found in Chuecos (2003) and Torres (2002). Overall, he resuls sugges ha moneary policy in Mexico has been consisen wih IT principles. Moreover, he resuls from he inclusion of addiional variables in he ineres rae rule reflec he concern by he moneary auhoriies in aspecs relaed o financial sabiliy. Marginal and posiive conribuions are observed in all variables being he mos significan he one from he lending boom indicaor. The value of he parameer when changes in he nominal exchange rae are included becomes quie small. This suggess ha a flexible exchange rae has indeed conribued o reduce inflaion. In he 24