NBER WORKING PAPER SERIES ESTABLISHING CREDIBILITY: EVOLVING PERCEPTIONS OF THE EUROPEAN CENTRAL BANK. Linda S. Goldberg Michael W.

Transcription

1 NBER WORKING PAPER SERIES ESTABLISHING CREDIBILITY: EVOLVING PERCEPTIONS OF THE EUROPEAN CENTRAL BANK Lnda S. Goldberg Mchael W. Klen Workng Paper hp:// NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachuses Avenue Cambrdge, MA November 2005 The vews expressed n hs paper are hose of he ndvdual auhors and do no necessarly reflec he poson of he Federal Reserve Bank of New York or he Federal Reserve Sysem. We have benefed from commens from, Donald Andrews, Roc Armener, Menze Chnn, Sylveser Ejffnger, Graham Ello, Aruro Esrella, Francesco Gavazz, Ulrch Müller, Zhongjun Qu, Rchard Pores, John Rogers, Jm Sock, Cedrc Tlle and parcpans a he January 2005 ASSA, March 2005 TAFI JIMF conference, NBER IFM meeng Sprng 2005 and parcpans n he Federal Reserve Sysem Commee Meeng on Inernaonal Economcs. Address correspondences o Lnda S. Goldberg, Federal Reserve Bank of NY, Research Deparmen, 33 Lbery S, New York, N.Y emal:lnda.goldberg@ny.frb.org, or Mchael W. Klen, Flecher School, Tufs Unversy, Medford MA emal Mchael.klen@ufs.edu. The auhors hank Da Ganguly and Eleanor Dllon for excellen research asssance. The vews expressed heren are hose of he auhors) and do no necessarly reflec he vews of he Naonal Bureau of Economc Research by Lnda S. Goldberg and Mchael W. Klen. All rghs reserved. Shor secons of ex, no o exceed wo paragraphs, may be quoed whou explc permsson provded ha full cred, ncludng noce, s gven o he source.

2 Esablshng Credbly: Evolvng Percepons of he European Cenral Bank Lnda S. Goldberg and Mchael W. Klen NBER Workng Paper No November 2005 JEL No. F3, E5, E6 ABSTRACT The percepons of a cenral bank's nflaon averson may reflec nsuonal srucure or, more dynamcally, he hsory of s polcy decsons. In hs paper, we presen a novel emprcal framework ha uses hgh frequency daa o es for perssen varaon n marke percepons of cenral bank nflaon averson. The frs years of he European Cenral Bank ECB) provde a naural expermen for hs model. Tess of he effec of news announcemens on he slope of yeld curves n he euro-area, and on he euro/dollar exchange rae, sugges ha he marke's percepon of he polcy sance of he ECB durng s frs sx years of operaon sgnfcanly evolved, wh a belef n s nflaon averson ncreasng n he wake of s moneary ghenng. In conras, ess based on he response of he slope of he Uned Saes yeld curve o news offer no comparable evdence of any change n marke percepons of he nflaon averson of he Federal Reserve. Lnda S. Goldberg Research Deparmen, 3rd Floor Federal Reserve Bank-New York 33 Lbery Sree New York, NY and NBER lnda.goldberg@ny.frb.org Mchael W. Klen 33 Cooldge Road Belmon, MA and NBER mchael.klen@ufs.edu

3 Esablshng Credbly, Goldberg and Klen 1. Inroducon The percepon of he nflaon averson of a cenral bank plays a key role n deermnng wheher s goal of low nflaon s aaned. Ths pon s, by now, a sandard heorecal resul. 1 I s also receved wsdom among praconers. In a survey of he heads of 84 cenral banks, as well as 52 promnen academc moneary economss, Blnder 2000) fnds ha an-nflaon credbly s consdered vally mporan and helps keep nflaon low. 2 Ths consensus on he mporance of he percepon of nflaon averson naurally leads o he queson of how s acheved, and wheher and how evolves over me. One vew s ha esablshng an approprae nsuonal srucure s he key elemen n nsulang he moneary auhory from polcal pressure and hereby convncng markes ha a cenral bank has srong averson o nflaon. A second, more dynamc, vew focuses on he role ha acual polcy conduc plays n buldng he repuaon of a cenral bank. These wo dfferen vews have dsnc mplcaons for he relave mporance of he nsuonal srucure of a cenral bank as compared o s conduc for aanng and mananng s credbly. A majory of respondens o Blnder s survey beleve ha cenral bank credbly s based more on s hsory of acons han on he consrucon of nsuonal srucures ha nsulae a cenral bank from polcal concerns and afford ndependence. Noneheless, here s also a consensus among respondens ha srucure maers. Ths laer vew s conssen wh emprcal research ha has found, n cross secons of counres, ha nsuonal srucure s assocaed wh economc performance, perhaps because ndcaes he ably of an nsuon o e s hands and comm o a polcy ha may cause shor-erm pan n he pursu of longer- 1 Semnal conrbuons on he role of credbly ncludes Kydland and Presco 1977), Calvo 1978), Barro and Gordon 1983). 2 Blnder 2000) pons ou ha he erm cenral bank credbly can mean nflaon averson, ncenve compably or pre-commmen. He repors ha, among hese hree conceps, cenral bankers denfy nflaon averson wh credbly far more closely han do [academc] economss. p. 1424) Usng a fve-pon scale, nearly 90 percen of hs cenral bank respondens denfed he conceps credbly and dedcaon o prce sably as que closely relaed or vrually he same, whle jus over half of he academc respondens repled ha hese wo erms were eher unrelaed, slghly relaed, or moderaely relaed. In he le and body of hs paper, we use he erm credbly o mean nflaon averson. Theorecal conrbuons n whch credbly s synonymous wh nflaon averson nclude Rogoff 1985, 1987) and Backus and Drffll 1985). 1

4 Esablshng Credbly, Goldberg and Klen run gan. 3 There s less evdence, however, on wheher and how he credbly of a parcular cenral bank evolves over me n response o he conduc of polcy. The quesons of he achevemen and he manenance of nflaon averson credbly are especally relevan for a new cenral bank. An analyss of he experence of he European Cenral Bank ECB) durng s early years of operaon provdes a naural expermen for consderng hs queson. The archecs of he nsuonal srucure of he ECB were mndful of lessons from economc heory concernng he mporance of ndependence from polcal consderaons. 4 The role of conduc was also clearly apparen. As ndcaed by he survey resuls n Blnder 2000), he drecors of cenral banks are vally aware ha her polces are closely scrunzed for ndcaons of general endences. Ths may be especally rue wh a new cenral bank where each polcy choce can lead o a larger updang of marke prors han would be he case for a long-esablshed cenral bank. Ths paper sars wh he nsgh ha he responses of asse prces o economc news embed marke percepons of he polcy reacon funcon of cenral banks. The relaonshp beween asse prces and news evolves wh he change n marke percepons of a cenral bank s moneary reacon funcon and s assocaed degree of nflaon averson. As argued by Bernanke 2004), successful moneary polces should sablze, or anchor, nflaon expecaons so as o preven hem from becomng a source of nsably n her own rgh. Therefore, wh asse prces as a sarng pon, n Secon II, we presen a framework for a novel es of he evoluon of marke percepons of cenral bank nflaon averson. Ths es uses hgh frequency asse prce daa on he slope of yeld curves and on exchange raes, wh he surprse componens of economc daa releases, o esmae wheher he marke percepon of he an- 3 For example, Cukerman 1992) analyzes he charers of cenral banks and shows, n a cross-counry panel, ha average nflaon s lower n counres n whch laws afford cenral banks greaer ndependence. Alesna and Summers 1993) also fnd cross-counry evdence ha he level of nflaon, as well as s varably, s negavely assocaed wh ndcaors of cenral bank ndependence, bu here s no assocaon beween cenral bank ndependence and real varables. Quesons have been rased, however, abou wheher he de jure srucure s closely lnked o he de faco behavor of nsuons Forder 1999). 4 Despe hese lessons, some polcans connued o ry o nfluence polcy drecon. For example, Oscar Lafonane, apponed Fnance Mnser of Germany n he Auumn of 1998, called for he new ECB o lower neres raes from he me of hs apponmen unl hs resgnaon n March In response, Wm Dusenberg, he frs presden of he ECB, saed n November 1998 ha was a normal phenomenon for polcans o offer her vews on he conduc of moneary polcy, bu would be very abnormal f hose suggesons were o be lsened o. See Wm Dusenberg, Banker o a New Europe, The Economs, November 26,

5 Esablshng Credbly, Goldberg and Klen nflaon credbly of a cenral bank changes over me. 5 The key nsgh from hs model s ha a gven surprse ncrease n nflaonary pressures wll resul n a greaer ncrease n a long neres rae relave o a shor neres rae, and a larger exchange rae deprecaon, when a cenral bank s perceved as beng more oleran of nflaon and less credble as an nflaon fgher. If unvaryng nsuonal srucure s he domnan deermnan of a new cenral bank s credbly, hen one would no expec o fnd a change n he hgh frequency relaonshp beween economc news and he slope of he yeld curve over me. Bu f credbly for a new cenral bank s earned hrough he conduc of s polcy, one would fnd a sgnfcan break n he relaonshp beween news and he yeld curve as credbly evolves. 6 In Secon III we apply hs es o sudy he evoluon of he credbly of he European Cenral Bank from he me began s operaons n January 1999 hrough md Usng hourly daa on he euro-dollar exchange rae and on he erm srucure of bonds of euro-area counres, we fnd evdence ha he marke s percepon of he nflaon averson of he ECB has evolved over me and responded o ECB polcy acons. As a benchmark for our analyss, and also o denfy wheher he resuls we found for he ECB could be arbued o changes n he economc envronmen raher han n specfc vews of s nflaon averson, we also es for changes n he marke s percepon of he moneary reacon funcon of he Federal Reserve over he same sample perod. In conras o our resuls for he ECB, we fnd no evdence of changng percepons of he reacon funcon of he Federal Reserve, a resul ha s no surprsng gven he Fed s long-sandng commmen o prce sably under he charmanshp of boh Alan Greenspan and Paul Volcker. 5 Forward marke nformaon has been used n oher ess of polcy regme credbly. For example, Svensson 1991) shows ha forward exchange raes were no whn he arge zone band of he European Moneary Sysem EMS) n he 1980s, a resul he nerpres as ndcang ha he EMS generally dd no offer credble bands on s members currences. Svensson 1993) presens a smlar se of ess o deermne wheher he nflaon arges of Canada, New Zealand and Sweden were conssen wh marke yelds. These ess, whle nformave, requre he presence of an explc arge, lke an exchange rae band or an nflaon arge, o judge credbly. Oher relaed emprcal analyses on he polcy credbly of an exchange rae arge zone use nervenon daa o esmae perceved arge zone bands Klen and Lews 1993 and Lews 1995). 6 Klen, Mzrach and Murphy 1991) develop a smlar ype of analyss concernng dfferences n he responsveness of asse prces o news as polcy evolves n her sudy of he changng responsveness of dollar exchange raes o news abou he Uned Saes curren accoun. They fnd he 1985 Plaza Accord alered percepons of he degree o whch Amercan polcy was concerned wh he U.S. curren accoun defc. 3

6 Esablshng Credbly, Goldberg and Klen 2. Cenral Bank Polcy and Marke Responses o News In hs secon we presen a model ha shows how changes n percepons abou a polcy sance can aler he response of asse prces o news. We begn wh he sandard framework used n emprcal works ha sudy he effec of news on asse prces, as n, for example, Andersen, Bollerslev, Debold and Vega 2003). We hen nroduce a polcy reacon funcon, and show how an evolvng vew of he credbly of cenral bank nflaon averson affecs he relaonshp beween news and asse prces. We dscuss he emprcal mplemenaon of hs model o he yeld curve and exchange raes, so ha we can use hgh frequency daa o solae he effecs of news on asse prces under dfferen marke percepons of cenral bank nflaon averson. 2.1 Emprcal Specfcaon The sandard lnear specfcaon lnkng he surprse componen of news o he change n an asse prce s q + q = α + γ x E x + 1) + + ) + where q + q s he change n an asse prce over he shor perod of me beween -, jus before an announcemen, and +, jus afer ha announcemen. x represens he announced value of a + varable, whch s known a me +, E + represens he expeced value of ha varable before x he announcemen, so ha x E x + s he surprse componen of he announcemen, and ε + s + a whe-nose error erm. As emphaszed n Andersen e al. 2003) hs parsmonous specfcaon s mos approprae when he me horzon beween - and + s shor, for example, when s measured n mnues raher han days, and when news abou he varable x does no become avalable a he same me ha s, whn he span - o + ) as announcemens abou some oher relevan varable. The acual se of varables ha consue x depends upon he asse suded bu, n general, any varable ha markes consrue as revealng nformaon abou curren and fuure economc acvy may be approprae for sudy. A more general verson of equaon 1) akes no accoun marke expecaons abou he polcy response o news. Consder he pah, from me forward, of a polcy M, whch has an effec on q. We can augmen 1) o nclude he effec on q + q of he change n he percepon, beween me - and +, of he pah of polcy. Ths specfcaon, ε 4

7 Esablshng Credbly, Goldberg and Klen q + q = + γ x E x + φ E M E M + α ε, 2) + + ) + + ) + capures he possbly ha q q responds o economc news drecly hrough he coeffcen + γ and ndrecly hrough he coeffcen φ due o he effec of he news on he expeced course of polcy, where he perceved polcy pah before he announcemen occurs s E and s M perceved pah afer he announcemen s E + +. The parameer φ may be posve or negave, M dependng upon he polcy and he asse. Over a shor wndow of me, he only reasonable source of a change n he perceved pah of he polcy over he shor me span - o + s he surprse componen of he daa announcemen durng hs wndow. Ths lnk arses because of a percepon of he exsence of a polcy reacon funcon, such as 3) M = V λ x x ) where V represens oher varables ha affec he choce of M and he subscrp on he coeffcen λ allows for he possbly of a dfferen levels of responsveness of he cenral bank o he value of x a dfferen mes n he sample perod. Lkewse, hs formulaon allows for more han one arge level of he varable, x, durng he sample perod. To make he dscusson of he polcy reacon funcon more concree, consder he case where 3) represens a Taylor Rule. In hs case, M represens moneary polcy such ha an ncrease n M represens a more expansonary moneary polcy), x s nflaon daa, x s an nflaon arge, and V represens an ndcaor of oher economc condons, for example he oupu gap or unemploymen. A hgher value of λ represens more nflaon averson n he reacon funcon of he cenral bank. 7 In he presence of he perceved polcy reacon funcon, he surprse componen of he perceved change n polcy, E M + E M + ) +, s defned by 4) E M E M ) = E V E V ) λ x E x ) + λ E x E x ) Whle hs s he mos common formulaon of he polcy reacon funcon, alernave formulaons can of course be specfed. For example, λ can be modeled as asymmerc n ha s larger when nflaon exceeds argeed values, or nflaon devaons from arge can be enered nonlnearly, so ha moneary reacons are sronges when nflaon s furhes from arge values. 5

8 Esablshng Credbly, Goldberg and Klen Ths expresson shows ha expeced polcy can depar from s pror pah due o a change n he oupu gap, due he surprse componen of an nflaon daa release, or due o a change n he arge value of nflaon. Provded ha here s some news n an nflaon repor relave o marke expecaons, + + x E x ) s no equal o zero. In a shor wndow of me around he nflaon repor, for example an hour wndow, n he absence of a smulaneous announcemen of a change n arge nflaon, or of news on real economc varables such as he oupu gap, he oher wo rgh hand sde erms of equaon 4) can be se equal o zero. 8 Whn hs equaon, he parameer λ assocaed wh marke vews of cenral bank nflaon averson may change over me, and, ndeed, esng for me varaon n λ s he cenral emprcal ask n hs paper. Consder he response of bond prces o economc announcemens. The effecs of news announcemens, such as nflaon repors, on bond prces have been exensvely suded. A each horzon, nomnal reurns are comprsed of he real reurn, an nflaon expecaon, and a rsk premum. Bond reurns and nflaon expecaons are lnked hrough he Fsher relaonshp. Wh news affecs on equlbrum real neres raes common o reurns a all horzons along he yeld curve, when we dfference across he reurns of long and shor-daed bonds we absrac from he effec of news on equlbrum real reurns. Over shor wndows around news announcemens, dfferencng also absracs from he effec of news on erm prema or lqudy prema. In hs case, he regressand q + q s he expeced change n he dfferenal n he long horzon for example, 10 years) and shor horzon for example 2 years) nflaon raes due o he news announcemen ) q q = E π E π ) E π E π ) We have E V ) = 0 + E V f x and V are uncorrelaed, bu, even f hs s no he case, he qualave effecs dscussed below are no affeced f x and V are negavely correlaed. If x and V are posvely correlaed, we would expec he sgn on λ o be posve n he polcy reacon funcon. We would no expec, n a sample wh many observaons, many nsances where λ E + x E x ) does no equal zero snce a nonzero value for hs erm would mean ha he news announcemen self alers he vew of he arge value x n he shor me nerval - o +. Even f, say, an unusually large value of he surprse componen of he news alers marke parcpans percepons of he arge value n one or wo nsances n a sample wh many observaons, hs would leave λ E x ) = 0 + E x for he vas majory of cases. 6

9 Esablshng Credbly, Goldberg and Klen where 10 π s he average expeced nflaon rae over 10 years and 2 π s he average expeced nflaon rae over 2 years. 9 The change n he slope of he yeld curve over shor horzons reflecs he evoluon of nflaon expecaons across he erm srucure. Subsung 4) no 2), and nerpreng q q + 6) q + q = + ) x + E x + ) + + α γ φλ ε. as he slope of he yeld curve, we ge Ths equaon suggess ha, n he presence of changng percepons abou he cenral bank s polcy, he esmaed coeffcen γ φλ ) on nflaon news n regressons of he slope of he yeld curve s unsable. 10 We can be more precse abou he nsably of γ φλ ) by consderng he smple case where a polcy acon underaken by he cenral bank, such as a major ghenng, changes he vews of marke parcpans. Suppose ha before hs acon, marke parcpans hough ha he cenral bank polcy was dovsh D), meanng one of accommodang nflaon whle afer he acon here was he vew ha he cenral bank would be more aggressve or hawksh H) n combang nflaon. These polces are dsngushed by he condon ha λ H > λ. Evolvng percepons of cenral bank nflaon averson would be refleced n he coeffcen on nflaon news. For a cenral bank ha ganed marke credbly as an nflaon fgher, we would expec o fnd a larger value for he esmaed coeffcen n he earler perod as compared o he laer perod snce γ φλ ) > γ φλ ). If percepons of nflaon averson were unvaryng, perhaps D H because hese percepons were solely drven by he nal and unvaryng nsuonal srucure, we would expec o fnd a sable relaonshp beween he slope of a cenral bank s yeld curve and nflaon news. D 9 Flemng and Remolona 1999) also argue ha he effecs of news on asse prces of dfferen maures reveals nformaon abou marke parcpan belefs abou cenral bank reacon funcons. Research by Esrella, Rodrguez and Schch 2002) focuses on he relaonshp beween he slope of he yeld curve and subsequen real acvy and nflaon. Usng monhly daa from 1967 o December 1998, he German yeld curves were found o be mos nformave a nflaon predcon horzons beyond wo years, whle he US erm srucure was mos nformave a somewha shorer horzons. Such analyses, buldng on Esrella and Mshkn 1997), also argue ha nversons n he slope of he yeld curve have been successful as recesson predcors. The measured srengh of hese relaonshps are no sable over me, and vary wh he maury horzons of he bonds examned. Early research by Huznga and Mshkn 1986) appled o monhly daa for he Uned Saes recognzes he sensvy of he slope of he yeld curve o percepons of moneary polcy regmes n he 1970s and 1980s. These sudes do no examne he effecs of news announcemens on he slope of he yeld curve, or use hgh frequency daa as n our sudy. 10 If news alers he percepon of he arge level of x, we could fnd evdence of a me-varyng nercep as well q q = α + φλ E x E x )) + γ φλ x E x ε. snce, n ha case, we would have + + ) + + ) + + 7

10 Esablshng Credbly, Goldberg and Klen Ths example, ponng o a dscree change n percepons of λ and he regresson coeffcen, offers a parcularly sark vew of shfs n marke percepons of he cenral bank reacon funcon. An evolvng vew of cenral bank polcy, one reflecng a gradual learnng process, may be more conssen wh realy and wh heores of cenral bank credbly formaon. 11 In eher case, a more complee depcon of hs model would specfy he way n whch he marke s vew of he sance of he cenral bank evolves over me n response o polcy. However, he man economerc echnque we use for denfyng he changes over me n he slope of he yeld curve does no requre us o specfy hs learnng process nor he assocaed evoluon of γ φλ ). Insead, hs economerc echnque, developed by Ello and Müller 2005) ess for a very general form of perssen parameer nsably over he sample perod. An assocaed mehod of esmang he smoohed parameer pah under very general assumpons, from Müller and Pealas 2005), enables us o assocae changes n he esmaed values of γ φλ ) wh polcy changes of he European Cenral Bank. 3. Evolvng Percepons of European Cenral Bank Polcy In hs secon we presen he daa, mehodology, and resuls of our ess for changes n he marke s percepon of he an-nflaon sance of he European Cenral Bank durng s frs sx-and-one-half years of operaon, from January 1999 hrough June We begn, n Secon 3.1, wh a descrpon of he daa we use for hese ess. Fve dfferen dependen varables are examned: he change n he erm spread alernavely called he change n he slope of he yeld curve) for German, French and Ialan governmen bonds, he change n he Euro/dollar exchange rae, and he change n he erm spread of Uned Saes governmen bonds. The ess for possble parameer nsably of he Uned Saes erm spread s offered as a benchmark; were we o fnd evdence of parameer nsably for regressons based on hs seres, we would be concerned ha evdence of parameer nsably usng European bond yelds may no, n fac, reflec an evolvng percepon of ECB nflaon averson bu, raher, some srucural change common o fnancal markes across all four of hese ndusral counres. Lkewse, an absence of parameer nsably n he euro-dollar exchange rae regresson could suppor a common srucural change across U.S. and euro-area markes However, as shown n Secon 3.3, we fnd 11 For example, see Backus and Drffll 1985) or, more recenly, Ahey, Akeson, and Kehoe 2005). 8

11 Esablshng Credbly, Goldberg and Klen no evdence of parameer nsably for he regressons usng he Uned Saes erm spread seres whle we do fnd sgnfcan evdence of a change n he erm spread for he ess usng he oher four seres usng he Ello Müller es. In Secon 3.4 we presen esmaes of he me pah of γ φλ ) usng he Müller Pealas procedure, and lnk hese me pahs o acual polcy changes underaken by he ECB. Fnally, n Secon 3.5, o demonsrae he robusness of our resuls, we presen sup-wald ess for a dscree break n he regresson relaonshp and he daes assocaed wh hose breaks. 3.1 Daa The wo ypes of daa used n our analyss are varous asse prces, where he asses are governmen bonds and foregn exchange, and nflaon announcemens and relaed marke expecaons. We begn hs secon wh a dscusson of he fve dfferen asse prces used as dependen varables n our esmaon. We hen descrbe our consrucon of nflaon surprses. Asse Prce Daa: Fve dfferen dependen varables wll be used n he regressons. In each case, he dependen varable, q + q, represens he change n q beween hry mnues before and hry mnues afer he monhly nflaon announcemen. Four of he dependen varables are he change n he erm spread beween 10-year and 2-year neres raes, q q = r r, for French, Ialan, German, or Uned Saes governmen bonds. The regressand q s he change n he erm prema, whch under he Fsher relaonshp would reflec he change n he dfferenal beween he expeced en-year and wo-year nflaon raes due o he nflaon news. Thus, when usng hese four bond seres, γ φλ ) can be nerpreed as a funcon of he drec effec represened by γ) and he ndrec effec va a polcy response, as represened by nflaon raes. φλ ) of curren nflaon news on he expeced long-run relave o shor-run There s a smlar nerpreaon of γ φλ ) n he regressons ha use he ffh dependen varable, where q + q represens he change n he logarhm of he euro per U.S. dollar exchange rae, hry mnues before and hry mnues afer he news announcemen. In hs case, a posve value of q q ndcaes a deprecaon of he euro, reflecng eher an + 9

12 Esablshng Credbly, Goldberg and Klen ncrease n expeced nflaon n Europe relave o he n he Uned Saes or an expecaon of relavely more ghenng of moneary polcy n he Uned Saes han n Europe. 12 In hs case, evdence ha γ φλ ) decreases over he sample perod reflecs a vew among marke parcpans ha here s relavely srenghenng of he an-nflaon sance of he ECB as compared o he U.S. Federal Reserve, due o a srenghenng of he an-nflaon polcy sance of he ECB, a weakenng of he an-nflaon sance of he Fed, or some combnaon of he wo. Consderaon of ess on bond yelds of he euro-area counres and he Uned Saes help solae he source of he change of he responsveness of he euro/dollar exchange rae o news. 13 Inflaon Announcemens and Expecaons: The news varable we use s he dfference beween he monhly announcemen of he Core Consumer Prce Index for he Uned Saes and he expeced value of hs announcemen pror o s release, as measured by survey responses. The closely wached core CPI s he bes nflaon measure for hs analyss, as evdenced by he mpac of relaed news on markes, he heorecal leraure on prces and moneary polcy, and Humphrey Hawkns esmony by Alan Greenspan n recen years, where he CPI excludng food and energy s ypcally he only measure of prce nflaon dscussed. 14 Alhough our prmary emphass s on evolvng credbly of he ECB, and we nroduce European bonds for consrucng yeld curves, our choce of nflaon repor s he U.S. core CPI release, no European nflaon seres. Takng he moneary reacon funcon lerally, hs choce would mply ha he ECB reacs o U.S. nflaon news. Our specfcaon does no preclude an ECB reacon funcon o oher more local nflaon seres as well. I does, however, make explc an assumpon ha U.S. nflaon repors conan nformaon perceved as relevan for Europe and euro-area moneary polcy. We choose U.S. nflaon daa because here s exensve evdence showng he rch news conen of hs daa for euro area asse prces. Ths evdence sands n sark conras wh weak effecs of nflaon repors from Germany, Ialy, France, or Europe as a whole. 15 Uned Saes macroeconomc news affecs boh Uned Saes and 12 The change n he exchange rae over hs shor me horzon s a reflecon of a change n he expeced value of he exchange rae a some longer horzon whch, n urn, reflecs some form of long-run purchasng power pary. 13 A smlar jon use of neres rae daa and exchange rae daa was used by Engel and Frankel 1984) o analyze he response of neres raes o moneary announcemens. 14 For a nce overvew of he evdence and relaed leraure, see Clark 2001). 15 Recen relevan sudes nclude Andersen e al 2003), Goldberg and Leonard 2003), Faus, Rogers, and Wrgh 2003), Chnn and Frankel 2004) and Ehrmann and Frascher 2004). 10

13 Esablshng Credbly, Goldberg and Klen European asse prces, wh lle or no sgnfcan effec of European macroeconomc news on asse prces n eher Europe or Amerca. Marke parcpans argue ha he news conen of some German and oher European prce announcemens has a mes been quesonable because of ssues of daa qualy and epsodes of daa leaks pror o offcal announcemen mes. 16 The news or surprse componen of an economc daa release s he dfference beween he acual release and he markes pror expecaon of he conens of he release. The expecaons daa we use are medan responses from weekly surveys of marke parcpans conduced by Money Marke Servces, a dvson of Sandard & Poor s, and more recenly from Acon Economcs. 17 A regresson of he 75 medan monhly survey responses on he acual monhly nflaon repors generaes a coeffcen of 0.68, wh p-value of 0.026, wh he regresson unable o rejec unbasedness of he survey as a predcor of he acual value of he nflaon repors. In creang he nflaon news varable, we normalze news by he sample sandard devaon of he dfference beween he repored and he expeced values of he announcemens so ha he varable news has mean 0 and sandard devaon Economerc Mehods New economerc ess developed by Ello and Müller 2005) allow one o es for he presence of perssen me varaon n one or more regresson coeffcens over he sample perod whou specfyng he exac breakng process, such as breaks ha occur n a random fashon, seral correlaon n he changes of coeffcens, or a cluserng of break pons. 18 Ths feaure of her es makes well sued for our purposes snce we do no need o es for a parcular ype of updang by marke parcpans of her vews on cenral bank nflaon averson. The Ello and Müller 2005) quas-local Level qll) sasc akes a negave value, and a value smaller more negave) han he crcal value mples a falure o rejec me 16 As robusness checks, we also examne he mplcaons of alernave U.S. prce news, and of a range of measures of European prce news. Our resuls are robus o oher measures, bu hese oher measures somemes have small and volale effecs on bond yelds or he slope of he yeld curve. 17 Money Marke Servces were he source of hese daa hrough December Haver Analycs provded connuous expecaons and announcemen daa hrough 2005 usng daa from Acon Economcs. Gurkaynek and Wolfers 2005) show ha hese daa have been among he bes performng expecaons seres for mporan macroeconomc varables over he sample perod ha we analyze. 18 Ello and Müller wre ha, for her ess, he precse form of he breakng process [of he coeffcens] s rrelevan for he asympoc power of he ess. p.10) An mplcaon of hs s ha From a praccal perspecve he researcher does no have o specfy he exac pah of he breakng process n order o be able o carry ou almos) effcen nference. p. 4) 11

14 Esablshng Credbly, Goldberg and Klen varaon n one or more coeffcens for he enre sample perod. Ths procedure ess for me varaon over he enre sample and, as such, does no denfy a parcular dae as he one mos lkely o represen a dscree break pon. Whle an unspecfed evoluon of γ φλ ) may be conssen wh noons of markes updang percepons of he nflaon averson of a new cenral bank, he denfcaon of a break dae would be useful for our purposes f, n fac, hs s how percepons changed) because our ess do no allow us o dsngush beween me varaon n γ φλ ) due o varaon over me n λ raher han, say, varaon over me n γ, he drec responsveness of he change n asse prces o he surprse componen of news. In he wake of he creaon of a new cenral bank, such as he ECB n January 1999, s reasonable o expec ha he mos lkely cause of a me varaon n γ φλ ) s changes n λ raher han changes n γ or φ. A change n γ would mply a change n he drec effec of nflaon on asse prces over he sample perod. Whle heorecally possble, we wll presen wo ypes of evdence more supporve of evolvng marke percepons of he ECB reacon funcon. Frs, we compare he es resuls across euro area counres, he euro-dollar exchange rae, and he benchmark case of he yeld curve for U.S. asses. Gven he negraon of European and U.S. fnancal markes, as recenly documened by Baroln, Hlon, and Pra 2005), would be unusual for a γ change o occur across euro area bond markes bu no for U.S. bond markes. The second way we aemp o denfy he source of changes n γ φλ ) s by esmang s pah over me and consderng wheher s movemens correlae o acual changes n ECB polcy. We do hs esmaon usng he mehod developed by Müller and Pealas 2005) o esmae he smoohed me pah of a me-varyng parameer. Fnally, we es for he robusness of boh he fndng of parameer nsably and he mng of changes n γ φλ ) by presenng sup-wald ess for parameer sably see Andrews 1993, 2003). These ess are predcaed on he exsence of one or more dscree break pons. They are conduced by runnng 0.7 T regressons, where T s he number of observaons n he daa se, ha ake he form 6) q + q = α + β x E x ) + β D x E x ) + ε I I

15 Esablshng Credbly, Goldberg and Klen where D I s a dummy varable ha equals 0 for he frs n observaons of he sample and equals 1 for he remanng T n observaons. 19 Ths generaes a se of 0.7 T β I s and 0.7 T assocaed es sascs. The sup-wald es compares he larges F-value for all of he β I s wh crcal values presened n Andrews 2003) and, f hs sup-wald sasc exceeds he crcal value, he dae assocaed wh ha β I s he sascally sgnfcan esmaed break dae Tme Varaon n he Effecs of News on he Slope of he Yeld Curve In hs secon we repor he resuls of he Ello and Müller 2005) qll sasc for he fve asse prce seres dscussed above. These sascs are negave, and a smaller.e. more negave) value of he sasc allows one o rejec he null hypohess of a lack of perssen me varaon n he effec of news on nflaon expecaons. Thus, suppose here was an evolvng vew of he polcy sance of he ECB over me, bu no of he Fed over hs same perod. We would expec o see a smaller qll sasc han some crcal value for regressons usng he change n he erm spread for German, French and Ialan governmen bonds, as well as for he change n he euro / dollar exchange rae, bu a qll sasc larger han hs crcal value for a regresson n whch he dependen varable s he change n he erm spread on Uned Saes governmen bonds. Resuls of hs es are presened n Table The frs row s a es of he general perssen varaon n he slope coeffcen only. The second row s a jon es of he general perssen varaon n boh he slope and he nercep coeffcens. Crcal values are ncluded n he boom row of he able. Enres n bold and alc represen a qll sasc ha s sgnfcan a beer han he 99 percen level of confdence, bold enres represen a qll sasc ha s sgnfcan a beween he 95 percen and 99 percen levels of confdence, and alc enres represen a qll sasc ha s sgnfcan a beween he 90 percen and 95 percen levels of confdence. 19 The sup-wald ess requre rmmng, ha s, fewer han T regressons are run for a sample wh T observaons. Andrews 1993) suggess he use of symmerc 15% rmmng and herefore, n hs case, he es would nvolve 0.7 T regressons. There s no comparable rmmng n he Ello and Müller es. 20 Esrella and Rodrguez 2005) propose an alernave echnque, where he hypohess esed s a one-sded dreconal change, ncrease or decrease, n he esmaed parameers. 21 As suggesed by Ello and Müller 2005), we allow for he possbly of heeroskedascy n he varance- x E x ) ε by usng he Newey-Wes 1987) correcon. covarance marx of he score seres { + } + 13

16 Esablshng Credbly, Goldberg and Klen Table 1: Ello-Müller Tes for Perssen Tme Varaon Tes of Tme Change n Term Spread of Governmen Bonds of Change n Varaon of Germany France Ialy Uned Saes Euro/$ slope slope and nercep number of observaons Crcal Values: 1 coeffcen Slope alone) 1% ; 5% -8.36; 10% coeffcens Slope & Inercep) 1% ; 5% ; 10% The resuls n Table 1 provde evdence of perssen me varaon n γ φλ ) n regressons of nflaon news on he change n he erm spread of German governmen bonds and French governmen bonds, and n he Euro / dollar exchange rae, a greaer han he 95 percen level of confdence, and on he change n he erm spread of Ialan bonds a beween he 90 percen and he 95 percen level of confdence. In conras, here s no sgnfcan evdence of perssen me varaon n he slope coeffcen n a regresson of news on he change n he erm spread of Uned Saes governmen bonds over hs same perod of me. All of hese resuls are conssen wh he model presened above n whch γ φλ ) vares as λ changes wh an evolvng vew of he nflaon averson of he European Cenral Bank n he perod afer s ncepon. There s no a correspondng evoluon n he vew of he nflaon preferences of he Federal Reserve durng hs perod, whch followed almos ffeen years of observaons of he polcy acons of he Federal Reserve Board of Governors under he leadershp of Charman Greenspan. The second row of qll sascs n Table 1 presen resuls of a es of he jon perssen me varaon of he slope and nercep erms of a regresson ha akes he form of specfcaon 6) foonoe 11 presens a dscusson of he possbly of perssen me varaon n he nercep as well as he slope). There s even sronger evdence of perssen me varaon n hs jon es for he change n he erm spread for German bonds and he change n he euro / dollar exchange rae, bu weaker evdence for perssen me varaon when he dependen varable s he change n he erm spread of eher French or Ialan governmen bonds. Agan, he 14

17 Esablshng Credbly, Goldberg and Klen benchmark regresson, of he change n he erm spread of Uned Saes governmen bonds, fals o offer evdence of perssen me varaon n he coeffcens of he regresson. The evdence n hs able s suggesve of an evolvng percepon of he polcy sance of he European Cenral Bank. Ths concluson s bolsered by he esmaed me pah of γ φλ ) presened n he nex secon. 3.4 Esmaed Pahs of γ φλ ) In hs secon we presen he esmaed parameer pahs of γ φλ ), usng he echnque developed by Müller and Pealas 2005). They show how o esmae he parameer pah for general unsable me seres models by mnmzng a weghed average rsk creron, a procedure ha s akn o a smoohng problem. Ths procedure requres only general assumpons abou he rue perssen me varaon of he coeffcens. 22 Fgures 1 presens he esmaed parameer pahs of γ φλ ) for he regressons usng he four bond erm spreads. Fgure 2 presens he parameer pah for he slope coeffcen n he regresson on he euro/dollar exchange rae and, o provde comparably o he frs fgure, he slope coeffcens for he German and Uned Saes erm spreads ha are presened n Fgure 1. The frs hng o noe from Fgures 1 and 2 s ha he esmaed value of γ φλ ) for each of he erm spreads for he hree European governmen bonds, as well as he euro/dollar exchange rae, s greaer han he esmaed value of γ φλ ) for he erm spread of he Uned Saes bond. Ths s conssen wh he vew ha, a he ouse of he operaon of he European Cenral Bank, he marke perceved he ECB as more wllng o olerae nflaon han he Federal Reserve. Anoher mmedaely apparen characersc of he four me pahs n Fgure 1, and he me pah of he euro/dollar exchange rae n Fgure 2, s he relave varably of he hree European γ φλ ) s as compared o ha of he Uned Saes Müller and Pealas 2005) descrbe her procedure as an exenson of he Kalman smoohng formulae wh he opmal smooher for he rue pah of he me varyng coeffcen a funcon of he score sequence { x+ E x ) ε + }. See her paper for deals, and for an oulne of how o mplemen her procedure. 23 The sandard devaons of he esmaed γ φλ ) s are for Ialy, for France, for Germany, and for he euro/dollar exchange rae, bu only for he Uned Saes, all of whch are conssen wh he resuls of he Ello Müller qll sascs presened n Table 1. 15

18 Esablshng Credbly, Goldberg and Klen 0.02 Fgure 1 Tme Profle of Yeld Curve Slope Response o News λ-øλ esmaes Feb-99 Oc-99 Jun-00 Feb-01 Oc-01 Jun-02 Feb-03 Oc-03 Jun-04 Feb year - 2 year bond spreads Ialy France Uned Saes Germany 0.06 Fgure 2 Tme Profle of Euro/Dollar Response o News 0.04 λ-øλ esmaes Feb-99 Oc-99 Jun-00 Feb-01 Oc-01 Jun-02 Feb-03 Oc-03 Jun-04 Feb-05 Euro/dollar response German yeld slope Uned Saes yeld slope 16

19 Esablshng Credbly, Goldberg and Klen The me varaon of he esmaed pahs of γ φλ ) n lgh of he acons underaken by he European Cenral Bank bolser our conenon ha he varaon n hs parameer s due o changng vews of s polcy sance as refleced n λ ) raher han, say, changng values n γ. The peak values of γ φλ ) occur a he me of he May 1999 core CPI announcemen for he French and Ialan bond yelds, he June 1999 announcemen for he German bond yelds, and he Aprl 2000 announcemen for he euro / dollar rae. The declne n γ φλ ), conssen wh markes updang her percepons oward a more hawksh vew of ECB nflaon averson, connued for he regressons usng he hree European bond yelds unl he lae auumn of 2001, and for he euro/dollar rae for he remander of he sample perod. Durng much of wo-year perod begnnng n he auumn of 1999, he ECB was ghenng moneary polcy. November 4, 1999 marked he frs me ha he ECB rased s key neres rae snce began operaons on January 1, A ha me, hs neres rae for man refnancng operaons was rased from 2.5 percen o 3.0 percen. Ths was followed by anoher 25 bass pon ncrease on February 3, 2000, addonal 25 bass pon ncreases on March 16 and Aprl 27, and a 50 pon bass pon ncrease o 4.25 percen on June 9, Over hs whole perod, he regresson coeffcen on he slope of he U.S. erm spread was sable. The smoohed esmae of γ φλ ) for he hree European erm spreads began o rse agan owards he end of 2001, up unl Sepember 2002 for he German erm spread), January 2003 for he Ialan erm spread) and June 2003 for he French erm spread). In our model, hs could occur f he marke parally correced her vews of he degree of nflaon averson characerzng he ECB reacon funcon. Ths occurred over a perod of me when he acons of he European Cenral Bank led owards a more accommodave moneary sance. On May 11, 2001 he ECB lowered he mnmum bd rae for he man refnancng operaons by 25 bass pons, o 4.50 percen. 25 Four addonal neres rae cus by he ECB occurred on Augus 30, 2001 a cu by 25 bass pons), on Sepember 17, 2001 a cu by 50 bass pons), on November 24 The key neres rae on fxed rae enders was a 3.00 percen from January 1, 1999 hrough Aprl 9, when dropped by 50 bass pons o 2.50 percen. On November 4, 2000 a perod of moneary ghenng sared. For man refnancng operaons, changes n he rae are effecve from he frs operaon followng he dae when changes were ndcaed. 25 On June he ECB announced ha, sarng June 28, 2000, he man refnancng operaons of he Eurosysem would swch from fxed rae enders o varable rae enders. Thereafer he key neres rae se by he ECB was he mnmum bd rae of he varable rae enders for he man refnancng operaons. See 17

20 Esablshng Credbly, Goldberg and Klen 8, 2001 a cu by 50 bass pons), and on December 5, 2002 a 50 bass pon cu). Early n 2003, here were a seres of addonal rae cus [March 6, 2003, 25 bass pons and June 6, 2003, 50 bass pons] and he ECB refned s wo pllar approach, wh he offcal mporance of M3 apparenly reduced and a more offcal role for an nflaon goal a or slghly below wo percen. The man refnancng neres rae remaned a 2.00 percen from June 2003 unl he end of he sample perod n June In conras o he ncrease, begnnng n he lae auumn of 2001, n he smoohed esmaed pah of γ φλ ) for he hree European governmen erm spreads, he esmaed pah of γ φλ ) for he euro / dollar exchange rae connued o decrease hrough hs perod, and, ndeed, hrough he res of he sample perod endng n June Of course, he behavor of he euro / dollar exchange rae depends upon he acons of he Fed as well as ha of he ECB and he esmaed value of γ φλ ) for he regresson usng he Uned Saes erm spread ncreased along wh he coeffcens for he European erm spreads hroughou 2003; so, for hs perod, a leas, he smoohed esmaed values of γ φλ ) s from he bond regressons s conssen wh a vew of somewha parallel evoluon of percepons for he Fed and he ECB. 3.5 Sup-Wald Sascs Fnally, o gauge he robusness of he Ello Müller qll ess, and of he smoohed pahs of he γ φλ ) coeffcens obaned hrough he Müller-Pealas mehod, hs secon presens sup-wald ess for a dscree change n γ φλ ), based on Andrews 1993, 2003). These sup-wald ess are based on a more resrced assumpon concernng he break pon han he qll es bu, snce a break pon raher han he overall sably of he parameer s esmaed, he sup-wald ess also provde a dae for he break. We compare hese daes o he smoohed parameer pahs presened n Secon 3.4. Table 2 presens he sup-wald sascs based on ses of fve dfferen regressons ha ake he form of 6), among whch four have as he dependen varable he change n one of he erm spreads, and one has as he dependen varable he change n he euro / dollar exchange rae. The sascs presened n he op secon of hs able show evdence of a sgnfcan break, a beer han he 99 percen level of confdence, for he regressons usng he change n he erm spread for German governmen bonds and for he euro / dollar exchange rae, and a beween he 18

21 Esablshng Credbly, Goldberg and Klen 95 and 99 percen level of confdence for he change n he erm spread of Ialan governmen bonds. There s no evdence of a sgnfcan dscree break for he regresson usng he change n he erm spread of French or Uned Saes governmen bonds. Table 2: sup-wald Tes for Dscree Break Pon Change n Term Spread of Governmen Bonds of Change n Break Pon n Germany France Ialy Uned Saes euro/dollar Sup-Wald Sasc Esmaed Nov.16,2000 June 15,2001 Feb. 21,2001 Break Dae number of observaons Crcal Values from Andrews 2003) 1% 12.16; 5% 8.68; 10% 7.12 Tess conduced wh 15 percen symmerc rmmng. I s neresng o compare he daes obaned hrough he sup-wald ess wh he smoohed parameer pahs obaned usng he Müller and Pealas mehod. The daes presened n Table 2 for he sgnfcan esmaed break pons for he erm spread regressons, November 16, 2000 for he German case and he June 15, 2001 for he Ialan case, occur abou md-way beween he peak and he rough of he respecve me pahs of γ φλ ) n he perod beween md-1999 and lae-2001, he me when hese coeffcens had her larges average value. There s also a conssency beween he wo esmaed break pons for he Euro / dollar regresson and he Müller Pealas esmaed me pahs snce he frs esmaed break, February 21, 2001, comes a he me jus before he smoohed parameer pah descends from a hgh average value and he second esmaed break, Aprl 16, 2003, occurs mmedaely pror o a large decrease n he value of he esmaed smoohed parameer pah. Thus, here s an overall conssency beween he sup-wald resuls and he Müller Pealas esmaed smoohed me pah, suggesng he robusness of hese resuls. 4. Conclusons The mporance of he repuaon of a cenral bank for he success of s operaons s sressed n heory and s evden from praccal experence. An mporan queson s wheher a 19

22 Esablshng Credbly, Goldberg and Klen cenral bank gans credbly n s nflaon averson hrough s nsuonal srucure or hrough he conduc of polcy. Ths queson s especally relevan for a newly esablshed cenral bank ha faces he challenge of esablshng s repuaon, somemes n he face of polcal conroversy over he approprae conduc of moneary polcy. The evoluon of he markes percepons of he nflaon averson of he European Cenral Bank snce began operaons n January 1999 s neresng for a number of reasons. One of hese reasons s he nheren neres of he economc experence of he euro-zone. A second reason s ha he esablshmen of he European Cenral bank provdes a naural expermen for consderng how he repuaon of a cenral bank evolves over me. Ths epsode s a parcularly rch ven o mne because of he conroversy surroundng he conduc of moneary polcy n Europe as he ECB began s operaons. In hs paper, we have proposed and execued a novel es for he sudy of he evoluon of marke percepons abou he nflaon averson of a cenral bank hrough he use of hghfrequency daa. Ths mehodology and he use of hgh frequency daa provdes a unque wndow no he evoluon of percepons of moneary polcy rules, an ssue more ypcally and less precsely addressed usng lower frequency daa. We fnd evdence of an evoluon of percepons of he polcy sance of he ECB, one lnked o s neres rae polcy. There s no a smlar shf n he marke s percepon of he polcy sance of he Federal Reserve, a perod marked by he sably n s leadershp, he conssency of s saed goals, and he broad suppor for s conduc of polcy. 20

23 Esablshng Credbly, Goldberg and Klen References Alesna, Albero and Lawrence H. Summers, Cenral Bank Independence and Macroeconomc Performance: Some Comparave Evdence, Journal of Money, Cred and Bankng, vol. 25, no. 2 May), pp Andersen, Torben G., Tm Bollerslev, Francs X. Debold and Clara Vega, Mcro Effecs of Macro Announcemens: Real-Tme Dscovery n Foregn Exchange, Amercan Economc Revew, vol. 93, no. 1 March), pp Andrews, Donald, Tess for Parameer Insably and Srucural Change wh Unknown Change Pon, Economerca, vol 61, no. 4 July), pp Andrews, Donald, Tess for Parameer Insably and Srucural Change wh Unknown Change Pon: A Corrgendum, Economerca, vol. 71, no. 1 January), pp Ahey, Susan, Andrew Akeson, and Parck Kehoe, The Opmal Degree of Moneary Polcy Dscreon, Economerca, vol. 73, no. 5 Sepember), pp Backus, Davd and John Drffll, Raonal Expecaons and Polcy Credbly Followng a Change n Regme, Revew of Economc Sudes, vol. 52, no. 2 Aprl), pp Barro, Rober J. and Davd B. Gordon, A Posve Theory of Moneary Polcy n a Naural Rae Model, The Journal of Polcal Economy, vol. 91, no. 4 Augus), pp Baroln, Leonardo, Spence Hlon, and Alessandro Pra, Money Marke Inegraon, Federal Reserve Bank of New York. Bernanke, Ben, Wha Polcymakers Can Learn from Asse Prces, Remarks before The Invesmen Analyss Socey of Chcago Aprl 15). Blnder, Alan, Cenral Bank Credbly: Why Do We Care? How Do We Buld I?, Amercan Economc Revew, vol. 90, no. 5 December), pp Calvo, Gullermo, On he Tme Conssency of Opmal Polcy n a Moneary Economy, Economerca, vol. 46, no. 6 November), pp Chnn, Menze and Jeffrey Frankel, The Euro Area and World Ineres Raes, Sana Cruz Cener for Inernaonal Economcs Paper November). Clark, Todd, Comparng Measures of Core Inflaon, Economc Revew, Federal Reserve Bank of Kansas Cy, Quarer 2 pp Cukerman, Alex, Cenral Bank Sraegy, Credbly and Independence: Theory and Evdence, The MIT Press, Cambrdge, Massachuses. 21

24 Esablshng Credbly, Goldberg and Klen Ello, Graham and Ulrch Müller, Opmally Tesng General Breakng Processes n Lnear Tme Seres Models, Manuscrp, Prnceon Unversy. Ello, Graham and Ulrch Müller, forhcomng. Effcen Tess for General Perssen Tme Varaon n Regresson Coeffcens Revew of Economc Sudes. Engel, Charles and Jeffrey Frankel, Why Ineres Raes Reac o Money Announcemens: An Explanaon from he Foregn Exchange Marke, Journal of Moneary Economcs, vol 13, pp Ehrmann, Mchael and Marcel Frazscher, Equal Sze, Equal Role: Ineres Rae Inerdependence beween he Euro Area and he Uned Saes, European Cenral Bank workng paper seres no. 342 Aprl). Esrella, Aruro and Frederc Mshkn, The Predcve Power of he Term Srucure of Ineres Raes n Europe and he Uned Saes: Implcaons for he European Cenral Bank, European Economc Revew, 41, pp Esrella, Aruro and Anhony Rodrguez, One-sded es for an unknown breakpon: Theory, compuaon, and applcaon o moneary polcy, Federal Reserve Bank of New York. Esrella, Aruro, Anhony Rodrguez and Sebasan Schch, How Sable s he Predcve Power of he Yeld Curve? Evdence from Germany and he Uned Saes Revew of Economcs and Sascs, vol. 85, no. 3 Augus), pp Faus, Jon, John Rogers, and Jonahan Wrgh, The Hgh Frequency Response of Exchange Raes and Ineres Raes o Macroeconomc Announcemens, Inernaonal Fnance Dscusson Papers 784, Board of Governors of he Federal Reserve Sysem. Flemng, Mchael and El Remolona, Prce Formaon and Lqudy n he U.S. Treasury Marke: The Response o Publc Informaon Journal of Fnance vol. LIV, no. 5 Ocober). Forder, James, Cenral Bank Independence and Credbly: Is here a shred of evdence?: Revew, Inernaonal Fnance, vol. 3, no. 1 Aprl), pp Goldberg, Lnda and Deborah Leonard, Wha Moves Soveregn Bond Markes? The Effecs of Economc News on U.S. and German Yelds, Federal Reserve Bank of New York, Curren Issues n Economcs and Fnance volume 9, no. 9 Sepember). Gurkaynek, Refe and Jusn Wolfers, An Inal Analyss of Marke-Based Macro Forecass, Uncerany, and Rsk, forhcomng n NBER Inernaonal Semnar on Macroeconomcs Huznga, John and Frederc Mshkn, Moneary Polcy Regme Shfs and he Unusual Behavor of Ineres Raes, Carnege-Rocheser Conference Seres on Publc Polcy, vol. 24 Sprng), pp