Comparative analysis of the exchange market pressure in Central European countries with the Eurozone membership perspective

Transcription

1 MPRA Munich Personal RePEc Archive Comparaive analysis of he exchange marke pressure in Cenral European counries wih he Eurozone membership perspecive Daniel Savarek Silesian Universiy - School of Business Adminisraion 8. June 007 Online a hp://mpra.ub.uni-muenchen.de/3906/ MPRA Paper No. 3906, posed 8. July 007

2 Comparaive Analysis of he Exchange Marke Pressure in Cenral European Counries wih he Eurozone Membership Perspecive Daniel Savárek Silesian Universiy in Opava School of Business Adminisraion in Karviná Deparmen of Finance Univerziní nám. 1934/ Karviná Czech Republic el: fax: Absrac This paper esimaes he exchange marke pressure (EMP) in four Cenral European counries (Czech Republic, Hungary, Poland, Slovakia) during he period Therefore, i is one of very few sudies focused on his region and he very firs paper applying concurrenly model-dependen as well as model-independen approach o he EMP esimaion on hese counries. The resuls obained sugges ha he approaches are no compaible and lead o absoluely inconsisen findings. They ofen differ in boh idenificaion of principal developmen rends and esimaed magniude and direcion of he pressure. Therefore, any general conclusion on hose issues is hard o draw. The paper provides evidence ha a shif in he exchange rae regime owards he quasi-fixed ERM II should no lead o increasing EMP. However, i is highly probable ha some episodes of he excessive EMP will make he fulfillmen of he exchange rae sabiliy crierion more difficul in all counries analyzed unless he crierion will have eased. Keywords exchange marke pressure; model-dependen approach; modelindependen approach; EU New Member Saes; exchange rae sabiliy crierion JEL classificaion C3; E4; F31; F36 Research behind his paper was suppored by he Czech Science Foundaion wihin he projec GAČR 40/05/758 Inegraion of he financial secor of he new EU member counries ino he EMU 1

3 1. Inroducion Eigh counries from Cenral and Easern Europe (hereafer EU8) joined he European Union (hereafer EU) in he spring of 004 and compleed he ransformaion from cenrally planned economies o marke economies. Moreover, i is expeced ha hey will also join he Eurozone and implemen he euro as heir legal ender. However, membership in he Eurozone is condiioned by fulfillmen of he Maasrich crieria. One of which is he crierion of he naional currency s sabiliy in he period preceding enry ino he Eurozone. This crierion is associaed wih specific exchange rae regime, ERM II, which mus be adaped by all counries wih regimes whose principles do no correspond wih he ERM II s spiri. 1 I means ha all EU8 counries excep for Esonia and Lihuania had or will have o modify heir exchange rae arrangemen when joining ERM II. The Czech Republic, Hungary and Poland currenly use flexible exchange rae arrangemens. Slovakia and o a lesser exen Slovenia also mainained a flexible regime before enry ino he ERM II. Such a change oward a less flexible exchange rae sysem could increase suscepibiliy of he counries o currency crises and pressures on he foreign exchange markes. Therefore, he aim of his paper is o esimae exchange marke pressure (EMP) in he Czech Republic, Hungary, Poland and Slovakia (hereafer EU4) during he period Since all counries applied boh a fixed and flexible exchange rae regime, he ime span chosen allows us o compare magniude of ensions on he foreign exchange marke in differen exchange rae environmens. This kind of analysis has imporan policy implicaions as Slovakia has already swiched o a less flexible regime and he remaining counries will make his unavoidable sep is in he near fuure. The paper is srucured so ha Secion describes he meaning and heoreical conceps of EMP and provides a review of he relevan lieraure. In Secion 3, he models and daa used are cied; Secion 4 repors he empirical resuls and he conclusions are presened in Secion 5. 1 The group of incompaible regimes includes crawling pegs, free floas or managed floas wihou a muually agreed cenral rae and pegs o anchors oher han he euro. As of 31 s December 006, five of he EU8 (Esonia, Lavia, Lihuania, Slovakia and Slovenia) joined he ERM II. Neverheless, he exchange rae regime in Lavia was very similar wih he ERM II, hus he coss of he regime s rearrangemen are raher marginal.

4 . Exchange Marke Pressure and Lieraure Review.1 Meaning and Conceps of he Exchange Marke Pressure The erm exchange marke pressure is usually relaed o changes of wo cardinal variables describing he exernal secor of any economy: official inernaional reserve holdings and he nominal exchange rae. However, he noion of EMP was more precisely defined, for he firs ime, in Giron and Roper (1977). In his seminal paper, he auhors uilized a simple moneary model of he balance of paymens o devise an index of he excess demand for money ha mus be relieved by eiher exchange rae or reserve changes o keep he money marke, and hence he balance of paymens, in equilibrium. The index was he simple sum of he rae of change in inernaional reserves and he rae of change in he exchange rae. For he firs ime, hey ermed his index as EMP. However, some shorcomings can be found in he model. Since he measure is derived from a highly resricive moneary model he formula canno be applied o oher models. Furhermore, a model-dependen definiion is used, hus, a unique formula for EMP canno be idenified wihin he Giron-Roper framework. The original concep of EMP has been modified and exended by many researchers. For example, Roper and Turnovsky (1980) and Turnovsky (1985) inroduced he idea of using a small open-economy model and exended he original model by subsiuing he simple moneary approach by an IS-LM framework wih perfec mobiliy of capial. They allowed inervenion o ake he form of changes in domesic credi as well as changes in reserves. The consequence of hese modificaions was ha he EMP was sill a linear combinaion of he rae of change of he exchange rae and money base bu hese wo componens were no longer equally weighed as in he Giron-Roper model. A noable conribuion o he EMP heory was provided by Weymark (1995, 1997a, 1997b, 1998). She revised he models menioned above and inroduced a more general framework in which he models are boh special cases of he generalized formula. She inroduced and esimaed a parameer (conversion facor) sanding for he relaive weigh of exchange rae changes and inervenion in he EMP index. Since all previous EMP definiions semmed from a specific model, Weymark also proposed a model-independen definiion of EMP as: The exchange rae change ha would have been required o remove he excess demand for he currency in he absence of exchange marke inervenion, given 3

5 he expecaions generaed by he exchange rae policy acually implemened (Weymark, 1995, p.78) An exension making he simple model oulined in Weymark (1995) more realisic was inroduced in Spolander (1999). He incorporaed ino he model a moneary policy reacion funcion and serilized foreign exchange inervenion. Many researchers have criicized he mos undesirable aspec of he EMP measure, dependency on a paricular model, and proposed some alernaive approaches. A simpler and model-independen EMP measure was originally consruced in Eichengreen e al. (1994, 1995). According o his approach EMP is a linear combinaion of a relevan ineres rae differenial, he percenage change in he bilaeral exchange rae and he percenage change in foreign exchange reserves. Conrary o Weymark s approach, he weighs are o be calculaed from sample variances of hose hree componens wih no need o esimae any model. The modified EMP index was inroduced in Sachs e al. (1996). This measure consiss of he same elemens bu addresses he quesion how o weigh he hree componens o avoid he dominance of he mos volaile variable. Each weigh in he EMP index is calculaed wih respec o sandard deviaions of all componens included insead of using only sandard deviaion of he respecive componen. The original model by Eichengreen e al. (1996) was modified in Kaminsky e al. (1998) and Kaminsky and Reinhar (1999). The ineres rae differenial is subsiued by relevan ineres rae in he counry analyzed. Furhermore, he weighs on he reserves and ineres rae erms are he raio of he sandard error of he percenage change of he exchange rae over he sandard error of he percenage change of reserves and he ineres rae differenial respecively. An approach semming from Eichengreen e al. (1996) was also followed by Penecos e al. (001). However, hey deermined he weighs using principle componens analysis. Because neiher he componens of he index nor he weighing scheme is derived from a srucural model of he economy he EMP indices obained in Eichengreen e al. (1996), Sachs e al. (1996), Kaminsky e al. (1998), Kaminsky and Reinhar (1999) and Penecos e al. (001) are model-independen.. Review of Relevan Empirical Lieraure Since is inroducion, EMP has araced he aenion of many researchers and a grea number of heoreical as well as empirical papers have been published. The empirical EMP 4

6 lieraure is bi-direcionally oriened. Whereas some of he papers are direcly focused on esimaion of EMP in a variey of regions and counries, he second caegory of sudies use he EMP measure as an elemen of subsequen analysis examining currency crises, moneary policy, foreign exchange inervenion, exchange rae regime and oher issues. In accordance wih he geographical orienaion of his paper, only sudies empirically analyzing EMP in EU4 are cied in he following lieraure review. Alhough EMP has been a frequenly discussed and examined opic in he lieraure 3 one can find a very limied number of papers focused on new EU Member Saes and EU4 in paricular. Only four consisen sudies esimaing EMP in all or some of EU4 have been published o dae. The firs sudy esimaing EMP in, among ohers, some of EU4 (Czech Republic and Poland) was Tanner (00). He applied he radiional Giron-Roper model on daa from 3 emerging counries and, consequenly, examined he relaionship beween EMP and moneary policy in a vecor auoregression (VAR) sysem. The aim was o re-examine currency crises in emerging markes in in a more radiional way by emphasizing he role of moneary policy a or around he ime of crisis. Regarding he EMP calculaed in he Czech Republic and Poland, hey were modes as compared o oher counries and very similar o each oher. However, EMP in Poland was welve imes higher han in he Czech Republic during he Asian in he second half of he 1990s. Tanner s paper also provided evidence ha here was a posiive relaionship beween EMP and domesic money supply in boh EU4 counries analyzed bu no as significan and sraigh as in oher counries. The only shocks o EMP can help explain moneary policy in Poland, possibly indicaing serilized inervenion. A more specific applicaion of he Tanner (00) approach is Bielecki (005). The paper concenraes only on Poland from The resuls from he VAR sysem analyzing he relaionship beween EMP and changes in domesic credi (moneary policy acions) indicae ha domesic credi reaced in a couner direcion o innovaions o EMP. Furhermore, in his paper, Bielecki compared wo EMP measures calculaed under alernaive mehodologies (using all foreign reserve changes and pure official foreign exchange inervenion daa). He came o he conclusion ha he appreciaion pressure prevailed during he sample period. However, he behavior of he wo indices differed o some exen, especially wih evens characerized by exreme EMP values (July 1997, Augus 1998, 3 Some of he recen empirical sudies examining EMP are Jeisman (005) in Ausralia, Gohoco-Bauisa and Bauisa (005) in Philippines, Bird and Mandrilas (005) in Lain America, Carribean, Asian and Pacific regions, Wyplosz (00) in a worldwide group of ransiion counries, Kamaly and Erbil (000) in he MENA region or Kohlscheen (000) in Chile. 5

7 February 1999 or July 001). Generally, using he pure inervenion daa in he EMP esimaion provided more realisic and robus resuls. Vannese e al. (005) used EMP as an indicaor of currency crisis and addressed he quesion wheher currency crises in EU8 have been more frequen in fixed, inermediae or flexible exchange rae arrangemens. The auhors found ha EMP was marginally smaller in counries and periods characerized by an inermediae exchange rae regime as compared o hose wih a floaing arrangemen. Regarding EU4, he mos crisis quarers (excessive EMP) occurred in Hungary during he fixed peg regime and in Poland when a crawling peg was being applied. Managed floaing proved o be a relaively sable regime from he EMP perspecive. In addiion o hese conclusions, he auhors also provided evidence of high correlaion beween several EMP measure specificaions wih which hey experimened. Very similar conclusions were drawn in Savárek (005) where EMP in he Czech Republic, Hungary, Poland and Slovenia in is esimaed. The sudy applied he EMP measure proposed in Eichengreen e al. (1995) and he resuls obained sugges ha he Czech Republic and Slovenia wen hrough considerably less volaile developmen of EMP han Hungary and Poland. Besides he focus on he sill overlooked EU4 region, his paper conribues o he EMP lieraure in wo basic aspecs. Firs, i uses he mos recen daa and prolongs he period analyzed o he end of 006. Second, his paper represens he very firs concurren applicaion of he model-dependen and model-independen approaches on EU4 counries. Thus, he suiabiliy of hese models for EU4 counries can be evaluaed. 3. Measuring he Exchange Marke Pressure: Model and Daa 3.1 Model-Dependen Approach As menioned previously, his sudy originally sems from Weymark (1995) where he following formula for EMP calculaion was defined: where EMP = e + η r, (1) e is he percenage change in exchange rae expressed in direc quoaion (domesic price for one uni of foreign currency), r is he change in foreign exchange reserves scaled by he one-period-lagged value of money base and η is he conversion facor which has o be esimaed from a srucural model of he economy and is defined as: 6

8 η = e. () r The conversion facor represens elasiciy ha convers observed reserve changes ino equivalen exchange rae unis. 4 For pracical esimaion of EMP, he mehodology inroduced in Spolander (1999) was applied. Similarly wih Weymark (1995), i is a model-consisen measure of EMP in he conex of small open economy moneary model. However, he cenral bank s moneary and foreign exchange policies are explicily defined, foreign exchange inervenion parly serilized, and expecaions raional in he Spolander model. The model is summarized in equaions (3) o (9): d m = 0 + p + β1 c β β i (3) p = 0 + α1 p + α α e (4) i = i + E ) + e ( e 1 (5) m = d + ( 1 λ ) r (6) s a r = p e (7) y (8) a rend d = γ 0 + y + ( 1 γ 1) p γ d s gap m = m (9) where p is domesic price level, p is foreign price level, e denoes exchange rae (in direc quoaion), m is nominal money sock (he superscrip d represens he demand and s he supply), c is real domesic income, i is nominal domesic ineres rae, i denoes nominal foreign ineres rae, E e ) is expeced exchange rae change and λ is proporion of ( + 1 serilized inervenion. All variables up o his poin are expressed in naural logarihm. Nex, a d is auonomous domesic lending by he cenral bank and r is he sock of foreign exchange reserves, boh divided by he one period lagged value of he money base. rend y is he long-run rend componen of real domesic oupu y and gap y is he difference beween y and rend y. The sign naurally denoes change in he respecive variable. Equaion (3) describes changes in money demand as a posiive funcion of domesic inflaion and changes in real domesic income and a negaive funcion of changes in he 4 There is an assumpion ha all inervenion akes he form of purchases or sales of foreign exchange reserves. When, in addiion o his ype of inervenion, domesic credi changes are used o influence exchange rae, he EMP formula generaed by log-linear models has he general form: EMP = e + η [ r + λ d ] where λ is he proporion of he observed domesic credi change ha is associaed wih indirec exchange marke inervenion. 7

9 domesic ineres rae. Equaion (4) defines he purchasing power pariy condiion aribuing he primary role in domesic inflaion deerminaion o exchange rae changes and foreign inflaion. Equaion (5) describes uncovered ineres rae pariy. Equaion (6) suggess ha changes in he money supply are posiively influenced by auonomous changes in domesic lending and unserilized changes in he sock of foreign reserves. Equaion (7) saes ha changes in foreign exchange reserves are a funcion of he exchange rae and a ime-varying response coefficien p. Equaion (8) describes he evoluion of he cenral bank s domesic lending. Whereas domesic inflaion and changes in rend real oupu changes are posiive deerminans of he domesic lending he gap beween real oupu and is rend has a negaive impac on domesic lending aciviy. Equaion (9) defines a money marke clearing condiion ha assumes money demand o be coninuously equal o money supply. By subsiuing equaions (4) and (5) ino equaion (3) and subsiuing equaion (8) ino equaion (6) and hen using he money marke clearing condiion in equaion (9) o se he resuling wo equaions equal o one anoher, i is possible o obain he following relaion: where β λ e, (10) X + E( e + 1) + (1 ) r = γ1α + β X = γ γ α β + y γ α p γ y β c + β i (11) rend gap and he elasiciy needed o calculae EMP in equaion (1) can be found as: e η = r (1 λ) =. (1) γ α + β 1 3. Model-Independen Approach As menioned above, Eichengreen e al. (1994, 1995) argued ha dependency on a paricular model was an undesirable feaure for an EMP index. As an alernaive, hey proposed he following measure of speculaive pressure: e rm EMP = e σ rm 1 + ( ( i σ i 1 i r rm rm where σ r is he sandard deviaion of he difference beween he relaive changes in he raio of foreign reserves and money (money base) in he analyzed counry and he reference )) (13) 8

10 counry rm rm and σ i is he sandard deviaion of he nominal ineres rae rm rm differenial ( ( i )). Oher variables are as defined in he previous specificaion. i However, for he pracical calculaion we ook an inspiraion from Sachs e al. (1996) and made some modificaions of he EMP formula. We changed he weighing scheme o avoid he EMP measure being driven by he mos volaile componen and abandoned he relaion beween foreign reserves and money in home and reference counry. Consequenly, he EMP formula based on model-independen approach can be wrien as follows: 1 σ e EMP = ((1 σ e ) + (1 σ rm 1 σ i + ((1 σ e ) + (1 σ rm e ) (1 i )) + σ e ( ( i ) (1 i )) + σ i 1 )) 1 σ ((1 σ e ) + (1 σ rm rm rm ) (1 i )) + σ rm where σ e is he sandard deviaion of he rae of change in he exchange rae variables are denoed consisenly wih (13). 1 e e 1 + (14) and oher 3.3 Daa The samples of daa used in his paper cover he period 1993:1 o 006:4 yielding 56 quarerly observaions for all EU4 counries. The daa were predominanly exraced from he IMF s Inernaional Financial Saisics and he Eurosa s Economy and Finance daabase. The missing observaions in he ime series were replenished from daabases accessible on he EU4 cenral banks websies. The deailed descripion of all daa series and heir sources is presened in Appendix 1. Neverheless, a brief overview of he daa used is provided here for beer undersanding of he models described above. We used nominal bilaeral EU4 naional currencies exchange raes agains he euro (e ). The exchange raes prior o 1999 were obained using he irrevocable conversion rae of he German mark o he euro. The domesic (i ) as well as foreign ineres rae ( i ) are represened by he 3-monh money marke raes in EU4 counries and he Eurozone. The M1 moneary aggregae was employed as he domesic money sock (m ). The domesic (p ) and foreign price levels ( p ) are proxied by he respecive consumer price index. As he level of domesic oupu (y ) we applied he gross domesic produc (GDP). The gross naional income (c ) was derived by adding he ne income from abroad o GDP. The domesic money base (B ) 9

11 and oal reserves minus gold (r ) were also included in he model-dependen EMP esimaion. The proporional raio of reserves (rm ), used in he model-independen approach, was yielded by raio of change in level of reserves and money base of previous period. 4. Esimaion of he Exchange Marke Pressure 4.1 Model-Dependen Approach As is eviden from he model presened in Secion 3.1, he EMP esimaion (1) mus be preceded by he calculaion of he conversion facor η (, 1). This sep is, however, required o obain values of he serilizaion coefficien λ (6), he elasiciy of he money base wih respec o he domesic price level γ 1 (8), he elasiciy of he domesic price level wih respec o he exchange rae α (4), and he elasiciy of he money demand wih respec o he domesic ineres rae β (3). equaions. More precisely, he parameer esimaes are obained by esimaing he following hree m p = β 0 + β1 c β i + ε1, (15) (16) p = α 0 + α1 p + α e + ε, B B 1 r y rend p = γ + λ r + γ p + γ y + ε 0 1 Equaions (15) and (16) are obained direcly from equaions (3) and (4). Equaion (17) is derived by subsiuion of (7) ino (5) and noing ha change in money supply equals he change in money base B B 1 gap 3, assuming he money muliplier o be consan. One can disinguish wo ypes of variables included in he model: endogenous and exogenous. The endogenous variables are exogenous variables are c, p, i,, p, e, i, m rend y and 1 (17) B and r. The B gap y. Despie he fac ha e does no appear on he lef-hand side of any of he equaions, i is he endogenous variable because he exchange rae is clearly he variable deermined by his model. The model is esimaed using he wo-sage leas square regression echnique (SLS). The main reason is ha he endogenous variables are on boh sides of equaions (3)-(9). I 10

12 means ha in each equaion having endogenous variables on he righ-hand side, hese variables are likely o correlae wih he disurbance erm. Thus, using he ordinary leas square mehod would lead o biased esimaes. On he oher hand, he hree-sage leas square mehod was no chosen because of he limied size of he daase used (small number of observaions). The SLS used requires he incorporaion of insrumens (variables uncorrelaed wih he disurbance erm) ino he esimaion. Thus, he firs empirical sep of he analysis was o find appropriae insrumens. For his purpose we run he firs sage regressions on endogenous variables having all possible insrumens as regressors. As possible insrumens we se he conemporaneous and one-quarer lagged values of exogenous variables and onequarer lagged values of all endogenous variables. Finally, he regressors wih sufficien saisical significance were seleced as insrumens. This procedure was carried ou for all counries and equaions of he model. The nex aspec which had o be assessed is he saionariy of regressors. This feaure is essenial for all regression models. We applied Phillips-Perron and Augmened Dickey- Fuller ess o examine he saionariy of he ime series used. Uniform oucomes of boh ess were necessary for he final conclusion abou he (non)saionariy of each ime series. According o he characer of each ime series we esed he saionariy wih a linear rend and/or inercep or none of hem. To conserve space he saionariy ess resuls are no repored here bu hey allow us o conclude ha he firs differences of all ime series are saionary. Thus, hey can be used in esimaion of all equaions of he model. 5 The SLS esimaion resuls are presened in Tables 1 o 3, individually for each equaion. The ables also conain he lis of insrumens and resuls of some diagnosic ess. We applied Jarque-Berra (J-B) indicaor o assess normaliy of he residuals disribuion, Breusch-Godfrey Langrange Muliplier (LM) o es serial correlaion and Whie es o check heeroscedasiciy. All LM ess were run wih four lags. The ess indicaed evidence of serial correlaion in residuals from he equaions and he poenial heeroscedasiciy was also idenified in some cases. Therefore, we correced he sandard errors of parameer esimaes by he Newey-Wes procedure. Even more frequenly, he residuals seem o be non-normally disribued. Therefore alhough he -saisics can be misleading, his does no reduce he 5 The percenage change in money base is a naurally flow variable and, hus, already differenced and saionary. Likewise, y gap is saionary on level in all counries because of is consrucion. 11

13 validiy of he parameer esimaes. 6 According o he model specificaion he parameers β 1, α 1, and α should be posiive and β, γ 1, γ, and λ should be negaive. Since λ is a fracion, is absolue value should be beween zero and one. Table 1: Esimaes of equaion (15) Czech Republic Hungary rend gap rend insrumens: y 1 r 1 i 1 y p insrumens: 1 c p i 1 y 1 param. esim. s.er. prob. param. esim. s.er. prob. β β β β β β R =0.0784, SEE=0.0094, DW=1.906 R =0.104, SEE=0.010, DW=1.633 J-B= (0.0000), LM=6.167 (0.1870) WHITE=4.917 (0.0001) J-B= (0.6448), LM=1.709 (0.000) WHITE= (0.078) Poland Slovakia rend insrumens: y 1 e 1 i 1 p c insrumens: 1 c 1 p 1 m p 1 i 1 param. esim. s.er. prob. param. esim. s.er. prob. β β β β β β R =-0.486, SEE=0.0101, DW=.397 R = , SEE=0.0165, DW= J-B= (0.7546), LM=8.077 (0.084) WHITE=3.585 (0.0001) J-B= (0.0000), LM=4.94 (0.3676) WHITE= (0.883) Source: Auhor s calculaions The esimaions of equaion (15) provide mediocre resuls. The parameers β necessary for he conversion facor calculaion are correcly signed in all EU4. However, he parameer is no saisically significan in Slovakia. One can see some evidence of nonnormal disribuion (Czech Republic, Slovakia), serial correlaion (Hungary) and heeroscedasiciy (Czech Republic, Hungary and Poland). In he esimaions of equaion (16) we obained very good resuls. The signs of all parameers are consisen wih he heoreical assumpions and imporan α parameers are significanly differen from zero in all counries. On he oher hand, however, only error erms in he Polish and Slovak equaions seem o pass he sandard diagnosic ess compleely. Furhermore, one can find a subsanially lower elasiciy of he domesic price level wih respec o he exchange rae ( α ) in Poland and, o a lesser degree, in Slovakia han in oher EU4. Alhough i is no direcly linked wih he EMP esimaion, i is worhwhile o poin ou 6 Since differen equaion specificaions have differen insrumens, R for SLS can be negaive even if a consan is used in he equaion. 1

14 a general feaure of relaively high elasiciy of he domesic price level wih respec o he foreign inflaion ( α 1). One can find ha quie common in small and open economies during he ransiion period. Table : Esimaes of equaion (16) Czech Republic Hungary rend rend insrumens: p y 1 i 1 e insrumens: 1 p y 1 c p 1 param. esim. s.er. prob. param. esim. s.er. prob. α α α α α α R =-.9594, SEE=0.0058, DW= R =0.1440, SEE=0.0053, DW=1.737 J-B=10.0 (0.0066), LM= (0.0665) WHITE= (0.0000) J-B= (0.9931), LM=10.01 (0.0401) WHITE= (0.0351) Poland Slovakia gap rend insrumens: p e 1 i 1 p 1 c 1 p insrumens: 1 i e 1 y y p 1 param. esim. s.er. prob. param. esim. s.er. prob. α α α α α α R = , SEE=0.0070, DW= J-B= (0.6070), LM= (0.0756) WHITE= (0.441) Source: Auhor s calculaions R = , SEE=0.0048, DW= J-B=3.50 (0.1719), LM=5.780 (0.16) WHITE= (0.1186) The resuls from he money supply equaion (17) are somewha poorer. This is rue because especially he esimaion of he Polish equaion led o confusing resuls. The parameer γ 1 has an opposie sign han he heory suggess and he absolue value of he serilizaion coefficien λ exceeded he upper margin of he poenial inerval from zero o one. Moreover, γ 1 in all EU4 excep for Hungary are saisically insignifican. Neiher he performance of he elasiciies of he money base wih respec o he domesic oupu gap ( γ ) are significan (again, Hungary is he excepion). According o Spolander (1999, p.7) his problem sems from differen specificaion of he equaion and, unforunaely, i is a common drawback of many sudies of moneary policy rules and reacion funcions. As saed in McCallum (1997, p.8), here has been much debae on he subjec of moneary policy rules bu he appropriae specificaion of a model suiable for he analysis of moneary policy rules does no exis. 13

15 Table 3: Esimaes of equaion (17) Czech Republic Hungary gap rend gap gap insrumens: p y r 1 i i 1 y 1 insrumens: y i 1 c 1 y 1 i 1 param. esim. s.er. prob. param. esim. s.er. prob. γ γ λ λ γ γ γ γ E R =0.3475, SEE=0.008, DW=.0566 R =0.6565, SEE=0.0119, DW=.5815 J-B=8.14 (0.0000), LM= (0.9739) WHITE= (0.0000) J-B=38.06 (0.0000), LM= (0.0095) WHITE= (0.7050) Poland Slovakia rend gap rend gap insrumens: m 1 p y 1 p 1 y insrumens: y 1 c i 1 y 1 i 1 param. esim. s.er. prob. param. esim. s.er. prob. γ γ λ λ γ γ γ γ R =0.7436, SEE=0.01, DW=.4557 J-B= (0.0007), LM= (0.0616) WHITE= (0.6780) Source: Auhor s calculaions R =0.951, SEE=0.04, DW=.400 J-B=.8966 (0.349), LM=9.11 (0.0558) WHITE=5.178 (0.8189) The parameer esimaes of he serilizaion coefficiens λ in all EU4 excep for Hungary do no significanly differ from minus uniy, which implies full serilizaion. 7 However, he EU4 cenral banks have never publicly declared ha all foreign exchange inervenion has no impac on he money base. Hence, we assume ha he parameer esimaes of λ indicae less han full serilizaion. This assumpion is in accordance wih he pracice of cenral banks from developed counries which usually serilize heir inervenion parially raher han fully. Table 4 summarizes esimaes of he conversion facors η calculaed for all counries using equaion (1). Due o non-sandard resuls of he esimaion of equaion (17) in Poland, he Polish conversion facor differs subsanially from oher facors in magniude as well as sign. The exraordinary value of Polish η is subsequenly ransmied o EMP whose exen will no correspond wih he EMP scale in oher EU4. 7 The Wald es of he null hypohesis λ =-1 resuled in he following F-saisics and probabiliies. Czech Republic:.7181 (0.1055), Hungary: 7.08 (0.0098), Poland: (0.1955), Slovakia: (0.4639). 14

16 Table 4: Esimaes of conversion facors Czech Republic Hungary Poland Slovakia Source: Auhor s calculaions The EMP developmen according o model-dependen approach is graphically presened for all counries analyzed in Appendix. To evaluae EMP correcly i is necessary o remember some elemenary facs. Firs, a negaive value of EMP indicaes ha he currency is under general pressure o appreciae. On he conrary, posiive EMP shows ha he currency is pressured o depreciae. Second, he value of EMP represens he magniude of he foreign exchange marke disequilibrium which should be removed by a respecive change of he exchange rae. The figures conain, besides he EMP curve, he lines represening 1.5 muliple of he sandard deviaion above and below he mean EMP value. A breach of he corridor is considered as an excessive EMP, alering o a poenial crisis. Furhermore, he graphs are divided ino several secions, hus allowing he disinguishing of differen exchange rae arrangemens applied in EU4 during he period examined. One can find he EMP developmen in EU4 as alike in many aspecs. The firs hree years were characerized by many episodes of excessive EMP and is high volailiy. The EMP esimaes sugges ha here was a general pressure on EU4 currencies o depreciae. The principal excepion was Poland whose EMP measuremens surpassed 60% on he appreciaion side in five quarers during I is very hard o believe ha he magniude of money marke disequilibrium would be so enormous ha he Polish zloy (PLN) should have appreciaed by 60% in order o remove ha disequilibrium noing he sill saring sage of he ransformaion process. Moreover, Vannese e al. (005) as well as Bielecki (005) obained considerably differen (and more realisic) esimaions of EMP in Poland in ha period. I is worhwhile o remember ha all EU4 counries applied some version of fixed exchange rae regime in Furhermore, he Czech Republic and Slovakia sared heir exisence in January 1993 afer he spli of former Czechoslovakia. The relaed currency separaion, launch of new currencies, esablishmen of new cenral banks, and formaion of new moneary policies had an obvious impac on daa used in he esimaion and consequenly on he EMP figures. Since 1996, EMP developed more smoohly and free of any abnormal flucuaions. There was only one example of breaching he corridor s margin afer In Hungary, EMP in 00:1 was -1.96% suggesing a pressure on he forin (HUF) o appreciae. A high (no 15

17 excessive) EMP also occurred a he end of 00. HUF was under speculaive aack on he upper edge of he band which culminaed in devaluaion of he cenral pariy. In he Czech Republic, he highes EMP was idenified in 00: when he pressure reached 1.4% forcing he koruna (CZK) o depreciae. This refleced he necessiy for a correcion afer he previous long-lasing appreciaion and peaking a he hisoric high. Whereas he depreciaion pressure prevailed on HUF and Slovak koruna (SKK) he proporion of appreciaion-pressure and depreciaion-pressure quarers was more balanced in he case of CZK in Model-Independen Approach The EMP values obained from he model-independen approach are subsanially differen from he model-dependen ones (see Appendix 3 for graphical illusraion). They differ in magniude as well as basic developmen endencies. Since he comprehensive comparison of alernaive resuls is provided in Secion 4.3 we only focus here on descripion of he mos noable feaures of he model-independen EMP. None of he counries analyzed experienced exraordinary volaile developmen of EMP in he fis hree or four years of he period examined. Far from i, he developmen in he Czech Republic and Poland a ha period of ime was he mos sable ever. Furhermore, one can find many episodes of he excessive EMP in all counries during he second half of he period analyzed. Generally, he crisis quarers (EMP surpassing upper or lower limi) seem o occur more frequenly in he model-independen han model-dependen approach. I is obvious as he no-crisis band in he model-independen approach is considerably igher han he model-dependen band in hree counries. However, all breaches of he limis are very emporary and, hus, he foreign exchange marke disequilibrium did no las more han one observaion (quarer). I is worh o menion a similariy in he very recen EMP developmen in hree counries (Hungary, Poland and Slovakia). The pressure exceeded or came near he lower limi a he end of 006 announcing he appreciaion pressure on he naional currencies. Whereas he appreciaion pressure prevailed during he enire period in he Czech Republic and Poland, he more balanced proporion of posiive and negaive EMP observaions was revealed in Slovakia. Hungary, on he oher side, had o face predominanly a depreciaion pressure on HUF. The mos exreme EMP in he Czech Republic (+13.39%) can be observed in 004:. Such a high depreciaion pressure was caused by increase of he Czech ineres rae above he 16

18 Eurozone level and subsequen change in he ineres rae differenial (+10%). In Poland, we deeced he mos exreme EMP in 005:4 (-0.56%). A separae analysis of he EMP componens allows us o deermine he principal cause. I is a subsanial change in he reserves-money raio (+10.9%) driven by a massive increase in reserve holdings. Slovakia is he counry wih he mos numerous escapes from he no-crisis band, mainly on he appreciaion side. However, he breaches of limis are raher marginal and he mos significan one was recorded in 005:1 (-7.75%) as a consequence of growing inernaional reserves. Slovakia also winessed a high depreciaion pressure (+9.91%) in 1998:4, jus afer he shif in he exchange rae arrangemen owards managed floaing. In Hungary, we can disinguish, ignoring he very early sage, wo cases of excessive depreciaion EMP. The firs one (+11.19%) occurred in 003:3 following culminaion of he speculaive aack on appreciaing HUF. In 005:1, EMP reached even higher level (+13.78%) foreseeing he coming period of massive HUF depreciaion. 4.3 Comparison of Alernaive Approaches As saed above, he alernaive empirical approaches o he EMP esimaion resuled in considerably differen findings. I can be documened by descripive saisics of he EMP ime series as well as correlaion analysis. The elemenary descripive saisics is presened in Table 5 and correlaion coefficiens of he EMP measures in Table 6. Table 5: Descripive saisics of exchange marke pressure Czech Republic Hungary Poland Slovakia m_dep m_ind m_dep m_ind m_dep m_ind m_dep m_ind mean median max min s. dev upper lower Source: Auhor s calculaions Noes: m_dep and m_ind denoe model-dependen and model-independen approach respecively The only counry wih resuls signaling some degree of consisency is Hungary. Means and medians of boh EMP indices have posiive signs and, furhermore, he correlaion coefficien is he highes among all counries. One can find a furher uniqueness in Hungarian 17

19 resuls. Developmen of he model-dependen EMP was significanly less volaile han developmen of he alernaive model-independen EMP. This is eviden in sandard deviaion, widh of he no-crisis band and spread beween maximum and minimum values. Toally opposie conclusions can be drawn on remaining counries. Their mos noable common aribues are higher volailiy of he model-dependen EMP and disharmonic developmen of he EMP measures mirrored in he reversely signed means and medians and low and/or negaive correlaion coefficiens. 8 We should remind ha he high sandard deviaions and wide bands sem from he varying developmen in he very early sage of he esimaion period. Table 6: Correlaion coefficiens of alernaive exchange marke pressure measures Czech Republic Hungary Poland Slovakia Source: Auhor s calculaions The consisency of he wo EMP indices can also be assessed by discrepancies in idenificaion of crisis. For ha purpose, Figures 1-4 puing developmen of he boh EMP measures ogeher are presened. Moreover, Table 7 shows how many quarers were idenified by he model-independen approach as crisis and how many of which would be similarly classified by he model-dependen approach if he model-independen no-crisis band applies. Table 7 also repors number of EMP crisis observaions ha were awarded wih same sign and similar magniude by boh approaches. The resuls presened confirm he negligible consisency and provide evidence ha he empirical ools used end o inerpre EMP developmen differenly. Table 7: Consisency of alernaive approaches in idenificaion of crises Czech Republic Hungary Poland Slovakia m_ind crises m_dep crises same sign 6 3 similar magniude Source: Auhor s calculaions Noes: m_dep and m_ind denoe model-dependen and model-independen approach respecively. Similar magniude means ha value of he m_dep EMP is wihin inerval 50%- 150% of he m_ind EMP value. 8 We also carried ou a cross-border correlaion analysis ha revealed generally higher correlaion beween model-dependen indices. The highes correlaion coefficien can be observed beween Czech and Slovak EMP (0.5456) and he mos negaive beween Hungarian and Polish EMP ( ). Correlaion analysis of he model-independen EMP yielded o considerably lower coefficiens. The highes one (0.867) was calculaed beween Hungary and Poland. 18

20 Figures 1-4: Developmen of exchange marke pressure based on model-dependen and model-independen approaches Czech Republic Hungary EMP_DEP Poland EMP_INDEP EMP_DEP Slovakia EMP_INDEP EMP_DEP EMP_INDEP EMP_DEP EMP_INDEP Source: Auhor s calculaions 19

21 One of he aims of he paper is o compare EMP in various exchange rae arrangemens in EU4 keeping in mind he necessiy o ener ino ERM II (a quasi-fixed regime wih a flucuaion band) and fulfill he exchange rae sabiliy crierion in EU4 in he near fuure. The comparison of he EMP sandard deviaions calculaed over he periods wih paricular exchange rae regime along wih number of crisis quarers is provided in Table 8. Table 8: Sandard deviaions of exchange marke pressure and number of crisis quarers Czech Republic Hungary Poland Slovakia m_dep m_ind m_dep m_ind m_dep m_ind m_dep m_ind fixed (crawl) peg crawling band floaing (0:0) (0:0) (0:0) (3:1) (:0) (0:0) (0:1) (3:0) (0:0) (:1) (0:) (0:0) (0:0) 0.01 (0:0) (1:0) (0:) ERM II (:0) (1:1) (0:0) (0:0) (1:3) (0:1) Source: Auhor s calculaions Noe: The raio in parenheses is (number of excessive depreciaion EMP : number of excessive appreciaion EMP) The resuls clearly sugges ha any conclusion abou he relaionship beween EMP and exchange rae regime is exremely sensiive on selecion of he EMP esimaion mehod. Model-dependen and model-independen approaches lead o absoluely conroversial findings on how EMP develop and flucuae in paricular exchange rae arrangemen. The modeldependen approach provides evidence ha EMP during he floaing-regime period was very sable in all EU4 and he excessive deviaions of EMP occurred sporadically. On he conrary, he periods of fixed arrangemen winessed numerous episodes of surpassing he 1.5 muliple of sandard deviaion level along wih he more volaile developmen. The resuls of he model-independen approach are oally opposie. Generally, any kind of he fixed regime paved he way for boh lower and less volaile EMP and fewer crisis periods. To deermine wheher he differences among EMP values in various exchange rae regimes are saisically significan we carried ou a single-facor Analysis of Variance (ANOVA). The EMP observaions from all EU4 were gahered in he single daase and grouped ino four caegories according o classificaion sysem applied in Table 8. The ANOVA es resuls for model-dependen as well as model-independen approach are repored in Table 9. 0

22 Table 9: ANOVA es resuls model-dependen approach model-independen approach no. observ. mean variance no. observ. mean variance fixed (crawl) peg crawling band floaing ERM II F-saisics: P-value: F-saisics: P-value: Source: Auhor s calculaions Noe: Criical value of F-saisics is The ANOVA ess show ha he exchange rae regime does no influence he average of he model-dependen EMP considerably as he F-saisics is small and insignifican. On he oher hand, he means of he grouped model-independen EMP are significanly differen a 1% level. Thus, one can consider he floaing arrangemen as he environmen conribuing o he volaile developmen and excessive values of EMP. The resuls obained allow us o derive some policy implicaions. There is no empirical jusificaion for he a priori concerns ha a shif in he exchange rae regime from floaing o he quasi-fixed ERM II would simulae EMP o increase. More likely, he basic characerisics of he EMP developmen will be reained afer he change. Thus, supposing ha he recen level of EMP volailiy and densiy of crisis observaion revealed by he modelindependen approach will remain unchanged, i would cas serious doub on he European Commission s requiremen ha EU4 mus paricipae in ERM II wihou subsanial ensions on he exchange raes. The doub gains imporance if he acual European auhoriies posiion o he fulfillmen of he exchange rae sabiliy crierion is considered as decisive. When all relevan saemens and declaraions of he ECB, European Commission and Ecofin Council are summarized and combined wih he approaches o he crierion assessmens applied in he pas, we can consider he use of he ERM II sandard flucuaion band of ± 15% as highly improbable. On he conrary, he auhorized flucuaion margin is likely o be asymmeric wih he limis of 15% on he appreciaion side and.5% on he depreciaion side. Alhough EMP flucuaed predominanly wihin his narrow band in EU4 in he las four years one can expec ha all EU4 will have o face excessive EMP from ime o ime when paricipaing in ERM II and fulfilling he exchange rae sabiliy crierion. Since he 1

23 depreciaion par of he asymmeric band is very igh he EMP developmen should be moniored carefully. The depreciaion EMP would principally require some policy acions aken by he cenral bank in he form of foreign exchange inervenions or any oher suiable insrumen. Owing o some facors he EMP esimaes presened and discussed previously mus be viewed wih some degree of skepicism. There are several drawbacks of he model-dependen approach which mus be aken ino accoun when inerpreing he resuls obained. Firs, many parameer esimaes required for calculaion of he conversion facor and EMP are saisically insignifican. I can be aribued o eiher wrong specificaion of he model or some problems wih esimaion procedure or daa used. Second, he parameer esimaes are sensiive o he choice of insrumens and even small changes in he parameers values have a considerable impac on EMP. Third, he serilizaion coefficien in all EU4 excep for Hungary is no significanly differen from minus uniy which indicaes full serilizaion. Fourh, modeldependen EMP in all counries behaved almos absoluely parallel o changes in reserves during he enire period. 9 I implies a frequen applicaion of he cenral bank official inervenion even in he environmen of floaing exchange rae regime. The realiy in many EU4 was, however, differen. These limiaions should be eliminaed in fuure research. We recommend use of he pure foreign exchange inervenion daa as an alernaive o he change in reserves. I could lead o more plausible resuls as eviden in Bielecki (005). The model could also be exended by he possibiliy of indirec inervenion operaing hrough changes in domesic lending or he domesic ineres rae. 5. Conclusion In his paper, we esimaed EMP for he EU4 currencies agains he euro exchange rae over he period from We concurrenly applied he Spolander (1999) model based on he Weymark (1995) model-dependen approach and model-independen approach based on Eichengreen e al. (1994, 1995) and Sachs e al. (1996). Fundamenal differences in spiri and consrucion of hese approaches are refleced in considerably differen resuls. Thus, he principal conclusion of his sudy is a clear saemen ha he wo empirical approaches applied are no compaible if daa from EU4 are used. The wo alernaives differ subsanially 9 This finding is based on resuls of a variance analysis of EMP and is componens.

24 in assessmen of he general developmen rends and endencies as well as magniude and direcion of EMP in paricular quarers. According o model-dependen approach, EMP in he Czech Republic, Hungary and Slovakia is of similar magniude. Whereas a depreciaion pressure prevailed on HUF and SKK, no dominance of any direcion of he pressure can be found in he case of CZK. The esimaes of he Polish EMP are burdened by subsanial saisical insignificance. The resuls obained sugges ha EMP in EU4 decreased over ime and was subsanially lower and less volaile during he periods of floaing exchange raes han in he environmen of fixed exchange rae regime. However, here are some concerns abou he validiy of he parameer esimaes and consequenly he EMP measures in all EU4. Therefore, we can conclude ha he Spolander (1999) model in is pure version is no fully suiable for EU4 and we recommend some exensions and propose furher seps for fuure research. The model-independen approach is no burdened by he above menioned drawbacks. Moreover, his approach pus greaer emphasis on he ineres rae differenial, which has ofen been idenified as a facor of exchange rae deerminaion in EU4. Thus, resuls of he model-independen approach can be considered as more reliable. The EMP developmen can be described as homogeneous during he enire period analyzed wihou any sage of abnormal volailiy or excepionally frequen occurrence of excessive EMP. While CZK and PLN were largely under appreciaion pressure, HUF was forced o depreciae and no dominance was revealed in Slovakia. However, he model-independen approach idenified more crises han model-dependen approach including he very recen excessive pressure on appreciaion of hree EU4 naional currencies. The sudy did no confirm he concerns ha he unavoidable shif in he exchange rae regime owards he quasi-fixed ERM II could evoke EMP o grow o excessive levels. Insead, he empirical ess sugges ha he regime change will have, wih high probabiliy, a negligible impac on he EMP developmen. However, he asymmeric flucuaion band which is likely o be applied for he assessmen of he fulfillmen of he exchange rae sabiliy crierion seems o be very igh on he depreciaion side. Semming from esimaions obained, he EU4 cenral banks will be probably confroned wih occasions of excessive EMP jeopardizing fulfillmen of he exchange rae sabiliy crierion. 3