TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES DURING THE CURRENT FLOAT

Transcription

1 JOURNAL OF ECONOMIC DEVELOPMENT 29 Volume 28, Number 2, December 2003 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES DURING THE CURRENT FLOAT Al-zayoonah Universiy of Jordan Previous sudies on purchasing power pariy (PPP), using uni roo ess, have esed eiher he null hypohesis of a uni roo or he null of saionary real exchange rae. I has been argued ha using eiher approach is insufficien o confirm he exisence or non-exisence of PPP. To srenghen inferences made abou a series, he wo approaches should be applied wihin he same sudy. In conras o previous sudies underaken o es PPP in developing counries, his paper ess PPP for a sample of developing counries in he Asian financial crisis counries during he curren floa. The paper applies he ADF and PP ess o es he null of a uni roo and he newly developed KPSS es o es he null of saionariy. The null of a uni roo can be rejeced for Indonesia, Korea and Thailand. We canno rejec he null of saionariy for all counries excep for Singapore. Join esing of boh nulls confirms saionariy for Indonesia and Korea. The impac of he Asian financial crisis on he behavior of he real exchange raes in he crisis counries is examined using Perron s uni roo es ha accouns for poenial srucural breaks. The resuls indicae evidence of saionariy for Indonesia, Korea, Malaysia, and Thailand. Keywords: PPP, The ADF Tes, The KPSS Tes, Confirmaory Analysis, Srucural Break JEL classificaion: F3. INTRODUCTION Purchasing power pariy (PPP) is an exchange rae deerminaion model formalized and developed by Gusav Cassel in he 920s. As an imporan heory in inernaional economics, PPP saes in i is absolue version ha he nominal exchange rae, defined as unis of he domesic currency per uni of he foreign currency, should be equal o he price raio of domesic o foreign counry. The relaive version saes ha changes in he nominal exchange rae should be equal o he inflaion differenial beween he home and foreign counries. Many sudies have been underaken o es he validiy of PPP especially afer he collapse of he Breon Woods sysem in 973 and he ransiion o he flexible exchange

2 30 rae sysem. While i is generally acceped ha relaive PPP fails o hold in he shor run, long run PPP is sill under invesigaion. Depending on he sample size, and ype of ess employed, some sudies are able o find some evidence in favor of long run PPP. Ye, he speed of convergence o PPP is exremely slow since deviaions appear o damp ou a a rae of roughly 5% per year implying a half-life for PPP deviaions beween 3 and 5 years. Mos of he early lieraure on PPP is based on he ime-series analysis of shor spans of daa for he pos-breon Woods era where he focus was developed counries. Mos of his lieraure did no find evidence in favor of PPP; and herefore, concluded ha PPP does no hold. Engle (2000), for insance, using quarerly daa from 970 o 995 for he U.S. and U.K. could no rejec he null of a uni roo and he null of no-coinegraion (oher sudies include Baum e al. (999), and Mark (990)). Recenly, however, some economiss have argued ha he reason for he failure o find evidence in favor of PPP is due o he shor sample size used for he curren floa and / or he lack of power in he sandard ess (he Augmened Dickey-Fuller es (ADF) and Phillip-Perron es (PP)). Frankel (986, 990) argued ha he reason ha some sudies could no rejec he random walk hypohesis in he real exchange rae was a lack of power. He poined ou ha if he speed of convergence o PPP is exremely slow, hen i migh require sufficienly long daa ses for one o be able o reliably rejec he random walk hypohesis in he real exchange rae. He esed a long daa se on he US$-DM exchange rae for he period , where he was able o rejec he random walk hypohesis wih a poin esimaes of 4% per year, implying a half-life for he real exchange rae deviaions of 4.6 years. Lohian and Taylor (996) using daa on he dollar-pound rae for he period and he franc-pound rae for he period were able o rejec he random walk model wih an esimaed half-life of abou six years for he dollar-pound rae and a lile under hree years for he franc-pound rae (oher sudies include Glen (992), Froo and Rogoff (994), and Kuo and Mikkola (999)). Using long run daa has been criicized since i combines daa from fixed and floaing exchange rae sysems. Therefore, rejecing he null of a uni roo or no-coinegraion does no provide direc evidence on he validiy of PPP under he curren floa. Inspired by he work of Levin and Lin (992 and 993), economiss have urned o panel daa ess o es he validiy of PPP for he curren floa. Levin and Lin show ha he power of uni roo ess in shor ime periods can be improved by increasing he number of counries. The resuls from he panel daa approach during he curren floa are mixed. Wu and Chen (999), for example, using wo panel uni roo ess on monhly daa from 980 o 996 for 9 Pacific Basin counries could no find evidence of saionariy. MacDonald (996), using annual daa for 40 OECD CPI and WPI-based real exchange raes for he period is able o rejec he null hypohesis of a uni roo (oher examples include Abuaf and Jorion (990), Frankel and Rose (996), Papell (997), and Heimonen (999)). O Connell (998, p. 2) argues ha previous panel uni roo ess ha

3 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES 3 are able o find evidence in favor of PPP are incorrecly sized, owing o heir failure o conrol for cross-secional dependence in real exchange raes. Bu afer conrolling for cross-secional dependence, he does no find evidence in favor of PPP. The mixed resuls from long-run and panel daa ess have led economiss o quesion he power of he sandard uni roo ess. I is now generally acceped ha in he sandard uni roo ess he null of non-saionariy for many aggregae economic ime series canno be rejeced unless here is srong evidence agains i. Therefore, he reason for he failure o rejec he null of non-saionariy is due o lack of power in hese ess. Diebold and Rudebusch (99) show ha he uni roo ess proposed by Dickey and Fuller have low power agains fracionally-inegraed processes. DeJong e al. (992) argue ha he Dickey-Fuller uni roo es has low power agains rend saionary processes. Perron (989) argue ha if he ime-series conains a srucural break, hen he sandard uni roo ess will lead o he accepance of he null of a uni roo, when in fac he series is saionary. Consequenly, many economiss have argued agains using he sandard uni roo ess and proposed using oher powerful ess, such as ess ha can be used o es he null of saionariy agains he alernaive of non-saionariy. A number of ess have been developed; he mos popular one is he KPSS es developed by Kwiakowski, Phillips, Schmid, and Shin (992). Kwiakowski e al. (992, p. 76) argue ha heir es is inended o complemen uni roo ess, such as he Dickey-Fuller ess. By esing boh he uni roo hypohesis and he saionariy hypohesis, we can disinguish beween series ha appear o be saionary, series ha appear o have uni roo, and series for which he daa (or he ess) are no sufficienly informaive o be sure wheher hey are saionary or inegraed. Henricsson and Lundbäck (995) poin ou ha a lo of informaion can be gained by comparing he oucomes of esing boh he null of saionariy and he null of non-saionariy. Join esing of boh nulls can srenghen inferences made abou he saionariy or non-saionariy of a ime series especially when he oucomes of he wo nulls corroborae each oher. This join esing has been known as confirmaory analysis. For example, if he null of saionariy is acceped (rejeced) and he null of non-saionariy is rejeced (acceped), we have confirmaion ha he series is saionary (non-saionary). Conversely, we can no have confirmaion if boh nulls are acceped or boh are rejeced. The possible oucomes of his join esing are summarized in Table. Anoher es is he LMC es developed by Leybourne and McCabe (993). The wo ess differ in heir reamen of auocorrelaion under he null hypohesis. The KPSS es adjuss for auocorrelaion nonparamerically by choosing a lag runcaion parameer, whereas he LMC es adjuss for auocorrelaion paramerically by choosing an appropriae lag lengh in he ARMA model.

4 32 Table. The Possible Oucomes of he ADF and KPSS Tess ADF Tes ) KPSS Tes 2) Accep Rejec Accep Inconclusive decision (Insufficien informaion) Conclusive decision (Saionary) Rejec Conclusive decision (Non-saionary) Inconclusive decision (Fracional inegraion) Noes: ) indicaes he null hypohesis of non-saionariy in he ADF es. 2) indicaes he null hypohesis of saionariy in he KPSS es. Oher ess proposed o es PPP are ess ha accoun for poenial srucural breaks in he series, such as he Augmened Perron s uni roo es. I is argued ha in he presence of srucural breaks sandard uni roo ess are biased owards acceping he null of a uni roo when in fac he series is saionary; herefore, allowing for srucural breaks in modeling processes behavior is imporan. 2 In conras o previous sudies, his paper ess PPP for a sample of developing counries in he Asian financial crisis counries during he curren floa. To srenghen inferences made abou he real exchange rae series, he paper conducs confirmaory analysis by applying he ADF and PP ess o es he null hypohesis of a uni roo and he newly developed KPSS es o es he null of saionary real exchange rae exchange rae. In addiion o using confirmaory analysis, he paper applies Perron s uni roo es ha accouns for poenial srucural breaks in he real exchange rae series arising from he Asian financial crisis. The res of he paper is organized as follows. Secion wo presens he heory of PPP and mehodology. Secion hree provides a brief lieraure review. Secion four describes he daa and is source. Secion five presens he resuls and analysis of he empirical work. Secion six sheds some ligh on he Asian financial crisis and provides he resuls of applying Perron es, and finally secion seven gives he summary and conclusions. 2. THE THEORY OF PPP AND METHODOLOGY The basic building block of PPP is he Law of One Price (LOP). The LOP saes ha in he absence of rade barriers, such as ransporaion coss, and ariffs, compeiion will equalize he price of an idenical and raded good across counries when he prices are expressed in he same currency. PPP is derived by exending LOP o n idenical 2 Oher ess used o es PPP are he variance raio analysis developed by Lo and MacKinley (988) o es he null of a random walk, and he mulivariae coinegraion procedures o es for r coinegraing vecors developed by Johansen and Juselius (990), and Johansen (99).

5 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES 33 and raded goods and assuming ha LOP holds for each of he n goods. The heory of PPP involves a relaionship beween he nominal exchange rae and he price raio of domesic o foreign counry. Thus; P E i =, () P i US where E i is he nominal exchange rae, defined as unis of domesic currency per uni of he foreign currency, for counry i a ime per U.S. dollar, P i is he domesic US price index (he CPI), P is he foreign price index (he U.S. in his case), and i is an index for counry i = Indonesia, Korea, Malaysia, Singapore, he Philippines, and Thailand. Using lowercase leers o denoe he naural logarihm of he variables in US Equaion () yields ei = pi p, he absolue PPP. Taking he firs difference of he US absolue PPP yields ei = pi p, he relaive PPP. The real exchange rae is defined as he nominal exchange rae adjused for changes in he home and foreign price levels. Using lowercase leers o denoe he variables in heir naural logarihm form yields r i = e + p p, (2) i US i where r i is he naural logarihm of he real exchange rae for counry i a ime. For PPP o hold he real exchange rae r i should be consan (saionary), ha is, a X % increase/decrease in relaive prices should be mached by a X % depreciaion/ appreciaion in he nominal exchange rae. Thus, if we can show ha he real exchange rae is saionary, we can provide evidence in favor of PPP. If he saionariy of he real exchange rae is no found, he heory of PPP will be rejeced. 2.. The ADF and PP Tess To es he null hypohesis of non-saionary real exchange rae using he ADF es, he following model is used r i i ki = µ + δ + ρ r + µ r + υ i i ij i j i, (3) j= where k is he number of lags in he ADF es chosen by minimizing Schwarz Informaion Crierion (SIC) o ensure ha he errors are whie noise. The coefficien of ineres is ρ ; a finding ha ρ < 0 is saisically significan implies ha he real exchange rae is saionary. The null hypohesis of a uni roo is ρ = 0 agains he

6 34 alernaive ρ < 0. A ime rend is included o allow for he possibiliy of deerminisic rend in he alernaive hypohesis. The ADF es assumes saisically independen residuals wih consan variance. The PP es relaxes hese assumpions and allows he residuals o be weakly dependen wih heerogeneous variance. The PP es applied using he following model wih a consan and a ime rend: r T θ ( + ςi (4) 2 i = i0 + θi ) + θi2r wih a es saisic Z( θ i2) for he null hypohesis θ = i2. The number of lag runcaion in he PP es is seleced auomaically by Newey and Wes Bandwidh using Barle Kernal Specral esimaion mehod The KPSS Tes To es he null hypohesis of saionariy using he KPSS es, we follow Kwiakowski e al. (992). They consider a series y ha can be decomposed ino he sum of deerminisic rend, a random walk, and a saionary error: y = ξ + + ε, (5) r where ε a saionary is process and r is a random walk given by: r = r + u 2 wih u ~ iid(0, σ u ). The iniial value r 0 is fixed and serves as he inercep. Under hese assumpions, he null hypohesis of saionariy is σ 2 u = 0. Since ε is assumed o be a saionary process, under he null hypohesis he series y is rend saionary. To es he null hypohesis of level saionariy ξ is se equal o zero (Culver and Papell, 999 es he null of level saionary real exchange rae). The null of rend saionary is esed by esimaing Equaion (5) on an inercep and rend. The KPSS es saisic is given by where T 2 LM ˆ 2 2 = T S s ( l), (6) = S is he parial sum of deviaions of residuals from he sample mean, T l T 2 2 l) = T ˆ = s= = s+ 2 ˆ ε + ( ) w( s, l) ˆ ε is a consisen esimaor of he long run variance s s ( ε T 2 ( σ ) of he regression error, l is a lag runcaion parameer, and w ( s, l) = [ s ( l + ) ] is an opional weighing funcion (Barle weighs) used o smooh he sample

7 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES 35 auocovariance funcion, which ensures ha s 2 ( l ) is non-negaive (Newey and Wes (987)). The number of lags runcaion in he KPSS es is seleced auomaically by Newey and Wes Bandwidh using Barle Kernal Specral esimaion mehod. The null hypohesis of saionariy is acceped if he value of he KPSS es saisic is less han i is criical value Perron s Tes One major drawback of sandard uni roo ess is ha hey implicily assume ha he deerminisic rend is correcly specified. Perron (989) argue ha if he ime-series conains a srucural break, hen sandard uni roo ess will lead o he accepance of he null of a uni roo, when in fac he series is saionary. There are a number of sudies showing he relevance of allowing srucural breaks in modeling he long-run behavior of real exchange raes (see, for example, Baum e al. (999)). For his purpose, he Perron (989) Augmened (PA) uni roo es is applied. The es allows under boh he null and alernaive hypoheses for he presence of a one-ime change in he level or in he slope of he series. The es consis of hree models Model A: he crash model which allows for a one-ime change in he level of he series y k = + βdmu + β2 + δdtb + αy + p= β 0 θ y + ε. (7) p p Model B: he changing growh model which allows for a one-ime change in he slope of he rend funcion y k = + βdmu + β2 + φdts + αy + θ p p= β 0 y + ε. (8) p Model C: his allows boh effecs a one-ime change in he level and he slope of he rend funcion y k 0 + βdmu + β2 + δdtb + ψdt + αy + θ p y p ε, (9) p= = β + where TB is he ime break, and DTB = if = TB + and 0 oherwise. Under Model A: DMU = 0 if <= TB, and if > TB. The null hypohesis is a uni roo wih a one-ime change in he level of he series agains he alernaive of a one-ime change in he level of a rend saionary series. Under Model B: DTS = 0 if <= TB, and TB if > TB. The null hypohesis

8 36 is a uni roo wih a one-ime change in he drif agains he alernaive of a one-ime change in he slope of a rend saionary series. Under Model C: DMU = DT if TB, and DMU =, DT = if > TB. The null hypohesis is a uni roo wih a one-ime change in boh level and drif agains he alernaive of a one-ime change in he inercep and slope of a rend saionary series. Perron s es assumes ha he break ime is given exogenously. The hree models are esimaed assuming he break ime is 997:2. The appropriae number of lags ( k ) is deermined by adding lags unil he Ljung-Box es fails o rejec no serial correlaion a he 5% significance level. The value on he coefficien α is compared agains is criical value calculaed by Perron for he appropriae value of λ = TB, which T represens he raio of he break sample size o he oal sample size a some significance level. 3. LITERATURE REVIEW A number of sudies have been underaken o es he validiy of PPP in he Asian counries. None of which has used confirmaory analysis, and mos of hem could no find evidence in favor of PPP. Luinel (2000), for example, using monhly panel daa from 958 o 989 for 8 Asian counries could no accep he null of a uni roo. Wang (2000), using monhly daa during he curren floa, examines PPP for seven Asian counries (Indonesia, Japan, Korea, Malaysia, Philippines, Singapore, and Thailand) agains he U.S. Using Johansen coinegraion procedure, he auhor finds evidence of coinegraion bu could no accep he symmery and proporionaliy resricions. Doğanlar (999), using quarerly ime-series from 980 o 995 for India, Indonesia, Pakisan, Turkey, and he Philippines could rejec he null of no-coinegraion. Moniel (997), using annual daa ess he absolue version of PPP for five Asian counries: Indonesia, Malaysia, he Philippines, Singapore, and Thailand. The main finding of he paper is lack of saionariy in he real exchange rae for all counries excep Singapore (oher sudies include Gan (994), Baharaumshah and Ariff (997), Weliwia (998), and Achy (2003)). 4. DATA We employ quarerly daa from 973:2-999:4 for six Asian counries. The counries are Indonesia, Korea, Malaysia, Singapore, he Philippines, and Thailand. The daa are obained from IMF s Inernaional Financial Saisics and conain he nominal exchange defined as he marke rae per U.S. dollar (line rf), and he consumer price index (CPI, line 63). The CPI is he sandard choice in he lieraure due o is availabiliy and consisency. The U.S. is reaed as he foreign or base counry o reflec is imporan

9 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES 37 role in he world economy and he availabiliy and consisency of is daa. 5. RESULTS AND ANALYSIS Table 2 shows he cross-correlaion beween changes in he log of he nominal exchange raes and changes in he log of he relaive prices from hree lags o hree leads. The resuls indicae ha changes in nominal exchange raes and he inflaion differenial are weakly correlaed. Table 3 shows he conemporaneous correlaion beween changes in he in he log of he nominal and log of he real exchange raes. The able indicaes ha he correlaion is almos perfec for all counries. This implies ha nominal exchange rae movemens dominae real exchange rae movemens. Tables 2 and 3 show ha changes in he nominal exchange raes are no offse by changes in relaive prices; raher mos of he change in he nominal exchange raes is refleced in he real exchange raes. This can be used o explain why relaive PPP fails o hold in he shor-run, and why real exchange raes can be represened by a maringale process. Since nominal exchange raes can be approximaed by a maringale process and mos of he changes in hem are refleced in real exchange raes, i is reasonable o say ha real exchange raes follow a maringale process. Table 2. Cross-correlaion of Changes in he Log of Nominal Exchange Raes and Relaive Price Levels Counry Lag 3 Lag 2 Lag 0 Lead Lead 2 Lead 3 Indonesia Korea Malaysia Philippines Singapore Thailand Noe: The correlaion coefficien is calculaed for he enire sample (973:2-999:4). Table 3. Correlaion beween Changes in Nominal and Real Exchange Raes Counry Correlaion Coefficien Indonesia Korea Malaysia Philippines 0.89 Singapore Thailand Noe: The correlaion coefficien is calculaed for he enire sample (973:2-999:4).

10 The Null of Non-saionariy As a preliminary sep, he saionariy of he nominal exchange raes and relaive prices is invesigaed using he ADF and PP ess. The null of a uni roo is esed wih and wihou ime rend. The rend erm is kep in he esimaion only if i is significan a he 0 percen significance level. The resuls of esing he variables in heir levels and firs difference are shown in Table 4. The resuls show ha he nominal exchange raes and relaive prices are non-saionary in heir levels for all counries; however, he firs difference of he variables is saionary. The null of non-saionary real exchange rae is esed by applying he ADF and PP uni roo ess wih a consan, a ime rend, and lagged firs difference of he real exchange rae. The resuls in Table 5 indicae ha he null of non-saionary real exchange rae can be rejeced by he ADF es for Indonesia and Korea a he 5 and 0 percen significance levels, respecively. The resuls from he PP uni roo es show evidence of saionariy for Indonesia, Korea and Thailand a he 0 percen significance level. These findings show very lile suppor for PPP in he Asian counries, which is line wih previous sudies. These findings may be due o lack of power in he ADF and PP uni roo ess. Therefore, in he nex secion we apply he newly developed KPSS es o es he null of saionary real exchange agains he alernaive of non-saionariy. Table 4. Uni Roo Tess for Level and Firs Difference of Nominal Exchange Raes and Relaive Prices Counry Variable ADF Tes ) PP Tes 2) Indonesia Ln(E) () (2) Korea Malaysia Philippine Ln(CPI) Ln(E) Ln(CPI) Ln(E) Ln(CPI) Ln(E) Ln(CPI) Ln(E) Ln(CPI) Ln(E) Ln(CPI) Ln(E) Ln(CPI) Ln(E) Ln(CPI) -2.75() (0) * (0) * (0) (3) (0) * (2) * (0) (3) (0) * (0) * (0) -3.39(3) () * (2) * -.59(5) () * () * (4) (6) (4) * (6) * (2) (5) (4) * (6) * (5) (6) (5) * (3) *

11 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES 39 Table 4. (Coninued) Counry Variable ADF Tes ) PP Tes 2) Singapore Ln(E) Ln(CPI) Ln(E) Ln(CPI) (0) (6) (0) * (0) * (3) (5) (3) * (7) * Thailand Ln(E) Ln(CPI) Ln(E) Ln(CPI) (0) () (0) * (0) * -.363(2) ) () * (2) * Noes: ) is he ADF es for he null hypohesis of a uni roo. The numbers in parenheses are he numbers of lags in he ADF es chosen by minimizing Schwarz Informaion Crieria (SIC) o ensure ha he residuals are whie-noise. A maximum of 2 lags are used. 2) is he Phillips-Perron (PP). The numbers in parenheses are he lags runcaion seleced auomaically by Newey and Wes Bandwidh using Barle Kernal Specral esimaion mehod. 3) *, **, *** denoes rejecion of he null hypohesis a he %, 5%, and 0% significance levels. No aserisk indicaes ha he series in no saionary. 4) All variables are ransformed by aking heir naural logarihm. The series are quarerly daa from 973:2-999:4 obained from he IMF s Inernaional Financial Saisics (CD-Ram). Ln(E) is naural logarihm of he nominal exchange rae defined as unis of he domesic currency per US dollar (line rf of he IMF s Inernaional Financial Saisics), Ln(CPI) is he naural logarihm of he Consumer Price Index (CPI) in he home counry relaive o he foreign counry (he USA) (line 63 of he IMF s Inernaional Financial Saisics). denoes he firs difference operaor. 5) No ime rend is included for he firs difference of he variables. indicaes insignifican ime rend a he 0 percen significance level. Table 5. Uni Roo Tess for he Real Exchange Raes (973:2-999:4) Counry ADF Tes PP Tes ADF ρ (θ ) β i Z i Indonesia () ** (8) *** Korea (0) *** (3) *** -0.4 Malaysia (0) (2) Philippines (0) (4) Singapore (0) (6) Thailand (0) (2) *** Noes: *, **, *** denoes rejecion of he null hypohesis of a uni roo a he %, 5%, and 0% significance level. No aserisk indicaes ha he series is non-saionary. The ADF equaions used are: ADF: i i ki iri + µ ij ri j υi. j= r = µ + δ + ρ + The Phillips-Perron (PP) es is applied using he following equaions wih four lag runcaions: T ri = β oi + β iri + β2i ( ) + ζ i wih a es saisic Z (β ) for he null hypohesis β i =. The numbers 2 in parenheses in he ADF es are he numbers of lags in he ADF es chosen by minimizing Schwarz Informaion Crieria (SIC). The numbers in parenheses in he PP es are he lags runcaion seleced auomaically by Newey and Wes Bandwidh using Barle Kernal Specral esimaion mehod. indicaes insignifican ime rend a he 0 percen significance level.

12 The Null of Saionariy To circumven he low power in he sandard uni roo ess, he newly developed KPSS es is applied o es he null of saionary real exchange agains he alernaive of non-saionariy. The resuls on he nominal and real exchange raes in Table 6 show ha he null of saionary nominal exchange rae could no be rejeced for all counries, which implies ha he nominal exchange raes are saionary. These resuls are no consisen wih hose obained from he sandard ess since he null of a uni roo is acceped and he null of saionariy is also acceped. Table 6. The KPSS Saionariy Tes for Nominal and Real Exchange Raes Counry The Nominal Exchange Rae The Real Exchange Rae The KPSS Tes Saisic The KPSS Tes Saisic Indonesia 0.002(6) *** (7) *** Korea 0.58(6) * 0.045(8) *** Malaysia 0.297(8) ** 0.07(8) *** Philippines 0.324(8) ** 0.68(8) * Singapore 0.62(8) *** 0.229(8) Thailand (5) * (8) *** Noes: ) The %, 5%, and 0% criical values for he KPSS es of rend saionariy are 0.26, 0.46, and 0.9, respecively. 2) The %, 5%, and 0% criical values for he KPSS es of level saionariy are 0.739, 0.463, and 0.347, respecively. 3) *, **, *** denoes accepance of he null hypohesis of rend saionariy a he %, 5%, and 0% significance levels, respecively. 4) The null hypohesis of saionariy is acceped if he value of he KPSS es saisics is less han i is criical value. 5) The numbers in parenheses are he lags runcaion seleced auomaically by Newey and Wes Bandwidh using Barle Kernal Specral esimaion mehod. 6) he null of level saionariy is esed. The resuls of applying he KPSS es on he real exchange raes show srong evidence of saionariy for four counries: Indonesia, Korea, Malaysia and Thailand since he null of saionariy is acceped a he 0 percen significance level. Weak evidence of saionariy is found for he Philippines since he null is acceped a he percen significance level. The null could no be acceped for Singapore, which implies ha he real exchange rae in Singapore is non-saionary. These findings differ markedly from hose of he sandard uni roo ess and give evidence in favor of long-run PPP in hese counries. As argued, sandard uni roo ess lack power; herefore, he null hypohesis of a uni roo can no be rejeced unless here is srong evidence agains i. In he nex secion we conduc confirmaory analysis o srenghen inferences made abou he saionariy or non-saionariy of he real exchange raes.

13 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES Confirmaory Analysis In confirmaory analysis, he null of a uni roo and he null of saionariy are joinly esed o confirm he exisence or non-exisence of PPP. Table 7 presens he resuls of he confirmaory analysis for he nominal and real exchange raes in he Asian counries. The resuls show no confirmaion on he saionariy or non-saionariy of he nominal exchange raes for all he counries since he null of a uni roo is acceped and he null of saionariy is acceped. This may be due o he daa being insufficienly informaive abou he behavior of he nominal exchange raes. Counry Table 7. Confirmaory Analysis for he Nominal and Real Exchange Raes The Nominal Exchange Rae The Real Exchange Rae ADF H Conclusion H ADF H KPSS Conclusion H KPSS Indonesia Accep Accep Inconclusive Rejec Accep Saionary Korea Accep Accep Inconclusive Rejec Accep Saionary Malaysia Accep Accep Inconclusive Accep Accep Inconclusive Philippines Accep Accep Inconclusive Accep Accep Inconclusive Singapore Accep Accep Inconclusive Accep Rejec Non-saionary Thailand Accep Accep Inconclusive Accep Accep Inconclusive Noes: H ADF is he null hypohesis of a uni roo under he ADF es. H KPSS is he null hypohesis of level saionariy under he KPSS es. As for he real exchange raes, saionariy is confirmed for only wo counries: Indonesia and Korea since he null of a uni roo is rejeced and he null of saionariy is acceped. Non-saionariy is confirmed for Singapore since he null of a uni roo is acceped and he null of saionariy is rejeced. We have no confirmaion for he remaining counries: Malaysia, he Philippines and Thailand since boh nulls are acceped. Thus, we are able o provide and confirm saionariy of real exchange raes for Indonesia and Korea wih a poin esimae for half-life deviaions of approximaely 20.3 and 4. percen per quarer, respecively. This implies a half-life of 0.76 years for Indonesia and.4 years for Korea. We furher explore he behavior of he real exchange raes in he Asian counries by examining he impac of he 997 Asian financial crisis on he real exchange raes in he crisis counries. The financial crisis should have changed he behavior of he real exchange raes in he crisis counries and caused srucural break. Perron (989) argues ha mos macroeconomic variables are no uni roo processes; raher hey are rend saionary processes wih srucural breaks. He argues ha if a series conains a srucural break, hen sandard uni roo ess will fail o rejec he null of a uni roo when in fac he series is saionary. Therefore, i seems relevan o allow for srucural breaks when esing real exchange raes for saionariy. In he nex secion we shed some ligh on he

14 42 Asian crisis and hen es he null of a uni roo allowing for a one-ime srucural break o capure he effec of he crisis. 6. THE ASIAN CRISIS Prior o he Asian financial crisis, banks and businesses in he crisis counries had borrowed heavily shor-erm in yen and dollars (Tobin (998)). Repors from he Inernaional Moneary Fund (IMF (998)) have indicaed he following major reasons for he crisis: high yields in he crisis counries relaive o he res of he world, high volumes of unproducive capial inflows ino he crisis counries, 3 inconsisen domesic macroeconomic and exchange rae policies, and weaknesses in he financial secors. These policies have resuled in overvaluaion of he currencies of he crisis counries, sronger han if lef o marke forces. Signs of weakness began o appear, crediors of he crisis counries began worrying abou heir money. Markes overreaced, causing massive capial ouflows. The resul was a financial panic; a quick and huge decrease in he value of he crisis counries currencies. The depreciaion was bigger han was required o offse he iniial over-valuaion, which resuled in deviaions from PPP. If PPP holds in he long-run, deviaions from PPP should be ransiory. Arbirage opporuniies will pu pressure on he exchange rae and price levels o adjus, eiher he exchange rae will depreciae or he price raio of domesic o foreign counry will decrease. Since he exchange raes in he crisis counries were pegged o he U.S. dollar under a managed floa sysem (excep he Philippines which operaed under an independen floa sysem), he price level was he only mechanism for adjusmen. Bu since prices are relaively sicky in he shor-run (Rogoff (996)), his caused he pegged sysem o be removed, which resuled in depreciaion of he crisis counries currencies wih respec o he U.S. dollar. Figure shows he cumulaive deviaions ( cd ) from PPP from 990: o 999:4. The choice of 990: as he base year is based on Figure 2, which shows he cumulaive deviaions from PPP from 973:2 o 999:4. Figure 2 shows ha he build-up pressure for he crisis counries currencies o devalue sared for mos of he counries a lile afer 990. Prior o he crisis, if he crisis counries currencies were overvalued, he slope of cd should have been negaive indicaing appreciaion in he real exchange rae. Figure shows ha he slope of cd prior o he crisis was indeed negaive for he crisis counries. The 997 Asian crisis should have caused a srucural break due o he adjusmen in he exchange rae (depreciaion). The depreciaion should have caused he slope of cd o change from negaive o posiive. The figure indicaes ha afer he crisis here was a change in he slope from negaive o posiive. This marks a srucural 3 Almos half of oal capial inflows o developing counries wen o Asia 996-nearly $00 billion (IMF, June 998).

15 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES 43 break in he slopes of he real exchange rae series for he crisis counries. Therefore, a uni roo es in he presence of a srucural break which allows for a change in he slope of he series is relevan. For his purpose, he Perron (989) augmened uni roo es is applied. The resuls of he esimaion are given in Table 8 for only Model B he changing growh model, which allows for a one-ime change in he slope of he series. 4 We can rejec he null of a uni roo for Indonesia and Korea a he 5 percen significance level and for Malaysia and Thailand a he 0 and percen significance levels, respecively. Thus, allowing for a one-ime break in he slope of he real exchange rae series provides srong evidence in favor of PPP for Indonesia, Korea and Thailand and weak evidence for Malaysia. The null of a uni roo could no be rejeced For he Philippines and Singapore, even wih srucural breaks aken ino accoun. Table 8. Uni Roo Tes in he Presence of Srucural Break for he Real Exchange Rae Series Counry k β 0 β β 2 φ α Indonesia (4.8) (5.056) (3.757) (-4.336) (-4.97) ** Korea (4.005) (4.07) (-0.590) (-2.758) (-4.020) ** Malaysia (3.66) (5.667) (2.796) (-3.884) (-3.533) * Philippines (2.275) (3.27) (0.72) (-2.752) (-2.272) Singapore (.570) (3.034) (-0.907) (-2.038) (-.322) Thailand (4.869) (5.723) (2.982) (-3.884) (-4.876) *** Noes: The %, 5%, and 0% criical values for λ = 0. 9 are -4.26, -3.68, and -3.35, respecively. *, **, and *** denoe significance a he 0%, 5%, and %, respecively. -values are parenheses. Model B: he changing growh model which allows for a one-ime change in he slope of he rend funcion. y k 0 + βdmu + β2 + φdts + αy + θ p y p ε. p= = β + The null hypohesis is a uni roo wih a one-ime change in he drif agains he alernaive of a one-ime change in he slope of a rend saionary series. 4 The resuls obained from esimaing Model A indicaed rejecion of he null for Indonesia and Korea a he 5 and percen significance levels, respecively. The resuls from Model C indicaed rejecion of he null for Indonesia and Korea a he percen significance levels. The hree models are also esimaed assuming he break dae is 997:3. For Model A, he null of a uni roo could no be rejeced for all cases. For Model B, he null is rejeced for Indonesia and Korea a he 5 percen significance level, and for Thailand a he 0 percen level. For Model C, he null is rejeced for Indonesia a he percen level.

16 cd_ind cd_kor cd_mal cd_phi cd_sin cd_tha Figure. Cumulaive Deviaions from PPP from 990:2 o 999: c d _ I N D O c d _ K O R c d _ M A L c d _ P H I c d _ S I N c d _ T H A Figure 2. Cumulaive Deviaions from PPP from 973:2 o 999:4

17 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES SUMMARY AND CONCLUSIONS This paper has aemped o es he validiy of PPP for a sample of developing counries in he Asian financial crisis counries during he curren floa. Using he ADF and PP uni roo ess, srong evidence of saionary real exchange is found for only Indonesia, and weak evidence a he 0 percen significance level is deeced for Korea and Thailand. Srong evidence in favor of PPP is deeced by he KPSS es for Indonesia, Korea, Malaysia, and Thailand. Combining he null of a uni roo wih he null of saionariy confirmed real exchange rae saionariy for Indonesia and Korea. Non-saionariy is confirmed for Singapore. The resuls from applying Perron s es ha accouns for poenial srucural breaks in he real exchange rae series shows evidence of saionariy for Indonesia and Korea a he 5 percen significance level, and for Malaysia and Thailand a he 0 and percen significance levels. Thus, we conclude ha PPP does hold in four ou of he six Asian counries under sudy, which implies ha deviaions from PPP are ransiory. REFERENCES Abuaf, N., and P. Jorion (990), Purchasing Power Pariy in he Long Run, Journal of Finance, 45(), Achy, L. (2003), Pariy Reversion Persisence in Real Exchange Raes: Middle Income Counry Case, Applied Economics, 35, Baharaumshah, A., and M. Ariff (997), Purchasing Power Pariy in Souh Eas Asian Counries Economies: A Coinegraion Approach, Asian Economic Journal, (2), Baum e al. (999), Long Memory or Srucural Breaks: Can Eiher Explain Nonsaionary Real Exchange Raes under he Curren Floa? Journal of Inernaional Financial Markes, Insiuions and Money, 9, Culver, S., and D. Papell (999), Long-run Purchasing Power Pariy wih Shor-run Daa: Evidence wih a Null Hypohesis of Saionariy, Journal of Inernaional Money and Finance, 8, DeJong, D. e al. (992), Inegraion Versus Trend Saionariy in Time Series, Economerica, 60(2), Diebold, F., and G. Rudebusch (99), On he Power of Dickey-Fuller Tess agains Fracional Alernaives, Economics Leers, 35, Doğanlar, M. (999), Tesing Long-run Validiy of Purchasing Power Pariy for Asian Counries, Applied Economics Leers, 6(3), Engle, C. (2000), Long-Run PPP May No Hold Afer All, Journal of Inernaional Economics, 5, Frankel, J. (986), Inernaional Capial Mobiliy and Crowding-ou in he US Economy: Imperfec Inegraion of Financial Markes or Goods Markes? In How Open is he

18 46 US Economy? ed. by W. Hafer, Lexingon: Lexingon Books, (990), Zen and he Ar of Modern Macroeconomics: A Commenary in: Moneary Policy for a Volaile Global Economy, eds. by William S. Haraf and Thomas D. Wille, Washingon, DC: American Enerprise Insiue for Public Policy Research, Frankel, J., and A. Rose (996), A Panel Projec on Purchasing Power Pariy: Mean Reversion wihin and beween Counries, Journal of Inernaional Economics, 40, Froo, K., and K. Rogoff (994), Perspecives on PPP and Long Run Real Exchange Raes, NBER working paper No Gan, W. (994), Characerizing Real Exchange Rae Behavior of Seleced Eas Asian Economies, Journal of Economic Developmen, 9(2), Glen, J. (992), Real Exchange Raes in he Shor, Medium, and Long Run, Journal of Inernaional Economics, 33, Heimonen, K. (999), Saionariy of he European Real Exchange Rae - Evidence from Panel Daa, Applied Economics, 3, Inernaional Moneary Fund (998), World Economic Oulook, May 998, IMF, Washingon, DC. Kuo, B., and A. Mikkola (999), Re-Examining Long-Run Purchasing Power Pariy, Journal of Inernaional Money and Finance, 8, Kwiakowski, D. e al. (992), Tesing he Null Hypohesis of Saionariy agains he Alernaive of a Uni Roo, Journal of Economics, 54, Levin, A., and Chen-Fu Lin (992), Uni Roo Tess in Panel Daa: Asympoic and Finie-sample Properies, Universiy of California, San Diego, Deparmen of Economics, Discussion paper (993), Uni Roo Tess in Panel Daa: New Resuls, Universiy of California, San Diego, Deparmen of Economics, Discussion paper Lohian, J., and M. Taylor (996), Real Exchange Rae Behavior: The Recen Floa from he Perspecive of he Pas Two Cenuries, Journal of Poliical Economy, 04(3), Luinel, K. (2000), Real Exchange Rae Behavior: Evidence from Black Markes, Journal of Applied Economerics, 5, MacDonald, R. (996), Panel Uni Roo Tess and Real Exchange Raes, Economics Leers, 50, 7-. Mark, N. (990), Real and Nominal Exchange Raes in he Long Run: An Empirical Invesigaion, Journal of Inernaional Economics, 28, Moniel, P. (997), Exchange Rae Policy and Macroeconomic Managemen in ASEAN Counries, in J. Hicklin e al.: Macroeconomic issues facing ASEAN counries, Washingon, DC, IMF. Newey, W., and K. Wes (987), A Simple, Posiive Semi-definie, Heeroskedasiciy and Auocorrelaion Consisen Covariance Marix, Economerica, 55(3), O Connell P. (998), The Overvaluaion of Purchasing Power Pariy, Journal of

19 TESTING THE VALIDITY OF PURCHASING POWER PARITY FOR ASIAN COUNTRIES 47 Inernaional Economics, 44, -9. Papell, D. (997), Searching for Saionariy: Purchasing Power Pariy under he Curren Floa, Journal of Inernaional Economics, 43, Perron, P. (989), The Grea Crash, he Oil Price Shock, and he Uni Roo Hypohesis, Economerics, 57, Rogoff, K. (996), The Purchasing Power Pariy Puzzle, Journal of Economic Lieraure, 34, Tobin, J. (998), Asian Financial Crisis, Japan and he World Economy, 0, Wang, P. (2000), Tesing PPP for Asian Economies during he Recen Floaing Period, Applied Economics Leers, 7, Weliwia, A. (998), Coinegraion Tess and he Long-run Purchasing Power Pariy: Examinaion of Six Currencies in Asia, Journal of Economic Developmen, 23(), Wu, J., and Chen (999), Are Real Exchange Raes Saionary Based on Panel Uni Roo Tess? Evidence from Pacific Basin Counries, Inernaional Journal of Finance and Economics, 4, Mailing Address: Assisan Professor, Deparmen of Finance and Banking, Al-zayoonah Universiy of Jordan, Amman, Jord. Tel: , ex. 28, Fax: Manuscrip received December, 2002; final revision received July, 2003.