ECONOMICS AND POLITICS: INTEREST RATE CONVERGENCE IN EUROPE AND EMU by Lvo Sracca * Absrac Ths paper ses ou o explan he move of long-erm neres raes n Europe oward convergence wh he Deusche Mark n he perod leadng up o he esablshmen of EMU. On he bass of an a-heorecal laen facor analyss approach, whch s ndependen of any arbrary assumpon wh regard o he neres rae spread prevalng n he non-emu scenaro, he convergence s broken down no nernaonal and domesc facors. The man concluson of he sudy s ha he pursu of polces of macroeconomc sablsaon a he sngle-counry level, alhough a necessary prerequse for he convergence process o ake place, would no have been suffcen had he whole EMU projec, for polcal reasons, been abandoned. Therefore, long-erm neres raes, and hus ulmaely moneary polcy credbly, n perpheral European counres were hghly sensve o he (essenally polcal) probably of he realsaon of a broad moneary unon n Europe. Boh economcs and polcs conrbued o he srucural change n he moneary regme mplc n EMU. Conen 1. Inroducon...2 2. An apprasal of he leraure: why more research s needed...6 3. An APT approach...11 3.1 The model...11 3.2 Idenfcaon of he laen facors: esable condons...12 4. The emprcal analyss...15 4.1 The emprcal model...15 4.2 Resuls...17 4.3 Hgh yelder counres: he forward spreads wh he DM...21 4.4 Dscusson of he resuls...24 5. Some caveas...29 6. Concludng remarks...30 * Address: European Cenral Bank, Moneary Polcy Dvson, Kasersrasse 29, 60311 Frankfur am Man. E- mal: lvo.sracca@ecb.n. I wsh o hank wo anonymous referees, as well as colleagues a he ECB, especally Vncen Brousseau and Francesco Mongell, for her helpful commens. I reman responsble for any ousandng errors and for he vews expressed n hs paper, whch are my personal vews and do no n any way nvolve he nsuon of whch I am a member.
1. Inroducon The move of he long-erm neres raes of he man European currences oward convergence wh he Deusche Mark (DM) n he perod precedng he realsaon of Economc and Moneary Unon (EMU) and he launch of he common currency n January 1999 has been mpressve. In some counres, once called he EU perphery or hgh yelders, he longerm (.e. 10-year) neres rae spread wh Germany declned by several percenage pons beween md-1995 and he end of 1998, reachng hsorcally low levels a he sar of 1999. Accordng o he expecaons heory of he erm srucure, long-erm neres raes are a (weghed) expecaon of fuure shor-erm neres raes. The move of long-erm neres raes oward convergence wh he DM hus refleced he growng consensus among marke parcpans ha, n he dsan fuure, moneary polces n sngle counres would have algned hemselves wh he moneary polcy of Germany. I can be clearly observed ha hs expecaon had aken frm roo n he fnancal markes n recen years; however, he precse movaon behnd he convergence process canno be easly denfed. A number of counres were se o share, n he fuure, he moneary polcy of Germany, owng eher o parcpaon n a moneary unon wh Germany, or o an ndependen decson by her cenral banks o follow he moneary polcy of he Deusche Bundesbank. In parcular, mos recen years have wnessed a process of growng economc convergence, as well as he adopon of nsuonal arrangemens ha closely resemble hose of he Bundesbank n erms of cenral bank ndependence and of a deermnaon o fgh nflaon n he counres of he EU perphery. I could be argued ha rrespecve of wheher or no moneary unon wh Germany were acheved moneary polces n hese counres would have followed he German cenral bank n s devoon o prce sably, hereby behavng as f moneary unon acually exsed. Conversely, accordng o an alernave scenaro, had EMU faled, he same counres would have reurned o he moneary regme prevalng n he pas or a leas hs would have been he prevalen expecaon n fnancal markes crcles. The objecve of hs paper s o asceran wheher he observed convergence n long-erm neres raes wh he DM, and herefore ulmaely n moneary polcy credbly, could have been possble whou he realsaon of a moneary unon (EMU) n whch hese counres and Germany ook par. Sudyng he relave mporance of purely domesc and EMU-relaed 2
facors o neres rae convergence does no necessarly call no queson he mporance of he economc polces pursued n he counres ha have joned EMU or wll probably jon n he near fuure. For example, many of hese counres have succeeded n subsanally reducng he nflaon rae and n correcng severe fscal mbalances. Ths paper ends o assume ha hese polces were a necessary prerequse enablng he convergence o ake place. The objecve of hs paper s raher o examne wheher hese polces would have been suffcen f he whole EMU projec, or raher EMU nvolvng a large number of counres and no resrced o he radonal core of Europe, had had o be abandoned for essenally polcal reasons. Expressed more drecly, could sngle (hgh yelder) counres have mananed he credbly ganed n he process of he convergence of economc fundamenals even f EMU had faled? Was a permanen gan n moneary polcy credbly for hese counres possble only for he group as a whole, or also for ndvdual counres? In he laer case could well be argued ha he facors underlyng he convergence n neres raes were prmarly domesc and lnked o he convergence of economc fundamenals, alhough he vruous economc polces were ceranly nfluenced o some exen by he requremens of he Treay on European Unon (Maasrch Treay). Ths s broadly he concluson reached by he only wo papers exsng on hs subjec, namely Angelon and Vol (1997) and Favero, Gavazz, Iacone and Tabelln (1997). Accordng o hese auhors, he reducon n long-erm dfferenals wh he DM was prmarly a resul of sound domesc economc polces and herefore no heavly nfluenced by he prospecs of he realsaon of EMU per se. Alernavely, a promnen role n he ncrease n moneary polcy credbly n hgh yelder counres has o be arbued drecly o EMU, wh domesc economc polces acng as a mere prerequse, bu by no means as a decsve elemen. The queson has grea neres no only for he specfc, alhough very mporan, case of EMU, ha s o asceran he relave mers of he long-erm neres rae convergence of domesc auhores essenally he governmens and he cenral banks compared o hose of he EMU projec. I s also neresng o examne how such subsanal gans n moneary polcy credbly can be reached n more general erms. Indeed, a sandard raonale adduced for he esablshmen of EMU, as well as for he European Moneary Sysem (EMS), was founded on he collecve acquson of moneary polcy credbly by a number of counres (Frowen, 1993). Here he focus s on he erm collecve : how mporan s he collecve dmenson o explan subsanal gans n credbly? How dependen s he acquson of 3
moneary polcy credbly by one counry on he (jon) acquson of moneary polcy credbly by oher counres? In order o ackle hs neresng bu demandng analycal ask, hs paper adops an approach smlar o ha employed by De Grauwe (1996), who red o esmae he probably of EMU from long-erm forward spreads beween seleced European currences and he DM. Forward spreads are parcularly useful n hs analyss because hey denfy he expecaon of he (dfference n) shor-erm neres raes prevalng n a gven perod n he fuure. De Grauwe assumed ha, gven ceran condons (for example rsk neuraly), he observed long-erm forward spread wh he DM for any (EU) currency can be expressed as a weghed average of he expecaon of wo saes: he frs beng ha EMU s realsed and he currency akes par n, whle he second s ha EMU s no realsed or ha he currency does no parcpae. The weghs of he average are gven by he probably of EMU beng acheved and he currency parcpang n, whch s ndeed he objecve of he esmaon. Usng hs approach, s also possble o handle a relaed bu dfferen queson, namely he conrbuon of domesc and EMU-relaed facors o he convergence n long-erm neres raes. Ths was emphassed by Favero e al (1997) and s also he pon of vew adoped n hs paper. De Grauwe s approach has, however, an obvous drawback. In he EMU scenaro, he shorerm dfferenal of any currency vs-à-vs he DM can be approxmaed o zero so ha forward spreads are also easly denfed wh zero. By conras, n he non-emu scenaro he forward spreads wh he DM are no precsely denfed. De Grauwe esmaed hem by usng he neres rae dfferenal prevalng n 1990, when EMU was only a remoe possbly. The assumpon made n respec of he non-emu dfferenal s crucal o hs approach and unavodably conans some arbrarness. Beng a couner-facual hypohess, canno be esed. For nsance, Favero e al (1997) esmaed he dfferenal for Ialy, assumng ha he Bank of Ialy followed a parcular polcy rule (namely he Taylor rule) n he even of EMU no beng realsed or Ialy no akng par. Angelon and Vol (1997) for Ialy, Span and Sweden consdered he dfferenal prevalng jus before he launch of he euro n 1999. Boh assumpons evdenly conras wh ha of De Grauwe (1996). The need for an arbrary assumpon on he non-emu dfferenal s nheren n he De Grauwe approach, because he behavour of he long-erm forward spreads (of he expecaon of fuure shor-erm neres rae spreads) mgh n fac reflec boh an ncreasng probably 4
of EMU beng realsed and/or a decreasng non-emu dfferenal. In fac, he wo drvng forces are dffcul o separae concepually bu above all hey are emprcally mpossble o dsngush whn hs framework. Ths paper res o fnd an answer o hs denfcaon problem by adopng a mulvarae approach, alhough sll n par based, as n De Grauwe (1996), on he analyss of long-erm forward spreads. The basc dea s very smple: he probably of EMU should affec all nvolved currences, whle purely domesc facors are, almos by defnon, mperfecly correlaed across counres. To handle he problem n an analycally sound manner, an APT model was esmaed on he long-erm forward neres raes n a large sample of currences, adopng a very general specfcaon n order o ake no accoun any possble dsurbng facor, such as he presence of a purely global componen, whch could have had asymmercal effecs on dfferen European currences. The model ook he form of a laen facor analyss of he daly changes n long-erm forward spreads. By usng hs mehod, hs paper denfed a laen facor whch has every lkelhood of represenng he probably of a broad EMU from he begnnng of 1996 o he end of 1998. The sample of currences used n he emprcal analyss covered hree man groups of counres. Frs, a group of European counres wh radonally low moneary polcy credbly (Ialy, Span, Porugal, Denmark, Fnland, Sweden and he Uned Kngdom). Second, a group of core European counres, he moneary polcy of whch had closely followed ha of he Bundesbank (Germany and s moneary saelles: Ausra, Belgum, France and he Neherlands). Thrd, a conrol group of oher currences no n any way nvolved n he esablshmen of EMU (he US dollar, he Japanese yen, he Swss franc, he Canadan dollar, he Ausralan dollar and he Norwegan krona). The nroducon of he conrol group was movaed by he need o augmen he explanaory power of he analyss and o beer denfy possble dsurbng facors. The use of a laen facors analyss was essenally movaed by s ably o spl he behavour of a sngle forward spread beween s relaon o he common facors and an orhogonal counry-specfc componen. The adopon of hs mulvarae mehod has wo noable advanages over he unvarae mehod proposed by De Grauwe (1996). Frs, does no requre he specfcaon of a necessarly arbrary assumpon n respec of he non- EMU spread, whch s, raher, a resul of he analyss. Second, akes explcly no accoun he dsncon, also nroduced by Wedmann (1996), beween he probably of EMU beng 5
esablshed (common facor) and he probably of a sngle currency parcpang n EMU, gven ha EMU s esablshed (counry-specfc facor). The man resul of he analyss s ha he forward spreads n fve currences (he Ialan lra, he Spansh pesea, he Poruguese escudo, he Fnnsh markka and he Swedsh krona) were closely relaed o a facor whch was clearly possble o nerpre as he probably of a broad EMU. The relaon was hgh and sgnfcan even on a day-o-day bass. Moreover, he movemens of forward spreads n sngle counres, ha were no explaned by he probably of a broad EMU, exered only a emporary nfluence. Thus, he rend componen of he move of he long-erm neres raes of hese currences oward convergence wh he levels prevalng n Germany, was also explaned by he EMU-relaed, and herefore collecve, facor. Consequenly, he evdence suggesed ha domesc economc polces a he level of sngle counres, whle beng a necessary requremen for he convergence, would no have suffced n he absence of a collecve, and essenally polcally, drve oward he formaon of a large moneary unon n Europe. The mplcaon of hs paper s ha, f he whole EMU projec had had o be abandoned for polcal reasons, he subsanal gans n moneary polcy credbly n EU perpheral counres would have been largely reversed. The paper s srucured as follows. Secon 2 conans an apprasal of he exsng leraure on he probably of EMU beng esablshed and of he convergence of long-erm neres raes n Europe n recen years. The model employed n he sudy s hen descrbed n Secon 3. The emprcal analyss s presened and dscussed n Secon 4. Secon 5 conans some caveas on he valdy of he analyss presened n hs paper. Fnally, Secon 6 presens he conclusons. 2. An apprasal of he leraure: why more research s needed The basc concep of De Grauwe (1996) s o express he forward spread (FS) of currency wh he DM afer January 1999 n he followng erms: (1) FS? P(? EMU)0? P(? EMU) FS 6
where FS s he spread n he absence of a moneary unon and FS s a forward spread of currency vs-à-vs he DM wh boh selemen and maury afer 1 January 1999. Ths relaon holds whaever choce were made abou he converson rae beween he ECU (or euro) and he DM. 1 The probably of EMU s obaned as: P (? EMU )? 1? FS / FS (2) whch converges o zero f FS goes o FS and o one f FS goes o zero. The man problem n esmang P (? EMU ) he objecve of De Grauwe from expresson (2), s gven by FS, whch s unobservable. For example, De Grauwe assumed ha FD,?, would have reurned o he levels prevalng n 1990, when EMU was only a remoe possbly and no a concree opon. The assumpon of De Grauwe was plausble a he me when he wroe hs paper, when a broad EMU had a very low probably and he vruous economc polces n hgh yelder counres had ye o be drawn up. Favero, Gavazz, Iacone and Tabelln (1997) clamed ha equaon (1) can be used o ackle a relaed, bu separae queson: he relave mporance of EMU and purely domesc componens n he convergence of he long-erm neres raes of hgh yelders wh he DM. If wo perods n me are consdered and +1 and?? 1 s defned as he frs dfference beween me +1 and me can be easly derved ha:?? 1? 1? 1? 1 FS? [( 1? P (? EMU ))? FS ]? [? P(? EMU ) FS, ] (3) In hs expresson he change of FS from me o me +1 s broken down no he dfference beween wo componens (n square brackes). Favero e al (1997) reaed he frs 1 Expresson (1) s obvously a smplfcaon as overlooks possble sources of dsoron. Frs of all, assumes rsk neuraly of nvesors (rulng ou he exsence of rsk prema). I also neglecs he convexy problem of he yeld curve (of he Jensen nequaly ype). Furhermore, assumes he overall valdy of he expecaons heory of he erm srucure, whch s frequenly rejeced n emprcal analyses (alhough (segue) 7
as a purely domesc facor, represened by he change n non-emu dfferenals weghed wh he probably of EMU a me +1. For nsance, f he probably of EMU s one, hen hs erm becomes rrelevan. The second componen s reaed as a purely nernaonal facor and drecly relaed o he change n he probably of EMU over me. The wegh s gven by he dfferenal a me. For example, f he dfferenal FS was already around zero a he sar (me ), any change n he probably of EMU should no nfluence he observed? FS (a good example mgh be gven by core Europe counres). Assumng,?1 whou loss of generaly, ha quanes a me are gven and can be reaed as consans and f we le A and B be, respecvely, P (? EMU ) and FS,, rearrangng erms from equaon (3), he followng s obaned: (4)???? 1FS? [( 1? A??? 1P(? EMU ))?? 1 FS ]? [? 1P( EMU ) B] Defnng?? 1P(? EMU ) as INTERNATIONAL /? 1 and? FS as DOMESTIC /? 1 s he followng fnally obaned:?1? 1FS? [( 1? A? INTERNATIONAL /? 1) DOMESTIC /? 1] [ INTERNATIONAL /? 1B]?? (5) I s shown ha he effecs of INTERNATIONAL and DOMESTIC are mulplcave and he change n forward spreads beween and +1 canno be drecly broken down addvely no wo componens whn hs approach. Therefore, he Favero e al (1997) breakdown s only an approxmaon of a rue addve spl no domesc and nernaonal componens. In fac, by neglecng he presence of INTERNATIONAL n he frs erm, and arbung o he domesc componen only, ha paper mgh have undervalued s role n he convergence of forward spreads. The concluson s ha he dsncon proposed by Favero e al, whle Gerlach and Smes, 1997, found encouragng evdence for European counres). Alhough mporan n prncple, hese facors are deemed neglgble n De Grauwe (1996) and hs approach s adoped here. 8
convenen and neresng as a frs sep o acklng he problem, s, when examned more closely, concepually unclear. Furhermore, and more mporanly, expresson (5) s a sngle equaon wh wo unknown parameers (INTERNATIONAL and DOMESTIC); hence an denfcaon problem exss. Favero e al (1997) resolved hs second problem by assumng a parcular value for FS (gven by he shor-erm neres rae mpled n 1999 onwards by a Taylor rule and ha of Germany). In he same ven, Angelon and Vol (1997), focusng on Ialy, Span and Sweden, esmaed he domesc componen of he convergence n forward spreads as he dfference beween curren spreads and (shor-erm forward) spreads prevalng jus before he launch of he euro. Boh papers concluded ha he domesc componen prevaled n he reducon of forward spreads wh he DM. The assumpons of Favero e al (1997) and Angelon and Vol (1997) on he non-emu spread ndubably and correcly refleced he greaer opmsm prevalng n 1997 on he economc polces conduced n hgh yelder counres, for example n erms of he reducon of fscal mbalances and of nflaon. Ths explans why her assumpon abou he crucal value of he non-emu spread conrass srkngly wh ha of De Grauwe (1996). Table A summarses he assumpons and fndngs of hese man conrbuons o he (lmed) leraure on he probably of EMU beng realsed and he convergence of long-erm neres raes n Europe. 9
Table A Synhec descrpon of he leraure on EMU probably FS n non-emu Ineres raes used Counres Man resul De Grauwe (1996) Ineres rae dfferenals prevalng n 1990 Forward raes 5-10 years prevalng n 1998-2001 Core Europe (Neherlands, Belgum, France) and perphery (UK, Denmark, Ialy, Span, Sweden) Core counres expeced o parcpae, UK s possbly n, Ialy, Sweden and Span ou; low probably of a broad EMU Favero e al (1997) The Bank of Ialy follows a Taylor rule from 1999 Insananeous forward neres raes prevalng n 1999-2002, calculaed wh he Nelson-Segel nerpolan Ialy Domesc fundamenals preval n explanng he convergence Wedmann (1996) (a) Weghed mean over he spreads beween sngle currences and remanng ECU baske currences Forward neres raes 5-10 years prevalng n 1999-2001 Core Europe (Neherlands, Belgum, France) and he perphery (UK, Denmark, Ialy, Span) Core counres expeced o be n, Ialy and Span ou Angelon and Vol (1997) Ineres rae dfferenal prevalng jus before January 1999 1-year forward raes prevalng n 1999 and n 2002 Ialy, Span, Sweden Domesc fundamenals preval n explanng he convergence (a) Wedmann (1996) dsngushed beween absolue and condonal probables of currences jonng EMU. The wo papers dscussed so far (namely Favero e al, 1997, and Angelon and Vol, 1997) represen mporan, and n fac he only exsng, conrbuons n he leraure on he assessmen of he reasons behnd he convergence of long-erm neres raes (of moneary polcy credbly) n Europe. However, seems clear ha more heorecal and emprcal research s needed on hs opc. The man reason s ha he approach adoped by Favero e al (1997) and by Angelon and Vol (1997) s (admedly) heavly condoned by he assumpon made by he auhors abou he spread prevalng n he non-emu case. In fac, by specfyng a parcular assumpon abou he behavour of he domesc componen (more precsely, abou he me seres behavour of he non-emu spread), hese papers mplcly denfed he breakdown of equaon (5). Therefore, he measuremen of he relave conrbuons of domesc and nernaonal facors o he convergence of long-erm spreads, 10
nsead of beng a real produc of he analyss, s more he oher sde of he con of he assumed varaon over me of he non-emu spread. 3. An APT approach 3.1 The model The objecve of hs paper s o develop a esable model whch s able o denfy he domesc and nernaonal conrbuons o he convergence of long-erm neres raes whou mposng a pror a parcular value for he non-emu spread. The basc nuon s ha whle he esablshmen of a broad EMU was an essenally mulvarae phenomenon nvolvng a ceran number of currences (counres), domesc economc polces, and he expecaons of hem, were a leas only parally correlaed across counres. Therefore, n a mulvarae seng, should be possble o separae, and measure, he purely domesc componen from he collecve prospec of he realsaon of a broad moneary unon n Europe. In par, hs was already aken no accoun n Wedmann (1996), bu here he model s explcly bul wh hs perspecve n mnd. Gven hese requremens, he mos drec way o proceed would be o esmae a mulvarae model for he forward spreads n hgh yelder counres vs-à-vs he DM n order o capure he common and he ndvdual componens. However, hs roue would be heorecally weak because oher facors, no necessarly relaed o he probably of a broad EMU, could have played a role n recen years n reducng he forward spreads n some currences vs-à-vs he DM. Some recen evdence, gahered, for example, a he OECD (Chrsansen and Pgo, 1997), has suggesed ha long-erm neres raes are hghly correlaed nernaonally; hey may hen parally reflec a purely global facor, whch, durng he perod precedng he launch of he common currency, mgh have been relavely ndependen of he probably of a broad EMU. If neres raes of dfferen currences dd no reac unformly o he developmen of a global facor, he forward spread vs-à-vs he DM may have been affeced by somehng unrelaed o EMU. Therefore, was decded o focus on a very general mulvarae model o allow alernave facors o be aken no accoun. The basc srucure of he model s an APT as nroduced by Ross (1976), n whch he dependen varables are he long-erm forward neres raes of a 11
number of currences. If we le he model s: FF be he long-erm forward rae of currency a me hen P j FF??? jl?? j?1 (6) The long-erm rae s broken down n s relaon, measured by he coeffcens? j, o P laen facors j L, plus a purely ndvdual componen?. Theorecally, hs can be nerpreed as he forward raes presenng separae rsks wh coeffcens gven by her exposure o such rsks. The model n (6) s appled o a number N of currences. A frs group of currences s comprsed of hose of he hgh yelder counres, he Ialan lra, he Spansh pesea, he Poruguese escudo, he Fnnsh markka, he Swedsh krona, he Fnnsh markka, he Dansh krona and he Brsh pound. A second group of currences s comprsed of he core European counres: he German mark, he Ausran schllng, he Duch gulder, he French franc and he Belgan franc. Fnally, a hrd group of conrol varables s made up of oher currences no nvolved n European Moneary Unon, noably he US dollar, he Japanese yen, he Canadan dollar, he Norwegan krona and he Swss franc. 3.2 Idenfcaon of he laen facors: esable condons The model n (6) s able o denfy laen facors and o measure her mporance n he developmen of long-erm forward raes n ndvdual currences, bu o be relevan for he purpose of hs paper he laen varables have o be unambguously nerpreed. Because he emprcal analyss was conduced on daa from he nnees, a perod of hgh negraon of world capal markes, one facor deemed mporan should be global. I makes a grea deal of sense for long-erm neres raes, and herefore forward raes, o move all ogeher o some exen. Ths corresponds o he facual observaon, whch can easly be drawn from he fnancal press, ha markes end o move globally. Moreover, because mos of he man mpulses o he world economy, whch ulmaely drve long-erm neres raes, come from he Uned Saes, hs global facor should be also hghly correlaed o he long-erm neres raes of he US dollar. 12
To summarse, a global facor may be reasonably denfed f: (1) he coeffcens? have all he same sgn, posve (or negave), and are possbly all of sgnfcan magnude (e.g. sascally sgnfcan); and (2) he coeffcen assocaed wh he US dollar s posve (or negave) and hgh. Snce he objecve of hs paper s o denfy a facor assocaed wh he probably of a broad EMU, we also need esable condons necessary o denfy ha facor among he laen varables n (6). The EMU facor for whch we are lookng may be defned qualavely as he probably ha hgh yelder counres n Europe would share n he foreseeable fuure he moneary polcy of Germany (broad EMU). For a laen varable o be denfed as he probably of a broad EMU, a leas hree necessary condons need o preval. Frs, longerm forward raes n hgh yelders should have been negavely correlaed wh hs facor. Ths seems reasonable because a hgher probably of EMU beng esablshed would have mpled a hgher lkelhood of he moneary polcy of Germany beng shared, ha s, parcpaon n an envronmen of enhanced moneary polcy credbly. Second, he sgn of he coeffcens of he core European currences should raher be oppose: gven ha hese currences were radonally consdered safe, hey were lkely o be he objec of safe haven flows when he perpheral currences were perceved as more rsky by he markes due o a decrease n he probably of a broad EMU. Ths corresponds o he facual observaon, almos a clché, ha he DM ended o srenghen, and he German (core Europe) long-erm neres rae o declne, n mes of hgh uncerany abou he realsaon of a moneary unon n Europe. As a consequence, a posve relaon beween he long-erm forward raes of hese currences and he broad EMU facor s o be expeced. The same sgn of he coeffcens s also compable wh a dfferen movaon, whch appeared repeaedly n he fnancal press, especally durng he early sage of he convergence leadng o Sage Three of EMU. Core European currences could have been perceved as more rsky by markes had here been a hgher probably of a broad EMU, due o he weakenng of he moneary dscplne of he Bundesbank. Ths would, n urn, have refleced he case of he Governng Councl of he ECB beng formed by a majory of counres wh a lower moneary polcy credbly. The qualave dfference beween he wo scenaros (core European currences subjec o negave safe haven flows or o a decrease of credbly wh he lkelhood of a broad moneary unon beng esablshed) s mporan. I 13
les n he marke s early apprasal of he common currency. However, he sgns of he core European currences n model (6) are no able o denfy he scenaro drvng he marke s expecaons. Neverheless, hs ssue may be resolved by usng he currences of he conrol group. Among hem, he US dollar, he Swss franc and he Japanese yen are wdely raded and radonally sable currences. If he safe-haven heory s correc, hen would also apply o hs group of currences, whch should hus dsplay a posve correlaon wh a broad EMU facor. If, nsead, he euro were perceved o be a weak currency f nvolved a broad number of currences, hen hese hree currences should dsplay a negave coeffcen wh a facor relaed o he probably of a broad EMU, smlar o ha found for he perpheral counres of Europe. Therefore, he sgn arsng from he esmaon of model (6) also represens a es of he early credbly of he euro on he nernaonal fnancal markes. Furhermore, wh regard o he hrd esable mplcaon, he ncluson of he Norwegan krona, he Canadan dollar and he Ausralan dollar s movaed by her raher secondary role n he world currency marke, wha makes lkely ha long-erm forward spreads of hese currences should be broadly unaffeced by he probably of EMU beng esablshed. Ths s essenal o guaranee ha a laen varable, denfed wh he probably of a broad EMU, s no absorbng an addonal facor and herefore dsorng he analyss. To sum up, he hree necessary condons o denfy a laen facor as he probably of a broad EMU are: (1) Coeffcens of he hgh yelders should be negave and sgnfcan; (2) Coeffcens of core European currences should be posve and probably sgnfcan; (3) Coeffcens of he conrol group currences should be () posve or negave, bu probably sgnfcan for he US dollar, he Japanese yen, he Swss franc and () nsgnfcan for he Canadan dollar, he Ausralan dollar and he Norwegan krona. In he followng secon he emprcal model semmng from he framework as n equaon (6) s se up n order o derve drec ess of he condons nroduced herewh. 14
4. The emprcal analyss 4.1 The emprcal model The model n (6) canno be used drecly n he emprcal analyss. The mos sraghforward objecon s ha, by usng levels of forward long-erm neres raes, poenally negraed varables are nroduced n he analyss. In such a seng, he laen facors analyss could be undermned by he same knd of dsorons popularsed by Granger and Newbold (1974) for regresson analyss. Moreover, as facor analyss s normally employed on a cross-secon bass, may no be suable for he reamen of daa characersed by he presence of a rend (see Morrson, 1990). Fnally, and perhaps more mporanly, as more horoughly dscussed n Secon 4.4, hs paper s more concerned wh he hgh frequency and shor-run relaon beween he varables han wh he rend or long-run relaonshps. 2 Ths s an addonal reason for focusng on he frs dfferences n he forward raes, raher han on he levels. Wh regard o he frequency of he esmaon, daly daa were used. Ths corresponds o he focus on he hgh frequency feaures of he daa. Therefore, he analyss was carred ou on he daly varaons of he long-erm forward neres raes n he wde group of currences nroduced n Secon 3. All forward raes employed n he emprcal analyss were bul on nerbank swap raes, avalable from Daasream. The mehod of calculaon of he forward raes s descrbed n Appendx 1. In he case of he Norwegan krona, due o a lack of daa n he sample under consderaon (from June 1995 o December 1998), forward raes were calculaed on he 5 o 10-year governmen bond yelds. In prncple, hs approxmaon s no correc because governmen bond yelds ncluded a defaul rsk, whch s no, or dfferenly, presen n nerbank swap raes. However, he correlaon beween he daly changes n he Norwegan forward raes calculaed on governmen bonds and on nerbank swap raes was as hgh as 0.99 n he overlappng perod (sarng from md-1997); he approxmaon used can herefore be regarded as adequae. Furhermore, all he emprcal analyses were repeaed elmnang he Norwegan krona from he sample, and no sgnfcan changes n he resuls were found, a leas from a qualave vewpon. 2 Ths s also he reason why oher esmaon echnques, e.g. conegraon, whch focused more on he longrun relaonshps, dd no appear sued n hs conex. 15
Raher han usng he exac formula, for he sake of smplfyng he compuaon, forward raes were calculaed usng an approxmaon. The relably of he approxmaon was checked by comparng wh forward raes derved wh he exac formula for some man currences (noably he US dollar, he German mark and he Japanese yen), and broad confrmaon was found for he perod from md-1995 o he end of 1998. All subsequen emprcal analyss s based on forward raes wh (mplc) selemen a 7 years and maury a 10 years (wh he sole excepon of Norway). Sarng he sample from 1996 (see more on hs laer), hs mean consderng forward raes wh selemen sarng from 2003 and maury from 2006. No dsncon was made beween he case of a currency jonng n 1999 or some me laer, because lmng maers o a precse dae would have undervalued he role of EMU n explanng long-erm neres rae convergence (see also Angelon and Vol, 1997 on hs pon). In hs way, a longer-erm perspecve was delberaely adoped as compared o he analyss of parcpaon from 1999 only. All he emprcal analyses conaned n hs paper were also repeaed usng 5 o 10-year forward neres raes, and no sgnfcan dfference n resuls was deeced. If he 7 o 10-year forward neres rae n currency s defned as and a me as F, and s absolue daly change as f, he emprcal model used n he esmaon was: P j f??? jl? j?1 e (7) where l j are he laen facors relaed o he changes n forward neres raes. The emprcal analyss ook he form of a prncpal componens analyss on he correlaon marx of he forward raes. As a resul, he denfed laen facors were orhogonal o each oher and wh he domesc componen e, whch s a desrable propery n vew of denfyng he ndependen conrbuons of he nernaonal and he domesc componens n he me seres behavour of he f, he man objecve of he mulvarae analyss of hs paper. 16
The laen facors are normalsed o have a zero mean and a un varance. They canno, herefore, be drecly nerpreed n economc erms as any lnear ransformaon of hem would be equally vald. Ther behavour over me, however, can be meanngfully analysed. 4.2 Resuls The daly daa on forward neres raes descrbed n he prevous secon were avalable from June 1995 o December 1998. Deleng mssng daa from he sample, slghly more han 1,000 observaons were made avalable. Regardng he reamen of oulers, he repored resuls dd no nclude any adjusmen. However, he soundness of he resuls was checked by excludng observaons (for every forward neres rae) whch vared, n absolue value, by more han hree sandard devaons from he mean. Ths changed he resuls o some degree, bu no n such a way as o fundamenally aler he qualave message arsng from he analyss. Therefore, only resuls obaned wh unadjused fgures are repored here. A very mporan ssue n hs esmaon concerned he choce of he sample perod. Snce wha s sough are he effecs of he probably of EMU on long-erm forward neres raes, he sample should sar n a perod n whch ha probably was sgnfcanly above zero. A convenen sar, for example, could be December 1995, whch would concde wh he Madrd meeng of prme mnsers, a whch he process of European moneary negraon ganed momenum. However, should be recalled ha he nenon was o denfy he probably of a broad EMU, ha s, a moneary unon comprsng a large number of noncore European currences. Wedmann (1996), for example, repored ha publc opnon surveys conduced a he sar of 1996 showed ha he parcpaon of Ialy and Span n he moneary unon was perceved as hghly unlkely, f no mpossble. Therefore, aempng o deec he nfluence of a broad EMU facor by he end of 1995 and he begnnng of 1996 s no sraghforward. To confrm hs nformal argumen, he facor analyss n (7) was run on dfferen perods from around June 1995 o around June 1996. The resuls of he facor analyss n he frs sx monhs of 1996 were also esed. Invarably, no laen facor could be denfed as he probably of a broad EMU. Insead, from Aprl 1996 a facor clearly relaed o EMU began 17
o emerge; as a consequence, 1 Aprl 1996 was chosen as he sarng dae of he man facor analyss. Table I (Secon Tables and Chars) repors he facor loadngs of a prncpal componens analyss as n (7), on a sample of 18 currences, coverng a me span from 1 Aprl 1996 o 31 December 1998. As already poned ou n he prevous secon, he facor analyss was conduced on he daly changes of he 7 o 10-year forward neres raes. The resuls here were raher neresng. The frs laen facor, accounng for 42% of all he varance (Table II), can be easly denfed as purely global. In fac, all long-erm forward neres raes dsplayed a posve and hghly sgnfcan coeffcen on hs common facor. The coeffcens of ndvdual currences showed dfferen absolue values, wh he Deusche Mark scorng he hghes (0.89) and he Japanese yen he lowes (0.15). The laer fndng seemed conssen wh anecdoal evdence ha he Japanese fnancal markes and he Japanese economy are relavely more ndependen from he developmen of global facors. The coeffcen of he US dollar, alhough hgh, was no among he hghes (0.62). Overall, hs common facor suggesed he (hghly reasonable) presence of a global facor, drvng he developmen of long-erm forward neres raes n all he man raded currences of he world. However, he wegh of hs common facor n ndvdual currences was hghly varable, as accouned for 79% of he explaned varance of he German neres raes and for 2% of he Japanese neres raes (Table II). To check he purely global naure of hs frs facor, he correlaon coeffcen was calculaed wh a smple average of he forward raes n he sample of 18 currences, obanng a value as hgh as 0.98 (Table III). Ths rose o 0.99 f oulers were excluded from he analyss. More neresngly, he hrd laen varable, accounng for a 6% of oal varance, clearly appeared o be hghly correlaed o he probably of a broad EMU beng esablshed. Ths nerpreaon sems, n he frs place, from he nspecon of he coeffcens of ndvdual currences. All hgh yelders, wh he excepon of he Brsh pound, dsplayed a negave and hghly sgnfcan relaon wh hs facor. The absolue values of hese negave coeffcens ranged from 0.l4 for he Dansh krona o 0.55 for he Poruguese escudo. The excepon of he Brsh pound s no very surprsng, gven he relave ndependence and almos averson of publc opnon n he Uned Kngdom o he process of European moneary unfcaon. The small coeffcen found for he Dansh krona s also no surprsng, 18
as Denmark may be consdered an nermedae case beween core European and hgh yelder counres. Moreover, s parcpaon n EMU has been always n queson. A second pece of evdence wh regard o he naure of he hrd laen facor came from he core European currences, he coeffcens of whch were nvarably posve and sgnfcan, as predced n Secon 4. Indeed, he magnude of he coeffcens was hghly smlar across core Europe, rangng from 0.18 for he French franc o 0.28 for he Deusche Mark. Ths resul s very nuve because he Deusche Mark has been radonally regarded as somewha more relable han he French franc among core European currences. The Belgan franc and he Duch gulder, reflecng her recen hsory of peggng o he Deusche Mark, showed coeffcens no sgnfcanly dfferen from ha of he German currency. Fnally, he coeffcens found for he conrol group currences broadly confrmed he prerequses se down n Secon 3 and provded some addonal nsghs. The coeffcens of he Swss franc and he US dollar were boh posve and sgnfcan, of he same order of magnude as he Deusche Mark. Ths s conssen wh he nerpreaon ha hese wo currences benefed from safe haven flows when he probably of a broad EMU declned, and he oppose movemen when he probably of a broad EMU recovered. To a lesser exen, hs seemed o be rue also of he Japanese yen. I s, however, more dffcul o nerpre he posve and sgnfcan coeffcen found for he Canadan dollar, a currency, whch s neher wdely raded nor regarded as parcularly sable n fnancal markes. The mos lkely nerpreaon leads o a hgh culural and geographcal proxmy wh he US dollar, whch may have caused an ndrec correlaon wh he broad EMU facor for he Canadan dollar. In lne wh he a pror of Secon 3, he coeffcens of he Ausralan dollar and he Norwegan krona were nsgnfcan. Oher evdence suppored he nerpreaon of he hrd laen facor as he probably of a broad EMU ha hs analyss was seekng. A frs, obvous es s o see wheher he coeffcens found for he whole perod dd collapse n 1998, when he probably of EMU was always close or equal o one and no nfluence should have been exered on long-erm forward raes. Ths requremen was esed by runnng he facor analyss n equaon (7) on he perod 1 March 31 December 1998 (Table IV). Any EMU-relaed regulary, prevously observed over he whole sample 1996-1998, collapsed and no laen facor could be nerpreed as conneced o EMU. Furher evdence may be gahered by he me seres 19
behavour of he laen facor. As expeced, he me seres behavour of hs varable whch was calculaed on he daly changes of forward raes dsplayed an nsgnfcan auocorrelaon durng 1996 and 1997, whle showng a hghly sgnfcan negave auocorrelaon durng 1998 (Table V). In oher words, he realsaons of he hrd laen facors were permanen (from he pon of vew of he levels of he forward raes) durng 1996 and 1997, becomng ransory n 1998. Ths, agan, s conssen wh a reasonable qualfcaon of a good broad EMU facor: he large, permanen ncreases n he probably of a broad EMU ook place n 1996 and 1997, and ceranly no n 1998. Fnally, he correlaon coeffcen of hs laen varable wh he smple average of daly changes n he spread wh he DM of fve currences was calculaed, he Ialan lra, he Spansh pesea, he Poruguese escudo, he Fnnsh markka and he Swedsh krona. The calculaon of forward spreads wh he DM s parcularly convenen because hs was he sarng pon of De Grauwe (1996) n assessng he probably of EMU. Ineresngly, he correlaon coeffcen found n hs exercse was as hgh as 0.90 (Table III); f he oulers n he sample were dsregarded, ouched 0.95. The second laen facor was raher more dffcul o nerpre. The mos lkely nerpreaon has o do wh he specfc developmens of he Pacfc basn area, because he coeffcens of counres n ha area have all a hgh magnude. The coeffcens of he Ausralan dollar, he Canadan dollar and he Japanese yen were parcularly sgnfcan. The coeffcen of he Swss franc, of he same sgn as he Pacfc basn counres, s however very hard o nerpre. Neverheless, here s no sgn ha hs second laen facor may be meanngfully relaed o he probably of a broad EMU. Therefore, gven ha hs laen varable was no cenral o our analyss, hs ssue can be lef for furher research. Taken ogeher, he frs hree laen facors accoun for 56% of oal varance n he sample, whch can be regarded as a hgh value f calculaed on de-rended daa. To evaluae he soundness of he resuls, sensvy analyss was made of he me range used for he esmaon. As already repored, he resuls were no sound f he me range sared n 1995, as he facor assocaed wh he probably of EMU was mpossble o denfy. Generally speakng, he exsence of he global facor was relavely ndependen of any me range and of he ncluson/excluson of oulers from he sample. However, he presence of an EMU facor was more roublesome. Noneheless, all he basc resuls repored n hs secon were confrmed n he perod beween Aprl 1996 and he end of 1997. Therefore, as 20
expeced, he excluson of 1998 dd no aler he man resuls concernng he probably of EMU. Esmaons made separaely on 1996 and 1997, no repored here, confrmed he presence of an EMU varable, alhough he resuls were less nea and elegan han hose repored above. The qualave message arsng from he emprcal analyss, however, dd no change. The sascal analyss of he daly changes n he forward neres raes showed ha fgures were affeced o a large exen by skewness; o ake hs no accoun, an aemp was made o calculae es sascs on facor loadngs by boosrappng. Ths changed he sgnfcance values only a he margn, whle he mporan coeffcens remaned unaffeced. Therefore, only es sascs based on classcal nference are repored n Tables I and IV. I should be noed ha he sandard es sascs apply o varables characersed by no auo-correlaon, whch was no generally a feaure of he daly changes of he consdered forward neres raes. However, he auo-correlaon coeffcens were so small ha her praccal relevance should be very lmed. 4.3 Hgh yelder counres: he forward spread wh he Deusche Mark From he prevous secon we have learn abou he exsence of a broad EMU facor n he deermnaon of daly changes of long-erm forward neres raes n he perod 1996-1998, bu no reference has ye been made o he relave mporance of nernaonal and domesc conrbuon o he reducon of he forward spreads wh he DM n hgh yelders whch s he man objecve of hs paper. Afer all, around half of he varance of daly changes n he forward raes n hgh yelders could no be explaned by he common and herefore nernaonal facors (see Table II). Therefore, furher seps had o be underaken o drecly assess he cenral ssue. The hrd facor has already been repored as hghly correlaed wh he daly changes n he average forward spread of fve currences wh he DM. In hs secon hs average s drecly subsued for he laen facor, for wo man reasons. The frs s ha he average has he rgh scale and s free of any normalsaon, whch affeced he laen facor. Ths has an obvous advanage n erms of nerpreably. Second, he average s heorecally very convenen because Ialy, Span, Porugal, Fnland and Sweden were precsely hose counres whch amed o parcpae sooner (he frs four) or laer (Sweden) n a broad EMU. Moreover, for hese counres EMU was deemed o represen a fundamenal change of 21
moneary regme. As a consequence, here are boh sascal and economc reasons for nerpreng her common behavour expressed as he average of he long-erm forward spreads wh he DM as a vald proxy of he probably of EMU. Le us defne fs as he average of he daly changes n forward spreads wh he DM n he aforemenoned fve currences. To assess he relave weghs of he common and he domesc componens n he move of forward spreads o around zero n 1996-1998, he followng regressons were run: fs???? fs? e where (8) fs represens he daly changes n he forward spreads wh he DM, wh =ITL, ESP, PTE, FIM, SEK. Of course, because every currency enered wh wegh 1/5 n he average, he? were expeced o be hghly sgnfcan. The conrbuon of he purely domesc facor. e represened, for every currency, he The resuls of hs regresson, calculaed for he perod 1 Aprl 1996 31 December 1998, are repored n Table VI, Panel A. The coeffcens? were all sgnfcanly above 1/5 and only hose of he Ialan lra and of he Swedsh krona were sgnfcanly smaller han one. The R- squared were also hgh, akng no accoun ha hese were daly changes n forward spreads. The hgh posve correlaon on a daly bass beween forward spreads n hese fve currences s also confrmed by calculang a smple correlaon marx (Table VII). The concluson here s ha forward spreads refleced he developmen of he probably of EMU even on a day-oday bass and o a sgnfcan exen. An obvous objecon o hs lne of reasonng s ha he daly correlaon of he forward spreads could smply be he effec of shor-erm movemens on he fnancal markes, whch may no mply a permanen effec on he longer-run behavour of he forward spreads of hese fve currences. Therefore, o assess he permanen or ransory naure of he movemens n hese forward spreads, he correlograms of he average and of he domesc componens resulng from equaon (8) were observed. These, no repored here, showed ha he domesc componens were nvarably characersed by a hgh degree of negave auo-correlaon, 22
whle he common facor was no. In Table VI, Panel B, some evdence wh regard o he sascal properes of he common and of he domesc componens n he fve currences was repored. The daly changes were regressed on 44 lags (namely, wo monhs of daly daa) of hemselves o allow for a longer-run dependence on pas values. The resuls were very clear. The nernaonal facor dsplayed no auo-correlaon, whereas ha of he domesc componen was generally close o, and no sgnfcanly dfferen from, mnus one. The sandard errors were hgh, and he R-squared low, bu he overall paern was unambguous and was also common o all he fve currences consdered. In sum, he resuls of Table VI (Panel B) ndcaed ha, whn wo monhs, any nnovaon n he domesc componens was reversed, whle hs dd no happen o he common facor. Loosely speakng, he behavour of he forward spreads wh he DM n 1996-1998 appears o be conssen wh he followng sochasc process: fs?? 1?? 2 fs? fs? w e (9) e? v? v? 1 where? 1 and? 2 are posve parameers for every, w and he nose componens. Transposed o he levels, hs drecly mples ha: (1) he nnovaons o he probably of a broad EMU arrved n random amouns; (2) hey were permanen o he forward spreads of he fve consdered currences; v are uncorrelaed whe (3) he purely domesc componens were ransory; her effec on he levels of he forward spreads were reversed whn a couple of monhs. Wrng down equaon (9) on levels of he forward spreads (FS), and defnng EMU as he probably of a broad EMU, he followng s obaned: 23
FS?? 1?? 2EMU? DOM EMU? w? EMU?1 DOM? v (10) n whch DOM represened, nuvely, he domesc componens of he forward spreads. The process of he probably of EMU s perssen, whereas ha of he developmen of he domesc componen s ransory. 4.4 Dscusson of he resuls Summarsng he man resuls found n he emprcal exercse of Secons 4.2 and 4.3 before assessng her analycal relevance,. hree man resuls emerged: (1) Daly changes n he long-erm forward neres raes n 18 currences, n he perod 1996-1998, were nfluenced by a common, global facor; (2) They were also, alhough o a less sgnfcan exen, affeced by he probably of a broad EMU, whch exered a downward nfluence on he forward raes n radonally regarded hgh yelder currences, such as he Ialan lra or he Spansh pesea, and a upward nfluence on safe currences lke he Deusche Mark, he Swss franc and he US dollar; (3) The laen facor assocaed o he probably of a broad EMU was, n urn, hghly correlaed o he average forward spread o he Deusche Mark of fve currences (he Ialan lra, he Spansh pesea, he Poruguese escudo, he Fnnsh markka, he Swedsh krona). Havng denfed he probably of a broad EMU n a general mulvarae, laen facors framework, he relaon beween he forward spreads n he fve hgh yelders was examned n order o assess he relave weghs of domesc and nernaonal componens o he move oward convergence wh German raes observed n 1996-1998. The man resuls here were: 24
(1) There was a close relaon beween he daly changes n he forward spreads and he average of he forward spreads n he fve currences (common facor represenave of he probably of a broad EMU); (2) More mporanly, whle he changes n he common facor were permanen, hereby conrbung o a perssen convergence wh German neres raes, he conrbuon of he domesc facors were, for all he fve currences consdered, o a large exen ransory. Taken ogeher, hese resuls sugges ha he convergence of long-erm neres raes n hgh yelders, whch s ulmaely convergence wh he moneary polcy credbly of he Deusche Bundesbank (of he ECB), was deermned o a sgnfcan exen by he probably of a broad EMU, ha s, by a collecve phenomenon. The mporance of purely domesc facors, for example he pursu of sound economc polces, appeared o be subdued. Ths nerpreaon should no, however, be pushed oo far. For example, long-erm neres raes denomnaed n Greek drachma were sll sgnfcanly above he German levels a he end of 1998, ndcang ha he wegh of he domesc polces was no neglgble and ndeed essenal. In erms of he model n (10), he domesc componen DOM could be expressed n he Greek case as: DOM??? w (11) where? represens he premum pad by he Greek ssuers of he non-complance of he Greek economy wh he requremens of he Maasrch Treay. If he fve currences examned n he prevous secon reached?? 0, hs mus be arbued o he vruous economc polces conduced n hose counres, whch permed hem o jon hose counres allowed o form a moneary unon wh Germany. 3 Ths, n urn, may be expressed as follows: domesc economc polces were a necessary bu by no means suffcen condon o oban a permanen gan n moneary polcy credbly. Raher, gven he presence of good economc fundamenals, he suffcen condon permng he achevemen of such mporan and 3 I would have been of grea neres o nclude long-erm forward raes on he drachma n he sample, bu hs was unforunaely mpossble due o daa lmaons. The ncluson of Greece n he sample would shed (segue) 25