Purchasing Power Parity Hypothesis in Developing Economies: Some Empirical Evidence from Sri Lanka

Transcription

1 Purchasing Power Pariy Hypohesis in Developing Economies: Some Empirical Evidence from Sri Lanka Gunerane B Wickremasinghe Deparmen of Economerics and Business Saisics Monash Universiy Caulfield Vicoria, gunerane.wickremasinghe@buseco.monash.edu.au

2 Absrac Purchasing power pariy (PPP) hypohesis has araced a lo of aenion from academics and policy-makers paricularly, during he recen floa. Mos previous sudies used daa from he developed world. This sudy examines he validiy of he PPP hypohesis using daa during he recen floa from Sri Lanka. In conras o previous sudies, we use uni roo ess which ake ino accoun unknown means and rends in he real exchange raes as well as graphical echniques. Boh hese echniques overwhelmingly rejec he empirical validiy of he PPP hypohesis for Sri Lanka. The resuls from widely-used uni roo ess, however, provide mixed evidence. We aribue hese inconclusive resuls o he low power of he widely-used uni roo ess and heir inabiliy o accoun for unknown rends and means in he real exchange raes. Key Words: Real exchange raes, Sri Lanka, Purchasing power pariy, ERS es, DF- GLS es, US dollar JEL Classificaions: F31, C22 1

3 Purchasing Power Pariy Hypohesis: Some Empirical Evidence from Sri Lanka 1. Inroducion Purchasing power pariy (PPP) is a cornersone of many heoreical models in inernaional finance. I assers ha he price of a paricular commodiy, when expressed in a common currency, should be he same in every counry. The PPP is an imporan concep for policy makers in developing counries for a leas wo reasons (Holmes, 2001b). Firs, PPP can be used as a model o predic exchange raes and deermine wheher a paricular currency is over or undervalued. Predicing exchange raes and deermining wheher a currency is over or undervalued is paricularly imporan for less developed counries and hose experiencing large differences beween domesic and foreign inflaion raes. Second, many heories of exchange rae deerminaion use some noion of PPP in heir consrucion. Therefore, he validiy of PPP is imporan o policy makers in developing counries who make heir recommendaions on he basis of PPP (Liu and Burke, 1995). Empirical evidence on PPP is abundan in relaion o developed as well as developing economies (see, Moosa, 1994; Papell, 1997; Frische and Wallace, 1997; Koureas, 1997; Heimonen, 1999; Gil-Alanla, 2000; Caporale e al., 2001; Esaka, 2002, for developed counries and Soofi, 1998; Choudhry, 1999; Azali e al., 2001; Nagayasu, 2002; Holmes, 2002; Achy, 2003, for developing counries). These sudies use differen daa ses and mehodologies. However, he resuls of empirical sudies, paricularly during he floaing exchange rae regime, have no been consisen providing mixed evidence on he validiy of PPP. Almos all he empirical sudies on PPP prior o he early par of he 1990s were conduced on he assumpion of linear adjusmen of real exchange raes owards 2

4 deviaions from PPP. However, several recen papers (see, Dumas, 1992; Uppal, 1993; Sercu e al., 1995; Coleman, 1995) develop models of real exchange rae deerminaion which ake ino accoun fricions exising in inernaional rade such as ransacion coss ha lead o nonlinear adjusmen of real exchange raes oward deviaions from PPP. Subsequenly, here have been several empirical sudies which model he adjusmen of real exchange raes oward PPP deviaions as a nonlinear process using recenly developed economeric echniques (see, Baum e al., 2001, Coakley and Fueres, 2001; Enders and Dibooglu, 2001 and Chen and Wu, 2000). Mos of hese papers provide evidence in favour of PPP hypohesis. To he knowledge of he auhor, here have been four recen sudies using Sri Lankan daa. These sudies used daa for bilaeral exchange raes (Aggarwal e al., 2000; Holmes, 2001a; Holmes, 2001b) and nominal and effecive exchange raes (Weliwia, 1998), and employed differen economeric echniques providing mixed resuls. However, none of he previous sudies has examined wheher he PPP exiss beween Sri Lanka and France, Germany, he Unied Kingdom and India on a bilaeral basis. The objecive of his paper is o exend he previous empirical lieraure on PPP in Sri Lanka by examining he mean-reversion of six bilaeral real exchange raes. In addiion o he graphical analysis used o gain a preliminary idea on he PPP, wo recen uni roo ess (GLS-derended Dickey-Fuller es and ERS poin opimal es (Ellio, Rohenberg and Sock, 1996)) ha assume linear adjusmen of real exchange raes oward PPP deviaions are used in he analysis. The GLS-derended Dickey- Fuller es has he bes overall performance in erms of small sample size and power dominaing he ordinary Dickey-Fuller es. The ERS poin opimal es is robus o 3

5 he presence of an unknown mean or rend in a ime series. For comparison, resuls from wo widely-used uni roo ess (Phillips-Perron (PP) (1988) es and Kwiakowski, Phillips, Schmid, and Shin (KPSS) (1992) es) are also repored. The paper is organized as follows. Mehodology and daa used are described in he second secion. The penulimae secion discusses he empirical resuls. The fourh secion presens he conclusion of he paper. 2. Mehodology and daa To invesigae he empirical validiy of PPP in Sri Lanka, wo mehodologies are used: graphical and economeric. In he firs graphical mehod, validiy of PPP is invesigaed by examining wheher here are deviaions beween acual and PPP exchange raes. If PPP holds, ime series plos of spo exchange raes should overlap wih hose for he PPP exchange raes. The PPP exchange raes are calculaed using he following formula: = P / P0 S S0 (1) * P / P * 0 where S is he PPP exchange rae for period, S 0 is he spo exchange rae in he base period, is he domesic price level in period, P is he domesic price level P 0 * * in he base period, is he foreign price level in period, and is he foreign price P level in he base period. If PPP holds, he lines for PPP exchange raes and hose for acual exchange raes should move close o each oher over ime. P 0 4

6 In he second graphical mehod, he validiy of he PPP is examined by ploing he graphs of real exchange raes for each currency. If he PPP holds real exchange raes should be saionary. This means ha alhough here are shor-run deviaions of he real exchanges raes from heir means, in he long-run hey end o rever o heir means. Real exchange raes are calculaed using he following equaion: r = s + p * p (2) * where, r is he naural logarihm of he real exchange rae, p and are as defined above. p Economeric analysis is underaken using wo recenly developed uni roo ess: Dickey-Fuller generalized leas squares, (DF-GLS), and Ellio, Rohenberg and Sock (ERS) poin opimal es. Two old uni roo ess (Phillips-Perron (PP) and Kwiakowski, Phillips, Schmid, and Shin (KPSS)) ha are widely used are used for comparison. Each of he uni roo ess used is briefly discussed below. Phillips-Perron (PP) es The Phillips-Perron (PP) (1988) es suggess a non-parameric mehod of conrolling for higher order auocorrelaion in a series. This es is based on he following firs order auo-regressive (AR(1)) process: y = α + β + ε (3) y 1 5

7 where y is he variable of ineres, is he difference operaor, α is he consan, β is he slope and is a subscrip for ime. The non-parameric correcion is made o he - raio of β coefficien from equaion (3) o accoun for he auocorrelaion of ε. This correcion is based on an esimae of he specrum of ε a zero frequency ha is robus o heeroskedasiciy and auocorrelaion of unknown form. In his paper, his esimaion is based on Barle kernel. The opimal bandwidh in he PP equaion is seleced using he Newey-Wes (1994) mehod. Criical values abulaed by MacKinnon (1996) are used in making inferences regarding he ime series properies of he variables. Kwiakowski, Phillips, Schmid, and Shin (KPSS) Tes KPSS es is based on he residuals from an ordinary leas squares regression of he variable of ineres on he exogenous variable or variables: consan or consan and ime rend. The residuals obained are used o calculae he Lagrange Muliplier (LM) saisic (see, Kwiakowski, Phillips, Schmid, and Shin, 1992 for deails) used in he es as follows: (uˆ) LM = T = S( ) /( T f ) (4) 0 where S() is uˆ r r= 1 and f0 is an esimaor of he residual specrum a zero frequency and T is he number of observaions. Unlike in oher uni roo ess, in his es i is assumed ha he series is saionary under he null. The criical values abulaed by KPSS are used in making inferences regarding saionariy. 6

8 Dickey-Fuller Generalised Leas Squares (DF-GLS) es This is a more powerful es han Dickey-Fuller ype ess. In he Augmened Dickey- Fuller (ADF) (1979,1981) es regression, eiher a consan or a consan and a linear ime rend is included o ake accoun of he deerminisic componens of daa. Ellio, Rohenberg and Sock (ERS), however, propose a modificaion o ADF regression in which daa are derended before he uni roo es is conduced. This de-rending is done by aking he explanaory variables ou he daa (see, Ellio, Rohenberg and Sock, 1996 for deails). Then he following equaion is esimaed o es for a uni roo in he variable: y = α y + β y β y + v d d d d 1 1 p p (5) d where is he difference operaor, is he generalised leas squares de-rended y value of he variable, α, β and β p are coefficiens o be esimaed and v is he independenly and idenically disribued error erm. As in he case of he ADF es, es for a uni roo of he variable y consiss of esing wheher he coefficien of he AR(1) erm, in his case α, in equaion (5) is zero agains i is less han zero. In making inferences, criical values abulaed by Ellio, Rohenberg and sock are used. Ellio, Rohenberg and Sock (ERS) Poin Opimal Tes The ERS poin opimal es has been found o dominae oher commonly used uni roo ess, paricularly, when a ime series has an unknown mean or a linear rend. This es is based on he following quasi-differencing regression: d( y a) = d( x a) δ ( a) + η (6) 7

9 where d( y a ) and d( x a) are quasi-differenced daa for y and x respecively and η is he error ha is independenly and idenically disribued. Deails on compuing quasi differences are given in Ellio, Rohenberg and Sock (1996). In equaion (6), y is he variable whose ime series properies are esed, x may conain a consan only or boh a consan and a ime rend and δ ( a) is he coefficien o be esimaed. ERS recommend he use of a for a in equaion (6) ha is compued as a = 1 7/ T and a = / T when x conains a consan and a consan and ime rend respecively. In he ERS poin opimal es, he null and alernaive hypoheses esed are α = 1 and α = a respecively. The relevan es saisic (P T ) o es he above null hypohesis is: P = ( SSR( a) ( a) SSR(1))/ f (7) T 0 where SSR is he sum of squared residuals from equaion (6) and f 0 is an esimaor for he residual a frequency zero. In making inferences, he es saisic calculaed is compared wih he simulaion based criical values of ERS. In he empirical analysis, he four uni roo ess are conduced wih a consan and a ime rend in he es equaions. Daa Daa used in he sudy consis of he average exchange raes expressed in erms of he amoun of Sri Lankan rupees per uni of German mark (DM), French franc (FF), he UK pound (GBP), Indian rupee (IR), Japanese yen (JY) and he US dollar (USD) and he consumer and wholesale price indices for Sri Lanka, Germany, France, he UK, 8

10 India, Japan and he USA on a monhly basis from January 1986 o November Monhly exchange raes were obained from he Cenral Bank of Sri Lanka while daa on consumer and wholesale/producer price indices wih base year 1995 for each counry excep for producer price index for France were obained from inernaional financial saisics CD-ROM. The producer price index for France wih base year 1995 was obained from he DX daabase. 3. Empirical resuls Take in Figure 1 Figure 1 depics he graphs for he CPI-based PPP exchange rae and he acual exchange rae for each foreign currency. According o he PPP, he spo exchange rae of a paricular exchange rae should be equal o he PPP exchange rae. If he line for he acual exchange rae is above/ below he line for he PPP exchange rae, he local currency is said o be under-valued and vice versa. According o he graphs, all bu in one case he lines for he acual exchange raes move closely wih he line for he PPP exchange rae alhough hey do no overlap. However, in some periods hey drif apar. This means ha he exchange rae of he Sri Lankan rupee does no follow he PPP exchange raes. In relaion o he Indian rupee, is acual exchange rae is significanly above he PPP exchange rae ill 1991, bu hereafer, wo exchange raes move close o each oher. However, hey do no overlap. Ou of he six exchange raes, he exchange rae for he US dollar moves very closely wih he PPP exchange rae during he sample period. 9

11 Take in Figure 2 Figure 2 depics he graphs for acual and PPP exchange raes when wholesale price index is used o proxy domesic and foreign price levels. As in he case of CPI-based PPP exchange raes, lines for acual exchange raes and hose for PPP exchange raes excep in he case of Indian rupee, show upward rends and move close o each oher during cerain periods of ime. The gap beween he PPP exchange raes and acual exchange raes are wider for Indian rupee. These resuls are consisen wih he resuls obained when CPI is used o proxy he price levels. Take in Figure 3 Figure 3 exhibis he ime series plos of he real exchange raes for he six currencies when CPI is used o represen he domesic and foreign price levels. Figure 1a and 1b for German mark and French franc respecively show idenical behavioural paerns over ime. However, he ime series for he oher real exchange raes behave differenly over he sample period. If he PPP holds for any of he currencies, he ime series plo for such a currency should be saionary. Saionariy means ha alhough here may be flucuaions in real exchange raes around heir mean in he shor-run, he real exchange raes rever o heir means over he long run. However, figure for each currency shows ha none of hem is saionary refuing he validiy of PPP. Take in Figure 4 The ime series plos of WPI-based real exchange raes are shown in Figure 4. As in he case of CPI-based real exchange raes, he ime series plos of German mark and French franc show similar behavioural paerns. However, he oher real exchange 10

12 raes exhibi behavioural paerns differen from each oher during he sample period. A perusal of he figure indicaes ha none of he real exchange raes is saionary. Therefore, he ime series behaviour of he six real exchange raes is no consisen wih he PPP hypohesis. Take in Table 1 Table 1 repors he resuls of he four uni roo ess for he CPI-based real exchange raes. Panel A of he able repors resuls when a consan is included in he es equaion. According o he resuls, he real exchange raes for he UK pound are saionary under he PP and he KPSS ess. The Japanese yen real exchange raes are saionary only under he PP and he KPSS ess. The real exchange raes for he oher currencies (German mark, French franc, Indian rupee and he US dollar) are nonsaionary under all he uni roo ess. Therefore, PPP hypohesis applies only o he Japanese yen and he UK pound. The DF-GLS and he ERS poin opimal ess provide consisen resuls indicaing ha all he real exchange raes are non-saionary. These resuls, herefore, refue he validiy of PPP hypohesis in Sri Lanka. Panel B of Table 1 repors resuls when a consan and a linear rend are included in he es equaions. All he four uni roo ess provide consisen resuls indicaing ha he real exchange raes are non-saionary. These resuls are inconsisen wih he PPP hypohesis. Take in Table 2 Table 2 repors resuls for WPI-based real exchange raes for he six currencies. Panel A of he able repors resuls when a consan only is included in he es equaion. 11

13 Resuls indicae ha he UK pound real exchange rae is saionary under he PP and he KPSS uni roo ess. The Japanese yen is saionary only under he PP es. These resuls are consisen wih he PPP hypohesis. However, he oher four real exchange raes are no saionary under any of he four uni roo ess used. These resuls are inconsisen wih he PPP hypohesis. Resuls when a consan and a ime rend are included in he es equaions are shown in Panel B of Table 2. Resuls show ha he real exchange raes for German mark, French franc and he UK pound are saionary under he PP es. Furher, he real exchange raes for he Japanese yen is saionary under he KPSS es and hose for he UK pound are saionary under he DF-GLS es. While he above resuls are consisen wih he PPP hypohesis, he resuls among he ess are no consisen. Overall here is some evidence in favour of PPP hypohesis in respec of he real exchange raes for he UK pound. In mos of he cases, he real exchange raes for he oher currencies are non-saionary. Furhermore, he resuls of DF-GLS and ERS poin opimal ess are consisen beween CPI-based and WPI-based real exchange raes excep ha he WPI-based UK pound real exchange rae is saionary under he DF- GLS es. 4. Conclusion This paper examines he validiy of he PPP hypohesis for Sri Lanka. Exchange raes for six foreign currencies in erms of he Sri Lankan rupee during he recen floaing exchange rae regime are used in he analysis. Graphical analysis indicaes ha here are deviaions beween acual and PPP exchange raes and real exchange raes are non- 12

14 saionary. Such resuls provide preliminary evidence agains he validiy of he PPP hypohesis for Sri Lanka. Economeric analysis performed using wo new uni roo ess, DF-GLS es and ERS poin opimal es, provide resuls consisen wih he graphical analysis. Resuls of wo widely-used uni roo ess, PP and KPSS ess, are also repored for comparison. These uni roo ess provide mixed resuls. These resuls may be due o he low power of hese ess and he presence of unknown means or ime rends in he real exchange raes of Sri Lanka ha hese ess fail o ake ino accoun. The uni roo ess used in his paper are based on he assumpion of symmeric adjusmen of real exchange raes o deviaions from PPP. Therefore, he failure of symmeric uni roo ess o provide evidence in favour of PPP may also be due o he marke fricions prevailing in inernaional rade such as ransacion coss. When marke fricions are presen, deviaions from PPP follow a nonlinear sochasic process which symmeric uni roo ess fail o ake ino accoun. Nonlinear adjusmen of real exchange raes o deviaions from he PPP can be modelled using nonlinear economeric echniques such as exponenial smooh ransiion auoregressive (ESTAR) and hreshold auoregressive (TAR) models. These aspecs are lef for fuure research. Anoher avenue of fuure research in he area of PPP in Sri Lanka would be o use mulivariae linear and nonlinear models o corroborae he resuls of univariae ess applied in his paper. 13

15 References Aggarwal, R., Monanes, A. and Ponz, M. (2000) Evidence of Long-Run Purchasing Power Pariy: Analysis of Real Exchange Raes in erms of he Japanese Yen, Japan and he World Economy, 12, Achy, L. (2003) Pariy Revision Persisence in Real Exchange Raes: Middle Income Counry Case, Applied Economics, 35, Azali, M. e al. (2001) Does PPP Hold beween Asian and Japanese Economies? Evidence Using Panel Uni Roos and Panel Coinegraion, Japan and he World Economy, 13, Baum, C. F., Barkoulas, J. T. and Caglayan, M. (2001) Nonlinear Adjusmen o Purchasing Power Pariy in he Pos-Breon Woods Era, Journal of Inernaional Money and Finance, 20, Caporale, G. M., Kalyviis, S. and Piis, N. (2001) Tesing for PPP and UIP in an FIML Framework: Some Evidence for Germany and Japan, Journal of Policy Modelling, 23, Chen, S. and Wu, J. (2000) A Re-Examinaion of Purchasing Power Pariy in Japan and Taiwan, Journal of Macroeconomics, 22, Choudhry, T. (1999) Purchasing Power Pariy in High-Inflaion Easern European Counries: Evidence from Fracional and Harris-Inder Coinegraion Tess, Journal of Macroeconomics, 21, Coakley, J. and Fueres, A. (2001) Border Coss and Real Exchange Rae Dynamics in Europe, Journal of Policy modeling, 23, Coleman, A. M. (1995) Arbirage, Sorage and he Law of One Price : New Theory for he Time Series Analysis of an Old Problem, Unpublished Working Paper, Princeon Universiy, NJ. Dickey, D.A., and Fuller, W.A. (1979) Disribuion for he Esimaes for Auo Regressive Time Series wih a Uni Roo, Journal of he American Saisical Associaion, 74, Dickey, D.A., and Fuller, W.A. (1981) Likelihood Raio Saisics for Auoregressive Time series wih a Uni Roo, Economerica, 49, Dumas, B., (1992) Dynamic Equilibrium and he Real Exchange Rae in a Spaially Separaed World, Review of Financial Sudies, 5, Ellio, G., Rohenberg, T. J. and Sock, J. H. (1996) Efficien Tess for an Auoregressive Uni Roo, Economerica, 64, Enders, W. and Dibooglu, S. (2001) Long-Run Purchasing Power Pariy wih Asymmeric Adjusmen, Souhern Economic Journal, 68,

16 Esaka, T. (2002) Panel Uni Roo Tess of Purchasing Power Pariy beween Japanese Ciies, : Disaggregaed Price Daa, Japan and he World Economy, 15, Frische, C. P. and Wallace, M. (1997) Forecasing he Exchange Rae PPP Versus a Random Walk, Economics Leers, 54, Gil-Alana, L. A. (2000) Mean Reversion in he Real Exchange Raes, Economics Leers, 69, Heimonen, K. (1999) Saionariy of he European Real Exchange Raes Evidence from Panel Daa, Applied Economics, 31, Holmes, M. J. (2001a) Principal Componens, Saionariy, and New Evidence of Purchasing Power Pariy in Developing Counries, Developing Economies 39: Holmes, M. J. (2001b) New Evidence on Real Exchange Rae Saionariy and Purchasing Power Pariy in less developed counries, Journal of Macroeconomics 23: Holmes, M. J. (2002) Purchasing Power Pariy and he Fracional Inegraion of Real Exchange Rae: New Evidence for Less Developed Counries, Journal of Economic Developmen, 27, Koureas, G. P. (1997) The Canadian Dollar and Purchasing Power Pariy during he Recen Floa, Review of Inernaional Economics, 5, Kwiakowski, D., Phillips, P. C. B., Schmid, P. and Shin, Y. (1992) Tesing he Null Hypohesis of Saionariy agains he Alernaive of a Uni Roo: How Sure are We ha he Economic Time Series Have a Uni Roo?, Journal of Economerics, 54, Liu, P. and Burke, P. (1995) Insabiliy in Shor-Run Adjusmen o Purchasing Power Pariy: Resuls for Seleced Lain American Counries, Applied Economics, 27, MacKinnon, J. G. (1996) Numerical Disribuion Funcions for Uni Roo and Coinegraion Tess, Journal of Applied Economerics, 11, Moosa, I. A. (1994) Tesing Proporionaliy, Symmery and Exclusiveness in Long-Run PPP, Journal of Economic Sudies, 21, Nagayasu, J. (2002) Does he Long-Run PPP Hypohesis Hold for Africa? Evidence from a Panel Coinegraion Sudy, Bullein of Economic Research, 54, Newey, W. and Wes, K. (1994) Auomaic Lag Selecion in Covariance Marix Esimaion, Review of Economic Sudies, 61, Papell, D. (1997) Searching for Saionariy: Purchasing Power Pariy under he Curren Floa, Journal of Inernaional Economics, 43,

17 Phillips, P. C. B. and Perron, P. (1988) Tesing for a Uni Roo in Time Series Regression, Biomerica, 75, Sercu, P., Uppal, R. and Van Hull, C. (1995) The Exchange Rae in he Presence of Transacion Coss: Implicaions for Tess of Purchasing Power Pariy, Journal of Finance, 50, Soofi, A. S. (1998) A Fracional Coinegraion Tes of Purchasing Power Pariy: The Case of Seleced Members of OPEC, Applied Financial Economics, 8, Uppal, R. (1993) A General Equilibrium Model of Inernaional Porfolio Choice, Journal of Finance, 48, Weliwia, A. (1998) Coinegraion Tess and he Long-Run Purchasing Power Pariy: Examinaion of Six Currencies in Asia, Journal of Economic Developmen, 23,

18 Figure 1. Acual and CPI-based PPP exchange raes PPP exchange rae PPP exchange rae Acual exchange rae Acual exchange rae Acual exchange rae Figure 1a. German mark Figure 1b. French franc 1.4 PPP exchange rae Figure 1c. Indian rupee Acual exchange rae PPP exchange rae Acual exchange rae PPP exchange rae Acual exchange rae PPP exchange rae.1 Figure 1d. Japanese yen 30 Figure 1e. UK pound 20 Figure 1f. US dollar 17

19 Figure 2. Acual and WPI-based PPP exchange raes PPP exchange rae PPP exchange rae Acual exchange rae Acual exchange rae Acual exchange rae PPP exchange rae 10 Figure 2a. German mark 2 Figure 2b. French franc 1.2 Figure 2c. Indian rupee PPP exchange rae Acual exchange rae PPP exchange rae Acual exchange rae PPP exchange rae Acual exchange rae.1 Figure 2d. Japanese yen 30 Figure 2e. UK pound 20 Figure 2f. US dollar 18

20 Figure 3. CPI-based real exchange raes Figure 1a. German mark 1.9 Figure 1b. French franc Figure 1c. Indian rupee Figure 1d. Japanese yen Figure 1e. UK pound Figure 1f. US dollar 19

21 Figure 4. WPI-based real exchange raes Figure 2a. German mark 1.9 Figure 2b. French franc Figure 2c. Indian rupee Figure 2d. Japanese yen Figure 2e. UK pound 3.76 Figure 2f. US dollar 20

22 Table 1 Uni roo ess of CPI-based real exchange raes Exchange PP DF-GLS KPSS ERS rae Panel A: Consan GM (8) (4) (10) b (1) FF (9) (5) (10) a (1) IR (2) (1) (10) a (1) JY (4) b (11) (10) (1) GBP (6) b (2) (10) (1) USD (25) (11) (10) a (3) Panel B: Consan and linear rend GM (10) (4) (10) a (1) FF (11) (5) (10) a (1) IR (2) (1) (10) a (1) JY (5) (11) (10) b (1) GBP (6) (2) (10) b (1) USD (23) (13) (10) b (3) Noes: 1. a and b imply significance a he 1% and 5% level, respecively. 2. GM, FF, IR, JY, GBP and USD denoe respecively he real exchange raes for German mark, French franc, Indian rupee, Japanese yen, he UK pound and he US dollar. 3. The numbers wihin brackes followed by DF-GLS and ERS saisics represens he lag lengh of he dependen variable used o obain whie noise residuals. 4. The lag lenghs for DF-GLS equaion were seleced using Akaike Informaion Crierion (AIC). 5. The numbers wihin brackes followed by PP and KPSS saisics represen he bandwidh seleced based on Newey-Wes mehod using Barle Kernel. 6. The numbers wihin brackes shown afer he ERS saisic indicae he specral OLS AR based on SIC. 21

23 Table 2 Uni roo ess of WPI-based real exchange raes Exchange rae PP DF-GLS KPSS ERS Panel A: Consan GM (5) (2) (10) a (0) FF (4) (2) (10) a (0) IR (3) (1) (10) a (0) JY (4) (1) (10) (1) GBP (0) a (2) (10) (0) USD (0) (4) (10) b (1) Panel B: Consan and linear rend GM (5) c (2) (10) a (0) FF (2) c (1) (10) b (0) IR (3) (1) (10) a (0) JY (4) (1) (10) (1) GBP (0) c (1) b (10) b (0) USD (2) (1) (10) b (1) Noes: 1. a, b and c imply significance a he 1%, 5% and & 10% level, respecively. 2. See noes for Table 1 for he definiions of variables in column 1. 22