9. Darvas Zsolt Simon András: NYITOTTSÁGÁNAK FELHASZNÁLÁSÁVAL december

MNB Füzetek 2000/9

Darvas Zsolt – Simon András:

A POTENCIÁLIS KIBOCSÁTÁS BECSLÉSE A GAZDASÁG NYITOTTSÁGÁNAK FELHASZNÁLÁSÁVAL

∗

2000. december

∗

Köszönettel tartozunk Kőrösi Gábornak, Jerome Henrynek, Neményi Juditnak, Vincze Jánosnak, valamint Magyar Nemzeti Bankban rendezett szeminárium és a Link Project 2000 konferencia résztvevőinek a hasznos észrevételekért. Az esetleges hibák kizárólag bennünket terhelnek. A tanulmányban kifejtett nézetek nem feltétlenül esnek egybe az MNB hivatalos véleményével.

ISSN 1219 9575 ISBN 963 9057 57 6 Online ISSN: 1585 5597

Darvas Zsolt: főmunkatárs, Közgazdasági és kutatási főosztály E-mail: [email protected] Simon András: főosztályvezető-helyettes, Közgazdasági és kutatási főosztály E-mail: [email protected]

E kiadványsorozat a Magyar Nemzeti Bankban készült elemző és kutató munkák eredményeit tartalmazza, és célja, hogy az olvasókat olyan észrevételekre ösztönözze, melyeket a szerzők felhasználhatnak további kutatásaikban. Az elemzések a szerzők véleményét tükrözik, s nem feltétlenül esnek egybe az MNB hivatalos véleményével. Magyar Nemzeti Bank 1850 Budapest Szabadság tér 8-9. http://www.mnb.hu

Darvas Zsolt – Simon András:

A potenciális kibocsátás becslése a gazdaság nyitottságának felhasználásával 2000. november Nem-technikai ismertet½ o Ez az ismertet½o annak az olvasónak készült, aki nem kívánja megismerni sem a számítások alapjául szolgáló modell részletes feltevéseit, sem a számítás módszerét, de intuitív módon szeretné megérteni, hogy mi volt a számítások célja, mit takar a címben szerepl½o fogalom és milyen következtetések levonására vagy gyakorlati felhasználásra alkalmasak a számítások eredményei. Írásunk tehát nem a szokásos összefoglaló, amely a tanulmány tartalmát fogalmazza meg tömören, és elhelyezi mondanivalónkat a közgazdasági irodalomban, rámutatva az új gondolatokra és annak különbségére a korábbi megközelítésekhez képest. Célunk csupán e speciális témában és a technikai részletek iránt kevésbé érdekl½od½o közgazdász olvasó tájékoztatása munkánkról, meggy½ozése arról, hogy értelmes feltevésekb½ol hasznos következtetésekre jutottunk. A fejlett ipari gazdaságok elemzésének egyik fontos eszköze a potenciális kibocsátás fogalma. A potenciális termelés hosszú távon fenntartható termelést jelent. Az 1960-1970-es években vált nyilvánvalóvá annak a nézetnek a helyessége, hogy a kibocsátást hosszú távon a termelési tényez½ok mennyisége és a technológia fejlettsége meghatározza. A hosszú távú meghatározottság több év átlagában való meghatározottságot jelent. A potenciális termelés fogalma lehet½ové teszi, hogy felismerjük a makrogazdasági keresleti politika lehet½oségeit és korlátait a gazdasági növekedésben. A kibocsátás a keresletre ható politikával rövid távon növelhet½o ugyan, de csak úgy, hogy az id½oleges növekedésnek ugyanakkora jöv½obeli csökkenés az ára. 1

Az aktuális és a potenciális kibocsátás különbsége a többletkereslet, vagy ”kibocsátási rés”. Ha ez pozitív akkor ”er½onkön felül” költekezünk, vagyis a jelenlegi magas kibocsátásnak meglesz a ”böjtje”. A potenciális kibocsátás mérésének hagyományos megközelítése abból indul ki, hogy a pozitív kibocsátási rés növeli az in‡ációst, és az in‡áció csökkentése csak ugyanolyan mérték½u negatív kibocsátási rés létrehozásával szüntethet½o meg. Ezt a jelenséget nevezzük Phillips-görbe hatásnak. Így – hacsak nem vállalunk végtelenül gyorsuló in‡ációt – a kibocsátás tartósan nem haladhatja meg a potenciálisat. Ez a megközelítés gyakran a munkanélküliség adatait használja fel, feltételezve, hogy a kibocsátási rés arányos a tényleges munkanélküliségi ráta és az ún. természetes munkanélküliségi ráta különbségével, azaz a munkanélküliségi réssel. A természetes ráta mérése ugyan sok problémát felvet, de abban a vélemények megegyeznek, hogy a munkanélküliségi rés az in‡ációval szoros kapcsolatban áll. Véleményük szerint ezzel a megközelítéssel nem kapunk kielégít½o magyarázatot a magyar in‡ációs folyamatra az elmúlt 10 évben. Nem vonjuk kétségbe, hogy a munkaer½opiac feszültségei hatással vannak az in‡ációra, de úgy gondoljuk, hogy ebben az id½oszakban Magyarországon más fontos tényez½ok is szerepet játszottak az in‡áció kialakulásában és kés½obbi ingadozásaiban. Gondoljunk csak arra, hogy 1990-1992 között az in‡áció növekedése és a munkanélküliség növekedése egymással párhuzamosan történt. Ráadásul azt is …gyelembe kell vennünk, hogy a Phillips-görbe jelenség meg…gyelésére legfeljebb az utolsó 10 év áll rendelkezésre, hiszen korábban a tervgazdaságban ilyen összefüggés nem létezett. Mindezen nehézségek a Phillips-görbe alkalmazásában és mérésében arra késztettek, hogy a fenntartható növekedésnek egy másik értelmezést adjunk. Az új fogalom segítségével ugyan nem jutunk majd el oda, hogy a kibocsátási rés nagyságából közvetlenül az in‡áció változására következtessünk, de azért az álatlunk számított kibocsátási résnek is van kapcsolata az in‡ációval. A fenntartható növekedés fogalmát ugyanarra a meg…gyelésre alapozzuk, mint a Phillips-görbe modellje. Rövid távon a kereslet serkenti a gazdasági növekedést. Ez azonban mindig csak ideiglenes, amit kés½obb ellenkez½o irányú ingadozás követ. A fentarthatatlansághoz vezet½o mechanizmus azonban nem közvetlenül az in‡áción, hanem a kereskedelmi mérlegen keresztül m½uködik. Ha a kibocsátás a belföldi kereslet hatására n½o, akkor n½o az import, változatlan vagy éppen csökken½o export mellett. A létrejöv½o de…cit azonban nem tartható fenn, mert minden hitelt el½obb utóbb vissza kell …zetni, és akkor a belföldi kereslet visszaszorítására van szükség, ami csökkenti a kibocsátást. 2

Hasonlóan a külföldi konjunktúra is növelheti a hazai kibocsátást, mert nagyobb lehet az exportunk. Ez azonban – mivel a konjunktúra csak ideiglenes és ellenkez½o irányba fordul, szintén nem hoz létre fenntartható kibocsátást. A rokonság az eredeti modell potenciális kibocsátás fogalmával kétségtelen. A Phillips-görbe összefüggés a külfölddel nem versenyz½o szektorra jellemz½o: ott a többletkereslet in‡ációs hatású. A mi felhasznált összefüggésünk a külfölddel versenyz½o szektorra jellemz½o, ahol a belföldi többletkereslet nem közvetlenül az in‡ációban tükröz½odik, hanem el½obb külkereskedelmi de…citet hoz létre. A kibocsátási rés mindkét esetben fenntarthatatlan. A de…cit és az in‡áció között is van kapcsolat, de nem olyan közvetlen, mint a munkapiaci többletkereslet és az in‡áció között a külfölddel nem versenyz½o szektorban. A gazdaság nyitottsága lehet½ové teszi, hogy az in‡ációs hatás elhalasztódjon. A de…cit valutaleértékelési várakozásokat ébreszthet, s½ot, tényleges leértékeléshez is vezethet. Akár a várakozások, akár a tényleges leértékelés ceteris paribus növeli az in‡ációt. Nehéz volna számszer½ uen meghatározni, hogy ezek a hatások milyen er½osek, illetve milyen késési strukturában érvényesülnek, mert feltehet½o, hogy mind a hatás intenzitása, mind annak késése nagy változékonyságot mutat. A változékonyság onnan származik, hogy egy gazdaságban sohasem egyértelm½u, hogy egy adott de…cit mikor tekinthet½o elfogadhatatlannak, vagyis olyannak, amelynek a vissza…zetése az ország számára elviselhetetlen1 . Amíg a de…cit ”elviselhet½o”, addig nem okoz in‡ációs várakozást Ebben a tanulmányban megbecsüljük a küls½o egyensúly szempontjából fenntartható kibocsátás volumenét. Fenntarthatónak vagy potenciálisnak nevezzük azt a kibocsátást, ami hosszú távon egyensúlyban tartja a termékek és szolgáltatások kereskedelmének mérlegét. A ”hosszú táv” azt jelenti, hogy a tényleges kibocsátás értékéb½ol kisz½ urjük mind a bels½o, mind a küls½o konjunktúra (rövid távú, átmeneti ingadozások) hatását. A becslés alapjául szolgáló modell tehát nem egyszer½ uen a tényleges kibocsátást korrigálja a külkereskedelmi egyenleggel, hanem annál ki…nomultabb, mert mind belföldi, mind a külföldi keresletet …gyelembe veszi. Ha a kibocsátás például azért nagy, mert jó a külföldi konjunktúra, akkor a kibocsátás akkor is fenntarthatatlan lehet, ha egyébként a kereskedelmi mérleg egyensúlyban van vagy pozitív. Összefoglalóan megállapíthatjuk, hogy a 1

A nem éppen pontos ”elviselhetetlen” jelz½ot csak az egyszer½ u megfogalmazás kedvéért használjuk. Elméletileg pontos de…níció adható arra, hogy mennyi az a de…cit, amelynek vállalása az ország növekedési kilátásai alapján optimális. Ez az optimum azonban új és új információk birtokában egyre változik. Ha a tényleges de…cit ett½ol eltér, akkor a belföldi kereslet korrekciójára van szükség. Ezek a korrekciók lehetnek in‡ációs hatásúak.

3

számított id½osor alapján kétféle területen vonhatók le következtetések. Ezek a növekedés kilátásai és az in‡ációs kilátások. (1) A pozitív rés azt jelzi, hogy a belföldi kibocsátást er½os kereslet támogatja: vagy a külföldi konjunktúra er½os, vagy a belföldi, vagy mindkett½o. Ez a kapacitások átlagosnál jobb kihasználását teszi lehet½ové. Hosszú távon ez nem tartható fenn, a kapacitások kihasználtsága csökkenni fog, amikor a konjunktúra romlik. (2) A pozitív rés a kapacitás-feszültségek miatt in‡ációs hatású is lehet. Ha a kibocsátást nem a külföldi, hanem a belföldi konjunktúra ”húzta”, akkor küls½o de…cit is keletkezik. Ha a piac úgy látja, hogy ez a de…cit túl kockázatos az ország növekedési kilátásaihoz képest, akkor az válsághoz és in‡ációhoz vezethet. Az általunk értelmezett fogalomnak adható egy versenyképességi értelmezése. Ha a potenciális magyar kibocsátás gyorsabban n½o, mint a külföldé, akkor exportunk részesedése n½o a világpiacon. Ezt nevezzük a versenyképesség növekedésének. Ebben a megközelítésben azt mondhatjuk, hogy hosszú távon gazdasági növekedésünk ütemét versenyképességünk határozza meg. Ideiglenesen növekedhetünk gyorsabban a belföldi kereslet élénkítésével is, de ekkor adósságot halmozunk fel, amit csak úgy tudunk vissza…zetni, ha a jöv½oben növeljük versenyképességünket. A potenciális növekedés ilyen értelmezése nem függ a gazdasági rendszerekt½ol. Az a tény, hogy ha a belföldi felhasználás túl nagy, akkor nagyobb lesz az import, vagy ha partnereinknél fordul el½o ugyanez, akkor nagyobb lesz az exportunk, független attól, hogy tervgazdaságban vagy piacgazdaságban vagyunk. Az is független a rendszerekt½ol, hogy ha egy országban nagyobb termékek kínálata, mint a kereslete, akkor ott a kereskedelmi mérleg javulni fog. Az export vagy az import expanziója természetesen más formában valósul meg a két esetben. Piacgazdaságban nagymértékben az árrendszer közvetíti a nagyobb kínálat érvényesülését a piacon, tervgazdaságban ennek kisebb a szerepe. Ez azonban a számítás során számunkra közömbös volt, mert nem volt szükségünk áradatokra, a kereslet-kínálat és a volumenek közötti kapcsolatot közvetlenül értelmeztük. Így lehet½ové vált, hogy a kibocsátási rést hosszú id½oszakra számítsuk ki, olyan id½oszakra, amelyben volt tervgazdaság annak többféle változatával, volt átmeneti gazdaság és volt piacgazdaság. A következ½o két ábra az 1960 és 1999 közötti id½oszakra tartalmazza a GDP tényadatait összehasonlítva a potenciális GDP adataival.

4

5%

kibocsátási rés

4%

tényleges és fenntartható GDP növekedési üteme közötti különbség

3% 2% 1% 0% -1% -2% -3%

2000

1995

1990

1985

1980

1975

1970

1965

1960

-4%

ábra 1: A kibocsátási rés, valamint a tényleges és fenntartható GDP növekedési üteme közötti ütemkülönbség Az els½o gra…kon a kibocsátási rést, valamint a potenciális és a tényleges növekedési ütemek különbségeit mutatja. A második gra…kon együtt mutatja a potenciális kibocsátás és a GDP növekedési ütemeit, a belföldi kereslet növekedését, valamint a külföldi konjunktúra indexét. Utóbbi változó nem egyszeru ½en a külföldi kereslet növekedési üteme, hanem azt mutatja, hogy a külföldi kereslet mennyivel növekszik gyorsabban vagy lassabban a külföldi kínálatnál.2 A gra…konokból látszik, hogy a tervgazdaságban is és az utána következ½o átmeneti majd piacgazdasági id½oszakban is voltak konjunktúra-ciklusok Magyarországon. A kibocsátási rés jól mutatja azokat a tervgazdasági stop-go periódusokat, 2

Az ábra potenciális kibocsátás növekedési ütem adatai jelentéktelen mértékben, de eltérnek a f½o tanulmányban közölt számoktól. Ennek az az oka, a két helyen a fogalom különböz½o értelmezését használtuk. A tanulmányban a potenciális kibocsátás egy becsült valószín½uségi változó várható értéke, míg itt, a történeti értékelésben a változónak egy realizációja. A két változó egy véletlen tagban különbözik. Ennek nagysága a gyakorlatban érdektelen.

5

amelyekre az id½osebbek még jól emlékeznek, az 1965-ös megszorítást, az olajválság el½otti keresleti felfutást majd a válság utáni visszaesést, az 1978-as nagy (beruházási) túlköltekezési hullámot, a 80-as évek közepi megszorításokat, majd a kiengedést a ”Gorenje-konjunktúra” idején. 12%

8% 4%

0%

-4%

fenntartható kibocsátás -8%

GDP belföldi kereslet

-12%

külföldi konjunktúra index 2000

1995

1990

1985

1980

1975

1970

1965

1960

-16%

ábra 2: Növekedési ütemek: fenntartható kibocsátás, GDP, belföldi kereslet, és külföldi konjunktúra index Érdekes az átmeneti id½oszak képe a modell tükrében. A legnagyobb gazdasági visszaesés idején 1990-1992 között a tényleges és a potenciális növekedés nagyon szorosan együtt mozog. Ha feltételezzük, hogy ebben az id½oszakban a gazdaságpolitika képes volt a kereslet alakítására, akkor az együttmozgásból arra a következtetésre juthatunk, hogy a politika semleges volt. Ez gyakorlatilag azt jelenti, hogy a monetáris és …skális hatóságok egyformán féltek, vagy legalábbis összességében úgy viselkedtek, mintha egyformán félnének az in‡ációtól és a kibocsátás visszaesését½ol. Ez az összegez½o megállapítás talán annak ellenére igaz, hogy a …nomabb elemzés valamelyest az in‡ációellenesség felé való hajlást igazol. A kibocsátási rés 1990 és 1992 között ugyanis negatív volt 1-3 százalékos mértékben. Ha ezt összehasonlítjuk a GDP 15 százalékos kilengésével ebben az id½oszakban, akkor 6

állíthatjuk, hogy a számításokban kimutatott megszorító hatás eltörpül a folyamatok akkori erejéhez képest.3 1993-1994-ben a lazítás er½oteljesebben mutatkozik. 1993-ban a potenciális növekedés üteme még a negatív tartományban mozgott, amikor a GDP már 3 százalékkal n½ott. 1994-ben a két növekedési ütem ugyan közelít (lásd 2. gra…kon) egymáshoz, de a kibocsátási rés er½osen pozitív, vagyis a kibocsátás szintje magasabb a potenciálisnál. Tudjuk, és a 2. gra…konból is látszik, hogy 1993-1994 nem volt kedvez½o a külföldi kereslet szempontjából. Ez rontotta a kereskedelmi mérlegünket, de a potenciális kibocsátás számított ütemét nem befolyásolta, mert – mint már korábban említettük – a számítás módszere olyan, hogy ezt kisz½uri, mint ideiglenes hatást. 1995-1996-ban a kereslet visszafogása helyreállítja a kereslet és a kínálat egyensúlyát. Ezekben az években a potenciális kibocsátás gyorsabban n½o, mint a tényleges. 1998-1999-ben a potenciális és a tényleges kibocsátás növekedési üteme lényegében azonos, a hosszú távon fenntartható szint azonban valamelyest, mintegy fél százalékkal a tényleges alatt van.

3

Az ábrán látható GDP növekedési adatok két okból eltérnek a GDP volumenindexét½ol: (1) a számításoknál a folyó áras adatokat a belföldi felhasználás árindexével de‡áltuk, (2) az ábrán látható növekedési ütemek a logaritmizált értékek növekményei.

7

Contents 1 Introduction

1

2 The Conceptual Framework 2.1 Supply as Competitiveness in an Open Economy . . . 2.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . 2.2.1 The Latent Integration Variable . . . . . . . . 2.2.2 Some Features of Demand and Supply . . . . . 2.2.3 Transferring to an Absolute Measure of Supply 2.3 Comparison with Other De…nitions and Models . . . . 2.4 Short-run E¤ects . . . . . . . . . . . . . . . . . . . . . 2.4.1 Income Elasticities and Coe¢cients . . . . . . . 2.4.2 Import Content of Exports . . . . . . . . . . . 2.4.3 Exchange Rate E¤ects . . . . . . . . . . . . . .

3 3 4 5 5 6 7 8 8 8 8

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

3 Statistical Representation and Results 9 3.1 Statistical Representation . . . . . . . . . . . . . . . . . . . . 9 3.2 Long-Run Results . . . . . . . . . . . . . . . . . . . . . . . . 11 3.3 The Error Correction Model . . . . . . . . . . . . . . . . . . . 15 4 Results for Poland

17

5 A Note on Time Series Decomposition and Model Choice

17

6 Conclusions

19

7 References

20

8 Appendix 8.1 Demand Index of the Rest of the World . . . . . . . . . . . 8.2 Preliminary Data Analysis . . . . . . . . . . . . . . . . . . . 8.2.1 Unit Root Tests . . . . . . . . . . . . . . . . . . . . 8.2.2 Harvey–Jaeger Arguments for the I(2) Speci…cation 8.3 State-Space Representation of the Model . . . . . . . . . . . 8.4 Estimating Rest of the World Supply . . . . . . . . . . . . . 8.5 Data Sources . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . .

22 22 23 23 23 25 27 28

9 Tables

29

10 Figures

32

1

List of Figures 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

Hungary: Trade balance, GDP growth, unemployment rate, in‡ation, 1960-99 . . . . . . . . . . . . . . . . . . . . . . . . . Hungary: Estimated output gap of world demand based on the two statistical models . . . . . . . . . . . . . . . . . . . . Hungary: Actual and estimated growth of world demand based on the I(2) model . . . . . . . . . . . . . . . . . . . . . Hungary: Actual and estimated growth of world demand based on the breaking I(1) model . . . . . . . . . . . . . . . . Hungary: Estimated state vectors based on the I(2) model . . Hungary: Estimated state vectors and their growth rates based on the breaking I(1) . . . . . . . . . . . . . . . . . . . . Hungarian output gap based on the I(2) model . . . . . . . . Hungarian output gap based on the breaking I(1) model . . . Hungary: Actual and potential growth and the level of output gap based on the I(2) model . . . . . . . . . . . . . . . . . . . Hungary: Actual and potential growth and the level of output gap based on the breaking I(1) model . . . . . . . . . . . . . Hungary: Chow breakpoint tests for the error correction model at each year between 1965-1995 . . . . . . . . . . . . . . . . . Hungary: Growth rate of world demand and of its components, 1960-99 . . . . . . . . . . . . . . . . . . . . . . . . . . . Hungary: Destination distribution of exports, 1976-99 . . . . Poland: Destination distribution of exports, 1980-99 . . . . . Poland: Estimated state vectors . . . . . . . . . . . . . . . . . Poland: Actual and potential growth and the level of output gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Poland: Trade balance, GDP growth, unemployment rate, in‡ation, 1980-99 . . . . . . . . . . . . . . . . . . . . . . . . .

2

33 33 34 34 35 36 37 37 38 38 39 39 40 40 41 42 42

Abstract This paper develops a model that gives a reasonable interpretation to the term ’potential output’ or ’permanent component of output’ in small open economies. Although the output gap is frequently de…ned as excess demand leading to in‡ationary pressures, in an open economy excess demand may simply result in increased imports without a short-run e¤ect on in‡ation. The increased imports may or may not have an impact on future in‡ation depending on policies adopted, therefore, both the magnitude and the lags of this impact is likely variable. That’s why we set up a new model to measure potential output. We rely on foreign trade data and develop trade equations that incorporate the interaction of supply and demand e¤ects. One of the objectives of the model is to give an interpretation of structural shifts in supply and demand, such as the process of transition in Hungary and Poland. The central element of our trade equations is the incorporation of two latent variables called ’relative supply’ and ’economic integration’. (1) Traditional trade models assume that supply of a country manifests itself in the relative price of its variety of the exported good. If supply increases it can be sold only at a lower relative price. This idea leads to the practice that relative price is considered the supply variable in econometric trade models. However, if supply increases, lower sales price is only one of the channels of expansion. The supplier may enter more and more micromarkets with the same price. In other words it, may increase the variants of the good supplied. Thus it is not adequate to represent supply by prices but better to have a variable that quanti…es planned sales directly. This idea is reinforced by the statistical di¢culty to …nd an aggregate relative price that is free from compositional distortions. (2) Exports and imports usually grow faster than domestic expenditures due to trade liberalization, decreasing transaction costs, utilization of economies of scale and scope in intraindustry trade. These e¤ects are captured by a latent variable, called economic integration. Estimation was carried out with the maximum likelihood method using the Kalman-…lter. The Kalman-…lter allows the modeling of latent variables and time-varying parameters that are inevitable in our case. Both statistical analysis of observed variables and economic arguments favor the assumption that our key variables are I(2) processes. Consequently, we have also estimated an error correction model consistent with the I(2)ness of the data.

1

Introduction

The concept of potential output is a useful tool in analyzing macroeconomic developments in industrial countries. The origin of the concept goes back to the traditional theory of business cycles that decomposed output into a deterministic trend and a stationary cycle component. With the development of theory and econometrics the decomposition was re…ned and permanent and transitory stochastic components were distinguished. The permanent component was called potential output.1 The decomposition cannot be accomplished without some assumptions about the structure of the time series. These assumptions may di¤er depending on the purpose of the decomposition or, in other words, on the aspect of the statistical process that is considered to be important. This means that any decomposition is model-dependent. Models of potential output range in a wide spectrum. There are simple univariate models that use some assumed properties of the data generating process2 . The production function approach analyzes factor inputs to production3 . There are also structural models that include information on in‡ation and unemployment4 . Our model belongs to the class of the structural models but it might seem to be an eccentric member of this class as it does not use either in‡ation or unemployment data as information. This model choice arises from our point of departure. Our primary interest was in estimating a time series of an indicator of in‡ationary pressures in Hungary for a period that gives a reasonable number of observations for the estimation. Several systemic changes took place in the recent history of Hungary. This is apparent from a look at the data on output, in‡ation and unemployment (Figure 15 ). The large drop in output coinciding with the surge of in‡ation during the system switch in the early nineties may explain why an atheoretic ”smoothing out” method would be misleading. ”Dummying out” these years might be an idea but the switch was gradual and lasting for several years. Starting the sample period after the consolidation from the transition recession was not an option due to the resulting very short sample period and the gradual nature of transformation which makes it di¢cult to set the date when the transition ended and the new system 1

Throughout this pap er potential output/trend output/permanent component of output and output gap/cycle/transitory comp onent are used as synonyms. None of them has a characteristically distinctive de…nition and the same methods were applied to recover all of them. Policy makers use the wording ’potential output’ and ’output gap’ more frequently while academics often use the other expressions. 2 See Canova (1998) for a comparison of several such methods. 3 See Giorno et al. (1995). 4 See Laxton–Tetlow (1992), Kuttner (1994), Dupasquier et al. (1997), Rasi–Viikari (1998), Gerlach–Smets (1999). 5 Trade balance also shown in the …gure refers to balance of trade and services according to national accounts statistics.

1

started. Therefore, we had to select a model that is able to capture potential output during a relatively long period. This period covers a planned economic system gradually changing in its nature, a period of switch of this system to a market economy accompanied by an unprecedented intensity of structural change, and a period of a market system with a high rate of growth that might lead to a catching up with the rest of Europe. The in‡ation – unemployment trade-o¤ apparently did not exist in the …xed-price – lifelong-employment system of the planned economy and in the early transition years. Capital stock …gures could not be used in estimating output capacity because of the intensity of unexpected structural changes during the switch of systems. In any model capital stock data are used to indicate capital as a factor of production. Whenever accumulated investments are used as capital input in the production process, data on these accumulated investments are a good approximation to capital. However, during the regime switch the old stock of capital was scrapped at once and data on accumulated investments became useless as a measurement of a factor of production.6 Discarding all these possible sources of information there remains one important relation that can be used, and this is the output – external de…cit trade-o¤. This is the piece of information that our model is based on. In an open economy excess demand may simply result in increased imports without any direct e¤ect on in‡ation. Increased imports may have an impact on future in‡ation but the e¤ects may be variable both in lags and magnitude, depending on (exchange rate) policies and a capricious business sentiment. In the analogy of the de…nition of potential output as an output sustainable from the aspect of in‡ation we will de…ne an output level sustainable from the aspect of external equilibrium. The derived output gap may be used to infer on in‡ation to the extent as external de…cits carry in‡ationary risks. By using this approach we do not discard the usual Phillips-curve based models. It is evident that the importance of the output – external de…cit trade-o¤ is a tradable sector phenomenon. The larger the share of the nontraded sector in a country the less the information that this trade-o¤ might give about in‡ationary pressures and the more is the signi…cance of the Phillips-curve e¤ect. However, even though our model does not tell the whole story for any country, it can be useful not only for Hungary but for any small open economy. In Section 2 we present the conceptual framework and set up the long-run equations. Section 3 describes the statistical representation of the model and 6

When Darvas–Simon (2000) construct capital stock …gures for 1980-98, for the transition years they have to estimate ”e¤ective” (productive) capital stock …gures derived from the output data. It would be meaningless to revert the estimation and use these data to get information on potential output.

2

presents the empirical results. Section 4 brie‡y presents results for Finland and Poland. Section 5 gives a further discussion on time series decomposition related to the model. Section 6 concludes. Appendices provide further details, such as the calculation of the world demand index, the preliminary data analysis, the state-space representation of the system, and data sources.

2

The Conceptual Framework

2.1

Supply as Competitiveness in an Open Economy

Traditional trade models7 assume that supply of a country manifests itself in the relative price of its variant of the exported good. If supply increases it can be sold only at a lower relative price. In this sense the relative price is considered the supply indicator. Although planned sales and relative prices (costs) correlate, they are not the same notions. If supply increases, lower sales price is only one of the channels of expansion. The supplier may enter more and more micromarkets with the same price. In other words it may increase the variants of the good supplied.8 In practice we say that the given supplier has increased its competitiveness, and this competitiveness does not show up in lower sales prices. Thus it is not satisfactory to represent supply by prices. It is better to have a variable that quanti…es planned sales directly. This idea is reinforced by the statistical di¢culty to …nd an aggregate relative price that is free from compositional distortions.9 We do not deny the relationship between prices and volumes but we believe that …rstly, relative price is not the only determinant of sales volume and secondly, in a small country it cannot be captured reliably by aggregate statistics. Supply is a function of prices and costs. We did not take the challenge to measure costs directly and model its determinants. Instead we used the information on sales volume directly by using the following properties of relative supply, de…ned as the logarithmic di¤erence of domestic and rest of the world supply. 1. If in a market any supplier increases its share relative to total demand, this indicates an increase in its relative supply. 2. In the long run relative supplies of countries are equal to their relative incomes. In other words, income and supply are cointegrated. 7

Trade blocks of macroeconometric models originating from the idea of Armington (1969) belong to this ”traditional” group. 8 Seminal papers of Dixit–Stiglitz (1977), Lancaster (1979), and Salop (1979) opened up the literature on ”economies of scope”. Simon (1992) estimated an econometric model that incorp orates this e¤ect. 9 In aggregate data price di¤erences arising from relative prices of variants by place of origin are mixed with price di¤erences arising from divergences of price indexes of goods in di¤erent baskets of exports.

3

The …rst assumption de…nes an ordinal relationship among relative supplies of countries. The second assumption de…nes a scale for this variable.

2.2

The Model

In our model we assume two markets and two producers, domestic and rest of the world. Planned purchase and sale is de…ned separately for these two. There are three goods: export goods, goods for domestic use, and import goods. Export goods are not substitutes for other goods in domestic consumption. Domestic goods and import goods are substitutes. In production supplies of export and domestic goods are substitutes. Sales volume depends on income and relative supply, which manifests itself in relative prices and the number of variants entering the market. Income abroad is measured by a time-varying weighted average of the partner countries’ GDP. For the domestic demand total expenditure is used instead of income in order to avoid modeling saving behavior. As mentioned before, prices as functions of supply and sales as function of prices are not modelled. Instead supply determines sales directly. Similarly to prices that enter trade equations in relative term, supply as a determinant of sales can be interpreted only in relative terms compared with trading partners. This is plausible if we consider that sales plans cannot be done without taking competitors plans into account. Higher share can be achieved only by outperforming competitors in costs, price, quality, and number of varieties. Taking all this into account we formulate export and import equations by substituting relative supply terms for the relative price variable of the ”traditional” trade equations. Equations (1) and (2) show the long-run relationships for trade ‡ows.10 xt = a + B dwt + ¡ rq¤t + ei¤t + "(x) t

mt = d + ª ddt + © rq¤t +

Xt¡1 ¤ (m) eit + "t Mt¡1

(1)

(2)

where lowercase letters indicate logarithms of capital letter variables, Xt; Mt : exports and imports of goods and services de‡ated by the de‡ator of domestic expenditures, ddt : total expenditures in the home country, dwt: world demand, measured as time-varying weighted average of 65 trading partners’ GDP (see Appendix 8.1), 10

Throughout the paper unobserved variables are denoted by asterisks.

4

rqt¤: relative supply, de…ned as the di¤erence between domestic (qt¤) and foreign (qwt¤) supplies, ei¤t : indicator of economic integration, "(x) and "(m) t t : residuals, a; B; ¡; d; ª; ©: parameters. By using a uniform domestic expenditure de‡ator for de‡ating trade ‡ows we implicitly interpret terms of trade changes in a special way. If the terms of trade improves, then it means that we supply something which is more valuable for the world. From a domestic point of view this is equivalent of a higher volume of output. This interpretation, besides being plausible, assures the exclusion of e¤ects of possible permanent changes in the terms of trade, that otherwise may disturb the cointegration of volume data. 2.2.1

The Latent Integration Variable

Exports and imports usually grow faster than domestic expenditures. Reasons of this are trade liberalization decreasing transaction costs, utilization of economies of scale and scope in intraindustry trade. These e¤ects are captured by one variable, called economic integration. This variable is not observed, we de…ne it by its attributes. It’s main feature is that it increases exports and imports by the same amount. As our model is log-linear, this can be approximated only. To improve the approximation, we make a correction Xt¡1 Mt¡1

to the integration variable in the import equation and use ln(EIt 2.2.2

).

Some Features of Demand and Supply

Traditional trade equation estimations frequently report demand elasticities that are larger than one. In our view this is the consequence of the omission of an important variable, the one which we label as economic integration. Otherwise, for the long run there is no satisfactory explanation for this result. There is no convincing argument that would suggest that goods of a given country have a higher income elasticity abroad while foreign goods have a higher elasticity at home than the elasticity of their competitors’ goods. Therefore, we set B = ª = 1. In the short run, however, the demand coe¢cient might be di¤erent from one. Equations (1-2) imply the following enforced features as a de…nition to relative supply: ² Increasing relative supply increases exports and decreases imports ceteris paribus. Therefore, ¡ > 0 and © < 0. ² If relative supply and relative demand increases at the same rate, the export/import ratio does not change. If income elasticities equal to 1, this requirement is ful…lled by the constraint ¡ ¡ © = 1. This feature 5

allows an equilibrium output interpretation of supply, because in this case we may say that along the path where supply equals demand excess demand is 0. The de…nition implies that if a country posts a higher growth of both supply and demand than the rest of the world, then its trade as a share of GDP will shrink (unless integration increases). 2.2.3

Transferring to an Absolute Measure of Supply

The two features described above de…ne supply only in a relative sense. The following de…nition maps it into country levels. ² In the long run supply equals demand both at home and in the rest of the world. The gap between them can de described by (stationary) cycle variables. We assume that world supply is exogenous and estimate it with a univariate model independently. That is, world demand is assumed to be the sum of world supply and a cycle variable: dwt = qw¤t + gapwt¤

(3)

where qw¤t is the supply and gapwt¤ is the stationary output gap of the rest of the world. For the sake of simplicity and for data availability reasons demand of the rest of the world is considered to be identical to GDP. In the domestic economy GDP is equal to the sum of domestic demand (de…ned as total domestic expenditures) plus net exports. As demand is cointegrated with supply and GDP is cointegrated with demand, GDP is also cointegrated with supply and can be decomposed as the sum of supply and a stationary cycle: gdpt = qt¤ + gap¤t

(4)

where gdpt is the logarithm of GDP, de…ned as the (log of) current price GDP de‡ated by the de‡ator of domestic expenditures, and gap¤t is the stationary output gap. Using rqt¤ ´ q¤t ¡ qw¤t and equations (3) and (4) GDP is decomposed as gdpt = rq¤t + qw¤t + gap ¤t

6

(5)

2.3

Comparison with Other De…nitions and Models

Our notion of supply does not imply anything about the volatility or smoothness of this variable. In this sense our de…nition di¤ers from the de…nition implied by the Hodrick–Prescott …lter, the most commonly applied univariate method, that is based on assumed smoothness. The di¤erence from the approach of the unemployment-in‡ation based model can be summarized in the following way: In the unemployment-in‡ation based model output is above sustainable – and in‡ationary – if unemployment is below the natural rate. High external de…cit is not considered to provide useful information in this respect, because it contains two components that are di¢cult to disentangle: it is partly a phenomenon of an equilibrium process that reallocates the use of production intertemporally, and only partly a by-product of domestic excess demand. Splitting unemployment data into equilibrium and disequilibrium components seems to be an easier task by using in‡ation data. In our model demand and the associated output is above sustainable – but not necessarily in‡ationary – if the rate of export growth to import growth is slower than justi…ed by relative demands. This sustainability de…nition is not equivalent with in‡ation-sustainability. It is based on the idea that any de…cit is transitory, however persistent it may be. Because any credit has to be repaid some time, any demand that generates debt has to be reverted some time to repay the debt. In principle the model could be extended to calculate in‡ation-sustainability. De…cits had to be decomposed into ”equilibrium de…cits” and ”excess demand” de…cits. The adequate information for this separation would be in‡ation data, just like in the case of the unemployment decomposition. However, the impact of de…cits to in‡ation is transmitted through various channels like exchange rate policy and investor’s business sentiments that probably result in a large variance both in the coe¢cient and the lag of this impact. We did not undertake the hopeless task of such an extension. However, as it was told in the introduction, the choice of the alternative, unemploymentbased, approach to in‡ation-sustainability was not open for the period and country considered. There is nothing left but to be satis…ed with the concept of our external-balance-sustainability. Even though the concept might be considered as a substitute, we are convinced that it leads to a useful indicator. It is true that its application as a predictor of in‡ation is limited, but by using additional information on the assessment of the market of the ”equilibrium de…cit”, it may give information on in‡ationary pressures in the economy. To avoid misunderstanding, it is probably better to call our competitivenessbased concept as sustainable output, thus distinguishing it from the popular in‡ation-neutrality-based concept called potential output. 7

2.4 2.4.1

Short-run E¤ects Income Elasticities and Coe¢cients

We argued that in the long-run demand elasticities are unity when the economic integration variable is incorporated into the model. In the short run, however, there is a possible explanation for a coe¢cient higher than one provided that we do not interpret it as an income elasticity. Import has a role of …lling out short-run excess demand gaps. If demand goes up too fast, then domestic suppliers who deliver on the basis of long-run customer relations may not follow demand and the excess has to be met from external sources. Later on domestic supply adjusts and trade shares are restored. To capture the short-run nature of this behavior we set up an error correction model where the short-run coe¢cient may exceed 1. The only plausibility test that the coe¢cient has to pass is that any change in demand should not generate more import than the change in demand itself. The estimated parameters turned out to be plausible in this sense. 2.4.2

Import Content of Exports

The integration variable is supposed to capture the two-way increase in trade. Trade growth due to both the increase of varieties and to a deepening of vertical cooperation is included in this category in principle. However, the latter has a special feature in the Hungarian economy because it is one-sided: it typically means an import of materials and semifabricates and an export of the processed items and not vice versa. Therefore, this trade is very much driven by external demand. As a consequence, it is not only exports but imports as well that is external demand dependent. In the long run, this e¤ect is captured by the integration variable through the cointegration relationship. In the short run, however, the explanation of imports has to include external demand, that’s why we included exports in the equation. 2.4.3

Exchange Rate E¤ects

When de…ning supply we implicitly assumed that supply behavior is indifferent towards the direction of trade; the share of exports and domestic use in sales is determined by relative demands. This indi¤erence prevails if profitability is the same in both directions. In the long run, arbitrage equalizes this pro…tability. However, arbitrage works slowly and in the short run – especially because of the volatility in exchange rates – relative price di¤erences may exist and suppliers may prefer one direction to the other. This short run ”discrimination” in supply behavior could be captured by including the real exchange rate into the error-correction formulation of the model.

8

In spite of the theoretical arguments the existence of a relative price e¤ect could not be convincingly con…rmed for Hungary. The real exchange rate e¤ect was signi…cant for 1991-199911 in an experimental estimation of the import equation but not in the export equation. We did not include it into the …nal speci…cation partly because the asymmetric behavior of exports and imports would remain a puzzle and partly because there are good arguments against its inclusion. Until 1990 the economic role of the exchange rate was very small and it should not be surprising that a market behavioral e¤ect could not be substantiated. From 1991 behavior has changed but there was hardly any observation suitable for testing the exchange rate e¤ect. Until 1994 the path of the real exchange rate did not indicate changes in pro…tability because of the restructuring of the price and subsidy system. Since 1995 there was not enough ‡uctuation in the exchange rate because of the crawling peg system.

3

Statistical Representation and Results

3.1

Statistical Representation

We used three types of unit root tests to check the integratedness of the observed series. These are (1) Augmented Dickey-Fuller t-test (ADF), (2) Phillips-Perron t-test (PP), (3) Kwiatkowsky-Phillips-Schmidt-Shin ¹ and ¿-tests (KPSS). For ADF and PP the null hypothesis is unit root while for KPSS the null is stationarity and trend-stationarity. Detailed statistics are reported in Appendix 8.2. Two series proved to be stationary by any tests: the trade balance and the export/import ratio. All other variables are integrated processes, but the choice between I(1) and I(2) has to be based on contradictory test results. Some tests suggest an I(1), others an I(2) conclusion. However, we do not trust much in unit root testing.12 In fact, by inspecting for example the growth rate of Hungarian GDP in Figure 1, it is clear that we have to choose between two candidate processes: I(2)13 or I(1) with shifts in the drift (we will refer the later as ”breaking I(1)”). In the …rst representation the growth rates of our variables are approximated as driftless random walks, in the second case they have constant means which shift occasionally. We decided for an I(2) speci…cation by comparing the pros and cons of both representations. 11

That is, we constrained its parameter to zero for earlier years. When we included the real exchange rate unconstrained for the full period then its coe¢cient was not signi…cant even for imports. 12 The literature is full of papers showing the power and size distortions of various unit root tests. As a particular importance for our model, Harvey–Jaeger (1993) show that conventional tests applied to I(2) processes are biased toward the I(1) …nding. See more details in App endix 8.2.2. 13 See Haldrup (1998) for a thorough discussion on working with I(2) variables.

9

The arguments used against the I(2) representation usually refer to the fact that it implies a non-stationary growth rate with a variance that goes into in…nity with time. The starting point of this argument is that estimations of variables with in…nite variance give no information. In our view this argument is weak. Any non-stationary variable ends up in in…nite variance, even the I(1) model implies in…nite variance for the level of the series. Even worse, if we allow the drift to shift in the I(1) case then the uncertainties of the timings and magnitudes of the shifts blow up the variance of the growth rate as well. Therefore, we do not regard in…nite variance at in…nity as an important factor in choosing between an I(2) and a breaking I(1) representation. It is natural to assume that economic growth rates are determined by the prevailing political-social-economic systems. In history certain countries or regions grew faster, others slower, and their position has changed with time. We may consider these changes as shifts in growth regimes, but why should we assume that the shifts are discrete? In fact, the transitions are always gradual. Discrete regime shifts of a growth process might have been represented by an I(1) process with several breaks in the drift. However, if shifts are frequent and gradual, regimes themselves are not persistent enough to de…ne them as a regime. In this case no drift parameters may be de…ned and the process is better described by considering the continuously changing drift as a random change in the growth rate. This is the situation when the I(2) representation is an ideal approximation to describe a process that shows persistent ‡uctuations in growth rates. In a country like Hungary where several regime shifts took place within a relatively short period a breaking I(1) model needs too many parameters to follow all the shifts and assumptions on discrete shifts have to be pulled out of thin air. Therefore the lack of a constant steady-state growth rate is rather an advantage of the model than a drawback. Even for major industrial countries the steady-state growth is reassessed from time to time indicating that there are indeed persistent changes in the growth rate. The fact that the I(2) model does not give us a constant long-run growth rate does not mean its inferiority in its forecasting capabilities against a breaking I(1) representation. Making forecasts based on the growth rates of the most recent years is not worse than a forecast based on the last regime of a breaking I(1) model. We have to say that using processes that have undesirable properties at in…nity is not our invention. In most of the unemployment based models of potential output the unemployment rate is assumed to be an I(1) process. Authors adopting this assumption certainly know that the unemployment rate can not be I(1) (an obvious reason is that it is between zero and one by de…nition), but for empirical estimation of a given time period this as-

10

sumption gives reasonable approximation.14 We adopt a similar approach to the rate of growth: we do know that a forecast on the basis of the information extracted from the observation period will lead to an ever widening con…dence band, but this does not constitute a major concern for us if our purpose is to describe the observed 40 years period. Therefore, we have a …rm preference for the I(2) representation. Consequently, we assume that exports, imports, domestic demand, world demand, relative supply, and economic integration follow I(2) processes. For the interested reader we estimated the breaking I(1) model as well for comparison. As it is rather di¢cult to test for the number and location of shifts we imposed the possible break points based on visual inspection. From this inspection it is clear that during the centrally planned area there was a marked decline in the growth rate at the end of the seventies. The transition period was very much di¤erent from both the previous and the subsequent regime. Therefore, we assumed 3 shifts in the drift and consequently four subperiods for Hungary: 1960-1978, 1979-1989, 1990-1993, and 1994-1999. Ideally, regime changes should be contemporaneous in Hungary and in the rest of the world. Since the former socialist countries constitute a signi…cant part of the rest of the world this assumption might be reasonable. Indeed, four breaks seemed to be necessary for the rest of the world, but the visual inspection suggested that two of them di¤er from the Hungarian breaks by a year (1977, 1994). As we have already indicated, we estimated world supply independently.15 The implied output gap of the rest of the world is displayed in Figure 2; actual growth of world demand and implied growth of world supply are shown in Figures 3-4. The I(2) model indicates a more or less balanced world demand up to the eighties and a signi…cant cyclical downturn in the early nineties, while the breaking I(1) model implies huge variation in the cycle in the sixties and seventies and just a small cyclical downturn in the nineties.

3.2

Long-Run Results

We have estimated the model in two steps: …rstly the long-run and secondly the short-run relationships. We adopted a state-space representation and used the Kalman-…lter for evaluating the likelihood function. Equations (1), (2), and (5) are the so called observation (or measure14

Another example is the time-varying parameter estimation. Here a random walk is frequently assumed for the stochastic behavior of time-varying parameters in order to be able to capture p ermanent changes, although the very nature of parameters b earing economic interpretation is that they are not I(1) at in…nity. 15 In the I(2) case we …tted a local linear trend plus cycle (LLTC) model for world demand (dwt ) to estimate world supply that assumes a univariate I(2) process. In the breaking I(1) case we assumed that world supply also follows a breaking I(1) process. Details of the decomposition are given in Appendix 8.4. As world supply was estimated independently before estimating the whole model, standard errors of parameters of domestic demand and cycle are estimated unprecisely.

11

ment) equations in the state-space. Note that having imposed the suggested parameter restrictions the only ’structural’ parameter to be estimated is ¡. (The ’non-structural’ or ’technical’ parameters are the standard errors of "(x) and "(m) t t ). The state (or transition) equations describe the dynamics of the latent variables. As the model is not stationary, we have to give initial values for the mean and variance-covariance matrix of the state vectors (latent variables). The following numbers were chosen16 : Mean: ² output gap: 0 ² level of relative supply: the …rst observation of relative demand minus average growth ² integration: ln(1) = 0 ² growth rate of relative supply: average growth of relative demand ² growth rate of integration: 1% Variance-covariance matrix: Identity matrix multiplied by 0,0001, which means that 1% standard error was assumed for the initial values of the state vectors. The state equations are the following. The output gap speci…cation shown in equations (6) is the same for both the I(2) and I(1) with shifts in the drift model:17 ·

gap¤t gg ap¤t

¸

=½

·

cos µ sin µ ¡ sin µ cos µ

¸·

gap¤t¡1 gg ap¤t¡1

¸

+

"

(g1)

"t (g2) "t

#

; gap¤0 = gg ap¤0 = 0 (6)

16 All …ve values are needed for the I(2) model but only the …rst three is needed for the breaking I(1) model. 17 In the literature two kinds of cycle sp eci…cation are used: (1) ARMA representation (eg. Kuttner 1994, Gerlach-Smets 1999), and (2) functions of trigonometric functions as in equation (6) (eg. Harvey 1989, Harvey-Jaeger 1993, Harvey-Koopman-Penzer 1999). The latter may be derived from the following equation: gap¤t = a cos (µt) +b sin (µt), where ¡ ¢ 1=2 µ is the frequency of the cycle in radians and a 2 + b 2 is the amplitude of the cycle. Equation (6) is a generalization of the latter (i) enabling a and b to change in time (ii) adding stochastic disturbances and (iii) adding a dampening factor (½). The gg ap¤t ”shadow ¤ variable” is constructed only to generate gapt . For identi…cation we should assume that (9) (10) "t and "t have the same variance and they are uncorrelated. The AR(1) cycle is a special case of this model when µ = k¼ (k 2 Z ). In this case sin(µ) = 0, so the generating (9) equation for g gap ¤t is redundant and the cycle is represented either as gap¤t = ½gap¤t¡1 +"t (9) (when µ = :::; ¡2¼; 0; 2¼; :::) or as gap¤t = ¡½gap¤t¡1 + "t (when µ = :::; ¡¼; ¼; 3¼; :::). See, e.g. Harvey (1989) pp.38-40).

12

where the gg ap¤t ’shadow variable’ is needed to generate gap¤t , ½ is a dampening factor, µ is the frequency of the output gap in radians, "(i) t are white noise processes. In case of the I(2) model the behavior of relative supply and economic integration is described as ¢¢rqt¤ = "(rq) t

(7)

(ei)

(8)

¢¢ei¤t = "t

In case of the I(1) with 3 shifts in drift model equations (7) and (8) is replaced by (rq¤ )

(rq ¤ )

(rq ¤)

(rq¤ )

(rq)

¢rqt¤ = ¹t;1 dt;1 + ¹t;2 dt;2 + ¹t;3 dt;3 + ¹t;4 dt;4 + "t

¤

¤

¤

¤

) (ei ) (ei ) (ei ) (ei) ¢ei¤t = ¹(ei t:1 dt;1 + ¹t:2 dt;2 + ¹t:3 dt;3 + ¹t:4 dt;4 + "t ¤

(9)

(10)

¤

) ) where ¹(rq and ¹(ei are the drift terms for the 4 periods and dt;i are t;i t;i dummy variables having value 1 during period i and zero otherwise.18 We note that in equations (7-8) the only parameter to be estimated is n o (i) "(i)) ( the standard error (¾ ) of the innovation "t . Note also that in the I(1) with 3 shifts in drift model the number of parameters to be estimated is larger by 8 than in the I(2) model, but the number of initial conditions are less by 2. We also underline that the I(2) model includes only two additional ’structural’ parameters: ½ and µ, while the I(1) with shifts in drift model needs further 8. In our view the small number of structural parameters and the unit restrictions on most other long-run parameters due to economic reasonings is a great advantage of our model: it can be applied to various economic systems. Both in socialist and market economies the increase in demand increased trade ‡ows and we do not expect this simple relationship to be changed at any point in time. On the other hand, the underlying economic structures of Hungary has changed substantially in the sample period. This is exactly what is captured by the I(2) nature of the levels of the series which allow the growth rates to change permanently over time. These features of the model give us con…dence in applying it to both a planned economy and to 18 In case of the I(2) model the interrelations of the above equations is described in detail by the state-space representation of the system in Appendix 8.3.

13

a market system, and also to many other small open economies undergoing large structural changes. Estimation results were reasonable with the exception of b ¡, which was highly sensitive to starting values and was frequently out of the 0-1 range. Therefore, we constrained its value to 0.5.19 The results are given in the following table. Most estimated parameters are signi…cant with the correct sign and their magnitudes are interpretable in economic terms. It is remarkable that the common coe¢cients of the two models are rather similar with one exception. The exception is the standard error of the relative supply in case of the breaking I(1) model, which is estimated to be zero. This means that relative supply has been estimated to be a broken deterministic trend in this case. The likelihood function to be maximized is written in terms of the innovations20 , which are assumed to be normally distributed. As we can seen from the table, the Jarque-Bera test does not reject the assumption of normality in cases of the …rst and the second innovation, but it does for the third. However, even if the innovations were not normally distributed, the Kalman–…lter would still allow quasi maximum likelihood estimation. KPSS and PP statistics shown in the table are test results for the stationarity and unit root of the cointegrating vectors. There are no critical values available for our models. Using critical values developed for OLS estimation the KPSS tests do not reject stationarity, and the PP test rejects unit root in the case of the export cointegration vectors. However, the test statistics for the vector of imports are also close to OLS critical values, so keeping in mind the weaknesses of unit root tests we conclude that our model indeed maps the I(2) series into I(0). Figures 5-6 show the estimated latent variables of relative supply and integration. As it can be seen, relative supply is rather di¤erent (I(2) versus breaking deterministic trend) but economic integration is virtually the same in the two models. Although con…dence bands are wide, the point estimates have plausible economic interpretation. For example, integration in Hungary deepened fast during the reform period of 1968-74, then decreased during the declining period of the planned system, bottomed out when the structural switch of trade took place after the collapse of the Soviet-dominated system, and rebound again after transition. The output gap is estimated to be stationary in both models since the dampening coe¢cients are signi…cantly less then one. Figures 7 and 8 show the estimated output gaps in their two times standard error bands. Finally, 19

We have checked several values b etween 0.3-0.7 and estimates showed negligible sensitivity to the actual choice within this range. 20 Innovations are de…ned as vt = yt ¡ E t¡1 [yt], where yt denotes the vector of left hand side observable variables: yt = [xt mt gdpt].

14

…gures 9 and 10 show the actual and potential growth rates and the output gap together. The output gaps derived from the two models for 1993-99 are rather similar. However, results for the transition period are di¤erent: the I(2) model suggests only a slight negative gap of 2 percent while the breaking I(1) model implies a much larger cyclical downturn. The later is a straightforward consequence of the assumptions enforced by the choice of the I(1) representation. In our view the gradual pick-up of supply as implied by the I(2) model after the transition is much plausible then the abrupt change in potential growth after 1993. (We underline that the potential growth rates shown in the …gures are the expected values of growth. The actual growth of potential output di¤ers from its expected value by the realization of a white noise process with zero mean.)

3.3

The Error Correction Model

In the second step we estimated an error-correction model consistent with the I(2)-ness assumption. There are several possible cointegration relationships between I(2) variables. However, since we regard equations (1-2) as long-run relationships, we assume that the error terms attached to them are stationary, that is, the linear combinations of (xt ¡ ei¤t ¡ dwt ¡ ¡rqt¤) and ³ ´ t¡1 ¤ ¤ mt ¡ X Mt¡1 eit ¡ ddt ¡ (¡ ¡ 1) rqt are stationary. These assumptions were tested and not rejected above. This implies that the error correction model should be formed in terms of only second di¤erences. The coe¢cient of the latent variable of economic integration is 1 by the de…nition of the concept (for imports the X=M term modi…es the coe¢cient somewhat). The constraint for the supply-coe¢cient is valid for the short-run as well. The total e¤ect of supply on exports and imports is equal to the long-run coe¢cient, i.e. ¡, and ¡ ¡ 1: As imports are export-dependent in the short run, the partial e¤ect will be (¡ ¡ 1) ¡ »¡.21 Similarly, in the import equation the double rate of change of integration has the coe¢cient (1 ¡ »). With these restrictions we set up the following error correction model:

21

¢¢xt = a + ¯¢¢dwt + ¡¢¢rqt¤ + ¢¢ei¤t ¡ ¢ (3) +¸ xt¡1 ¡ ei¤t¡1 ¡ dwt¡1 ¡ ¡rq¤t¡1 + "t

If ¢¢xt = ¡¢¢rqt¤ and ¢¢mt = K¢¢rq ¤t + »¢¢xt ; then d¢¢xt d¢¢m t @¢¢mt d¢¢xt = K + »¡: d¢¢rqt¤ = ¡ and d¢¢rqt¤ = K + @¢¢x t d¢¢rq ¤ t d¢¢m t As d¢¢rq ¤ = ¡ ¡ 1 by assumption, K = ¡ ¡ 1 ¡ »¡. t

15

by di¤erentiation

(11)

¤ ¢¢mt = d + ³»¢¢xt + !¢¢ddt + ((¡ ¡ 1) ¡ »¡) ¢¢rq ´ t Xt¡1 ¤ Xt¡3 ¤ t¡2 ¤ + (1 ¡ ») M eit ¡ 2 ¤ X t¡2 Mt¡2 eit¡1 + Mt¡3 ei´ t¡1 ³ Xt¡2 ¤ (4) ¤ +¼ mt¡1 ¡ Mt¡2 eit¡1 ¡ ddt¡1 ¡ (¡ ¡ 1) rqt¡1 + "t

(12)

In the alternative model the error correction formulation is written in …rst di¤erences.22 For comparison, we estimated the ECM model for both the …rst and for the second di¤erences in both cases. The Table 2 shows the results. Every parameter has the expected sign and most of them proved to be signi…cant. Including export growth into the import equation seems to be justi…ed. In case of the exports equation estimate for the second di¤erences, the point estimate of the error correction coe¢cient was larger than 1 in absolute terms. This coe¢cient is not consistent with the underlying model. However, the estimated standard error was so large that we could not reject the null hypothesis that this parameter equals -0.8, the error correction coe¢cient of the imports equation. Therefore, we estimated the models imposing this constraint. Results con…rm our suspicion that demand coe¢cients might be larger than one in the short run. Fits of the equations are reasonably good even for the second di¤erences. The I(2) model seems to have better explanatory power than the breaking I(1) model in case of imports, while for exports the two models achieve virtually the same …t. We argued that the long-run relationships of our model include parameters restrictions derived from theory, therefore, we do not expect any break in these parameters. However, the short-run parameters might have changed. For example, we cannot exclude that the short-run response of trade ‡ows to changes in demands or the error correction coe¢cient changed in time. We tested this hypothesis by applying Chow tests for each date of the sample between 1965-95. The p-values of these tests are shown in Figure 11. As we can see the tests are not signi…cant, indicating – somehow contrary to our expectations – that even these parameters are stable. 22

Since the change of the drift of world demand and supply di¤ers from that of Hungarian variables at two break p oints we should have correct it in the model for exports. However, as the change in the drift di¤ers only by a year in both cases, we did not adjust the regression but assumed that this shock show up in the residuals.

16

4

Results for Poland

We have estimated the model – adopting the I(2) representation – for Poland as well. The necessary national accounts data were available only since 1980, and data availability further constrained us to use the GDP de‡ator instead of the de‡ator of domestic expenditures in case of all four variables (Xt ; Mt ; DDt ; GDPt ). Figures 15-16 show estimated state vectors, output gaps, and the growth rate of potential output. Estimated coe¢cients were signi…cant, they had correct signs and reasonable magnitudes with the exception of the dampening factor of the cycle (½): it was estimated to 1.04 with a standard error 0.06. This clearly indicates the nonstationarity of the estimated output gap, which would imply that GDP and supply are not cointegrated23 , as we have assumed in forming equation (4). Probably the small sample size (20 annual data) is responsible for this result. If we do not pay much attention to this problem, we may …nd some remarkable results. In 1981-82, during the crisis and martial law period, not just GDP declined substantially, but supply and economic integration as well. In the early nineties, during the system transition, the estimated supply and output gap is quite similar to the Hungarian case: the transition was accompanied by a signi…cant fall of supply and the resulting negative gap was small. Actual growth was higher than sustainable until 1998 and consequently the output gap reached a relatively high positive value (5%) by the end of the nineties. It is also remarkable that sustainable growth was estimated to be very stable in 1995-99.

5

A Note on Time Series Decomposition and Model Choice

All methods of decomposition are based on some preconceptions, even our model is not an exception from this claim. The Hungarian economy has experienced a dramatic transformation when it moved from a relatively ’liberal’ planned economy to a market one. Can a ”standard” model, which decomposes output into supply and demand shocks, help us in the analysis of the transition period? Due to the transfer to a system of market pricing the economy was shocked by enormous relative price changes which resulted in a structural change in demand during transition. The structure of supply could not adjust to this change quickly. Therefore, capacities became redundant while new capacities were established only in a gradual evolutionary process. 23 In statistical terms they still can b e cointergated, since these variables are I(2), so an I(1) linear combination implies cointegration. In economic terms, however, this is uninteresting.

17

Meanwhile excess demand and excess supply existed side by side and aggregate output decreased because the short-side rule prevailed in each micromarket. Decrease in output brought about unemployment. What was the nature of this shock, supply, or demand? The de…nition based on the concept of demand as aggregate planned purchases and supply as aggregate planned sales does not help much in the answer as it was the structure and not the aggregate amount that di¤ered. A demand shock creates excess demand in the short run but does not e¤ect output in the long run. A supply shock related mostly to technical changes has permanent e¤ects. Generally, supply shocks may have transitory e¤ects as well. A monetary squeeze e¤ects both supply and demand temporarily, as supply is constrained through the credit channel. Similarly, a structural shock may have transitory e¤ects. This variety of the nature of e¤ects in supply and demand shocks has led analysts to the reinterpretation of demand and supply shocks: in several context they do not mean changes in planned purchases or planned sales but only the temporary or permanent nature of the shock. The question may arise then, whether the recession of the early 1990-ies in the transition economies was the result of a temporary or a permanent negative shock. On one hand, the high rate of unemployment that have arisen would suggest that the shock was temporary. On the other hand, it is clear that the persistence of the crisis is longer than the usual excess-demand driven business cycle recessions. If output drops below equilibrium because of a lack of aggregate demand, then it is the speed of price adjustment that determines the length of the impact of the shock. However, if output drops because of a structural mismatch, not only prices have to adjust, but the structure of supply. This is presumably much slower than price adjustment, because it requires the establishment of whole new production cultures. The inertia in this process is too large to be explained by pure construction costs: uncertainties owing to limited information constrain the speed of adjustment decisively.24 How long does the e¤ect of a transitory shock last? In practice in …nite samples it is di¢cult to separate shocks which have an autoregressive representation with dominant (inverted) roots that are 1 from those which have roots less than 1. Sometimes it is useful to consider some roots to be 1 even though theory would tell that they are less than 1. This way some shocks that are transitory in theory may be considered as permanent in some models. This has been done implicitly in our model when we disregarded unemployment as an information on transitory shocks. In Hungary, even though unemployment have risen from 0 to13 percent after the system change in 1990 and it is still around 7 percent, we did not assume unemployment to be an important indicator of transitory changes in output. 24

See Stiglitz (1992) for a thorough development of this argument

18

6

Conclusions

The motivation of our model is twofold: (1) the unemployment-in‡ation based decomposition of output into a potential and a transitory component is unfeasible in the case of Hungary, and (2) in all small open economies excess demand may lead to increases in imports without an e¤ect on in‡ation. Therefore, we set up a model for measuring potential output that uses information given by external trade. The heart of the model is a set of trade equations augmented with two latent variables, relative supply and economic integration. We argue that in general, trade equations should incorporate a supply variable directly and should also include a variable capturing the e¤ect of intraindustry trade and other consequence of economic integration. We have treated these variables as unobserved and de…ned them by their properties. The Kalman-…lter was adopted for inference on these variables. The long-run relationships de…ned by the trade equations are general enough to be applied to both planned and market economic systems and even a system in transition between the two. For statistical estimation of the model we argued for the I(2) representation, that provides a ‡exible approximation of the case when growth rates of the underlying processes change permanently and gradually. Indeed, the I(2)-ness of supply can easily capture the large structural changes in the economy occurred during our sample period 1960-99. Empirical results indicated that the huge fall of GDP in the early nineties in Hungary was attributable mostly to a fall in supply even though demand has contributed somewhat to the fall as well. For comparison, we have also estimated an alternative statistical representation in which our variables are approximated as I(1) with occasional shifts in the drifts. Although many coe¢cients were remarkably similar in the two statistical models, in our interpretation the I(2) model described much better the transition period. In the second half of the nineties, when the Hungarian economy showed a relatively stable growth rate, both models delivered reasonable and similar results. The error correction models were slightly better for the I(2) representation. Even though we argued that the relation between the output gap de…ned by our model and in‡ation is rather remote, a visual analysis of the output gap data shows that some relation exists. For example, in 1993-94 in Hungary the gap took large positive values which were followed by a pick up in in‡ation a year later. As the gap moved toward zero in the second half of the nineties, the economy experienced a steady decline in in‡ation.

19

7

References

Armington, Paul S. (1969): A Theory of Demand for Products Distinguished by Place of Production. IMF Sta¤ Papers, Vol.16, No 1, pp. 159-178. Canova, Fabio (1998): Detrending and Business Cycle Facts, Journal of Monetary Economics, Vol. 41, pp. 475-512. Darvas, Zsolt – Simon, András (2000): Capital Stock and Economic Development in Hungary, Economics of Transition, Vol. 8, No. 1, pp. 197-224. Dupasquier, Chantal – Guay, Alain – St-Amant, Pierre (1997): A Comparison of Alternative Methodologies for Estimating Potential Output and the Output Gap. Bank of Canada Working Paper No. 97-5. Dixit, Avinash K. – Stiglitz, Joseph E. (1977): Monopolistic Competition and Optimum Product Diversity. American Economic Review, Vol. 67, No. 3, pp. 297-308. Gerlach, Stefan – Smets, Frank (1999): Output Gaps and Monetary Policy in the EMU Area. European Economic Review, Vol. 43, pp. 801-812. Giorno, Claude – Richardson, Pete – Roseveare, Deborah – van den Noord, Paul (1995): Estimating Potential Output, Output Gaps and Structural Budget Balances. OECD Economics Department Working Papers No. 152. Haldrup, Niels (1998): An Econometric Analysis of I(2) variables. Journal of Economic Surveys, Vol. 12, No. 5, pp. 595-650. Harvey, Andrew C. (1989): Forecasting, Structural Time Series Models, and the Kalman Filter. Cambridge U.K.: Cambridge University Press. Harvey, Andrew C. – Jaeger, Albert (1993): Detrending, Stylized Facts and the Business Cycle. Journal of Applied Econometrics, Vol. 8, pp. 231-47. Harvey, Andrew C. – Koopman, Siem-Jan – Penzer, Jeremy (1998): Messy Time Series: A Uni…ed Approach. Advances in Econometrics, Vol. 13. Kuttner, Kenneth N. (1994): Estimating Potential Output as a Latent Variable. Journal of Business & Economic Statistics, Vol. 12, No. 3, pp. 361-368. 20

Lancaster, Kelvin (1979): Variety, Equity, and E¢ciency. New York, Columbia Press. Laxton, Douglas – Tetlow, Robert (1992): A Simple Multivariate Filter for the Measurement of Potential Output. Bank of Canada Technical Report No. 59. Rasi, Chris-Marie – Viikari, Jan-Markus (1998): The Time-Varying NAIRU and Potential Output in Finland. Bank of Finland Discussion Papers 6/98. Salop, Steven C. (1979): Monopolistic Competition with Outside Goods. Bell Journal of Economics, Vol. 10, No. 1, pp. 141-156. Simon, András (1992): An East Asian Model of Price and Variety Competition. In Dutta, M. (ed): Economics, Econometrics, and the LINK. Essays in Honor of Lawrence R. Klein. North Holland. Stiglitz, Joseph E. (1992): Capital Markets and Economic Fluctuations in Capitalist Economies. European Economic Review, Vol. 36, pp. 269-306.

21

8

Appendix

8.1

Demand Index of the Rest of the World

For the period 1960-1999 a time-varying weighted average of trading partners’ GDP indices was used to represent world demand. 66 countries were divided into 4 groups: (1) Industrial countries: Austria, Australia, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, United States. (2) Former Soviet Union: USSR, Russia, Ukraine, Belarus, Estonia, Kazakhstan, Kyrgyz Republic, Latvia, Lithuania. (3) Former non-soviet socialist countries: Albania, Bulgaria, Croatia, Czechoslovakia, Czech Republic, Eastern Germany, Hungary, Poland, Romania, Slovenia, Slovak Republic, Yugoslavia. (4) Others: Algeria, Argentina, Brazil, Chile, China, Colombia, Egypt, Hong Kong, India, Indonesia, Iran, Israel, Korea, Malaysia, Mexico, Phillipines, Saudi Arabia, Singapore, South Africa, Thailand, Turkey, Venezuela. We used export data for calculating weights which are available for 19761999. Up to 1976 we used constant 1976 weights and for the rest of the years weights were based on data lagging one year. The reason for using previous years’s weights is that current years weights would include substitution e¤ects in response to demand developments.25 For each year we weighted GDP growth rates of trading partners. World demand index is the chained index of world demand growths. Since country representation in the …rst three group is (almost) perfect, we derived the weight of the fourth group as one minus the weight of the other three. Individual country weights within a group were calculated as to sum up to their group weight. The total representation of the 65 trading partners on average was 92.9% for Hungary and 92.0% for Poland. Figures 13-14 show the distribution of exports among the four group of countries. As GDP data for some countries for certain years were missing, the gaps were …lled with the average of the same group. Figure 12 shows the growth rate of the calculated world demand index and its components for Hungary. 25

An intuitive example: suppose we have exported only to Germany and USSR and exports to USSR disappeared completely in year t because of the huge soviet output fall. If we used weights derived from year t , then the demand index would not show any fall in demand since it would compose 100 percent of German growth.

22

8.2 8.2.1

Preliminary Data Analysis Unit Root Tests

This appendix reports on the tests analyzed in section 3.1. For the ADF and PP tests we do not report results for the case when a time trend is included in the test equation because we could never reject the unit root hypothesis in this case. Critical values for the applied tests are shown in the Table 3. Tables 4-5 show test statistics for the level, …rst di¤erence, and second di¤erence of the series using lags from 0 to 5 in the test equation (ADF) and truncation lags from 1 to 5 in the calculation of heteroskedasticity and autocorrelation consistent covariance (HAC) matrix (PP and KPSS). There are several methods of selecting the appropriate legs in the ADF test equation. These methods suggested lags between 0-3 in most cases, but not every method suggested the same number of lags for a given series. For the PP and KPSS tests (that adopt the Newey-West HAC estimator) there is no consensus on how to determine the appropriate lags. In many cases rule of thumb methods are used, such as the formula 4 (T=100)2=9 . This would give a choice of 3 in our case. Instead of choosing one number we look for various variants: 0-5 lags in the ADF case, and 1-5 truncated lags in the HAC estimation in the PP and KPSS cases. In the tables of the ADF and PP tests we shaded the cases when the unit root hypothesis could not be rejected at the 5 percent signi…cance level. In the KPSS table we shaded those cases where stationarity could be rejected at the signi…cance level of 5 percent. This means that in both tables the shaded areas indicate on the existence of unit root.26 8.2.2

Harvey–Jaeger Arguments for the I(2) Speci…cation

Harvey–Jaeger (1993) adopt an I(2) speci…cation for modelling US and Austrian GNP. In justifying the I(2) assumption they …rst note that both the Box-Jenkins identi…cation and formal unit root tests usually found that US GNP is I(1). However, they claim that since the variance of the growth component is relatively small, the I(2) component may be di¢cult to detect by ARIMA methodology. A convincing demonstration is that they simulate the local linear trend plus cycle model using the estimated parameters for sample sizes 100 and 500 and calculate the autocorrelation function of the …rst di¤erences and test for unit root in it. Both experiments indicate serious biases, for example, the size of the ADF test at T = 100 and 26

Unit root tests for world GDP (denoted as dw) behave strangely: according to ADF and PP tests its level is stationary (although it has a clear upward moving trend) while its …rst di¤erence is not. We cannot explain these results. On the other hand, KPSS tests suggested that the level of this series is not stationary.

23

k = 8 is 74 percent, that is, using the 5 percent critical value it rejects the true null hypothesis of unit root in the …rst di¤erences in 74 percent of the experiments. Economic arguments for two unit roots are also persuasive: ”A trend plus cycle model of the form (1) with ¾2´ = 0 has stationary components with no persistence and a smooth I(2) trend with in…nite persistence. But since the trend is re‡ecting slow long-term changes in growth rates, perhaps arising from demographic changes, innovations in technology, changes in savings behavior, or increasing integration of capital and goods markets, the shock which drive the smooth trend may have no connection with short-term economic policy. Following the extensive literature on the productivity slowdown phenomenon, we may well argue that understanding the reasons for persistent changes in growth rates is one of the key problems in macroeconomics.” (Harvey–Jaeger (1993) pp.242-43.)

24

8.3

State-Space Representation of the Model

Unobserved components (UC) or latent variable models are frequently applied to potential output estimates. These models can be conveniently represented in a state-space form. The state-space representation of the dynamics of an (n £ 1) observed vector time series yt consists of a state equation (or transition equation ) describing the dynamics of the unobserved (r £ 1) state vector ®¤t , and an observation equation ( or measurement equation) describing yt as a function of the state vector and possibly other exogenous variables xt . See Hamilton (1994), Chapter 13 for an excellent discussion. Harvey (1989) gives an extensive and splendid guide through speci…cation, estimation, and testing issues of state-space models. state equations: ®¤t = Tt ®¤t¡1 + ct + Rt v t (r£r)(r£1)

(r£1)

(r£1)

xt + Zt ®¤t (n£k) (k£1) (n£r) (r£1)

observation equations: yt = At (n£1)

(13)

(r£g) (g£1)

+ dt + wt (n£1)

(n£1)

(14)

where E (vt ) = 0 , E (wt ) = 0 for all t

(15)

for t = ¿

=

½ Qt

(16)

for t = ¿

=

½ H t

E

E

¡

¡

vt v0¿

¢

wt w¿0

¢

(g£g)

0 otherwise

(n£n)

0 otherwise

(17)

It is assumed that the disturbance vectors vt and w¿ are not correlated with each other and the state and the observed variables for contemporaneously and with all lead and lags as well. Exogeneity of xt in the observation equation (14) means that xt provides no information about ®¤t+s and wt+s for s = 0; 1; 2; ::: beyond that contained in yt¡1; yt¡2 ; :::; y1 . This system can be generalized to a system in which there is contemporaneous correlation between v t and w0¿ . Given parameters, the unobserved state vector and its variance-covariance matrix can be calculated by the Kalman–…lter. Apart from the parameters, we might be interested in three types of inferences for the unobserved state vector: ®¤tjt¡1 , ®¤tjt , ®¤tjT , that is, the forecast from the previous period, inference for current period t based on all information up to t, and inference 25

for t using the full sample. ®¤tjt is called as …ltered estimate, while ®¤tjT is called as smoothed estimate of the state vector. Similar magnitudes can be calculated for the variance-covariance matrix of the state vector. When the parameters are unknown, Kalman–…lter also allows the evaluation of the likelihood function, therefore, permits maximum likelihood (or quasi maximum likelihood) estimation of the parameters regardless whether yt and ®¤t are stationary or not. We used TSM for GAUSS for estimating the state-space form of our model. Since there is no At matrix in this software we incorporated exogenous variables into dt . The vectors and matrices of the state-space representation of the long-run model in the I(2) case is the following. 2 3 2 3 2 rqt¤ 0 0 0 0 0 6 ei¤ 7 6 7 6 t 6 7 60 7 6 0 0 0 0 6 (rq¤ ) 7 6 7 6 1 0 0 0 ¹ 0 6 7 7 ®¤t = 6 t(ei¤ ) 7 ct = c = 6 Rt = R = 6 6 7 0 6 ¹t 7 (6£1) (6£1) 6 7 (6£4) (6£4) 6 (6£1) 6 0 1 0 0 6 7 4 5 4 0 0 1 0 0 4 ct 5 0 0 0 0 1 ct e 2 3 1 0 1 0 0 0 6 0 1 0 1 7 0 0 6 7 6 0 0 1 0 7 0 0 7 Tt = T = 6 6 7 0 0 (6£6) (6£6) 6 0 0 0 1 7 4 0 0 0 0 ½ cos f c ½ sin fc 5 0 0 0 0 ¡½ sin fc ½ cos fc 2 ³ 3 ´2 (rq) ¾(" ) 0 0 0 6 7 ³ (ei) ´2 6 7 (" ) 6 7 0 ¾ 0 0 6 7 ³ ´2 Qt = Q = 6 7 (c) 6 7 (4£4) (4£4) 0 0 ¾(" ) 0 6 7 4 ³ (c) ´2 5 0 0 0 ¾(" ) 2

3 xt yt = 4 mt 5 (3£1) gdp t 2

¡ 1 Xt¡1 Zt = 4 ¡ ¡ 1 M t¡1 (3£6) 1 0 2 3 a + dwt dt = 4 d + ddt 5 (3£1) qwt

3 0 0 0 0 0 0 0 0 5 0 0 1 0

26

3 7 7 7 7 7 7 5

2 ³ (x) ´2 3 ¾(" ) 0 0 6 7 ³ (m) ´2 7 Ht = H = 6 (" ) 4 5 0 ¾ 0 (3£3) (3£3) 0 0 0

8.4

Estimating Rest of the World Supply

For estimating world supply, we …tted an I(2) plus stationary cycle model, the local linear trend (LLT) + cycle (C) model, which is the following: Local linear trend plus cycle (LLTC):

·

dwt = qwt¤ + cw¤t + "t

(18)

¢¢qw¤t = ³t

(19)

¸ · ¸· ¤ ¸ · ¸ cwt¡1 cwt¤ ·t cos fcw sin fcw + ¤ = ½w ¤ ¡ sin f cos f cf wt cf wt¡1 ·t e cw cw

(20)

where ³t , ·t and e ·t are white noise processes with variances ¾2³ , ¾2·, ¾2·; ½w is a dampening parameter; and fcw the frequency of the world cycle in radians. For the breaking I(1) model, of course, world supply follows an I(1) model with 3 shifts in the drift. Figures 3-4 show the actual growth of world demand and their estimated growth components in a two times standard error band.

27

8.5

Data Sources

We have used various data sources: ² Hungarian national accounts: publications of the Hungarian Central Statistical O¢ce ² Finnish national accounts: Statistics Finland ² Polish national accounts: IMF – International Financial Statistics ² Exports for calculating the weights of the world demand: IMF – Directions of Trade Statistics ² GDP …gures of rest of the world countries: three sources were used: (1) IMF – International Financial Statistics, (2) OECD – World Development Indicators, (3) PENN World Tables, for Eastern Germany GDP …gures for 1980-89 are from the Statistisches Bundesamt ² Nominal exchange rates: IMF – International Financial Statistics (line rf) ² Consumer prices: IMF – International Financial Statistics (lines 64 and 64a)

28

9

Tables Table 1: Estimation results of the long-run model

σ _x σ_m σ_rq σ_ei a d ρ fc σ_gap µ1_rq µ2_rq µ3_rq µ4_rq µ 1_ei µ 2_ei µ 3_ei µ4_ei Diagnostics L T Ncoef JB_x JB_m JB_gdp KPSS coiv_x KPSS coiv_m PP coiv_x PP coiv_m

I(2) model estimate t 0.052 5.4 0.065 6.3 0.014 2.6 0.033 4.4 0.605 17.0 0.597 16.3 0.678 5.2 1.150 2.4 0.014 2.7

value 170.3 40 9 0.46 5.81 15.46 0.077 0.120 -4.22 -3.41

5% CV

5.99 5.99 5.99

Breaking I(1) model estimate t 0.043 4.1 0.073 6.5 0.000 0.0 0.033 3.1 0.607 12.8 0.599 12.2 0.685 6.4 1.066 6.5 0.015 7.1 -0.003 -5.6 -0.021 -17.8 -0.052 -11.9 0.026 7.1 0.031 3.9 -0.025 -2.2 -0.049 -2.3 0.119 7.8 value 201.5 40 17 1.14 0.91 36.61 0.068 0.110 -4.35 -3.21

5% CV

5.99 5.99 5.99

Notes. L: value of the log-likelihood function, T: number of observations per equation, Ncoef : total number of estimated coe¢cients, JB_i: JarqueBera normality test for the innovations of the three observation equations, KPSS coiv_i: KPSS test for stationarity of the cointegrating vectors, PP coiv_i: PP test for unit root of the cointegrating vectors.

29

Table 2: Estimation results of the short-run model I(2) model Breaking I(1) model estimate t estimate t Imports, first differences 0.10 0.89 0.16 1.16 ∆xt ∆ddt 1.14 7.96 1.07 6.16 ECMt-1 -0.48 -3.40 -0.45 -3.31 R2-adjusted 0.811 0.717 SE 0.046 0.057 Imports, second differences ∆∆xt 0.35 2.72 0.36 2.31 1.52 7.12 1.38 5.40 ∆∆ddt ECMt-1 -0.82 -4.69 -0.82 -4.76 R2-adjusted 0.838 0.742 SE 0.056 0.070 Exports, first differences 1.13 8.84 1.11 9.21 ∆dwt ECMt-1 -0.67 -4.52 -0.65 -4.38 R2-adjusted 0.873 0.889 SE 0.034 0.031 Exports, second differences ∆∆dwt 1.12 1.82 0.85 1.36 ECMt-1 -1.20 -6.14 -1.20 -5.62 R2-adjusted 0.842 0.829 SE 0.042 0.043 Exports, second differences, ECM is constrained to be –0.8 ∆∆dwt 1.38 2.21 0.98 1.52 R2-adjusted 0.828 0.817 SE 0.044 0.045

Notes. See equations (11-12) on page 15 for the speci…cation of the short-run model. Estimation results irrelevant for the speci…c model are shaded. Table 3: Critical values of ADF, PP, and KPSS tests critical ADF t & PP t KPSS ¹ KPSS ¿ value 1% -3.62 0.739 0.216 5% -2.94 0.463 0.146 10% -2.61 0.347 0.119 Note. The critical values represent lower tail values in case of ADF and PP tests and upper tail values in case of KPSS.

30

Table 4: ADF and PP tests for unit root lags ADF

PP x

0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5

-2.09 -2.09 -1.50 -1.96 -1.35 -1.90 -1.21 -1.87 -1.31 -1.86 -1.97 -1.86 gdp -3.47 -3.47 -2.30 -3.03 -2.28 -2.90 -2.15 -2.85 -2.18 -2.84 -2.55 -2.82 dw -7.27 -7.27 -2.51 -5.62 -2.99 -4.95 -3.07 -4.60 -2.68 -4.41 -3.33 -4.31 TSB -3.46 -3.46 -3.74 -3.59 -2.18 -3.42 -2.45 -3.40 -2.50 -3.46 -2.00 -3.42

ADF

PP

d(x) -4.06 -4.06 -2.89 -4.01 -2.02 -4.02 -1.86 -4.09 -2.64 -4.18 -1.91 -4.16 d(gdp) -3.23 -3.23 -2.95 -3.25 -2.32 -3.19 -2.00 -3.17 -1.37 -3.15 -1.66 -3.20 d(dw) -1.50 -1.50 -1.29 -1.48 -1.15 -1.45 -1.23 -1.44 -1.09 -1.44 -1.05 -1.41 dTSB -6.45 -6.45 -7.54 -6.46 -4.26 -6.72 -3.46 -6.97 -3.58 -6.96 -3.51 -7.26

ADF

PP

d(d(x)) -9.00 -9.00 -6.83 -9.28 -4.67 -10.1 -2.87 -10.2 -3.36 -10.2 -2.80 -10.8 d(d(gdp)) -7.17 -7.17 -5.98 -7.21 -4.80 -7.47 -4.94 -7.85 -3.28 -8.48 -2.80 -8.65 d(d(dw)) -6.56 -6.56 -4.79 -6.56 -3.75 -6.58 -3.14 -6.61 -3.21 -6.64 -2.22 -6.70 ddTSB -8.01 -8.01 -9.93 -8.29 -6.82 -10.4 -4.59 -12.1 -4.06 -11.2 -4.71 -12.5

ADF

PP

m -1.62 -1.66 -1.49 -1.02 -0.66 -1.00

-1.62 -1.60 -1.60 -1.58 -1.57 -1.57 dd -2.84 -2.84 -2.51 -2.66 -3.01 -2.77 -1.99 -2.78 -1.72 -2.67 -2.65 -2.65 rd -0.89 -0.89 -0.93 -0.94 -0.51 -0.85 -1.01 -0.83 -1.10 -0.86 -0.85 -0.87 X/M -3.76 -3.76 -4.06 -3.88 -2.35 -3.72 -2.51 -3.69 -2.67 -3.73 -1.99 -3.69

ADF

PP

d(m) -5.02 -5.02 -4.00 -5.03 -2.15 -4.99 -1.48 -5.04 -1.75 -5.14 -1.60 -5.17 d(dd) -4.21 -4.21 -4.47 -4.28 -1.93 -4.10 -1.77 -4.10 -1.58 -4.24 -1.45 -4.29 d(rd) -5.27 -5.27 -6.89 -5.30 -2.74 -5.19 -2.45 -5.20 -2.01 -5.19 -1.61 -5.19 d(X/M) -6.57 -6.57 -7.74 -6.58 -4.52 -6.91 -3.51 -7.25 -3.81 -7.23 -3.64 -7.64

ADF

PP

d(d(m)) -8.80 -8.80 -8.98 -9.25 -6.92 -11.6 -3.52 -12.7 -2.96 -11.4 -2.62 -12.2 d(d(dd)) -6.91 -6.91 -10.6 -6.98 -5.77 -8.12 -3.68 -9.32 -3.45 -8.48 -3.19 -8.93 d(d(rd)) -6.98 -6.98 -11.4 -7.06 -6.08 -8.43 -4.56 -10.0 -4.15 -8.98 -3.80 -9.66 d(d(X/M)) -8.02 -8.02 -9.83 -8.30 -7.16 -10.5 -4.59 -12.4 -4.28 -11.3 -4.86 -12.8

Notes. Test equations include intercept but no time trend. Shaded area show non-rejection of the null hypothesis of unit root at 5%.

31

Table 5: KPSS tests for stationarity and trend-stationarity lags 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

µ

τ

x 2.820 0.552 1.522 0.298 1.078 0.213 0.853 0.172 0.720 0.150 gdp 3.208 0.872 1.670 0.451 1.151 0.313 0.892 0.246 0.737 0.207 dw 3.720 0.942 1.931 0.493 1.326 0.343 1.024 0.269 0.844 0.225 X/M 0.156 0.124 0.108 0.087 0.099 0.080 0.092 0.075 0.087 0.071

µ

τ

d(x) 0.213 0.224 0.157 0.165 0.130 0.135 0.112 0.116 0.103 0.106 d(gdp) 0.513 0.132 0.348 0.094 0.300 0.085 0.277 0.083 0.265 0.085 d(dw) 2.772 0.203 1.487 0.125 1.042 0.100 0.816 0.090 0.681 0.087 d(X/M) 0.031 0.023 0.033 0.025 0.055 0.042 0.073 0.056 0.070 0.054

µ

τ

d(d(x)) 0.062 0.031 0.123 0.063 0.163 0.086 0.187 0.103 0.159 0.088 d(d(gdp)) 0.042 0.035 0.050 0.042 0.068 0.057 0.085 0.072 0.124 0.106 d(d(dw)) 0.055 0.053 0.060 0.058 0.068 0.066 0.074 0.072 0.080 0.078 d(d(X/M)) 0.015 0.014 0.021 0.020 0.052 0.050 0.078 0.074 0.060 0.057

µ

τ

m 2.790 0.518 1.517 0.285 1.076 0.205 0.851 0.165 0.716 0.141 dd 3.238 0.884 1.690 0.461 1.167 0.321 0.904 0.251 0.746 0.210 rd 3.238 0.884 1.690 0.461 1.167 0.321 0.904 0.251 0.746 0.210 TSB 0.167 0.144 0.113 0.097 0.099 0.086 0.090 0.078 0.084 0.073

µ

τ

d(m) 0.213 0.210 0.175 0.172 0.164 0.159 0.146 0.140 0.129 0.123 d(dd) 0.356 0.107 0.280 0.088 0.292 0.099 0.279 0.101 0.248 0.092 d(rd) 0.356 0.107 0.280 0.088 0.291 0.099 0.279 0.101 0.248 0.092 d(TSB) 0.032 0.025 0.035 0.027 0.056 0.044 0.071 0.056 0.069 0.055

µ

τ

d(d(m)) 0.064 0.032 0.105 0.054 0.191 0.102 0.235 0.130 0.178 0.098 d(d(dd)) 0.039 0.031 0.045 0.036 0.093 0.075 0.134 0.109 0.110 0.089 d(d(rd)) 0.039 0.031 0.045 0.036 0.093 0.075 0.134 0.109 0.110 0.089 d(d(TSB)) 0.014 0.014 0.020 0.020 0.050 0.050 0.070 0.070 0.056 0.056

Note. Shaded area show rejection of the null hypothesis of stationarity (¹) and trend-stationarity (¿) at 5%.

10

Figures

32

9

30

6

20

3

10

0

0

-3

CPI

40

GDP, TB, U

12

-10

TB CPI (right scale) GDP U

-6

-9

-20

-30

2000

1995

1990

1985

1980

1975

1970

1965

-40 1960

-12

Figure 1: Hungary: Trade balance, GDP growth, unemployment rate, in‡ation, 1960-99

Figure 2: Hungary: Estimated output gap of world demand based on the two statistical models 33

Figure 3: Hungary: Actual and estimated growth of world demand based on the I(2) model

Figure 4: Hungary: Actual and estimated growth of world demand based on the breaking I(1) model 34

Figure 5: Hungary: Estimated state vectors based on the I(2) model

35

Figure 6: Hungary: Estimated state vectors and their growth rates based on the breaking I(1)

36

Figure 7: Hungarian output gap based on the I(2) model

Figure 8: Hungarian output gap based on the breaking I(1) model 37

Figure 9: Hungary: Actual and potential growth and the level of output gap based on the I(2) model

Figure 10: Hungary: Actual and potential growth and the level of output gap based on the breaking I(1) model 38

Exports, I(2) Imports, I(2) Exports, breaking I(1) Imports, breaking I(1)

1.0 0.9

p-value of Chow test

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 1965

1970

1975

1980

1985

1990

1995

break point

Figure 11: Hungary: Chow breakpoint tests for the error correction model at each year between 1965-1995

1.15

previous year = 1

1.10

1.05

1.00

Foreign demand index Group 1: industrial countries

0.95

Group 2: former USSR 0.90

Group 3: other former socialist countries Group 4: others 1995

1990

1985

1980

1975

1970

1965

1960

0.85

Figure 12: Hungary: Growth rate of world demand and of its components, 1960-99

39

100%

Other countries 90% 80%

Other former socialist countries

70% 60% 50%

Former USSR

40% 30% 20%

Industrial countries

10%

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

0%

Figure 13: Hungary: Destination distribution of exports, 1976-99

100%

Other countries 90% 80% 70%

Other former socialist countries

60% 50%

Former USSR

40% 30% 20%

Industrial countries

10%

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

0%

Figure 14: Poland: Destination distribution of exports, 1980-99

40

Figure 15: Poland: Estimated state vectors

41

60

15

50

12

40

9

30

6

20

3

10

0

0

GDP, TB, U

18

-3

-10

TB CPI (right scale) GDP U

-6 -9

CPI

Figure 16: Poland: Actual and potential growth and the level of output gap

-20 -30

2000

1995

1990

1985

-40 1980

-12

Note: Inflation was 104% in 1982 and exceeded 500% in 1990.

Figure 17: Poland: Trade balance, GDP growth, unemployment rate, in‡ation, 1980-99 42

MNB Füzetek / NBH Working Papers: 1995/1 (november) Simon András: Aggregált kereslet és kínálat, termelés és külkereskedelem a magyar gazdaságban 19901994 1995/2 (november) Neményi Judit: A Magyar Nemzeti Bank devizaadósságán felhalmozódó árfolyamveszteség kérdései 1995/3 (február) Dr. Kun János: Seignorage és az államadóság terhei 1996/1 (március) Simon András: Az infláció tényezõi 1990-1995-ben 1996/2 (június) Neményi Judit: A tõkebeáramlás, a makrogazdasági egyensúly és az eladósodási folyamat összefüggései a Magyar Nemzeti Bank eredményének alakulásával. 1996/3 (június) Simon András: Sterilizáció, kamatpolitika, az államháztartás és a fizetési mérleg 1996/4 (július) Darvas Zsolt: Kamatkülönbség és árfolyam-várakozások 1996/5 (augusztus) Vincze János - Zsoldos István: A fogyasztói árak struktúrája, szintje és alakulása Magyarországon 1991-1996-ban Ökonometriai vizsgálat a részletes fogyasztói árindex alapján 1996/6 (augusztus) Csermely Ágnes: A vállalkozások banki finanszírozása Magyarországon 1991-1994 1996/7 (szeptember) Dr. Balassa Ákos: A vállalkozói szektor hosszú távú finanszírozásának helyzete és fejlõdési irányai 1997/1 (január) Csermely Ágnes: Az inflációs célkitûzés rendszere 1997/2 (március) Vincze János: A stabilizáció hatása az árakra, és az árak és a termelés (értékesítés) közötti összefüggésekre 1997/3 (április) Barabás Gyula - Hamecz István: Tõkebeáramlás, sterilizáció és pénzmennyiség 1997/4 (május) Zsoldos István: A lakosság megtakarítási és portfolió döntései Magyarországon 1980-1996. 1997/5 (június) Árvai Zsófia: A sterilizáció és tõkebeáramlás ökonometriai elemzése 1997/6 (augusztus) Zsoldos István: A lakosság Divisia-pénz tartási viselkedése Magyarországon 1998/1 (január) Árvai Zsófia - Vincze János: Valuták sebezhetõsége: Pénzügyi válságok a ‘90-es években

1998/2 (március) Csajbók Attila: Zéró-kupon hozamgörbe becslés jegybanki szemszögbõl ZERO-COUPON YIELD CURVE ESTIMATION FROM A CENTRAL BANK PERSPECTIVE 1998/ 3 (március) Kovács Mihály András - Simon András: A reálárfolyam összetevõi THE COMPONENTS OF THE REAL EXCHAGE RATE IN HUNGARY 1998/4 (március) P.Kiss Gábor: Az államháztartás szerepe Magyarországon THE ROLE OF GENERAL GOVERNMENT IN HUNGARY 1998/5 (április) Barabás Gyula - Hamecz István - Neményi Judit: A költségvetés finanszírozási rendszerének átalakítása és az eladósodás megfékezése Magyarország tapasztalatai a piacgazdaság átmeneti idõszakában FISCAL CONSOLIDATION, PUBLIC DEBT CONTAINMENT AND DISINFLATION ; HUNGARY’S EXPERIENCE IN TRANSITION 1998/6 (augusztus) Jakab M. Zoltán-Szapáry György: A csúszó leértékelés tapasztalatai Magyarországon 1998/7 (október) Tóth István János - Vincze János: Magyar vállalatok árképzési gyakorlata 1998/8 (október) Kovács Mihály András: Mit mutatnak? Különféle reálárfolyam-mutatók áttekintése és a magyar gazdaság ár- és költség-versenyképességének értékelése 1998/9 (október) Darvas Zsolt: Moderált inflációk csökkentése Összehasonlító vizsgálat a nyolcvanas-kilencvenes évek dezinflációit kísérõ folyamatokról 1998/10 (november) Árvai Zsófia: A piaci és kereskedelmi banki kamatok közötti transzmisszió 1992 és 1998 között THE INTEREST RATE TRANSMISSION MECHANISM BETWEEN MARKET AND COMMERCIAL BANK RATES 1998/11 (november) P. Kiss Gábor: A költségvetés tervezése és a fiskális átláthatóság aktuális problémái 1998/12 (november) Jakab M. Zoltán: A valutakosár megválasztásának szempontjai Magyarországon 1999/1 (January) ÁGNES CSERMELY-JÁNOS VINCZE: LEVERAGE AND FOREIGN OWNERSHIP IN HUNGARY 1999/2 (március) Tóth Áron: Kísérlet a hatékonyság empirikus elemzésére a magyar bankrendszerben 1999/3 (március) Darvas Zsolt-Simon András: A növekedés makrogazdasági feltételei Gazdaságpolitikai alternatívák CAPITAL STOCK AND ECONOMIC DEVELOPMENT IN HUNGARY (May 1999) 1999/4 (április) Lieli Róbert: Idõsormodelleken alapuló inflációs elõrejelzések Egyváltozós módszerek

1999/5 (április) Ferenczi Barnabás: A hazai munkaerõpiaci folyamatok Jegybanki szemszögbõl Stilizált tények LABOUR MARKET DEVELOPMENTS IN HUNGARY FROM A CENTRAL BANK PERSPECTIVE – Stylized Facts 1999/6 (május) Jakab M. Zoltán – Kovács Mihály András: A reálárfolyam-ingadozások fõbb meghatározói Magyarországon DETERMINANTS OF REAL-EXCHANGE RATE FLUCTUATIONS IN HUNGARY 1999/7 (July) ATTILA CSAJBÓK: INFORMATION IN T-BILL AUCTION BID DISTRIBUTIONS 1999/8 (július) Benczúr Péter: A magyar nyugdíjrendszerben rejlõ implicit államadósság-állomány változásának becslése CHANGES IN THE IMPLICIT DEBT BURDEN OF THE HUNGARIAN SOCIAL SECURITY SYSTEM 1999/9 (augusztus) Vígh-Mikle Szabolcs–Zsámboki Balázs: A bankrendszer mérlegének denominációs összetétele 19911998 között 1999/10 (szeptember) Darvas Zsolt – Szapáry György: A nemzetközi pénzügyi válságok tova terjedése különbözõ árfolyamrendszerekben FINANCIAL CONTAGION UNDER DIFFERENT EXCHANGE RATE REGIMES 1999/11 (szeptember) Oszlay András: Elméletek és tények a külföldi működõtõke-befektetésekrõl 2000/1 (január) Jakab M. Zoltán – Kovács Mihály András – Oszlay András: Hová tart a külkereskedelmi integráció? Becslések három kelet.közép-európai ország egyensúlyi külkereskedelmére HOW FAR HAS TRADE INTEGRATIONADVANCED? AN ANALYSIS OF ACTUAL AND POTENTIAL TRADE OF THREE CENTRAL AND EASTERN EUROPEAN COUNTRIES 2000/2 (February) SÁNDOR VALKOVSZKY – JÁNOS VINCZE: ESTIMATES OF AND PROBLEMS WITH CORE INFLATION IN HUNGARY 2000/3 (március) Valkovszky Sándor: A magyar lakáspiac helyzete 2000/4 (május) Jakab M. Zoltán – Kovács Mihály András – Lőrincz Szabolcs: Az export előrejelzése ökonometriai módszerekkel FORECASTING HUNGARIAN EXPORT VOLUME 2000/5 (augusztus) Ferenczi Barnabás – Valkovszky Sándor – Vincze János: Mire jó a fogyasztói-ár statisztika? WHAT ARE CONSUMER PRICE STATISTICS GOOD FOR? 2000/6 (August) ZSÓFIA ÁRVAI – JÁNOS VINCZE: FINANCIAL CRIESES IN TRANSITION COUNTRIES: MODELS AND FACTS 2000/7 (Oktober) GYÖRGY SZAPÁRY: MAASTRICHT AND THE CHIOCE OF EXCHANGE RATE REGIME IN TRANSITION COUNTRIES DURING THE RUN-UP TO EMU

2000/8 (november) Árvai Zsófia – Menczel Péter: A magyar háztartások megtakarításai 1995 és 2000 között 2000/9 (November) ANDRÁS SIMON – ZSOLT DARVAS: POTENTIAL OUTPUT AND FOREIGN TRADE IN SMALL OPEN ECONOMIES A potenciális kibocsátás becslése a gazdaság nyitottságának felhasználásával 2001/1 Simon András – Várpalotai Viktor: Eladósodás és óvatos viselkedés