Munich Personal RePEc Archive
Exchange Market Pressure in Central European Countries from the Eurozone Membership Perspective
Stavarek, Daniel
Silesian University - School of Business Administration
10 November 2008
Online at https://mpra.ub.uni-muenchen.de/12079/
MPRA Paper No. 12079, posted 12 Dec 2008 10:43 UTC
November 2008 7
E xchange M arket P ressure in C entral E uropean C ountries from the E urozone M embership Perspective*
Exchange Market Pressure in Central European Countries from the Eurozone Membership Perspective
Daniel Stavarek*
Abstract:
Thispaperestimatestheexchangemarketpressure (E M P ) infour Central Europeancountries (Czech Republic, Hun- gary, Poland, Slovakia) overtheperiod 1 993 -2 006. Therefore, itisoneofveryfewstudiesfocusedonthisregionand theveryfirstpaperapplyingconcurrentlymodel-dependentaswellasmodel-independentapproachesto E M P esti- mationtothesecountries. Theresultsobtainedsuggestthattheapproachesleadtoinconsistentfindings. Theyoften differinidentificationoftheprincipaldevelopmenttrends, aswellasthemagnitudeanddirectionofthepressure. Thepaperprovidesevidencethatashiftintheexchangerateregimetowardsthequasi-fixed E R M II shouldnotstimu-
late E M P growth. However, itishighlyprobablethatsomeepisodesoftheexcessive E M P willmakethefulfillmentof
theexchangeratestabilitycriterionmoredifficultinallofthecountriesanalyzed.
Keywords: Exchange Market Pressure, Model-dependent Approach, Model-independent Approach, European Union, Euro-candidate Countries JEL: C32, E42, F31, F36
*Stavárek:
Silesian University in Opava, School of Business Administrationin Karviná,
Department of Finance Univerzitnínám. 1934/3, 733 40 Karviná,
Czech Republic
e-mail: stavarek@opf.slu.cz
1. Introduction
Tencountriesfrom Centraland Eastern Europejoined the European Union (hereafter EU) inthespringof 2004 andin 2007 completedthetransformationfromcentrally plannedeconomiestomarketeconomies. Moreover, itis expectedthattheywillalsojointhe Eurozoneandimple- mentthe Euroastheirlegal tender. However, member- shipinthe Eurozoneisconditionedbyfulfillmentofthe Maastrichtcriteria. Onecriterionofwhichisthenational currency’sstabilityintheperiodprecedingentryintothe Eurozone.
Thiscriterionisassociatedwithspecificexchangeratere- gime, ERM II, whichmustbeadaptedbyallcountrieswith regimeswhoseprinciplesdonotcorrespondwiththe ERM II’s spirit. The group of incompatible regimes includes crawlingpegs, freefloatsormanagedfloatswithoutamu- tuallyagreedcentralrateandpegstoanchorsotherthan the Euro. Itmeansthatall EU new Member Statesexceptfor
Bulgaria, Estonia and Lithuania had or will have to modify their exchange rate arrangement when joining ERM II. The Czech Republic, Hungary, Poland and Romania currently use flexible exchange rate arrangements. Slovakia and to a lesser extent Slovenia also maintained a flexible regime before entry into the ERM II. Such a change toward a less flexible exchange rate system could increase susceptibility of the countries to currency crises and pressures in foreign exchange markets.
Acknowledgement
Research behind this paper was supported by the Czech Science Foundation within the project GAČR 402/08/0067
“Financial Integration of the New EU Member States with Eurozone”.
DOI: 10.2478/v10033-008-0010-z
Therefore, the aim of this paper is to estimate exchange market pressure (EMP) in the Czech Republic, Hungary, Poland and Slovakia (hereafter EU4) during the period 1993-2006. Since all countries applied both a fixed and flexible exchange rate regime, the time span chosen al- lows us to compare magnitude of tensions in the foreign exchange market in differentexchange rate environ- ments. This kind of analysis has important policy implica- tions as Slovakia has already switched to a less flexible regime and the remaining countries will make this un- avoidable step in the near future.
The paper is structured so that Section 2 describes the meaning and theoretical concepts of EMP and provides a review of the relevant literature. In Section 3, the models and data used are cited. Section 4 reports the empirical results and the conclusions are presented in Section 5.
2. E xchange M arket P ressure and L iterature R eview
2.1 M eaning and C oncepts of E xchange M arket P ressure
Theterm “exchangemarketpressure” usuallyrefersto changesintwocardinalvariablesdescribingtheexternal sectorofanyeconomy: officialinternationalreservehold- ingsandthenominalexchangerate. However, thenotion of EMP wasdefinedexplicitlyforthefirsttimein Girtonand Roper (1977). The EMP indexinthisstudyisthesimplesum oftherateofchangeininternationalreservesandtherate ofchangeintheexchangerate. However, sincethemea- sureisderivedfromahighlyrestrictivemonetarymodel, theformulacannotbeappliedtoothermodels.
Theoriginal conceptof EMP hasbeen modified and extended bymanyresearchers. Forexample, Roperand Turnovsky (1980) and Turnovsky (1985) introduced the ideaofusingasmallopen-economymodelandextended the originalmodelby substitutingthesimplemonetary approachbyan IS-LM frameworkwithperfectmobilityof capital. Furthermore, thetwo EMP componentswereno longerequallyweightedasinthe Girton-Ropermodel. A notablecontributiontothe EMP theorywasprovided by Weymark (1995, 1997a, 1997b, 1998). Sherevisedthe modelsmentionedaboveandintroducedamoregeneral frameworkinwhichthemodelsarebothspecialcasesof the generalized formula. She introduced and estimated aparameter (conversionfactor) standingfortherelative weightofexchangeratechangesandinterventioninthe EMP index. Since all previous EMP definitionsstemmed fromaspecificmodel, Weymarkalsoproposedamodel- independentdefinitionof EMP as:
Theexchange ratechangethatwould havebeen re- quiredtoremovetheexcessdemandforthecurrencyin theabsenceofexchangemarketintervention, giventhe
expectations generated by the exchange rate policy actu- ally implemented (Weymark 1995, p.278).
Many researchers have criticized the most undesirable aspect of the EMP measure, dependency on a particular model, and proposed some alternative approaches called model-independent. A simpler and model-independent EMP measure was originally constructed in Eichengreen et al. (1994, 1995). According to this approach, EMP is a lin- ear combination of a relevantinterest rate differential, the percentage change in the bilateral exchange rate and the percentage change in foreign exchange reserves. Contrary to Weymark’s approach, the weights are to be calculated from sample variances ofthose three components with no need to estimate any model.
The measure by Sachs et al. (1996) consists of the same elements, but each weight in the EMP index is calculated with respect to standard deviations of all components in- cluded instead of using only standard deviation of the re- spective component.
Kaminsky etal. (1998) and Kaminsky and Reinhart (1999) substituted the interest rate differential by a relevant interest rate in the country analyzed. Furthermore, the weights on the reserves and interest rate terms are the ratio of the standard error of the percentage change of the exchange rate over the standard error of the percent- age change of reserves and the interest rate differential, respectively. An approach stemming from Eichengreen et al. (1996) was also followed by Pentecost et al. (2001).
However, they determined the weights using principle components analysis.
2.2 R eview of R elevant E mpirical L iterature
Sinceitsintroduction, EMP hasattractedtheattention ofmanyresearchersandagreatnumberoftheoreticalas wellasempirical papershavebeen published. Whereas someempiricalpapersarefocusedstraightonestimation of EMP inavarietyofregionsandcountries, otherstudies usethe EMP measureasanelementofasubsequentanal- ysisexamining currencycrises, monetary policy, foreign exchangeintervention, exchangerateregimeandother issues. Weonlyrefertostudiesanalyzing EMP in EU4 inthe followingliteraturereview.
The firststudy estimating EMP in, amongothers, the Czech Republicand Poland, wasby Tanner (2002). Using the Girton-Ropermodel, heexaminedtherelationshipbe- tween EMP andmonetarypolicyinavectorautoregression system. Regardingthe EMP calculatedinthe Czech Repub- licand Poland, theyweremodestincomparisontoother countriesandverysimilartoeachother. However, EMP in Polandwastwelvetimeshigherthaninthe Czech Republic duringthe Asiancrisisinthesecondhalfofthe 1990s. Al- thoughapositiverelationshipbetween EMP anddomestic moneysupplywasrevealedinbothcountries, theywere notassignificantandstraightasinothercountries.
Exchange Market Pressure in Central European Countries from the Eurozone Membership Perspective
November 2008 9
A more specific application of the Tanner (2002) approach is Bielecki (2005). The paper concentrates only on Poland from 1994-2002. The results indicate that domestic credit reacted in a direction counter to innovations to EMP. Fur- thermore, Bielecki compared two EMP measures calculated under alternative methodologies (using all foreign reserve changes and pure official foreign exchange intervention data) and came to the conclusion that the appreciation pressure prevailed over the sample period. However, us- ing the pure intervention data in the EMP estimation pro- vided more realistic and robust results.
Van Poeck et al. (2007) used EMP as an indicator of cur- rency crisis and addressed the question whether currency crises in the Euro-candidate countries have been more fre- quent in fixed, intermediate or flexible exchange rate ar- rangements. The authors found that EMP was marginally smaller in countries and periods characterized by an inter- mediate exchange rate regime as compared to those with a floating arrangement. Regarding EU4, the most critical quarters (excessive EMP) occurred in Hungary during the fixed peg regime and in Poland when a crawling peg was being applied.
Very similar conclusions were drawn in Stavárek (2005) where EMP in the Czech Republic, Hungary, Poland and Slovenia in 1993-2004 are estimated. The study applied the EMP measure proposed in Eichengreen et al. (1995) and the results obtained suggest that the Czech Republic and Slovenia went through considerably less volatile de- velopmentof EMP than Hungary and Poland.
3. M easuring the E xchange M arket P ressure : M odel and D ata
3.1 M odel -D ependent A pproach
This study originally stems from Weymark (1995) and Spolander (1999) and applies the following formula for EMP calculation:
t t
t
e r
EMP = Δ + ( 1 − ) Δ
(1) whereΔ e
t is the percentage change in exchange rate expressed in direct quotation (domestic price for one unit offoreign currency),Δ r
t is the change in foreign exchange reserves scaled by the one-period-lagged value ofmoneybaseand
istheconversionfactorwhichhasto beestimatedfromastructuralmodeloftheeconomyandistheproportionofforeignexchangeinterventionthat issterilizedbyachangeofdomesticcredit.
Theconversionfactorrepresentselasticitythatconverts observedreservechangesintoequivalentexchangerate units. This EMP formulaassumesthatthecentral bank’s monetary policyis completelyindependentofdemand andsupply conditionsforthe domesticcurrency inthe
international foreign exchange market. This means that autonomous money market interventions, i.e. changes in domestic credit not due to sterilization operations, are not assumed to be an instrument of exchange rate policy (Spolander 1999, p. 23).
For practical estimation of EMP the small open economy monetary model summarized in equations (2)-(8) was applied.
t t
t d
t
p c i
m = + Δ + Δ − Δ
Δ
0
1
2 (2)t t
t
p e
p = + Δ + Δ
Δ
0
1 *
2 (3)t t
t
t
E e e
i
i = Δ + Δ − Δ
Δ
*(
+1)
(4)t a
t s
t
d r
m = Δ + − Δ
Δ ( 1 )
(5)t t
t
p e
r = − Δ
Δ
(6)gap t t
trend t a
t
y p y
d =
0+ Δ + ( 1 −
1) Δ −
2Δ
(7)s t d
t
m
m = Δ
Δ
(8) wherep
tis domesticpricelevel,p
t*
isforeign price level,
e
tdenotes exchange rate (in direct quotation),
m
tisnominalmoneystock (thesuperscript
d
representsthe demand ands
thesupply),c
t isrealdomesticincome,i
t is nominal domestic interest rate,i
t* denotes nominal foreign interest rate,( Δ
+1)
t
t
e
E
is expected exchangerate change and
is proportion ofsterilized intervention.All variablesup to thispoint are expressed in natural
logarithm. Next, dta is autonomous domestic lending by thecentralbankand
r
t isthestockofforeignexchange reserves, bothdividedbytheoneperiodlaggedvalueof the money base.y
ttrend is the long-run trend component of real domestic outputy
t and ytgap is the difference betweeny
t andy
ttrend. The signΔ
naturally denotes change in the respective variable.Equation (2) describes changes in money demand as a positive function of domestic inflation and changes in real domestic income and a negative function of changes in the domestic interest rate. Equation (3) defines the purchasing power parity condition attributing the primary role in domestic inflation determination to exchange rate changes and foreign inflation. Equation (4) describes uncovered interest rate parity. Equation (5) suggests that changes in the money supply are positively influenced by autonomous changes in domestic lending and non- sterilized changes in the stock of foreign reserves. Equation (6) states that changes in foreign exchange reserves are a function of the exchange rate and a time-varying response coefficient. Equation (7) describes the evolution
Exchange Market Pressure in Central European Countries from the Eurozone Membership Perspective
of the central bank’s domestic lending. Whereas domestic inflation and changes in trend real output changes are positive determinants of the domestic lending the gap between real output and its trend has a negative impact on domesticlending activity. Equation (8) definesa money marketclearing condition thatassumes money demand to be continuously equal to money supply.
By substituting equations (3) and (4) into equation (2) and substituting equation (7) into equation (5) and then using the money marketclearing condition in equation (8) to set the resulting two equations equal to one another, it is possible to obtain the following relation:
2 2 1
1
2
( ) ( 1 )
+
Δ
− + Δ
= +
Δ
tX
tE e
t+r
te
(9)where
* 2 1 2
* 1 1 0
0 1
0 t t
gap t t trend
t
t y p y c i
X = − − +Δ −Δ − −Δ +Δ (10)
andtheelasticityneededtocalculate EMP inequation (1) canbefoundas:
2 2 1
) 1 (
+
− − Δ =
∂ Δ
− ∂
=
t t
r
e
(11)3.2 M odel -I ndependent A pproach
As mentioned above, Eichengreen etal. (1994, 1995) argued that dependency on a particular model wasan undesirable feature for the EMP index. As an alternative, they proposed the following measure of a speculative pressure:
)) ( 1 (
1
**
*
t t i t
t t
t r t
t
t
i i
rm rm rm
rm e
EMP e ⎟ ⎟ + Δ −
⎠
⎞
⎜ ⎜
⎝
⎛ Δ
Δ − Δ −
=
(12) where
ris the standard deviation of the difference between the relative changes in the ratio of foreign reserves and money (money base) in the analyzed countryand the reference country
⎟⎟⎠
⎞
⎜⎜⎝
⎛ Δ
Δ −
*
*
t t t
t
rm rm rm
rm and
i is thestandard deviation of the nominal interest rate differential
( Δ ( i
t− i
t*))
. Othervariablesareasdefinedin thepreviousspecification.However, forthepracticalcalculationwetookinspiration from Sachs etal. (1996) and made somemodifications ofthe EMP formula. Inordertoavoid the EMP measure
being driven by the most volatile component we changed the weighting scheme. We also abandoned the relation between foreign reserves and money at the home and reference country. Consequently, the EMP formula based on a model-independent approach can be written as follows:
(13) where
eis the standard deviation ofthe rate of change in the exchange rate and other variables are denoted consistently with (12).The samples of data used in this paper cover the period 1993:1 to 2006:4, yielding 56 quarterly observations for all EU4 countries. The data were predominantly extracted from the IMF’s International Financial Statistics and the Eurostat’s Economy and Finance database. The missing observations in the time series were replenished from databases accessible on the EU4 central banks’ websites.
The detailed description ofall data series and their sources is presented in Appendix 1.
4. E stimation of E xchange M arket P ressure 4.1 M odel -D ependent A pproach
As is evid ent fromthe m od el p resented in Section 3.1, the EMP estim ation (1) m ust b e p reced ed b y the calculation of the conversion factor
(11). However, this step is required to ob tain values ofthe sterilization coefficient
(5), theelasticityofthem oneyb asewithresp ecttothe d om estic p rice level 1 (7), the elasticity of the d om estic p rice level with resp ect to the exchange rate 2 (3), and the elasticity of the m oney d em and with resp ect to the d om esticinterestrate 2 (2).Morep recisely, thep aram eterestim atesareob tainedb y estim atingthefollowingthreeequations.
t t t
t
t
p c i
m − Δ =
0+
1Δ −
2Δ +
1,Δ
(1 4)t t t
t p e
p =0+1Δ *+2Δ +2,
Δ (1 5)
t gap t t t t
trend t t t
t r y p r p y
B B
, 3 2
1 0
1
+ Δ + Δ + +
= Δ
− Δ
− Δ Δ −
−
(1 6)
Equations (14) and (15) are ob tained d irectly from equations (2) and (3). Equation (16) is d erived b y sub stitutionof (6) into (4) andnotingthatchangeinm oney
Exchange Market Pressure in Central Europ ean Countries from the Eurozone Mem b ership Persp ective
)) ( )) ( 1 ( ) 1 ( ) 1 ((
1
)) 1 ( ) 1 ( ) 1 ((
1 ))
1 ( ) 1 ( ) 1 ((
1
*
1 1
t t i rm e
i
t t i rm e
rm t
t i rm e
e t
i i
rm rm e
EMP e
−
⎟⎟Δ
⎠
⎞
⎜⎜⎝
⎛
+ + +
Δ +
⎟⎟⎠
⎞
⎜⎜⎝
⎛
+
− + Δ
⎟⎟⎠
⎞
⎜⎜⎝
⎛
+
= +
−
−
−1
Δ
t t
e e
Novem b er 2008 11
sup p ly equals the change in m oney b ase assum ing the m oney m ultip lier tob e constant.
One can d istinguish two typ es of variab les includ edin the m od el: end ogenous and exogenous. The end ogenous variab les are Δmt ,
Δ p
t ,Δ e
t ,Δ i
t ,−1
Δ
t t
B
B
andΔ r
t . Theexogenous variab les are
Δ c
t, *p
tΔ
,Δ i
t*, trendyt
Δ and
gap
yt
Δ . Desp itethefactthat
Δ e
t d oesnotap p earontheleft-handsid eofanyoftheequations, itistheend ogenous variab leb ecausetheexchangerateisclearlythevariab le d eterm inedb ythism od el.
Them od elisestim atedusingthetwo-stageleastsquare regression technique (2SLS). The m ain reason isthatthe end ogenousvariab lesareon b othsid esofequations (2)- (8). It m eansthat in each equation having end ogenous variab les on the right-hand sid e, these variab les are likelyto correlatewith thed isturb anceterm. Thus, using the ord inary least square m ethod would lead to b iased estim ates.
The 2SLS usedrequirestheincorp orationofinstrum ents (variab lesuncorrelatedwiththed isturb anceterm) intothe estim ation. To find ap p rop riate instrum ents we run the firststageregressionsonend ogenousvariab leshavingall p ossib leinstrum entsasregressors. Asp ossib leinstrum ents we set the contem p oraneous and one-quarter lagged values of exogenous variab les and one-quarter lagged valuesofallend ogenousvariab les. Finally, theregressors with sufficient statistical significance were selected as instrum ents.
Weap p lied Augm ented Dickey-Fullerteststoexam ine thestationarityofthe tim eseriesused. Accord ingto the characterofeachtim eserieswetestedthestationaritywith alineartrendand/orintercep tornoneofthem. Tests’ results allowusto conclud ethatthefirstd ifferencesofalltim e seriesarestationary. Thus, theycanb eusedinestim ation ofallequationsofthe m od el. The p ercentagechangein m oney b ase is a naturally flow variab le and so alread y d ifferenced andstationary. Likewise, ytgap isstationaryon levelinallcountriesb ecauseofitsconstruction.
The 2SLS estim ationresultsarep resentedin Ap p end ix 2, ind ivid uallyforeachequation. The tab lesalso contain the list of instrum ents and results of som e d iagnostic tests. Weap p lied a Jarque-Berra (J-B) ind icatorto assess norm alityoftheresid ualsd istrib ution, a Breusch-God frey Langrange Multip lier (LM) to test serial correlation and a White test to check heterosced asticity. All LM tests were run with four lags. The tests ind icated evid ence ofserial correlation in resid ualsfrom the equations and the p otential heterosced asticity was also id entified in som ecases. Therefore, we corrected the stand ard errors of p aram eter estim ates b y the Newey-West p roced ure. Even m ore frequently, the resid uals seem to b e non- norm ally d istrib uted. Therefore, although the t-statistics can b e m islead ing, this d oes not red uce the valid ity
of the p aram eter estim ates. Since d ifferent equation sp ecifications have d ifferent instrum ents, R2for 2SLS can b e negative even if a constant is used in the equation. Accord ing to them od elsp ecificationthep aram eters
1,
1, and
2 should b e p ositive and 2,
1,
2, and
shouldb enegative. Since
is a fraction, its ab solute value shouldb e b etweenzero and one.The estim ations of equation (14) p rovid e m ed iocre results. The p aram eters 2 are correctly signed in all EU4.
However, the p aram eter is not statistically significant in Slovakia. One can see som e evid ence of non-norm al d istrib ution (Czech Rep ub lic, Slovakia), serial correlation (Hungary) and heterosced asticity (Czech Rep ub lic, Hungary and Poland).
In the estim ations of equation (15) we ob tained very good results. The signs of all p aram eters are consistent with the theoretical assum p tions and im p ortant 2 p aram eters are significantly d ifferent from zero in all countries. On the other hand, only error term s in the Polish and Slovak equations seemto p ass the stand ard d iagnostic tests com p letely. Furtherm ore, onecanfinda sub stantially lower elasticity of the d om estic p rice level with resp ect to theexchangerate (2) in Poland and, toalesserd egree, in Slovakiathaninother EU4.
The results from the m oney sup p ly equation (16) are som ewhat p oorer. This is true b ecause esp ecially the estim ation of the Polish equation led to confusing results. The p aram eter
1 has an op p osite sign than the theory suggests and the ab solute value of the sterilization coefficient
exceed edthe up p er m argin of the p otential intervalfrom zerotoone. Moreover,
1 inall EU4 excep t for Hungary are statistically insignificant. Neither the p erform ance of the elasticitiesof the m oney b ase with resp ect to the d om esticoutp utgap (2) are significant (again, Hungaryistheexcep tion). Accord ingto Sp oland er (1999, p.72) thisp rob lemstem sfromd ifferentsp ecification of the equation and, unfortunately, it is a com m on d rawb ack ofm any stud iesofm onetary p olicy rules and reactionfunctions.Theparam eterestim atesofthesterilizationcoefficients
inall EU4 exceptfor Hungarydonotsignificantlydiffer from m inus unity, which im plies full sterilization. This statem entisbased on resultsofthe Waldtestofthe null hypothesisλ =-1 However, the EU4 centralbankshavenever publiclydeclaredthatallforeignexchangeinterventionhas noim pactonthem oneybase. Hence, weassum ethatthe param eterestim atesof
indicatelessthanfullsterilization. Thisassum ptionisinaccordancewiththepracticeofcentral banksfromdevelopedcountries, whichusuallysterilizetheir interventionpartiallyratherthanfully.Exchange Market Pressure in Central Europ ean Countries from the Eurozone Mem b ership Persp ective
−1
Δ
t t
B B
Tab le 1 sum m arizes estim ates of the conversion factors
calculated forall countries using equation (11). Due to non-stand ard results of the estim ation of equation (16) in Poland, the Polish conversion factor d iffers sub stantially from otherfactors in m agnitud e as well as sign. The extraord inary value of Polish
is sub sequently transm ittedto EMP, whose extent will not corresp ond with the EMP scale in other EU4.Czech Rep ub lic Hungary Poland Slovakia 3.227 41 9 0.963507 -6.207 1 88 1 .1 67 87 4
Source: Author’s calculations
Table 1 : Estimates of conversion factors
The EMP d evelop m ent accord ing toa m od el-d ep end ent ap p roach is grap hically p resented for all countries analyzed in Ap p end ix 3. To evaluate EMP correctly it is necessary to rem em b er som e elem entary facts. First, a negative value of EMP ind icates that the currency is und er general p ressure to ap p reciate. On the contrary, p ositive EMP shows that the currency is p ressured to d ep reciate.
Second, thevalue of EMP rep resents the m agnitud e of the foreignexchange m arket d iseq uilib rium, which should b e rem ovedb y a resp ective change of the exchange rate.
The figures contain, b esid es the EMP curve, the lines rep resenting 1.5 m ultip le of the stand ard d eviation ab ove and b elow the m ean EMP value. A b reach of the corrid or is consid ered an excessive EMP, and alerts of a p otential crisis. Furtherm ore, the grap hs are d ivid ed into several sections, thus allowing one to d istinguish b etween d ifferent exchange rate arrangem ents ap p lied in EU4 d uringthep eriod exam ined.
One canfind EMP d evelop m ent in EU4 sim ilar in m any asp ects. The first three years were characterized b y m any ep isod es of excessive EMP andits high volatility. The EMP estim ates suggest that there was a general p ressure on EU4 currencies tod ep reciate. The p rincip al excep tionwas Poland, whose EMP m easurem ents surp assed 60% on the ap p reciation sid e in five q uarters d uring 1993-1995. It is very hardto b elieve that the m agnitud e of m oney m arket d iseq uilib riumwould b e soenorm ous that the Polish zloty (PLN) shouldhave ap p reciatedb y 60% in ord er to rem ove that d iseq uilib rium, noting the transform ation p rocess was still at a b eginningstage. Moreover, Van Poecket al.
(2007) and Bielecki (2005) ob tained consid erab ly d ifferent (and m ore realistic) estim ations of EMP in Poland in that p eriod.
It is worthwhile to rem em b er that all EU4 countries ap p lied som e version of fixed exchange rate regim e in 1993-1995. Furtherm ore, the Czech Rep ub lic and Slovakia started their existencein January 1993 after the sp lit of the form er Czechoslovakia. The related currency sep aration, launch of new currencies, estab lishm ent of new central
b anks, and form ation of new m onetary p olicies had an ob vious im p act on d ata used in the estim ation and conseq uently on the EMP figures.
Since 1996, EMP d evelop ed m ore sm oothly andfree of any ab norm al fluctuations. There was only one exam p le of b reaching the corrid or’s m arginafter 1995. In Hungary, EMP in 2002:1 was -1.96%, suggesting a p ressure on the forint (HUF) toap p reciate. A high (not excessive) EMP also occurred at the end of 2002. HUF was und er sp eculative attack on the up p er ed ge of the b and, whichculm inated in d evaluation of the central p arity. Inthe Czech Rep ub lic, the highest EMP was id entified in 2002:2 when the p ressure reached 12.24%, forcing the koruna (CZK) to d ep reciate. This reflected the necessity for a correction after thep revious long-lasting ap p reciation and p eaking at the historic high. Whereas the d ep reciation p ressure p revailed on HUF and the Slovak koruna (SKK), the p rop ortion of ap p reciation-p ressure and d ep reciation- p ressure q uarters was m ore b alanced in the case of CZK in 1996-2006.
4.2 M odel-Independent A pproach
The EMP valuesob tainedfromthem od el-ind ep end ent ap p roach are sub stantially d ifferent from those of m od el-d ep end ent ones (see Ap p end ix 4 for grap hical illustration). They d iffer in m agnitud e as well as b asic d evelop m enttend encies.
None of the countries analyzed exp erienced extraord inarily volatile d evelop m entof EMP in the first threeorfouryearsofthep eriodexam ined. Farfromit, the d evelop m entinthe Czech Rep ub licand Polandoverthat p eriod oftim e wasthe m ost stab le ever. Furtherm ore, one can find m any ep isod es of excessive EMP in all countriesd uringthesecondhalfofthep eriodanalyzed. Generally, the “crisisq uarters” (EMP surp assing up p eror lowerlim it) seemtooccurm orefreq uentlyinthem od el- ind ep end ent than m od el-d ep end ent ap p roach. This should b eob vious, asthe “no-crisis” b andin them od el- ind ep end entap p roach isconsid erab ly tighterthan the m od el-d ep end entb and in threecountries. However, all b reachesofthelim itshadatem p orarycharacter. Hence, theforeign exchangem arketd iseq uilib rium d id notlast m orethan oneob servation (q uarter). Itisworthwhileto m entionasim ilarityintheveryrecent EMP d evelop m ent thatwassharedb ythreecountries (Hungary, Polandand Slovakia). Thep ressureexceed edorap p roachedthelower lim it at the end of 2006 announcing the ap p reciation p ressureonthenationalcurrencies.
Whereas the ap p reciation p ressure p revailed over the entire p eriod in the Czech Rep ub lic and Poland, the m ore b alanced p rop ortion ofp ositiveand negative EMP ob servationswasrevealedin Slovakia. By contrast, Hungary had to face p red om inantly a d ep reciation p ressureon HUF.
Exchange Market Pressure in Central Europ ean Countries from the Eurozone Mem b ership Persp ective
Novem b er 2008 13
The m ost extrem e EMP in the Czech Rep ub lic (+13.39%) can b e ob served in 2004:2. Such a high d ep reciation p ressure was caused b y theincrease of the Czech interest rate ab ove the Eurozone level and the sub seq uent change in the interest rate d ifferential (+210%). In Poland, we id entified the m ost extrem e EMP in 2005:4 (-20.56%). A sep arate analysis of the EMP com p onents allows us to d eterm ine the p rincip al cause: a sub stantial change in the reserves-m oney ratio (+12202.9%) d riven b y a m assive increase in the reserve hold ings.
Slovakia is the country with the m ost escap es fromthe no-crisis b and, m ainly on theap p reciation sid e. However, the b reaches of the corrid or were rather m arginal and the m ost significant one was record edin 2005:1 (-7.75%) as a conseq uence of growing international reserves. Slovakia also witnessed a high d ep reciation p ressure (+9.91%) in 1998:4, just after the shift in the exchange rate arrangem ent toward s a m anaged floating. In Hungary, ignoring the very early p eriod, we can d istinguish two cases of the excessive d ep reciation EMP. The first (+11.19%) occurred in 2003:3 following culm ination of the sp eculative attack on ap p reciating HUF. In 2005:1, EMP reached an even higher level (+13.78%) foreseeing the com ingp eriod of a m assive HUF d ep reciation.
4.3 Comparison of A lternative A pproaches
The alternative em p irical ap p roaches to the EMP estim ation resulted in consid erab ly d ifferent find ings. This can b e d ocum ented b y d escrip tive statistics of the EMP tim eseriesas wellascorrelation analysis. The elem entaryd escrip tivestatisticsare p resented in Tab le 2 and correlation coefficients ofthe EMP m easures in Tab le 3.
The only country with results signaling som e d egree ofconsistency is Hungary. The m eans and m ed ians of b oth EMP ind iceshavep ositivesignsandthecorrelation coefficientisthe highestam ong all countries. One can find further uniq ueness in the results from Hungary. The d evelop m ent of the m od el-d ep end ent EMP was significantly less volatile than d evelop m ent of the
alternative m od el-ind ep end ent EMP. This is evid ent in all of the following ind icators: stand ard d eviation, wid th of the no-crisis b and, and sp read b etween m axim um and m inim um values.
Totally op p osite conclusions can b e d rawn from the rem aining countries. Their m ost notab le com m on attrib ute is the higher volatilityof the m od el-d ep end ent EMP. Moreover, they also share a d isharm onic d evelop m ent of the EMP m easures m irrored in the reversely signed m eans and m ed ians and low and/or negative correlation coefficients. It should b e rem ind ed here that the high stand ard d eviations and wid e b and s stem from the varying d evelop m ent in the very early stage of the estim ation p eriod.
Czech Rep ub lic Hungary Poland Slovakia
0.086080 0.462380 -0.292988 0.1 92232
Source: Author’s calculations
Table 3: Correlation coefficients of alternative exchange market pressure measures
The consistency of the two EMP ind ices can b e also assessed b y d iscrep ancies in the id entification of the crisis q uarters. For that p urp ose, the d evelop m ents of b oth EMP m easures in each country are p ut together andp resented in Figure 1. Moreover, Tab le 4 shows how m any q uarters were id entified b y the m od el-ind ep end ent ap p roach as a crisis occurrence, and how m any of these are sim ilarly classified b y the m od el-d ep end ent ap p roach if the m od el-ind ep end ent no-crisis b and ap p lies. Tab le 4 also rep orts the num b er of EMP crisis ob servations that ob tained the sam e sign and sim ilar m agnitud e in b oth ap p roaches. The results p resentedconfirmthenegligib le consistency and p rovid e evid ence that the em p irical tools usedtend to interp ret EMP d evelop m ent d ifferently.
Czech Republic Hungary Poland Slovakia
m_dep m_ind m_dep m_ind m_dep m_ind m_dep m_ind mean 0.0361 -0.0015 0.0091 0.0285 -0.1075 -0.0080 0.0352 0.0025 median 0.0036 -0.0011 0.0068 0.0228 -0.0362 0.0009 0.0062 0.0030
max 0.5544 0.1339 0.0578 0.1504 0.2150 0.1110 0.7466 0.0991
min -0.0577 -0.1003 -0.0199 -0.0531 -0.6992 -0.2056 -0.0891 -0.0786 st. dev. 0.0948 0.0371 0.0165 0.0478 0.2005 0.0554 0.1170 0.0337 upper 0.1783 0.0488 0.0339 0.1002 0.1933 0.0751 0.2107 0.0532 lower -0.1061 -0.0612 -0.0157 -0.0432 -0.4083 -0.0912 -0.1403 -0.0481 Source: Author’s calculations
Notes: m_d ep andm_ind d enote m od el-d ep end ent andm od el-ind ep end ent ap p roach resp ectively Table 2: Descriptive statistics of exchange market pressure
Exchange Market Pressure in Central Europ ean Countries from the Eurozone Mem b ership Persp ective
Czech
Rep ub lic Hungary Poland Slovakia
m_ind crises 4 6 3 7
m_d epcrises 1 0 1 0
sam e sign 2 6 2 3
sim ilar
m agnitud e 2 0 1 0
Source: Author’s calculations
Notes: m_d ep andm_ind d enote m od el-d ep end ent and m od el-ind ep end ent ap p roach, resp ectively. Sim ilar m agnitud e m eans that the value of the m_d ep EMP is within interval 50%- 150% of the m_ind EMP value.
Table 4: Consistency of alternative approaches in identification of crises
-.2 -.1 .0 .1 .2 .3
1994 1996 1998 2000 2002 2004 2006 EMP_DEP EMP_INDEP
-.04 .00 .04 .08 .12 .16
1994 1996 1998 2000 2002 2004 2006 EMP_DEP EMP_INDEP
-.6 -.4 -.2 .0 .2
1994 1996 1998 2000 2002 2004 2006 EMP_DEP EMP_INDEP
-.1 .0 .1 .2 .3
1994 1996 1998 2000 2002 2004 2006 EMP_DEP EMP_INDEP
Czech Rep ub lic
Poland
Hungary
Slovakia
Source: Author’s calculations
Figure 1 : Develop m ent of exchange m arket p ressure b ased on m od el-d ep end ent andm od el-ind ep end ent ap p roaches
Czech
Rep ub lic Hungary Poland Slovakia
m_dep m_ind m_dep m_ind m_dep m_ind m_dep m_ind
fixed (crawl)
p eg 0.1448 (0:0) 0.0148 (0:0) 0.0181 (2:0) 0.0458 (3:0) 0.3342 (0:2) 0.0212 (0:0) 0.1767 (2:0) 0.0336 (1:1) craw-
ling b and
--- --- 0.0165 (0:0) 0.0317 (0:0) 0.1291 (0:0) 0.0369 (1:0) --- ---
floa-
ting 0.0382 (0:0) 0.0438 (3:1) 0.0127 (0:1) 0.0483 (2:1) 0.0613 (0:0) 0.0605 (0:2) 0.0245 (0:0) 0.0339 (1:3) ERM II --- --- --- --- --- --- 0.0164 (0:0) 0.0380 (0:1)
Source: Author’s calculations
Note: The ratio in p arentheses is (num b er of excessive d ep reciation EMP: num b er of excessive ap p reciation EMP) Table 5: Standard deviations of exchange market pressure and number of crisis quarters
One of the aim s of the p ap er is tocom p are EMP in various exchange rate arrangem ents in EU4. The com p arison of the EMP stand ard d eviations calculated over the p eriod s with the p articular exchange rate regim e along with the num b ers of the crisis quarters are p rovid edin Tab le 5.
The results clearly suggest that any conclusion ab out the relationship b etween EMP and exchange rate regim e is extrem ely sensitive to the selection of the EMP estim ation m ethod. The m od el-d ep end ent and m od el-ind ep end ent ap p roaches lead to ab solutely controversial find ings on how EMP d evelop and fluctuate in the p articular exchange rate arrangem ent. The m od el-d ep end ent ap p roach
m od el-d ep end ent ap p roach m od el-ind ep end ent ap p roach no.
ob s. m ean vari-
ance no.
ob s. m ean vari-ance fixed
(crawl) p eg
56 0.00236 0.06289 56 0.02115 0.00179
crawling
b and 44 -0.04109 0.01097 44 0.01596 0.00118
floating 115 0.00232 0.00161 115 -0.00610 0.00244
ERM II 5 -0.01622 0.00026 5 0.00292 0.00144
F-statistics: 1.162374 P-value: 0.325034
F-statistics: 5.646847 P-value: 0.000963
Source: Author’s calculations
Note: Critical value of F-statistics is 2.646402.
Table 6: A N O V A test results
Exchange Market Pressure in Central Europ ean Countries from the Eurozone Mem b ership Persp ective
p rovid es evid ence that EMP was very stab le in all EU4 d uring the floating-regim e p eriod and the excessive d eviations of EMP occurred sp orad ically at that tim e. By contrast, the p eriod s of fixed arrangem ent witnessed m any ep isod es surp assing the level of 1.5 m ultip le of the stand ard d eviation as well as sub stantially m ore volatile d evelop m ent. The results of the m od el-ind ep end ent ap p roach are totally op p osite. Generally, any kind of the fixedregim e p aved the way for lower and less volatile EMP and also fewer crisis p eriod s.
In ord er to d eterm ine whether the d ifferences am ong EMP values in various exchange rate regim es are statistically significant we carried out a single-factor Analysis of Variance (ANOVA). The EMP ob servations from all EU4 were gathered in the single d ataset and group ed intofour categories accord ing to the classification system used in Tab le 5. The ANOVA test results for b oth ap p roaches are rep orted in Tab le 6.
Novem b er 2008 15
The ANOVA tests show that the exchange rate regim e d oes not influence the average of the m od el-d ep end ent EMP consid erab ly, as the F-statistic is sm all and insignificant.
On the other hand, the m eans of the group ed m od el- ind ep end ent EMP are significantly d ifferent at the 1% level.
Thus, one can consid er the floating arrangem ent to b e the environm ent contrib uting to the volatile d evelop m ent and excessive values of EMP.
The results ob tained allow us to d erive som e p olicy im p lications. There is no em p irical justification for the a p riori concerns that a shift in the exchange rate regim e from floating to the quasi-fixed ERM II will stim ulate EMP to increase. More likely, the b asic characteristics of EMP d evelop m ent will b e retained after the change. Hence, sup p osing that the recent level of EMP volatility and d ensity of the crisis ob servations revealed b y the m od el- ind ep end ent ap p roach rem ain unchanged, it will cast serious d oub t on the Europ ean Com m ission’s requirem ent that EU4 m ust p articip ate in ERM II without sub stantial tensions on the exchange rates.
The d oub t gains im p ortance if the authorizedfluctuation m argin is likely to b e asym m etric with the lim its of 15%
on the ap p reciation sid e and 2.25% on the d ep reciation sid e. Although EMP fluctuated p red om inantly within this narrow b andin EU4 in the last four years, the d ep reciation p art of the asym m etric b and is very tight and the EMP d evelop m ent shouldb e m onitoredclosely.
Owing to som e factors the EMP estim ates p resented and d iscussed p reviously m ust b e viewed with som e d egree of skep ticism. Besid es the d rawb acks alread y d iscussed, the m od el-d ep end ent EMP in all countries d evelop ed alm ost in p arallel with the changes in reserves over the entire p eriod. This im p lies a frequent ap p lication of the central b ank official intervention even in the environm ent of the floating exchange rate regim e. However, the reality in m any EU4 was d ifferent. These lim itations should b e ad d ressed and elim inated in future research. We recom m end use of the p ure foreign exchange intervention d ata as the alternative to the change in reserves. The m od el can also b e extend ed b y the p ossib ility of ind irect intervention op erating through changes in the d om estic lend ing or interest rate.
5. Conclusion
In this p ap er, we estim ated EMP for the EU4 currencies against the Euro exchange rate over the p eriod from 1993-2006. Fund am ental d ifferences in the sp irit and construction of the ap p roaches ap p lied are reflected in consid erab ly d ifferent results. Thus, the two alternatives are not com p atib le if the d ata from EU4 are used.
Accord ing to the m od el-d ep end ent ap p roach, EMP in the Czech Rep ub lic, Hungary and Slovakia are of sim ilar m agnitud e. Whereas a d ep reciation p ressure p revailed on HUF and SKK, no d om inance of any d irection of the
p ressure can b e found in the case of CZK. The estim ates of the Polish EMP are b urd ened b y a sub stantial statistical insignificance. The results ob tained suggest that EMP in EU4 d ecreased over tim e and was rem arkab ly lower and less volatile d uring the p eriod s of floating exchange rates than in the environm ent of the fixed exchange rate regim e.
However, there are som e concerns ab out the valid ity of the p aram eter estim ates and consequently the EMP m easures in all EU4.
The m od el-ind ep end ent ap p roach p uts greater em p hasis onthe interest rate d ifferential, which has often b een id entified as one of the factors of the exchange rate d eterm ination in EU4. EMP d evelop m ent can b e d escrib ed as hom ogeneous d uring the entire p eriod analyzed, with no ep isod e of an ab norm al volatility or excep tionally frequent occurrence of excessive EMP. While CZK and PLN were largely und er ap p reciation p ressure, HUF was forced to d ep reciate and no d om inance was revealedin Slovakia.
However, the m od el-ind ep end ent ap p roach id entified m ore crises than the m od el-d ep end ent ap p roach, includ ing the very recent excessive ap p reciation p ressure onthree EU4 national currencies.
The stud y d oes not confirm the concerns that the unavoid ab le shift in the exchange rate regim e toward s the quasi-fixed ERM II will p rovoke EMP to grow to excessive levels. Instead, the em p irical tests suggest that the regim e change will have, with a high p rob ab ility, a negligib le im p act on EMP d evelop m ent. Stem m ing from the estim ations ob tained, the EU4 central b anks will p rob ab ly b e confronted with som e occasions of excessive EMP jeop ard izingfulfillm ent of the exchange rate stab ility criterion.
Exchange Market Pressure in Central Europ ean Countries from the Eurozone Mem b ership Persp ective
A ppendix 1: D ata description
Alld ataareonquarterlyb asisandcoverthep eriod 1993:1 – 2005:4
B
t EU4 Obtained national from IMF’s International Financial Statistics (IFS) line 14 m oney base (Reserve m oney) and then logged.c
tEU4 Gross national incom e
Derived by adding the net incom e from abroad to Gross dom estic product (IFS line 99B). In national accounts statistics, the total of rents, interest, profits and dividends plus net current transfers is shown as “net incom e from abroad”. It was obtained from IFS by differencing current account balance (IFS line 78ALD) and balance on goods andservices (IFS line 78AFD). Logged values.
e
tNom inal bilateral exchange rate of EU4 currencies vis-à-vis Euro in direct quotation (num ber of EU4 currency units for one Euro)
Obtained from Eurostat’s Econom y and finance database (EEF) section Exchange rates and Interest rates, line Euro/ECU exchange rates – Quarterly data. Logged values.
i
t*Eurozone 3-m onth m oney m arket interest rate
Obtained from EEF section Exchange rates and Interest rates, line Money m arket interest rates – Quarterly data,
series MAT_M03
i
t EU4 national 3-m onth m oney m arket interest rateObtained from EEF section Exchange rates and Interest rates, line Money m arket interest rates – Quarterly data, series MAT_M03
m
t EU4 Obtained from IFS line 34..B (Mnational M1 m onetaryaggregateoney, Seasonally Adjusted) and then logged.p
t*Eurozone Harm onized indices of consum er prices
Obtained from EEF section Prices, line Harm onized indices of consum er prices – Monthly data (index 2005=100). Converted from m onthly to quarterly data by averaging the three m onthly figures and then logged.
p
tEU4 national Harm onized indices of consum er prices
Obtainedfrom EEF section Prices, line Harm onized indices of consum er prices – Monthly data (index 2005=100). Converted fromm onthly to quarterly data by averaging the three m onthly figures and then logged.
r
tEU4 national official reserves holdings
Obtained from IFS line 1L.D (Total Reserves Minus Gold) converted to national currency using nom inal bilateral exchange rate vis-à-vis US dollar (IFS line AE) and then logged.
rm
tProportional change in dom estic international reserves Obtained by ratio of change in the level of reserves (IFS line 79DAD) and m oney base of previous period (IFS line 14).
y
tEU4 national Gross dom estic product
Obtained from IFS line 99B (Gross Dom estic Product) and then logged.
y
t trendLong-runcom p onent of yt
Ob tained using the Hod rick-Prescott filter anda sm oothing p aram eter of 1600, as recom m end ed for quarterly d ata.
A ppendix 2: Estimations of equations (14)-(16)
Source: Author’s calculations Czech Republic
Poland
Poland Czech Republic
Hungary
Slovakia
Slovakia Hungary Equation (14)
Equation (15)
Novem b er 2008 17
A ppendix 2 (continued): Estimations of equations (14)-(16)
Source: Author’s calculations Source: Author’s calculations
Source: Author’s calculations
A ppendix 3: E xchange market pressure in EU 4 countries (model-dependent approach)
A ppendix 4: Exchange market pressure in EU 4 countries (model-independent approach)
Czech Republic
Poland
Hungary
Slovakia Equation (16)
Exchange Market Pressure in Central Europ ean Countries from the Eurozone Mem b ership Persp ective
References
Bielecki, S. 2005. Exchange m arket pressure anddom estic credit evidence from Poland. The Poznan University of Econom ics Review 5 (1): 20-36.
Eichengreen, B., A. K. Rose, C. Wyplosz. 1994. Speculative Attacks on Pegged Exchange Rates: an Em pirical Exploration with Special Reference to the European Monetary System. Working Paper No. 4898. National Bureau of Econom ic Research.
1995. Exchange Market Mayhem: The Antecedents and Afterm ath of Speculative Attacks. Econom ic Policy 10 (21):
249-312.
Girton, L., D. Roper. 1977. A Monetary Model of Exchange Market Pressure Applied to the Postwar Canadian Experience. Am erican Econom ic Review 67 (4): 537-548.
Kam inski, G., S. Lizondo, C. Reinhart. 1998. Leading indicators of currency crises. IMF Staff Papers 45 (1): 1-48.
Kam inski, G., C. Reinhart. 1999. The twin crises: The causes of banking and balance-of-paym ents problem s. Am erican Econom ic Review 89 (3): 473-500.
Pentecost, E. J., C. Van Hoydonk, A. Van Poeck. 2001.
Measuring and Estim ating Exchange Market Pressure in the EU. Journal of International Money and Finance 20 (3): 401- 418.
Roper, D., S. J. Turnovsky. 1980. Optim al Exchange Market Intervention in a Sim ple Stochastic Macro Model. Canadian Journal of Econom ics 13 (2): 296-309.
Sachs, J. D., A. Tornel, A. Velasco. 1996. Financial crises in em erging m arkets: The lessons from 1995. Brooking Papers on Econom ic Activity 1996 (1): 147-215.
Spolander, M. 1999. Measuring Exchange Market Pressure and Central Bank Intervention. Bank of Finland Studies E:17.
Stavárek, D. 2005. Exchange Market Pressure in New EU-m em ber Countries. In Money, Banking and Financial Markets. Proceedings from the International Conference, 244-248. Bratislava: Ekonóm.
Tanner, E. 2002. Exchange Market Pressure, Currency Crises, and Monetary Policy: Additional Evidence from Em erging Markets. Working Paper WP/02/14. International Monetary Fund.
Turnovsky, S. J. 1985. Optim al Exchange Market Intervention: Two Alternative Classes of Rules. In J.
S. Bhandari (ed) Exchange Rate Managem ent Under Uncertainty. Cam bridge: MIT Press.
Van Poeck, A., J. Vanneste, M. Veiner. 2007. Exchange Rate Regim es and Exchange Market Pressure in the New EU Mem ber States. Journal of Com m on Market Studies 45 (2):
459-485.
Weym ark, D. N. 1995. Estim ating Exchange Market Pressure and the Degree of Exchange Market Intervention for Canada. Journal of International Econom ics 39 (3-4): 273- 295.
1997a. Measuring the Degree of Exchange Market Intervention in a Sm all Open Econom y. Journal of International Money and Finance 16 (1): 55-79.
1997b. Measuring Exchange Market Pressure and Intervention in Interdependent Econom y: A Two-Country Model. Review of International Econom ics 5 (1): 72-82.
1998. A General Approach to Measuring Exchange Market Pressure. Oxford Econom ic Papers 50 (1): 106-121.
