WORKING PAPER NO. 92 A VAR DESCRIPTION OF THE EFFECTS OF MONETARY POLICY IN THE INDIVIDUAL COUNTRIES OF THE EURO AREA

Transcription

1 EUROPEAN CENTRAL BANK WORKING PAPER SERIES E C B E Z B E K T B C E E K P WORKING PAPER NO. 92 EUROSYSTEM MONETARY TRANSMISSION NETWORK A VAR DESCRIPTION OF THE EFFECTS OF MONETARY POLICY IN THE INDIVIDUAL COUNTRIES OF THE EURO AREA BY BENOÎT MOJON AND GERT PEERSMAN December 2001

2 EUROPEAN CENTRAL BANK WORKING PAPER SERIES WORKING PAPER NO. 92 A VAR DESCRIPTION OF THE EFFECTS OF MONETARY POLICY IN THE INDIVIDUAL COUNTRIES OF THE EURO AREA * EUROSYSTEM MONETARY TRANSMISSION NETWORK BY BENOÎT MOJON AND GERT PEERSMAN December 2001 * Benoî Mojon: European Cenral Bank, benoi.mojon@ecb.in. Ger Peersman: Bank of England and Ghen Universiy, ger.peersman@bankofengland.co.uk. This paper was prepared in he conex of he ECB s Moneary Transmission Nework. We hank he members of he MTN for valuable inpus. Ger Peersman worked on his paper while being in he ECB s Graduae Research Programme. We hank Andres Manzanares for ousanding research assisance and Frank Smes, Don Bredin and Luz Kilian for helpful discussions and Ignazio Angeloni, Paul Buzen, Fabio Canova, Fiorella de Fiore, Caherine Fuss, Ignacio Hernando, Carlos Robalo, Daniele Terlizzese and Raf Wouers for commens on a previous draf of his paper.

3 The Eurosysem Moneary Transmission Nework This issue of he ECB Working Paper Series conains research presened a a conference on Moneary Policy Transmission in he Euro Area held a he European Cenral Bank on 18 and 19 December This research was conduced wihin he Moneary Transmission Nework, a group of economiss affiliaed wih he ECB and he Naional Cenral Banks of he Eurosysem chaired by Ignazio Angeloni. Anil Kashyap (Universiy of Chicago) aced as exernal consulan and Benoî Mojon as secreary o he Nework. The papers presened a he conference examine he euro area moneary ransmission process using differen daa and mehodologies: srucural and VAR macro-models for he euro area and he naional economies, panel micro daa analyses of he invesmen behaviour of non-financial firms and panel micro daa analyses of he behaviour of commercial banks. Ediorial suppor on all papers was provided by Briony Rose and Susana Sommaggio. European Cenral Bank, 2001 Address Kaisersrasse 29 D Frankfur am Main Germany Posal address Posfach D Frankfur am Main Germany Telephone Inerne hp:// Fax Telex ecb d All righs reserved. Reproducion for educaional and non-commercial purposes is permied provided ha he source is acknowledged. The views expressed in his paper are hose of he auhors and do no necessarily reflec hose of he European Cenral Bank. ISSN

4 Conens Absrac 4 Non-echnical summary 5 1 Inroducion 7 2 The empirical analysis of he effecs of moneary policy in he individual counries of he Euro area 9 3 Idenificaion VAR models for he individual counries in he euro area The specificaion of he benchmark VAR-model for he euro area Resuls Esimaion Discussion of he resuls Pre-EMU euro area moneary policy and naional moneary policy hisories Furher evidence on he effecs of moneary policy shocks Concluding remarks 23 References 25 Appendix: The effecs of moneary policy on componens of M3, bank loans and bank reail ineres raes. 28 Graphics 30 European Cenral Bank Working Paper Series 42 ECB Working Paper No 92 December

5 Absrac: This paper presens a complee se of resuls describing he effecs of moneary policy in 10 counries of he euro area for he pre-emu period. For each counry, we impose one of hree idenificaion schemes depending on is moneary inegraion wih Germany, he nominal anchor of he ERM. The firs idenificaion scheme applies o Germany, he second o counries of he core EMS (Ausria, Belgium and he Neherlands) and he hird o all he oher counries. An unexpeced rise in he shor-erm ineres rae leads o a decrease in GDP, (wih invesmen and expors falling more han consumpion) and a gradual decrease in prices for all counries. We also show ha, given he widh of he error bands around he esimae, we canno rejec ha he effecs of moneary policy on GDP and on prices are broadly similar in he individual counries of he euro area. JEL classificaion sysem: E52. Keywords: Euro area counries, moneary policy, VARs. 4 ECB Working Paper No 92 December 2001

6 Non-echnical summary This paper analyses he ransmission mechanism of moneary policy in 10 counries (Ausria, Belgium, Finland, France, Germany, Greece, Ireland, Ialy, he Neherland and Spain) ha are now members of he euro area. We use VAR models, which is he mos widely used empirical mehodology o analyse he ransmission mechanism. In Europe, he perspecive of EMU led a large par of he lieraure o use VARs o evaluae crosscounry differences in he ransmission mechanism. The ypical paper in his lieraure imposes he same idenificaion of moneary policy shocks across counries, in spie of he differences in moneary policy regime of each counry wihin he European Moneary Sysem (EMS). For insance, Germany could follow an independen moneary policy as he de faco anchor of he EMS while he ineres rae of a counry like Ausria was ied by he fixed exchange rae wih he Deusche Mark during mos of he pas 20 years. We propose insead o esimae for each counry a model ha accouns for he EMS consrain o which he counry was subjec during he sample period. We form hree groups of moneary policy regime-like counries depending on heir moneary inegraion wih Germany. As he anchor of he EMS, Germany is a group on is own for obvious reasons. Ausria, Belgium and he Neherlands form he second group. We model heir moneary policy as if hey were in a "fixed" exchange rae regime wih respec o Germany. I follows ha hey do no have auonomous moneary policy shocks. All he oher counries are modelled as open economies wih a flexible exchange rae vis-à-vis Germany. The influence of German moneary policy is aken ino accoun by including he German ineres rae and he bilaeral Deusche Mark exchange rae in he model. Alhough he German ineres rae is hen a major argumen of he reacion funcion of he cenral bank, here is room for auonomous domesic moneary policy in he adjusmen of domesic ineres rae around he German ineres rae. The conribuion of he paper wih respec o he exising VAR lieraure on he ransmission mechanism in euro area counries is hreefold. Firs, we show ha imposing one of hree relaively simple idenificaion schemes, depending on our knowledge of he moneary policy decision process in he EMS, obain well-behaved and qualiaively consisen effecs of he moneary policy shocks in all he counries. While our models are fairly similar and comparable, we avoid he implausible uniformiy of approaches ha characerises mos of he VAR lieraure on inernaional comparisons of he ransmission mechanism. We also avoid he muliplicaion of models ha confuses he cross-counry comparison. The resuls of our model are consisen wih he consensus view on he ransmission mechanism. A conracionary moneary policy shock is defined as a posiive ECB Working Paper No 92 December

7 deviaion of he ineres rae from he average reacion funcion of he cenral bank for he sample period. I leads o a emporary fall in GDP ha peaks ypically around four quarers afer he shock and o a gradual decrease in he price level. We also show ha for mos counries, M1 iniially decreases and ha he response of invesmen is larger han he one of consumpion. Second, he resuls of our esimaions a he counry level are compared o he resuls obained in a VAR esimaed wih euro area aggregaes by Peersman and Smes (2001). We show ha he effecs of moneary policy on prices and on oupu esimaed for each counry are usually qualiaively similar o he ones obained when he model is esimaed on he aggregae euro area economy. Third, we compare he arificial moneary policy shocks measured for he euro area wih he ones of he individual counries. The euro area shocks seem o be dominaed by he French, he Ialian and he Spanish moneary policy shocks. In paricular, he early 1990s appear as a conracion for he euro area and for mos of he individual counries while he Deusche Bundesbank was jus reacing o he reunificaion boom. Finally, he paper sresses ha, given ha he confidence bands around he responses are generally large, one can no use VAR models o conclude ha some counries are characerised by larger effecs of moneary policy han ohers. 6 ECB Working Paper No 92 December 2001

8 1. Inroducion Undersanding he ransmission mechanism of moneary policy in he euro area is of primary imporance for he implemenaion of he Eurosysem s moneary policy sraegy. Alhough he Eurosysem arges price sabiliy for he area as a whole, counry level evidence should no be negleced for wo reasons. Firs, he analysis of he ransmission mechanism is usually carried ou wihin models ha assume a reacion funcion of he cenral bank. This assumpion is somewha arificial when considering he euro area economy as a whole before he sar of EMU. Therefore esimaing cenral bank reacion funcions a he counry level seems more appropriae. Second, economiss are sill uncerain abou he effecs of moneary policy on economic aciviy and prices. Recen empirical and heoreical sudies, mainly focused on he US economy, end o converge on he view ha conracionary moneary policy shocks lead o a emporary decrease in oupu and o a gradual decline in prices. These resuls are convincing, and herefore policy relevan, mainly because hey are derived from models ha imbed a plausible descripion of he moneary policy decision process. While uncerainy on he ransmission mechanism is even more criical in he conex of EMU, he experience of each counry before EMU provides addiional empirical evidence, which can be used o reduce uncerainy abou he ransmission of moneary policy. This paper analyses he ransmission mechanism of moneary policy in 10 counries ha are now members of he euro area 1. We use VAR models, which is he mos widely used empirical mehodology o analyse he ransmission mechanism. The use of VARs for he analysis of moneary policy sared wih he seminal work of Sims (1980). Recenly, Leeper, Sims and Zha (1998) and Chrisiano, Eichenbaum and Evans (1998) have reviewed wha one has learned from his exensive lieraure regarding he moneary ransmission mechanism in he Unied Saes. In Europe, he perspecive of EMU led a large par of he lieraure o use VARs o evaluae cross-counry differences in he ransmission mechanism. The ypical paper in his lieraure imposes he same idenificaion of moneary policy shocks across counries, in spie of he differences in moneary policy regime of each counry wihin he European Moneary Sysem (EMS). For insance, Germany could follow an independen moneary policy as he de faco anchor of he EMS while he ineres rae of a counry like Ausria was ied by he fixed exchange rae wih he German Mark during mos of he pas 20 years. We propose insead o esimae for each counry a model ha accouns for he EMS consrain o 1 Luxembourg is no modelled because i formed a moneary union wih Belgium, and had no independen moneary policy. Porugal is also excluded because of daa limiaions. ECB Working Paper No 92 December

9 which he counry was subjec during he sample period. We form hree groups of moneary policy regime-like counries depending on heir moneary inegraion wih Germany. As he anchor of he EMS, Germany is a group on is own for obvious reasons. Ausria, Belgium and he Neherlands form he second group. We model heir moneary policy as if hey were in a "fixed" exchange rae regime wih respec o Germany. I follows ha hey do no have auonomous moneary policy shocks. All he oher counries are modelled as open economies wih a flexible exchange rae vis-à-vis Germany. The influence of German moneary policy is aken ino accoun by including he German ineres rae and he bilaeral Deusche Mark (DM) exchange rae in he model. Alhough he German ineres rae is hen a major argumen of he reacion funcion of he cenral bank, here is room for auonomous domesic moneary policy in he adjusmen of domesic ineres rae around he German ineres rae. The conribuion of he paper wih respec o he exising VAR lieraure on he ransmission mechanism in euro area counries is hreefold. Firs, we show ha imposing one of hree relaively simple idenificaion schemes, depending on our knowledge of he moneary policy decision process in he EMS, 2 obain well-behaved and qualiaively consisen effecs of he moneary policy shocks in all he counries. While our models are fairly similar and comparable, we avoid he implausible uniformiy of approaches ha characerises mos of he VAR lieraure on inernaional comparisons of he ransmission mechanism. We also avoid he muliplicaion of models ha confuses he cross-counry comparison. The resuls of our model are consisen wih he consensus view on he ransmission mechanism. A conracionary moneary policy shock is defined as a posiive deviaion of he ineres rae from he average reacion funcion of he cenral bank for he sample period. I leads o a emporary fall in GDP ha peaks ypically around four quarers afer he shock and o a gradual decrease in he price level We also show ha for mos counries, M1 iniially decreases and ha he response of invesmen is larger han he one of consumpion. Second, he resuls of our esimaions a he counry level are compared o he resuls obained in a VAR esimaed wih euro area aggregaes by Peersman and Smes (2001). We show ha he effecs of moneary policy on prices and on oupu esimaed for each counry are usually qualiaively similar o he ones obained when he model is esimaed on he aggregae euro area economy. Third, we show how he moneary policy shocks defined a he euro area level relae o he paricular episodes of he domesic moneary policy shocks of he differen counries. 2 See also Clemens, Konomelis and Levy (2001) for a complemenary poin of view on he role of he EMS in he ransmission mechanism. Their sudy focuses on he effecs of he German moneary policy shock on all he oher counries of he euro area. 8 ECB Working Paper No 92 December 2001

10 The paper is srucured as follows. Secion 2 provides an overview of he lieraure. Secion 3 describes he hree idenificaion schemes chosen and secion 4 he resuls of heir implemenaion. Finally, secion 5 concludes. 2. The empirical analysis of he effecs of moneary policy in he individual counries of he Euro area In his secion, we review he empirical lieraure on he macroeconomic effecs of moneary policy shocks in he euro area counries. We briefly describe he modelling framework chosen in he differen sudies. I appears ha mos of he cross-counry comparison sudies based on VARs do no model he EMS conex of moneary policy. Firs, he BIS (1995) repors simulaion exercises from macroeconomeric models, eiher counry models developed by he cenral banks or he muli-counry model of he Federal Reserve Board. The repor presens he effecs of a emporary 100 basis poins increase in he policy rae for eigh quarers, afer which he policy rae would reurn o baseline. Second, Brion and Whiley (1997) simulae a varian of he Mundell-Flemming model o analyse he ransmission mechanism in he UK, France and Germany. Third, Dornbusch, Favero and Giavazzi (1998) use a small simulaneous model for oupu in differen counries. The impac of moneary policy on oupu is simulaed wih fixed exchange raes wihin Europe. In he same spiri, Peersman and Smes (1999) simulae he effecs of a German moneary policy shock while allowing for he ineracion effecs among he counries. Fourh, Sala (2001) implemens a dynamic facor model for six counries of he euro area. He esimaes hree common componens o he six counries and idenifies one of hem as a common moneary policy shock. He hen repors he effecs of his shock on individual counries. Fifh, a vas majoriy of sudies used VAR models. These sudies differ in erms of he variables included in he VAR, he number of euro area counries covered and he idenificaion sraegy chosen. Gerlach and Smes (1995) used a VAR approach wih a combinaion of shor run and long run resricions for he G7-counries. Only hree variables are included in heir VAR: oupu, he price level and he ineres rae. Ramaswamy and Sloek (1997) esimae a simple hree-variable (oupu, he price level and he ineres rae) VAR for 12 EU-counries 3. Dedola and Lippi (2000) esimae he effecs of a one-percenage poin increase in he shor-erm ineres rae using an VAR conaining indusrial producion, commodiy prices, he price level, he money sock and 3 See also he recen paper by Mihov (2001) for a comparable model esimaed for 5 counries of he euro area. ECB Working Paper No 92 December

11 he shor-erm ineres rae. Barran, Couder and Mojon (1997) esimae several VARs including combinaions of GDP, prices, DM exchange rae, world price index, money, credi and he long erm ineres rae for nine European counries. Ehrmann (2000) esimaes a VAR of indusrial producion, prices, an exchange rae, an ineres rae and, when appropriae, anoher variable ha is relevan for he specific moneary policy conex of he counry. One imporan shorcoming of he VAR papers above is ha he EMS conex of moneary policy is eiher no modelled or limied o he inclusion of he DM exchange rae in he model (Barran e al. or Ehrmann). Mos of he VAR sudies ha have expliciely modelled he EMS conex focus on one or on a small number of counries. 4 Smes (1997) focuses on Germany, Ialy and France, Kim (1998) invesigaes he cases of Spain and France, De Arcangelis and Di Giorgio (1998) and Gaioi (1999) concenrae on Ialy, Shioji (1997) on Spain while Levy and Halikias (1997) and Mojon (1999) focus on France. These sudies 5 show ha over he las wo decades, he moneary policy ineres rae of Spain, France and Ialy was largely, albei no enirely driven by he German ineres rae. While he room for a domesic moneary policy exised for hese hree counries, no accouning for he German leading role in he EMS leads o mis-idenificaion of moneary policy shocks. For insance, periods of rising German ineres raes could be wrongly inerpreed as periods of domesic moneary policy ighening. In his paper, we implemen a moneary policy idenificaion scheme ha accouns for he EMS conex for all he counries of he euro area, for which quarerly naional accoun daa are available. This is he purpose of he nex secion. 3. Idenificaion This secion discusses in urn he hree idenificaion schemes we propose o describe moneary policy in each of he hree groups of counries menioned in he inroducion. I hen recalls he specificaion of he benchmark Peersman and Smes (2001) model of he euro area. 4 Two relaed recen sudies use VAR models o evaluae he effec of he single moneary policy in each of he counries. Clemens e al. (2001) repor he effec of moneary policy shocks when he reacion funcion is consrained o be similar across counries and he inra-eu exchange raes are fixed. Peersman (2001) esimaes he effecs of area-wide moneary policy shocks on he individual counries. 5 See Mojon (1999) for a survey of hese sudies. 10 ECB Working Paper No 92 December 2001

12 3.1 VAR models for he individual counries in he euro area In his secion we presen VAR models for all euro area counries excep Luxembourg and Porugal. We discuss he feaures of he model ha are necessary o fi he individual counry experiences. In doing so, our objecive is o limi he differences in he models ha apply o he differen counries, so as o preserve comparabiliy in he oucome of he esimaes. We disinguish hree groups. The firs group conains Germany, which played a special role as he de faco anchor wihin he EMS sysem. The second group of counries consiss of Ausria, Belgium and he Neherlands. These counries have mainained heir fixed exchange rae pariy agains he DM during mos of he sample period. All he oher counries (Finland, France, Greece, Ireland, Ialy and Spain) can be described by a similar VAR model. Mos of hese counries have paricipaed in fixed, bu adjusable exchange rae regimes during large pars of he sample period, bu neverheless experienced quie large pariy changes. Wih he excepion of France and Ireland, each of hese counries also wen hrough a floaing exchange rae period during he sample period Germany For Germany, we esimae he following VAR model, comparable wih he benchmark model of Peersman and Smes (2001). [1] Y = A( L) Y 1 + B( L) X + µ The variables included in he model can be divided ino wo groups. 6 The firs group of US variables, X, conains a world commodiy price index ( cp ), US real GDP ( y ), and he US shor-erm nominal ineres rae ( s US ). These variables are included o conrol for changes in world demand and inflaion. Moreover, he inclusion of hese variables helps o solve he so-called price puzzle (i.e. he empirical finding in he VAR lieraure ha prices rise following an ineres rae ighening). 7 In all of he resuls repored below, we assume ha his group of variables is exogenous o he res of he VAR-model. In oher words, hese variables influence he oher variables of he model, Y, bu here is no feedback from he oher variables o hese variables. Furher, we also allow for a conemporaneous impac of he exogenous variables on he endogenous variables. [2] X = [ cp y s ] ' US US 6 Each of he VAR models conains also a consan and a linear rend. 7 See for example he resuls of Sims (1992). ECB Working Paper No 92 December

13 The endogenous variables of he benchmark model, Y, consis of real GDP ( y ), consumer prices ( p ), he domesic shor-erm nominal ineres rae ( s ) and he real effecive exchange rae ( x ): [3] Y ' = [ y p s x ] The main difference of his model wih he sandard VAR model used o idenify moneary shocks eiher for he US, bu also for Germany, is ha we do no include money in he model. This omission is mainly moivaed by our aim o esimae models ha would be as similar as possible across counries. Because mos counries now in EMU had a DM exchange rae arge during he period preceding he inroducion of he single currency, moneary aggregaes have had a secondary role in he moneary policy sraegy of hese counries. This is why we do no include a money aggregae in our benchmark model, neiher for hese counries, nor for Germany, nor for he Peersman and Smes (2001) euro area resuls repored in his paper 8. Sensiiviy analyses indicaed ha he inclusion of money in he model did no affec he impac of he German ineres rae shock on oupu and prices. In addiion, we also show ha he idenificaion of he benchmark model implies ha a conracionary moneary policy shock is followed by a fall in money for mos of he counries. In oher words, our moneary shocks idenified in a model wihou money do no produce a liquidiy puzzle. Turning o he idenificaion of a moneary policy shock, we allow for a conemporaneous ineracion beween he German ineres rae and he real effecive exchange rae. Assuming ha here is no conemporaneous reacion of he cenral bank o an exchange rae shock may be appropriae for relaively closed economies such as he euro area and he US, bu is less jusifiable for an open economy such as Germany. For example, boh Bernanke and Mihov (1997) and Clarida and Gerler (1997) have found a significan conemporaneous response of German ineres raes o changes in he exchange rae. Similarly, Smes and Wouers (1999) show ha allowing for such a response helps o avoid a price puzzle. Following Smes and Wouers (1999), we solve he simulaneiy problem by esimaing he reacion coefficien on he exchange rae using he spread beween he French and he German long-erm ineres rae and US dollar/yen exchange rae as insrumens. 9 8 See also in Peersman and Smes (2001) for a comparison of he response of GDP and prices using alernaive idenificaion sraegies. 9 See Smes and Wouers (1999) for an explanaion on he implemenaion of his wo-sep mehodology. 12 ECB Working Paper No 92 December 2001

14 3.1.2 Ausria, Belgium and he Neherlands During mos of he sample period, Ausria, he Neherlands and, o a lesser exen, Belgium, have mainained a fixed cenral exchange rae pariy vis-à-vis he DM. 10 This implies ha in hese counries he scope for an independen moneary policy was exremely limied and ha i is unlikely ha we are able o ge precise esimaes of he effecs of domesic moneary policy shocks. Insead, mos of he policy shocks are likely o be driven by policy innovaions in he German ineres rae. Moreover, hese counries are relaively small neighbours of Germany and srongly influenced by economic condiions in Germany. In his case, we herefore modify he benchmark model by including German oupu, prices, real effecive exchange rae and shor-erm ineres rae in he lis of endogenous variables and replacing he effecive exchange rae wih he bilaeral rae versus he DM. In addiion, we assume ha here is no feedback from he smaller counry o Germany. The moneary policy shock is idenified as he shock o he German ineres rae. We can represen his as follows: [4] Y Y DE j A(L) = C(L) DE 0 Y DE + + j j - 1 B( L) ε X σ D( L) Y -1 E( L) ε wih [5] [6] Y = DE j y j DE j p DE j s DE j x DE Y = y p x s, j = AT, NL or BE ] ] ] ] This implies ha we esimae he same moneary policy shocks as in he German case. The response of he German variables o his moneary policy shock is also unchanged. The block-recursive srucure of his wo counries model closely resembles he one applied by Cushman and Zha (1997) o model he influence of he US economy on Canada. Cushman and Zha consider he effecs of an independen moneary policy because of he flexible exchange rae regime ha characerises Canada. In conras, because here was hardly any variaion of he DM exchange rae during mos of he sample period, we focus on he effecs of he German moneary policy shock in Auria, Belgium and he Neherlands. 10 The cenral pariy wihin he EMS changed only in he early 1980s for he Neherlands, and in 1982, 1983, 1986 and 1987 for Belgium. The Ausrian exchange rae did flucuae in a very narrow band for he whole sample period. ECB Working Paper No 92 December

15 3.1.3 The oher counries: Finland, France, Greece, Ireland, Ialy and Spain For all he oher counries, we modify he German model in wo respecs. Firs, we include he German shor-erm ineres rae in he block of endogenous variables. Second, we replace he real effecive exchange rae wih he nominal bilaeral DM exchange rae given is prominence in he EMS. 11 This leads o he following se of endogenous variables: DE [7] Y ' = y p s x s ] ] The German ineres rae is included in addiion o he bilaeral DM exchange rae in order o describe he role of Germany as an anchor of he ERM. The domesic ineres rae may respond o he German one wihou any impac on he exchange rae, as i would be he case in a fixed exchange rae regime. This can urn ou o be imporan when esimaing a model a a quarerly frequency, while exchange raes and ineres raes inerac coninuously. Basically, we assume ha domesic moneary policy shocks are defined as deviaions of he domesic ineres rae from he German ineres rae 12. In oher words, we describe he effecs of moneary shocks aken in he process of nominal convergence, which ook place before EMU. We obain ha he esimaed domesic policy shocks lead o a decrease in prices. I is ineresing o noe ha neglecing he German ineres rae in he model resuls in a price puzzle for many of hese counries. The reason is ha if one does no conrol for increases in he domesic ineres rae ha are a response o increases in he German rae, such changes may be associaed wih a depreciaion of he exchange rae. This in urn pus upward pressure on prices. As before, he domesic policy shock is idenified using a sandard recursive idenificaion scheme, which corresponds o he ordering of he variables in [7]. This means ha here is a conemporaneous impac of all he endogenous variables on he moneary policy variable. On he oher hand, here is no immediae impac of a moneary policy shock on he oher variables. The ordering of he bilaeral DM exchange rae before he domesic ineres rae is plausible since here was a fixed exchange rae regime during mos of he esimaion period. However, he resuls are robus wih respec o a reverse ordering of he domesic ineres rae and he bilaeral DM exchange rae. Also allowing for a wo-way ineracion beween he exchange rae and he domesic ineres rae did no significanly affec he resuls. 11 Using an effecive exchange rae does no change he esimaed effec of moneary policy shocks on prices and on oupu in any significan way. We prefer o include he bilaeral DM exchange rae in order o model he specific siuaion of he EMS. 12 The macro-economeric model of he Banque de France also defines he reacion funcion in erms of deviaions from he German ineres rae. 14 ECB Working Paper No 92 December 2001

16 I is imporan o bear in mind ha in he case of Greece, moneary policy was implemened hrough quaniy raioning of he banks so ha he hree monh ineres rae was lef unchanged for long periods during he 1980 s. The resuls for his counry should herefore be aken wih exra cauion. 3.2 The specificaion of he benchmark VAR-model for he euro area The VAR-model from Peersman and Smes (2001) ha we use o analyse he effecs of a moneary policy shock in he euro area has he same represenaion as he one presened for Germany above. The only difference is ha he policy shock for he euro area is idenified hrough a sandard Choleski-decomposiion wih he variables ordered as menioned above. 13 The underlying assumpion is ha policy shocks have no conemporaneous impac on oupu and prices, bu may affec he exchange rae immediaely. However, he policy ineres rae does no respond o conemporaneous changes in he effecive exchange rae. The laer assumpion is appropriae for a large, relaively closed, economy such as he euro area as a whole Resuls 4.1 Esimaion Unless oherwise menioned, each of he VAR-models is esimaed in levels over he period , which corresponds approximaely o he sar of he EMS. 15 In he case of Germany, for which we consider ha he moneary policy decision process is independen of he EMS, we esimaed he model for he longes period of daa availabiliy, i.e In his paper we do no perform an explici analysis of he long run behaviour of he economy. By doing he analysis in levels we allow for implici coinegraing relaionships in he daa (Chap. 18 in Hamilon, 1994). A more explici analysis of he long-run behaviour of he various variables is limied by he relaively shor sample available. 16 The daa are seasonally adjused logs, excep he ineres raes, which are in levels. We use he hree-monh ineres rae 17 as he moneary policy rae as his is he only shor-erm ineres rae ha is available for all counries over he whole 13 As in Sims (1980) and Chrisiano e al (1998). 14 Eichenbaum and Evans (1995) make he same assumpion for he US. One can argue ha he euro area as a whole is more like he US in erms of openness han like any of is individual members. 15 Also, we ook 1980 as a saring dae because some of he daa series used are only available from ha year. 16 See Ehrmann (2000) for an explici coinegraion analysis of VAR models for he counries of he euro area. 17 Excep in Greece where we use he hree-monh average of he overnigh rae. ECB Working Paper No 92 December

17 sample period. Sandard likelihood raio ess are used o deermine he lag-order of he VARs, which usually urns ou o be of order wo or hree. Chow break ess did no rejec he overall sabiliy of he various VARs a he 5% confidence level. However, in some counries insabiliy was deeced in some of he equaions of he VAR. This was in paricular he case for Ialy where overall sabiliy is rejeced a he 10% confidence level. Boh he Ialian oupu and exchange rae equaion appear o be subjec o a significan break in he hird quarer of 1992 coinciding wih he oubreak of he EMS crisis and he floaing of he Ialian lira. Also in Finland here is some evidence of insabiliy in he exchange rae equaion in he early 1990s. In he case of Germany, i urns ou ha a longer sample period helps o reduce he weigh of he reunificaion episode, when he moneary policy ighening in he early 1990s coincided wih a big posiive demand shock due o direc governmen spending and ax incenives. Esimaing he model jus for he 1980 s and 1990 s renders he idenificaion of plausible effecs of German moneary policy exremely edious. The resuls shown for Ausria, Belgium and he Neherlands are compued by simulaing he effecs of he idenified German moneary policy shock, i.e. he impulse responses of he variables in he German VAR esimaed of , on he Ausrian, Belgian and Duch variables respecively as in [4]. The laer equaions are esimaed for he period of he EMS ( ), when hese counries where in a de faco fixed exchange rae regime wih Germany. Finally, boh in Ireland and Greece we include a dummy variable for unusual ineres rae increases of more ha 10 %. The Irish spike ook place in he fourh quarer of 1992 and he Greek one in he firs quarer of Discussion of he resuls The resuls of he idenificaion schemes described in secion 3 for each of he individual counries and he euro area are shown in Graph 1. These graphs summarise for each of he counries he effecs of a one-sandard deviaion moneary policy shock on domesic real GDP, domesic consumer prices, he exchange rae (effecive real exchange rae in he case of he euro area and Germany, he bilaeral DM exchange rae in he case of he oher counries), and he domesic shor-erm ineres rae. We repor he OLS esimae based impulse response funcion ogeher wih 90 percen confidence bands 18. Graph 1 shows 18 These confidence bands repored are consruced as he impulse response plus or minus 1.65 sandard deviaions of 200 boosrapped replicaions of he impulse response. These confidence bands are very similar, hough somewha narrower han he ones obained by Mone Carlo simulaions as proposed by Sims and Zha (1999). 16 ECB Working Paper No 92 December 2001

18 ha in each of he counries, a moneary policy ighening evenually leads o a fall in oupu and prices 19. I is remarkable ha hese fairly simple idenificaion schemes allowed us o compile well-behaved responses of GDP and of prices o a domesic moneary policy shock. Moreover, hese responses are consisen wih he area-wide resuls. However, some addiional feaures are worh noing. Firs, he effecs on oupu and on prices end o be larger when measured a he counry level han wha hey appear o be a he euro area level. This may reflec ha he effecs on oupu and prices riggered by moneary policy in he period before EMU worked hrough he quaniy or price adjusmen of rade wih oher members of EMU. These effecs are expeced o disappear in he new moneary policy regime. Second, while he overall paern for oupu and prices is quie similar across counries, he effecs on he exchange rae are less consisen across counries and generally less significan. For Belgium and he Neherlands he lack of response of he exchange rae o he German moneary policy shock reflecs he full credibiliy of he EMS for hose counries. In France, Ireland, Greece, Germany and Ausria he moneary policy shock riggers an appreciaion of he exchange rae. Finally, in Ialy and Spain, we find he so-called exchange rae puzzle, i.e. a ighening of he moneary policy sance leads o a (only slighly significan) depreciaion of he exchange rae (Grilli and Roubini, 1995). Given he shifs in exchange rae regime in hese counries, he finding of erraic exchange rae responses should no be oo much of a surprise. 20 More ineresing is ha he differen paerns in he exchange rae responses are no refleced in he responses of prices and oupu. I is likely ha he exchange rae response for one counry has ofen coincided wih a similar change of he exchange rae of oher European counries in a similar direcion so ha he effecive exchange rae of he counry was less affeced. Moreover, he las wo decades were characerised by a negaive correlaion beween he DM exchange rae and he dollar exchange rae of he European currencies. Third, we show ha in he counries of he hird group, domesic moneary policy shocks ha were orhogonal o he German ineres rae have had he ypical effecs of moneary policy boh on oupu and on prices. These effecs reflec he acions aken by cenral banks in hese counries o simulae nominal convergence wih Germany. They 19 The excepion is he impac of moneary policy on oupu in Ireland. This can be due o he specific siuaion of ha counry wih respec o he UK. Moreover, we do no find his erraic behaviour if we use indusrial producion insead of GDP and we observe ha boh consumpion and invesmen decrease afer a moneary ighening. Bredin and O Reilly (2000) show ha aking ino accoun he influence of he UK on he Irish business cycle helps solving his oupu puzzle. 20 While some sudies (Smes, 1997; Gaioi, 1999; Hernando, 2000) have shown ha oher idenificaion schemes can alleviae his exchange rae puzzle, we prefer o sick o our "simple" model ha performs well in erms of GDP and prices, for he sake of comparabiliy wih he oher counries. ECB Working Paper No 92 December

19 may also pick up he effecs of he EMS crisis on he 1993 recession and dis-inflaion ha followed. Fourh, he comparison of our resuls o he esimaes of a represenaive se of previous sudies indicaes ha aken globally, he lieraure does no poin o any counry as experiencing eiher weaker or larger effecs of moneary policy han he loose average of he counries. This is consisen wih our (and Kieler and Saarenheimo s, 1998) finding of qualiaively similar resuls across counries and high uncerainy on he size of he effecs. There is no black sheep. Mos sudies repor ha, overall, a majoriy of counries experience a fall in oupu (eiher GDP or indusrial producion) afer a conracionary moneary policy shock. An overview of he maximum impac on oupu is provided in Table 1. The sudies presen, however, differen rankings of he poency of moneary policy across counries. Some counries are documened o be more sensiive o a moneary policy shock in one sudy, bu less in anoher. For example, Barran e al. (1997) find he larges impac in Germany and he weakes impac in Ialy, while Peersman and Smes (1999) find he opposie. The problem wih mos of hese cross-counry comparisons is ha he size of he esimaed moneary policy shock differs across counries. In his paper, for example, a one sandard deviaion moneary policy shock is a rise in he shor-erm ineres rae of 48 basis poins for Germany, while i is 59 basis poins for Ialy, and 84 basis poins for Spain. Moreover, his also implies ha each individual counry has is own moneary policy reacion funcion. Thus, even if he same iniial disurbance is analysed (e.g. as done by Dedola and Lippi, 2000), he moneary policy responses would no be harmonised. 21 This is illusraed by he wo varians of Gerlach and Smes (1995). In he firs case (a one sandard deviaion moneary policy shock), he response of oupu looks similar across Germany, France and Ialy, while in he second case (a one percenage poin, eigh-quarers susained increase of he ineres rae), German GDP moves by almos wice as much as ha of France and Ialy. To jusify his laer ype of analysis, however, we have o assume ha he esimaed parameers of he model are invarian o he specificaion of he policy rule, and we are confroned wih he Lucas criique. Again, one should also noe ha, for any of hese sudies, he confidence bands around hese responses are such ha he differences across counries are likely o be no significan. In addiion, Kieler and Saarenheimo (1998) show ha screening he full space of observaionally equivalen idenificaions of moneary policy shocks for Germany, 21 See also Guiso e al. (2000) for a discussion of his problem. 18 ECB Working Paper No 92 December 2001

20 France and he Unied Kingdom, one can build very similar impulse responses of GDP and prices o moneary policy shocks. Given hese criicisms on he limied power of quaniaive comparisons of impulse responses o moneary policy shocks across counries, one can conclude ha he evidence of cross-counry differences in he ransmission brough abou by any single paper is no robus across sudies. This is parially confirmed by a recen sudy ha has aemped o es formally he cross-counry differences in he ransmission mechanism. Ciccarelli and Rebucci (2001) esimae a dynamic heerogeneous panel on a sample ha pools macroeconomic daa for Germany, France, Ialy and Spain. They show ha he cumulaive impac of moneary policy on economic aciviy is no significanly differen across counries. Table 1: Effec of moneary policy on oupu DE FR IT ES FI BE NL AT IR PT Mojon and Peersman BIS: Naional cenral banks b BIS: FRB muli-counry b Gerlach and Smes Gerlach and Smes 2 b Barran e al Brion and Whiley Ramaswamy and Sloek Ehrmann a Dedola and Lippi a c Dornbusch e al Mihov b Peersman and Smes Clemens e al Peersman Noe: maximum impac; daa no comparable across sudies. DE = Germany, FR = France, IT = Ialy, ES = Spain, FI = Finland, BE = Belgium, NL = he Neherlands, AT = Ausria, IR = Ireland, PT = Porugal. a : effec of moneary policy on indusrial producion. b : effec of a 100 basis poins, eigh quarers susained increase of he ineres rae. c : effec of a 1 percenage poin increase in he shor-erm rae. ECB Working Paper No 92 December

21 4.3 Pre-EMU euro area moneary policy and naional moneary policy hisories One ineresing oucome of a moneary policy idenificaion exercise is ha i allows a rerospecive view on when he moneary policy shock conribued o he evoluion of he variables of he model. More imporanly, we wan o be able o relae he evoluion of he moneary policy sance of he euro area in he pre-emu period of he sample o he moneary policy hisories of he individual counries. Graph 2 plos he esimaed policy shocks and heir hisorical conribuion o he domesic ineres rae for each of he counries. In he lef panel of each graph, he full line is he firs difference of he ineres rae, and he bars are he moneary policy shocks. In he righ panel, he bars are he conribuion of he moneary policy shocks o he (domesic) shor erm ineres rae, whereas he full line is he conribuion of he accumulaion of all shocks o he shor erm ineres rae. The conribuion of hose shocks o ineres rae developmens differs across counries. For he euro area as a whole, i is clear ha he years 1982, 1987, 1990 and are idenified as periods of relaively igh moneary policy, whereas in 1984 and 1991 policy is esimaed o be relaively loose. The mixed experience of he individual counries during he EMS crisis is sriking. Moneary policy ighening in France, in Ialy, and also in oher counries excep in Germany add up ino a sharp ighening a he euro area level. On he conrary, German moneary policy was no oo resricive during he period This means ha he high ineres rae in Germany afer he re-unificaion can be explained by he endogenous response of he Bundesbank o a booming economy. However, he sance of moneary policy for he oher counries, which kep a fixed exchange rae vis-à-vis Germany, was oo resricive in view of he recession hey were in. Anoher remarkable feaure is ha he proporion of he moneary policy shocks o he accumulaed shocks of all he variables is very low in Germany compared o he oher main counries of he Eurosysem. The reacion of moneary policy o price and oupu shocks was much more imporan in Germany. On he oher hand, his proporion is also very low for Belgium, he Neherlands and Ausria. For hese counries, he reacion o exchange rae shocks is very imporan. The cross-counry comparison of he shocks is pu ino sharper focus in Table 2, which repors he correlaions of he idenified moneary policy shocks, and he firs difference of he ineres rae, across counries and wih he euro area. The moneary policy of Ialy, France and Spain conribues largely o he euro area aggregae. The correlaion of he domesic moneary policy shock wih he area-wide shock is respecively 0.51, 9, and 6. On he conrary, he German moneary policy shock, which is idenified on a sample 20 ECB Working Paper No 92 December 2001

22 period which also includes he 1970s, is much less correlaed o he euro area shock. I is worh noing ha looking a he correlaion of he firs difference of he policy rae shows ha boh Ialy, France and Spain, bu also counries of he core EMS see heir ineres rae highly correlaed o he euro area one. Table 2: Correlaions of moneary policy shocks and ineres raes across counries Domesic moneary policy shocks Ea De I Fr Es Fi Gr Ir Euro area Germany Ialy France Spain Finland Greece Ireland 1.00 Policy ineres raes Ea De I Fr Es Fi Nl Be A Gr Ir Euro area Germany Ialy France Spain Finland Neherlands Belgium Ausria Greece Ireland Furher evidence on he effecs of moneary policy shocks In his secion, we discuss he influence of a moneary policy shock on oher macroeconomic variables ha are no included in he basic model. We have done his by esimaing he following equaions: ECB Working Paper No 92 December

23 [8] Y A( L) = Z C( L) 0 Y D( L) Z -1-1 B(L) + X E(L) Y µ + Y cµ + µ Z The sysem of equaions [8] is very similar o [1]. X is sill he block of exogenous variables and Y he endogenous block. Le Z be anoher macro-economic variable (for example invesmen). Again, we suppose ha neiher conemporaneous nor lagged values of he endogenous variables have an influence on he exogenous block. This is also he case for our variable of ineres. However, we suppose he same for our macro-economic variable, Z, wih respec o our endogenous block of variables, Y, i.e. here is no impac of he variable under invesigaion on he oher variables in he sysem. This assumpion ensures ha he shocks are invarian o he choice of he macroeconomic variable added o he original model. We show in Graph 3 for a few examples ha his modelling assumpion obains he same impulse response as if he variable was included in he VAR The impac on GDP componens Graphs 4a, 4b and 4c presen for each individual counry he response of invesmen, consumpion and expors respecively o a moneary policy shock. Wih he noable excepion of Greece and Ireland, he response of invesmen is a leas wice as large as he response of consumpion. Invesmen response is, however, insignifican in boh counries, and he response of consumpion is insignifican for Greece. These resuls are broadly consisen wih he ones of Barran e al. (1997) for models esimaed on a sample period spanning from 1975 o They reflec he fac ha consumpion is smooher han invesmen over he business cycle, bu also possibly he income effecs of moneary policy, whereby ne-debor invesors revise heir expendiure plans more han necrediors consumers. These resuls are also consisen wih he area-wide resuls obained in Peersman and Smes (2001). For mos counries, we find a srong impac of a moneary policy shock on expors. For insance, Ausrian, Belgian and Duch expors are affeced by he appreciaion of he real effecive German exchange rae, while heir nominal exchange rae vis-à-vis Germany hardly moves (Belgium and he Neherlands) or slighly depreciaes (Ausria). In hese hree counries, as well as in France, he dampening effec of conracionary moneary policy shocks on expors is larger han he one observed on GDP. In Finland, expors decrease as well and his fall is slower and smaller han he one observed for GDP. Finally, for Ialy and Spain, we do no find a negaive impac of conracionary moneary policy shocks on expors. The impac is even significanly posiive in he laer counry. This finding is no necessarily surprising given he exchange rae puzzle in hese 22 ECB Working Paper No 92 December 2001

24 counries. However, he posiive response of expors is no large enough o make he response of Spanish and Ialian GDP deviae from he ypical responses of GDP observed in mos counries and a he Euro area level No liquidiy puzzle We have no considered money in our idenificaion of moneary policy shocks because i had a much less imporan role han he exchange rae for all he counries, which were argeing a fixed exchange rae wih he DM wihin he EMS. We neverheless check ha our idenificaion is no characerised by a liquidiy puzzle. This puzzle sresses he risk of confusing money demand shocks and money supply shocks. A posiive shock o money, which would be accompanied by a rise in he ineres rae, is more likely o correspond o a money demand shock han o a moneary policy shock. We esimaed he response of M1 o he moneary policy shock idenified in secion 3. The cenral bank can beer conrol his narrow aggregae han M3 because he yield on ime deposis and on money muual funds, which are he main componens of M3-M1 is correlaed o he shor-erm ineres rae. Hence a conracionary moneary policy shock riggers an increase in he yield of ime deposis and money muual funds so ha M3 decreases less han M1 22. As expeced, he moneary policy shock riggers an immediae fall of M1 in all counries excep France, Greece and Ireland (Graph 5). This furher confirms he correcness of our idenificaion scheme. 5. Concluding remarks In his paper, we have used VAR models o analyse he effecs of a moneary policy shock in he individual counries of he euro area in he pre-emu period. Firs, we show ha hree, relaively simple idenificaion schemes, depending on he moneary policy decision process in he ERM, obain well behaved and qualiaively consisen effecs of he moneary policy shocks in all he individual counries of he euro area. We confirm ha, for hese counries, he qualiaive effecs of moneary policy are quie similar o he ones described in a large lieraure for he US and by Peersman and Smes (2001) for he euro area aggregaes. A conracionary moneary policy shock leads o a emporary fall in GDP ha peaks ypically around four quarers afer he shock and a gradual decrease in he price level. The invesmen response and he expor response are generally larger han he one of GDP while he response of consumpion is smaller. We show also ha he shocks 22 See he appendix where he response of M3-M1 o moneary policy shocks is repored. ECB Working Paper No 92 December