RESULTS USING STRUCTURAL CROECONOMIC MODELS

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 93 THE MONETAR ARY TRANSMISSION MECHANISM AT THE EURO-AREA LEVEL: ISSUES AND RESULTS USING STRUCTURAL MACR CROECONOMIC MODELS EUROSYSTEM MONETARY TRANSMISSION NETWORK BY PETER McAD ADAM AM AND JULIAN MORGAN December 2

2 EUROPEAN CENTRAL BANK WORKING PAPER SERIES WORKING PAPER NO 93 THE MONETAR ARY TRANSMISSION MECHANISM AT THE EURO-AREA LEVEL: ISSUES AND RESULTS USING STRUCTURAL MACR CROECONOMIC MODELS EUROSYSTEM MONETARY TRANSMISSION NETWORK BY PETER McAD ADAM AM AND JULIAN MORGAN * December 2 * Direcorae General Research, European Cenral Bank We hank, wihou implicaing, Gabriel Fagan, members of he Moneary Transmission Nework and, especially, Carlos Robalo-Marques and Ricardo Mesre for commens The opinions expressed are no necessarily hose of he ECB

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 8 and 9 December 2 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, 2 Address Kaisersrasse 29 D-63 Frankfur am Main Germany Posal address Posfach D-666 Frankfur am Main Germany Telephone Inerne hp://wwwecbin Fax Telex 4 44 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 56-8

4 Conens Absrac 4 Non-echnical summary 5 Inroducion 6 II The Moneary Transmission Mechanism 7 III Decomposing he Channels of Transmission 4 IV Design of Moneary Policy Transmission Simulaion Experimens 7 V Simulaion Experimens Using he AWM and he NiGEM Models 2 VI Conclusions 38 References 4 Appendix : Maquee Noaion 42 Appendix 2: Model Descripions 43 Appendix 3: Comparison wih he WGEM Experimen 45 European Cenral Bank Working Paper Series 49 ECB Working Paper No 93 December 2 3

5 Absrac: This paper addresses some of he issues faced by macroeconomic model builders in analysing he moneary ransmission mechanism These include he sensiiviy of he policy simulaion resuls o changes in he moneary and fiscal policy rule and he inroducion of forward-looking behaviour in he model To illusrae he imporance of hese issues he paper repors he resuls of varian moneary policy simulaions a he euro-area level using he AWM and NiGEM models JEL classificaion: C5, C52, E5 Keywords: Moneary Transmission Mechanism, Euro Area, Macro Models 4 ECB Working Paper No 93 December 2

6 Non echnical summary This paper looks a he moneary ransmission mechanism a he euro-area level using macroeconomic models and examines some of he issues raised by such an exercise The aim is o consider how imporan various aspecs of model and simulaion design are in deermining he resuls from moneary policy simulaions In his paper we examine he effecs of using moneary and fiscal policy rules and he imporance of forward-looking behaviour To illusrae he significance of hese issues he paper repors he resuls of a number of varian simulaions using he ECB s Area Wide Model and he Naional Insiue Global Economic Model The paper akes as is saring poin, he las major sudy of comparaive properies of cenral bank models, which was carried ou by he Bank of Inernaional Selemens (BIS) in 995 The BIS sudy examined cross-counry differences in he ransmission mechanism of moneary policy involving a hundred basis-poin increase in he shor-erm policy ineres rae for 2 years In his paper we examine some of he issues faced by macro model builders in examining he ransmission mechanism by underaking a number of varian simulaions similar o ha underaken in he BIS exercise A fundamenal difference, however, is ha we underake hese simulaions a he euro-area level Our resuls highligh he imporance of various aspecs of model and simulaion design in affecing he resuls from such moneary policy exercises We find ha he NiGEM and AWM models yield broadly similar resuls for he effec of moneary policy on oupu for he firs 2-3 years when he forward-looking elemens of he model are no used This also holds rue when moneary and fiscal policy rules are inroduced ino he simulaions Allowing forward-looking behaviour in he models ends o increase he iniial impac of he moneary policy exercise bu also hasens he reurn o baseline values, albei wih imporan differences beween he wo models in erms of he magniude and iming of he effecs In boh models, he user cos of capial ends o be he dominan channel of ransmission in erms of is impac on GDP when a common moneary policy is implemened across all channels ECB Working Paper No 93 December 2 5

7 I Inroducion This paper looks a he moneary ransmission mechanism a he euro-area level using macroeconomic models and examines some of he issues raised by such an exercise The aim is o consider how imporan various aspecs of model and simulaion design are in deermining he resuls from moneary policy simulaions for example, he effecs of using moneary and fiscal policy rules and forward-looking behaviour To illusrae he imporance of hese issues he paper repors he resuls of a number of varian simulaions using he ECB s Area Wide Model (AWM) as described in Fagan e al (2) - and he Naional Insiue Global Economic Model (NiGEM) 2 The paper akes as is saring poin, he las major sudy of comparaive properies of cenral bank models, which was carried ou by BIS (995) The BIS sudy examined cross-counry differences in he ransmission mechanism of moneary policy and considered he exen o which hese could be due o differences in financial srucure Simulaion experimens were underaken on he models involving a hundred basis-poin increase in he shor-erm policy ineres rae for 2 years 3 In his paper we examine some of he issues faced by macro model builders in examining he ransmission mechanism by underaking a number of varian simulaions similar o ha underaken in he BIS (995) exercise A fundamenal difference, however, is ha we underake hese simulaions a he euro-area level Alhough here have been many sudies of he moneary ransmission exercise a naional levels (see for insance Erhmann (2)) here has be no corresponding exercise carried ou a he euro aggregae level The paper is organised as follows In secion II we consider he naure of he moneary ransmission mechanism in macro models in general and specifically in he AWM and NiGEM models In secion III we address he issue of how o decompose he resuls ino individual channels of ransmission In secion IV, we consider he issues around he design of simulaion experimens whils in secion V here is a comparison of four varian simulaions beginning wih a compleely backward-looking simulaion wih no policy rules and hen gradually inroducing policy rules and forward-looking behaviour The resuls of he final simulaion are also decomposed ino he various channels of ransmission Secion VI concludes 2 The April 2 release of he NiGEM model was used 3 The resuls were summarised by Smes (995) 6 ECB Working Paper No 93 December 2

8 II The Moneary Transmission Mechanism This secion discusses which channels of moneary policy ransmission can be idenified in srucural macroeconomic models In general, in mos large-scale macroeconomic models he ransmission mechanism of moneary policy akes place primarily hrough he ineres rae The cenral bank chooses he shor-erm policy ineres rae, which has a pass hrough o oher marke yields, asse prices and he exchange rae Oher han hese financial linkages, he main effecs ypically emerge from oupu and prices via he impac on domesic spending (privae invesmen and consumpion) and on he exernal secor hrough expor and impor volumes These effecs are highlighed o differing degrees in boh models and accordingly, we will firs ry o give an overview of he general ransmission channels and hen discuss hose specific o our chosen models 4 When considering ypical moneary ransmissions (and comparing hem across wo separae models) i is useful o have somehing like a core model in mind To his end, we begin by considering an illusraive encompassing maquee (adaped from McAdam, 999) The idea behind presening his maquee is ha i can serve o illusrae he sandard channels in which moneary policy operaes in large scale macroeconomic models Indeed, he maquee is inended o be a broad approximaion o no only he models used in his paper bu also for oher models in he class (such as he IMF s Mulimod model or he Federal Reserve's Muli-Counry Model) Of course he maquee can only provide a saring poin as here are specific feaures of he AWM and NiGEM which differ from he maquee and have an imporan bearing on he impac of moneary policy We discuss hese specific feaures a he end of his secion The maquee iself is sraighforward o moivae: is long run is supply deermined (B8) bu he nominal ineria (B5) and lead expecaions (B5, B7, B2) cause dis-equilibrium and overshooing resuls in he shor run Roughly half of he model comprises ideniies Besides hese, he model has sandard feaures: classical opimaliy in invesmen (B3, B), Blanchard-Yaari (Blanchard and Fisher, 989) ype consumpion (B2), money demand (B7), a simple uncovered ineres pariy (B2) and erm srucure (B7), convenional rade equaions (B9, B2, B22) and inegral conrol policy rules (B5, B6) 4 An overview of boh models is given in Appendix 2 ECB Working Paper No 93 December 2 7

9 Box: An Illusraive Encompassing Maquee AGGREGATE DEMAND B Y = C + I + G + ( X IM ) : Demand Summaion B2 c = c + c2 ( wealh / c) c3rlr + c4 ( y ( TX )) : Consumpion B3 K d = ζ (( MPK PY ) / PK UCOC ) : Invesmen B4 d * l = µ ( y y ) : Labour Demand e * B5 wages = a + a2π+ + a3π + a4( y y ) a5( w p prod) : Wages B6 WEALTH = WEALTH( K, B, M, NFA ER ) : Wealh d d B7 ( m p) = β + β2 y + β3 y + β4( m p) β5rl : Money Demand AGGREGATE SUPPLY B8 Y = Y( K, L ) : Producion Funcion B9 I = K + ( δ ) K : Capial Accumulaion e B UCOC ( RL + Π )( TX ) B PK = zp + ( z) ER PIM : Invesmen Deflaor B2 PY = jp ( j) ER PIM : Value-Added Deflaor B3 P = n( W / MPL) + ( n) ER PIM : Oupu Price POLICY SECTOR = δ : User Coss of Capial B4 B = ( G TX Y ) + RL B M : Deb Accumulaion B5 TX = TX ( DEFICIT / Y ) : Tax Rae Rule T T B6 R = R( Y Y, Π Π ) : Moneary Policy Rule I B7 I + RL / = ( + R + i /) : Term srucure i= e B8 RLR = RL Π : Real Rae OVERSEAS SECTOR B9 B2 B2 B22? / Y NFA / Y = (( X PX IM PIM ) + R NFA ** er = er + + ( R R ) + qnfa : Exchange rae im = ε y ε er + pim p) : Impors CA ) / Y = : Curren Accoun ( 2 ** ** im + φ 2( er + p p) + φ 3 ( er + im + pim x = φ p) : Expors Noe: Capial leers symbolise variables in levels and lower-case for logarihms; sarred (double sarred) indicaes full capaciy (foreign) values The superscrip d indicaes demand, T indicaes arge values and e denoes expecaions Appendix liss variable names I = erm srucure lengh All parameers are posiive We omi he oher counry 8 ECB Working Paper No 93 December 2

10 The only direc effec of shor-erm ineres raes is hus on he exchange rae (B2) and he erm srucure (B7) 5 Leaving hese aside, indirec effecs dominae in his paricular model To illusrae, consider he corresponding change in he nominal long rae given he erm srucure (B7) and in urn he real rae (B8) This change (negaively) affecs capial accumulaion hrough he increased user cos of financing new invesmen (B, B3), which in urn affecs oupu (B), employmen (B4) and he rade balance (B9, B2, B22) ec This long rae will also direcly affec consumpion (B2) since he real lending rae proxies is opporuniy cos 6 There is also an indirec effec of long raes on consumpion hrough changes in ne wealh (B2, B6) The ne-wealh effecs ypically sem from such sources as changes in he sock of public deb, he capial sock, he moneary aggregae and, possibly, changes in equiy prices In he case of equiy prices, a rise in ineres raes is convenionally considered o lead o lower equiy prices and hus a de-cumulaion of wealh On he sock of governmen deb, his is revalued in line wih changes in he long rae: a rise in long raes will lead o a downward revaluaion of holdings of governmen deb and hence lower ne wealh Finally, here is a role for ineres raes o affec he moneary aggregae (B7), which feeds ino governmen deb sock wih a negaive coefficien (B4) and hereby affecs ne wealh (B6) The shorerm ineres rae eners money demand negaively (B7) Noice ha a wealh effec can come hrough ne foreign asses (B9) If he home economy is large (small) enough he relevan ineres rae here will effecively be he home (large counry) one (B9) Finally, here are a number of ineres sensiive income effecs Governmen ineres paymens ypically depend on long raes (B4) reflecing he erm srucure of governmen deb Governmen ineres paymens feed ino personal income (B6) (hereby affecing consumers expendiure) and ino he debis of ineres, profis and dividends (hereby affecing ne overseas asse accumulaion a componen of ne wealh, B9) in addiion o affecing he governmen budge balance (B4, B5) Comparing he AWM and NiGEM An imporan feaure of his paper like ha of Bryan e al (993) and BIS (995) is ha we do no specifically seek o harmonise he models Alhough many valuable insighs have been made from such exercises, we insead use each model wih he baseline 5 The complee feedback marix of moneary policy is hrough equaions R: 6, 7, 2, RL: 7,, 4, 7, 8, 9, RLR: 2, 8 Boldface indicaes normalisaion 6 As indeed in convenional macro-heory, he marginal uiliy of consumpion is given by he consumer s discoun rae, which is bounded by he real seady sae ineres rae, eg, Blanchard and Fisher (989) ECB Working Paper No 93 December 2 9

11 and equaions provided This allows us o make a pracical comparison of how he models in sandard mode compare The AWM is a single counry model of he euro area using aggregaed euro area daa a full descripion of he model is provided by Fagan e al (2) Thus, here is no counry dis-aggregaion, essenially no (modelled) inernaional linkages and no naional moneary or fiscal rules oher han aggregae ones In oher words, he AWM reas EMU members as one counry NiGEM, by conras, models each individual counry separaely and he euro area resuls ha we repor are based on a saic aggregaion of individual counry resuls Neverheless, i is possible o run he model consisen wih a moneary union in he euro area and hereby ensure common shor and long-erm nominal ineres raes, no changes in he exchange raes beween he residual currencies of he euro area and a common movemen of he euro agains hird currencies Forward-Looking Behaviour Lead relaionships are ypically always among he mos imporan ones in a macro-model since hey end o advance he effecs of shocks In boh models here is he possibiliy o allow for forward-looking behaviour in boh he exchange rae and in long-erm ineres raes In NiGEM, he sandard model also allows for forward-looking behaviour in equiy prices, and in he inflaion erms used in moneary policy rules and in he wage equaions The role of hese forward-looking erms is discussed in some deail in he remainder of his secion One general poin is ha wih he excepion of he forward looking inflaion erms in some of he NiGEM wage equaions forward-looking elemens have no been inroduced ino behavioural equaions bu ino calibraed ones Therefore here is lile difficuly in adaping hese equaions for forward-looking behaviour and essenially no need o re-esimae hem I is also worh noing ha, expecaions may also maer for oher behavioural equaions see for example Sgherri (2) who examines he imporance of forward-looking behaviour of consumers However, such forward-looking elemens have no been inroduced ino he sandard versions of hese models Turning firs o he poenial for forward-looking behaviour in he exchange rae As we know from he maquee, he link from ineres raes o he exchange rae is a common one I is also ypically a very powerful one The precise effec of ha ransmission depends on he exchange rae model If he nominal exchange rae is fixed hen of course here is no effec If he exchange rae follows, say, a porfolio balance approach hen he subsequen effec of moneary policy on real aciviy and poenial oupu will maer If he exchange rae follows a PPP closure hen he effec of he change on relaive prices ECB Working Paper No 93 December 2

12 will be he key and so on Here, however, and largely in line wih many oher models, we only consider a fixed nominal exchange rae or one following forward-looking UIP The laer implies ha he expeced appreciaion of he home currency exchange rae is se equal o he shor-erm ineres differenial in favour of he home currency This is ofen modified o include a erm in eiher ne foreign asses (NFA) or he curren accoun o GDP raio which proxies a risk premium: 7 er = er + qnfa () e * + ( r r ) + This equaion (being forward looking) needs a erminal condiion o ensure a unique soluion ofen given by a ne foreign asse closure Solving () for he firs period: T = T * er = ert + ( r r ) + qnfa (2) = This defines he exchange rae s iniial jump defined by is erminal value (er T ) and he sums of presen and fuure ineres rae differenials and ne foreign asses Afer his iniial jump he exchange rae evolves as, * er = [( r r ) + qnfa ] (3) Noice, herefore, ha modelling exchange raes as modified uncovered ineres pariy implies ha he exchange rae jumps in response o any change in exogenous insrumens wih ha change sufficien o clear any effec on ne foreign asses brough abou by he shock The oher area of he financial markes where here is a clear poenial for forward-looking behaviour is in he deerminaion of long raes As we saw from he maquee, long raes can be an imporan condui of he moneary ransmission Boh models have backwardand forward-looking opions The backward-looking opions ha we use in boh models are relaively similar in ha hey boh posi a moving-average weigh of boh shor and long raes For he AWM and NiGEM respecively hey are (using our earlier noaion): 8 RL = 25 R i= RL i (4) 7 The UIP condiion can of course be backward looking in which case here is simply a coninuous depreciaion 8 The backward looking long rae equaion used in he AWM is no a sandard feaure of he model bu has been used for illusraive purposes ECB Working Paper No 93 December 2

13 RL = RL + 8 R + 2( R RL + 5) 9 However, in boh models he forward-looking deerminaion of long-erm ineres raes is: (5) log(+ RL /) = 4 39 j= log( + R + j ) (6) Tha is o say, boh models direcly embody a -year bond erm srucure Policy Rules Boh models enable users o implemen he same Taylor rule: ~ ~ + α 2 Π α3 R = α + Y (7) Where R is (as before) he shor-erm ineres rae, π is curren inflaion, y is real GDP, a T ilde indicaes deviaion from (baseline) arge (eg, Π ) and where α,α 2 3 are se a heir sandard weighs The NiGEM model also allows for he possibiliy of a forwardlooking inflaion erm in he Taylor Rule In his, he inflaion erm π becomes fuure raher han curren inflaion The NiGEM model also offers users a variey of oher moneary policy rules including arges for () nominal GDP and inflaion, (2) nominal GDP, (3) inflaion and in addiion i is possible o selec (4) fixed real ineres raes, (5) fixed nominal ineres raes or differen combinaions of hese rules for differen counries and/or differen ime periods wihin he same counry Boh models incorporae a fiscal closure rule o mainain a defici-oupu raio o baseline by changes in he direc ax rae Furher Specific Feaures of he Moneary Policy Transmission in he AWM and NiGEM Models The maquee describes he ransmission channels and srucure involved in he AWM fairly well, alhough ineviably some excepions apply Firs, shor (raher han long) raes ener consumpion and drive he user cos of capial Second, here is no endogenous foreign rae and hus he exchange rae (hough modelled as UIP) is purely driven by movemens in shor-run raes relaive o baseline Also here is no deliberae coding of a 9 The risk premium from holding bonds is assumed o be 5% Of course, he AWM could, in principle, also accommodae all hese rules bu a he curren juncure, he model properies using such rules have no been examined 2 ECB Working Paper No 93 December 2

14 ne foreign asse closure Third, and more specifically, he income effec is deermined via he impac on governmen ineres paymens which are linked o changes in long-erm ineres raes Fourh, he wealh effec is embodied hrough he capial sock and public deb For public deb, as before, ineres paymens are linked o long erm ineres raes In he case of wealh hrough capial accumulaion, he accumulaion of invesmen defines he capial sock Finally, ne foreign asses do no direcly depend on any domesic ineres rae in he curren version of he model In relaion o he domesic secor in NiGEM i is imporan o noe ha, alhough very similar, he se-up in each of he naional economies wihin he euro area can be somewha differen To an exen, his reflecs he deliberae design of he model as he larger economies are modelled in somewha more deail han he smaller ones However, i also reflecs he economeric evidence as since he behavioural equaions are esimaed ineres rae effecs have been found o be presen in some equaions in some counries bu no in ohers A good example of his is a direc ineres rae effec in consumpion, which is only presen in he consumpion funcions of Ialy, he Neherlands and Ireland The ineres rae used is also differen, being he shor-rae in Ialy and he long-rae in he Neherlands an Ireland alhough in all hree cases he coefficien is negaive implying ha a rise in ineres raes has a direc effec in erms of lowering consumers expendiure However, indirec effecs of ineres raes on consumpion via changes in ne wealh as described in he maquee are presen in each of he domesic secors of hese economies The effecs on ne wealh sem from changes in equiy prices, he sock of public deb and in some cases also changes in he moneary aggregae M I is worh making a few remarks on he equiy price effecs, as hey are somewha richer han oulined in he maquee The model allows for backward or forward looking equiy prices In boh cases a rise in ineres raes should lead o lower equiy prices alhough in he backward looking case his is via changes in he long-rae whils in he forward looking case in is via he shor rae In he four larges euro area economies (Germany, France, Ialy and Spain) here is a role for ineres raes o affec M, which feeds ino he governmen deb sock wih a negaive coefficien and hereby affecs ne wealh The shor-erm ineres rae eners he M equaion wih a negaive coefficien in hese counries Turning now o invesmen, for mos euro area counries in NiGEM, ineres raes play a role hrough heir impac on he capial sock via he user cos of capial variable as oulined in he maquee However, in hree of he euro area counries Greece, Ireland In he AWM he sabilisaion of ne foreign asse comes hrough he real exchange rae, Fagan e al (2) ECB Working Paper No 93 December 2 3

15 and Porugal he se up is a lile differen Privae secor invesmen is direcly ied o he long-rae (wih a negaive co-efficien) and he capial sock is deermined based on he las period s capial sock plus new invesmen minus depreciaion Finally, in NiGEM, he domesic shor-erm ineres rae (and also he shor raes of he US, UK, Japan and Swizerland) feeds ino he rae of reurn on foreign liabiliies This in urn affecs he curren accoun by influencing he debis of ineres, profis and dividends The final forward-looking elemen in NiGEM emerges because in he wage equaions of some counries i is possible o inroduce forward-looking inflaion erms When hese equaions were esimaed hey ypically allowed for he possibiliy ha pas, curren and fuure inflaion developmens could have an impac on curren wages 2 In esimaion a significan role for expeced inflaion (wih a lead of -quarer and insrumened by some backward-looking erms) was found in some bu no all counries When he model is used in simulaion mode i is possible o choose beween a erm for expeced inflaion based on backward-looking variables or alernaively o allow for a ruly forward-looking expeced inflaion erm based on he model generaed values for inflaion in he nex quarer This choice has an impac on he dynamics of wages bu leaves heir long-run level unaffeced III Decomposing he Channels of Transmission I is sandard pracice when reporing moneary ransmission exercises o decompose he oal effecs ino heir various (ransmission) channels Tha is o say, we consider all he impacs shor-erm moneary policy changes have, hen caegorise each in an economically inerpreable manner Thus, for example, we migh consider he sole effec of ineres rae changes on invesmen/user cos (B3), exchange raes (B2) and consumpion (B2) as sufficienly imporan o be separaely repored Oher such channels include income and wealh effecs as discussed above Such a channel decomposiion was also a feaure of he BIS exercise In his secion we firs discuss some of he heory behind he decomposiion ino channels before addressing he pracical issues surrounding heir implemenaion in he wo models 3 In principle, he effec of each of hese channels should sum o he oal effec To illusrae, consider he srucural macro-model: 2 In he long-run real wages are deermined by labour produciviy and he unemploymen rae 3 Mauskopf and Siviero (995) and Alissimo e al (2) also examine he issue of decomposing he channels of ransmission 4 ECB Working Paper No 93 December 2

16 ECB Working Paper No 93 December 2 5 V R X Y Y = 3 2 π π π (8) Where Y represens endogenous elemens, X exogenous variables, R he exogenous moneary insrumen and V a vecor of residuals Le us assume here for simpliciy and wihou loss of generaliy in conras o he maquee ha R represens he only available moneary insrumen If we backward subsiue his model o an arbirary sar, we more clearly see he inerdependencies involved: = = = = 3 2 j j j j j j j j j V R X Y Y π π π π π π (9) Equivalenly, in marix form, (2) becomes: Y V V I I R R X X Y Y T T T T T T T T = π π π π π π π π π π π π π π π π π π π π π π () From his, we can derive key moneary mulipliers The Impac Muliplier of he change in he moneary insrumen is 3 =π R Y, he Inerim Muliplier, 3 π π j j R Y = + and he Dynamic (or oal) Mulipliers: = T R Y τ τ Marix 3 π hus defines he pure feedback whereas 3 π π j defines he oal effec To reurn o our maquee example: if he ineres rae only affecs money demand (B7), his means ha 3 π has a single non-zero elemen ( 5 β ) 4 Since money demand affecs Wealh (B6) which in urn affecs Consumpion (B2), Oupu (B), Impor demand (B2) and so on, hese subsequen effecs are embodied in j π Undoubedly his will have some zeros (ie, he moneary insrumen will no affec every endogenous variable) Specifically, closing off all he channels relegaes 3 π o be a null marix Sysemaic opening of one channel decreases is sparsiy by one elemen or one block of elemens In a linear model, he sum of hese effecs is commonly undersood o equal he global effec In all of his, he linkages hrough j π need no be conrolled Alhough, on one level, sraighforward o undersand, channel decomposiions can become complicaed Firs, because some channels of ineres ha are presen in some models may be absen in ohers: his makes model comparisons (such as ours) difficul and imperfec Second, since mos macro-models are non-linear, he individual effecs of 4 Effecively his would mean dropping B7 from he maquee and aking RL as our moneary policy insrumen

17 he channels need no necessarily sum o he global one Third, in some exercises, he summaion of he channels is no logically defined This is in fac a feaure of some of our own resuls when using endogenous policy rules and lead expecaions In such circumsances, i may make more sense o view he channel decomposiions and summaions are as very broad comparisons and local approximaions To illusrae, le us say we perform a global simulaion where all channels operae (ie π 3 has minimum sparsiy), his generaes T τ = Y R = A R τ as well as he rajecory, { } T = If we allow moneary policy o operae afer he iniial wo-year shock hen { R } T = will necessarily be channel-dependen (as indeed our laer simulaions show) Le us say we have I decomposiions or channels The firs one is idenified by, T Yτ Y,, Y A 2 τ Iτ = which in urn generaes { R } T = R Bu i where { } T R { i R }T, = = τ = hen I i= i A i A In oher words, if he moneary response is differen in each channel, he complee aggregaion is no defined In fac, he same hing could be said for any fiscal policy response Thus, we can consider hree decomposiions worh performing Firs, T T T T allowing { R} { Ri}, i, second, imposing { R } { R } i = = = i, and, finally, for = = T T T T each decomposiion j, imposing { R} = { Ri},{ TX } = { TX i}, i j For = = = = reference, we show resuls for he firs wo of hese Idenifying he Channels in Pracice In he following we follow he channels idenified in he BIS (995) exercise: An income/cash flow channel, which measures he effec of ineres raes on ne ineres paymens A wealh channel o capure he impac of ineres raes on wealh A direc ineres channel on consumpion A cos of capial channel o capure he effec of ineres raes on invesmen An exchange rae channel o capure he effecs of changes in he exchange rae due o changes in ineres raes The approach aken was o simulae he model wih all channels swiched off excep he one of ineres To idenify he channels a somewha eclecic approach was aken Eiher 6 ECB Working Paper No 93 December 2

18 we could exogenise whole equaions such as he Invesmen equaion (B3) o capure he user-cos effec or we could fix he ineres rae effecs enering individual equaions o base value In he AWM, he laer approach was adoped which involved subsanial recoding of he model In NiGEM, he channels were removed by exogenising he equaions where he shor-or long-erm ineres rae firs appear This provided a relaively racable way of idenifying he channels wihou he need o significanly aler he model by removing he ineres rae effecs in paricular equaions In some cases he resuls would have been idenical o he approach underaken in he AWM for example as a separae user cos of capial variable was available which only affeced invesmen via he capial sock i was possible o exogenise he user cos variable and hereby effecively remove he ineres rae effec on invesmen wihou swiching off he invesmen equaion However, in oher cases such as for Greece, Ireland and Porugal where here were direc ineres rae effecs in invesmen i was necessary o exogenise invesmen o remove he effec of changes in ineres raes Finally, in relaion o he channels, one poin ha should be borne in mind is ha he models used canno accoun for a credi channel as hey have neiher credi equaions nor lending ineres raes IV Design of Moneary Policy Transmission Simulaion Experimens This secion discusses how key aspecs of boh model consrucion and he design of simulaion experimens have an imporan bearing on he esimaed size of moneary policy effecs in srucural macroeconomic models We begin by considering wheher moneary and fiscal policies can really be reaed as exogenous in model simulaions I has been common for model users o underake wha-if exercises, which examine he consequences of changes in exogenous policy seings However, i is now cusomary for modellers o include closure-rules in heir models, which explain he behaviour of moneary and fiscal policy Reflecing his, he firs issue we address is wheher moneary policy can be reaed as exogenous or wheher i is more appropriae o rea policy as endogenous via a policy rule In he former case, moneary policy experimens can involve a shock o he policy ineres rae In he laer case, i is common o underake moneary policy experimens via a shock o he policy rule eg a change in he arge for he inflaion rae or he money sock In principle, he shock o he policy rule can be calibraed in such a way as o yield ECB Working Paper No 93 December 2 7

19 he same change in he policy ineres rae ha could be imposed if he ineres rae were reaed as exogenous For example, in he case of he BIS (995) exercise of a basis poin increase in he policy ineres raes in a model wih a policy rule, i would be possible o idenify a change in he policy rule ha would induce a change in he policy ineres rae of he same magniude However, i is imporan o bear in mind ha when comparing resuls across models, a common shock o a policy rule can produce very differen reacions of policy ineres raes, a poin ha is illusraed in he following examples Michell e al (998) compared hree global models MULTIMOD, MSG2 and NiGEM underaking a number of comparaive simulaion exercises involving a number of varian simulaions wih changes in fiscal and moneary policy The laer was achieved via a permanen 5% increase in he money arge variable in he models, which induced a reducion in he policy shor-erm ineres rae However, he reacion was raher differen across models This can be seen in he case of Germany where he same change in he money arge lead o a fall in German shor-erm ineres raes of jus over % in MULTIMOD and NiGEM bu over 2% in MSG2 in he firs year In all cases, he ineres raes gradually reurned o baseline hereafer, alhough he speed of he reurn noiceably varied across he models Similarly, Church e al (2) examined moneary policy in models of he UK economy via a 5 percenage poin reducion in he inflaion arge In hree of he five models examined, he change in inflaion arge iniially induces a rise in he shor-erm ineres rae ranging from 6 o 28 percenage poins However, in he oher wo models here is an immediae fall in he nominal ineres rae of beween 3 and 37 percenage poins The explanaion for he difference is ha he achievemen of he new lower inflaion arge is achieved much more quickly in he laer models (inflaion falls by around 4% in he firs year) and hence ineres raes need o be lower in order o reurn he real ineres rae o is baseline value 5 An ineresing relaed quesion is wheher o keep moneary policy endogenous or exogenous afer he compleion of he iniial moneary policy shock If moneary policy is kep exogenous hen here may be prolonged periods of dis-equilibria in many macroeconomic models This poin can easily be seen wih reference o he example of a wo-year hike in policy ineres raes followed by a reurn o base For example, in he iniial period, oupu falls below base and a negaive oupu gap emerges (measured 5 Noably, in forward-looking models, wih full credibiliy and no price-level sickiness, ineres raes would fall immediaely by 5 basis poins 8 ECB Working Paper No 93 December 2

20 relaive o base) Even when ineres raes reurn o baseline a year hree, he negaive oupu gap is likely o mean ha inflaion is also below base and hence real ineres raes remain above base This is likely o induce a furher fall in oupu and in he absence of any promp equilibraing mechanism in he model a downward spiral of falling oupu and inflaion and rising real ineres raes may coninue for some ime For his reason, here may be a case for allowing moneary policy o be endogenous following compleion of he iniial (fixed) moneary policy experimen The precise response of moneary policy would depend on he ype of rule in place In he previous example, wih a Taylor Rule, he negaive oupu gap and lower inflaion would be expeced o resul in ineres raes falling below base for a period o allow he negaive oupu gap o be closed and inflaion o reurn o base The drawback of such an approach is ha i renders he resuls sensiive o he policy rule used and herefore limis he comparabiliy of resuls across models if each model incorporaes a differen moneary policy response The use of fiscal policy rules in srucural macroeconomic models raises some similar issues I has long been recognised eg, Chris (968) ha he governmen budge consrain is imporan If a governmen defici emerges in a simulaion, i is necessary o have some financing assumpion Therefore, many models now incorporae fiscal closure rules, which aim o mainain some level of fiscal solvency hrough adjusing fiscal variables (ofen he direc ax rae bu someimes governmen expendiure) o achieve a arge specified eiher in erms of he defici or deb sock Boh models incorporae such a rule o mainain a defici raio o baseline by changes in he direc ax rae In some siuaions, paricularly in models wih many forward-looking elemens, such rules can help sabilise he model In he above example, a decline in oupu semming from he moneary policy ighening would be likely o resul in a worsening of he fiscal posiion due o lower ax receips and higher ransfers In such circumsances, a fiscal policy rule would seek o resore he original fiscal posiion hrough a ighening of policy (eiher on he revenue or expendiure side) and would herefore exacerbae he original fall in oupu 6 The main argumen agains incorporaing fiscal reacion funcions is ha hey would undermine he comparabiliy of he resuls across models An imporan feaure of model design ha will significanly affec he resuls of simulaions is he reamen of expecaions of variables such as long-erm ineres raes, he exchange rae and inflaion A radiional way of dealing wih expecaions in macro 6 The exen o which he operaion of he fiscal rule would reinforce he downurn would depend on he parameerisaion of he rule ECB Working Paper No 93 December 2 9

21 models was o assume ha hey are deermined as a funcion of curren and lagged values of some observed variables ofen in he form of adapive expecaions However, reflecing he increased populariy of he noion of raional expecaions in recen decades here has been a move o including expecaions ha are genuinely forward-looking in he sense ha hey are consisen wih he fuure oucomes generaed by he model For his reason hey are ofen called model-consisen expecaions 7 As already discussed he AWM and NiGEM models allow, albei o varying degrees, for forward-looking behaviour In he nex secion we explore he implicaions of changing he exen of forward-looking behaviour in he models V Simulaion Experimens Using he AWM and he NiGEM Models To illusrae he imporance of he issues raised in he previous secion, we now urn o a comparison of he resuls of a number of simulaion exercises using he AWM and NiGEM The Comparaive Simulaion Experimens () As a saring poin we followed BIS (995) and underook a moneary policy experimen involving an increase of he shor-erm policy ineres raes by percenage poin for wo years (in our case 2Q-22Q4 inclusive) Thereafer (ie from 23Q) a reurn o baseline values was assumed and no moneary policy rules were implemened No fiscal rules were in operaion eiher and he models were run in an enirely backwardlooking mode wih a fixed nominal exchange rae (2) The nex simulaion experimen was idenical o simulaion, excep for he fac ha a moneary policy rule was implemened a he end of he wo-year iniial shock (ie from 23Q onwards) The form of he moneary reacion is he Taylor rule specified in d equaion (7) wih Π and Y d se a heir baseline values (3) The nex simulaion was idenical o simulaion 2 bu a fiscal policy rule was allowed o operae from he sar of he simulaion As was discussed in he previous secion he fiscal rule in boh models adjuss he direc ax rae o achieve a arge for he governmen budge balance as a proporion of GDP 7 There is, of course, a large lieraure on wheher expecaion formaion can really be described as raional Robers (998) examines he evidence on inflaion expecaions and examines he raionaliy of US surveys of inflaion expecaions He concludes ha he expecaions repored by hese surveys represen an inermediae degree of raionaliy As he issue of he appropriae degree of raionaliy in expecaions formaion is an open one we illusrae he imporance of his quesion in he nex secion when we will underake he moneary policy exercise wih and wihou model-consisen expecaions erms 2 ECB Working Paper No 93 December 2

22 (4) The final simulaion was idenical o 3 bu i allowed all he forward-looking elemens of he models o operae 8 For he NiGEM model, his mean ha here was forward behaviour in he deerminaion of he exchange rae, long raes, equiy prices, wage formaion and he inflaion rae enering ino he Taylor rule In he AWM he forwardlooking elemens are he exchange rae and he long-erm rae 9 For informaion, he GDP and inflaion responses from one addiional simulaion are repored in Appendix 3 of his paper These correspond o a simulaion wih he same movemen in shor-erm ineres raes, no moneary or fiscal policy rules and a predeermined exogenous pah for he exchange rae and he long rae This experimen reflecs he moneary policy experimen using naional cenral banks models and he AWM repored in Van Els e al (2) The Resuls The resuls of he simulaion experimens are shown in Tables -4 In he firs simulaion, he rise in shor-raes induces a rise in long-raes in boh models bu he paern is raher differen reflecing he differences in he backward looking equaions for he long rae (see equaions 4-6) In NiGEM, he long rae increases by nearly as much as he shor rae in he firs wo years and hen immediaely reurns o close o he baseline hereafer In he AWM, he immediae rise in he long rae is much smaller han he rise in he shor rae In erms of he impac on aciviy, boh he maximum and he average loss of oupu over he firs hree years are quie similar, bu here are differences in he iming of he oupu loss Wih he NiGEM simulaion, oupu falls by 9% in he firs year, dropping by 36% and 44% in he second and hird years respecively In he AWM simulaion, he iniial impac is somewha larger wih a fall in oupu of 23% in year reaching o 46% in year 2 before moderaing o 39% in he hird year Thereafer, in boh models, here is a endency for oupu o reurn o he baseline, and hen o remain above he baseline for someime However, whils boh models show such a endency, he reversion 8 For reference, NiGEM was simulaed using a (Gauss-Seidel) Fair-Taylor (eg, Fisher, 992) rouine and he AWM wih he Sack-Time algorihm (Juillard e al, 998, McAdam and Hughes-Halle, 999) 9 We do no repor any simulaions wih endogenous moneary policy for he whole period ie from 2Q onwards However, a number of addiional simulaions were underaken on he NiGEM in which he policy rule was shocked in order o generae a rise in ineres raes of around percenage poin for wo years For example, i was found ha in a simulaion idenical o number (4) above excep ha moneary policy was deermined by an inflaion arge rule a fall in he inflaion arge for he firs 9 quarers of 35 percenage poins would deliver a rise in shor erm ineres raes of around percenage poin in he firs wo years (The inflaion arge needed o be alered for 9 periods ie one in excess of he desired moneary policy reacion due o he forward-looking naure of he rule) ECB Working Paper No 93 December 2 2

23 o base is more proraced in he AWM due o he fac ha he AWM involves relaively weaker feedbacks and error correcions leading o a slower speed of adjusmen 2 Table : Simulaion : GDP PCE U S-Rae L-Rae Nom XR Real XR TBR GBR NiGEM AWM Noes: GDP is real GDP, PCE is he consumer s expendiure deflaor, U is he sandardised unemploymen rae, S-Rae is he 3-monh ineres rae, L-Rae is he -year ineres rae, Nom XR is he nominal exchange rae, Real XR is he real exchange rae, TBR is he Real Trade Balance and GBR is he governmen budge balance o GDP raio GDP, PCE, Nom XR and Real XR are all expressed as a percenage difference from baseline The remaining variables are expressed as an absolue difference from baseline A fall in he nominal or real exchange rae is a depreciaion The governmen budge balance worsens in boh cases as lower oupu leads o lower ax receips and higher governmen ransfers due o higher unemploymen However, he deerioraion of he fiscal posiion in he firs wo years is more pronounced in NiGEM despie a more modes oupu loss over his period compared wih he AWM This reflecs a greaer sensiiviy of he fiscal variables o changes in economic aciviy in NiGEM han in he AWM I is also he case ha changes in economic aciviy exer a larger impac on unemploymen in he AWM, albei wih a somewha longer lag han in NiGEM Afer hree years, he unemploymen rae is 35 of a percenage poin higher in he AWM compared wih a rise of 9 of a percenage poin in he NiGEM simulaion This is despie he fac ha in year 3 oupu is furher below is baseline in he NiGEM simulaion han in he AWM one However, he longer lag in he impac on unemploymen is indicaed by he fac ha he peak in he rise in unemploymen occurs afer he rough in he fall in oupu in he AWM, in conras wih he NiGEM resuls when boh effecs occur in he same year 2 For insance, see McAdam (2) for a frequency-domain analysis of euro-area daa There, i is suggesed ha he ypical periodiciy of euro-area oupu is around 37 quarers o 24 quarers in he US 22 ECB Working Paper No 93 December 2

24 The response of prices reflecs he developmens in oupu, albei wih a marked lag due o convenional sicky price mechanisms in models In NiGEM, prices do no fall significanly in he firs couple of years, bu by year 3 hey are 2% below base and in year 5 hey are 34% below and hereafer hey reurn o base In he AWM he fall in prices is more marked in he firs wo years (reflecing he larger iniial oupu loss) and coninues o gaher pace during he reporing period for he simulaion and in year 5 prices are 4% below base In boh models, by year, prices have moved close o heir baseline levels However, i is noeworhy ha in he AWM simulaion oupu is above is baseline level whils he unemploymen rae is sill markedly below is baseline level This indicaes ha here is sill some furher adjusmen o ake place before all variables have susainably reurned o heir baseline levels Therefore afer year here are sill some oscillaions in he AWM resuls, albei of a gradually diminishing magniude 2 This effec is no presen o a noiceable exen in he NiGEM model The fac ha he simulaions show a endency o reurn o base parly reflecs he fac ha here is a fixed nominal exchange rae The lower domesic prices mean ha here is depreciaion in he real exchange rae, which ulimaely booss oupu hrough rade This is refleced in he rade balance, which improves as lower domesic aciviy reduces he demand for impors and lower domesic prices improve boh impor and expor compeiiveness 22 This boos o ne rade ulimaely resores oupu and prices o heir baseline levels In he second simulaion, he firs wo years are idenical bu in he hird year, he Taylor Rule begins o operae Because oupu and prices are boh below base here is a decline in he shor-erm ineres rae in he hird year in boh models In NiGEM he Taylor Rule leads o a decline in shor erm ineres raes of 38 basis poins whils in he AWM he 2 The precise naure of he apparen oscillaory behaviour in some of he AWM resuls (and o a lesser exen NiGEM) is beyond he scope and purpose of he presen paper since we use boh models in heir given form Technically, oscillaory behaviour can be raced and analysed using he linearised eigenvalues of he model equaions For insance, given a simple linear difference equaion, Y - αy - - βy -2 = V, we have he general soluion, Y = A λ + A 2 λ 2 + Y SS As is well known (eg Chiang, 974) roos (ie, he λ i s) less han one wihou (wih) imaginary componens generae sable non-oscillaory (sable oscillaions) dynamics owards he seady sae, and one sable and unsable roo wihou (wih) imaginary componens yields unsable nonoscillaory (unsable oscillaory) dynamics In more inuiive economic erms, he fac ha in he AWM, employmen is obained by invering he Cobb-Douglas producion essenially gives rise o a larger han one employmen-o-gdp elasiciy and hus poenially o some over reacion of employmen o changes in GDP in he shor run, only compensaed in he long run by changes in he capial sock Similarly, he lack of a smoohing parameer (in boh models) in he Taylor rule may engender a relaively high degree of ineres-rae volailiy To illusrae, mos empirical esimaes of Taylor rules give esimaes of such smoohing parameers of around The curren accoun balance (which is no shown in he able) in NiGEM acually worsens in he iniial period as he rise in ineres raes increases he magniude of debis of ineres profis and dividends ECB Working Paper No 93 December 2 23

25 decline is more modes As he parameers of he Taylor rules are idenical he difference in oucomes is due o he fac ha when he rule begins o operae he oupu loss and fall in inflaion are boh greaer in NiGEM han in he AWM This is no immediaely apparen from he able as in year 3 he change in prices is he same in boh ses of resuls However, o some exen his reflecs he fac ha he annual average figures mask inrayear changes and he larger fall in ineres raes induced by he Taylor Rule in NiGEM has already begun o have an effec in boosing oupu and prices When he rule is implemened in he firs quarer of year 3 boh oupu and inflaion are furher below base in he NiGEM simulaion han hey are in he AWM Ineviably he lower ineres raes from year 3 onwards reduce he magniude of he subsequen oupu losses alhough he iniial impac is no ha large In year 3 he oupu loss is reduced by abou 5% in boh he NiGEM and AWM simulaions when compared wih he firs simulaion Some ineresing conrass hen emerge beween he resuls from he wo models In NiGEM shor erm ineres raes are close o baseline by year 5, bu his reurn o baseline akes longer in he AWM, reflecing he fac ha oupu and prices are below base for longer In boh models he use of he Taylor Rule speeds he reurn of GDP owards he baseline As boh also generae a period of above baseline GDP, he Taylor Rule hen works o reduce owards he baseline by raising shor-erm ineres raes This paern can be seen quie clearly in Figures a and 2a 24 ECB Working Paper No 93 December 2