1 Inroducion In counries ha adop inermediae exchange regimes, policy auhoriies conduc exchange marke inervenion o mainain he nominal exchange rae sabi

Transcription

1 Esimaing Exchange Marke Pressure for Asian Counries Λ Nasuki Arai January 2005 Absrac Exchange marke pressure (EMP), he sum of he exchange rae depreciaion and reserve ou- Λows, summarizes he exen of pressure exered on foreign exchange marke. This paper analyzes exchange marke pressure in crisis affeced Asian counries which conains Indonesia, Japan, Korea, Malaysia, Philippines, Singapore and Thailand from 1990 o We calculae he exchange marke pressure in hese counries on he basis of simple small open economy framework proposed by Weymark (1995) using coineraing mehod and nd ha here exiss coinegraing relaionship in money demand in several Asian counries. In he lieraure of exchange marke pressure, he componen of he reserve ouλows is measured differenly. We also calculae he exchange marke pressure de ned by Giron and Roper (1977), Kaminsky and Reinhar (1999) respecively and compare hese resuls. We can infer ha he exchange marke pressure index speci ed by Weymark can idenify a currency crisis period more properly han indices speci ed by Giron and Roper and Kaminsky and Reinhar. Λ I am graeful o Professor Masahiro Kawai for helpful commens and suggesions. I also hank he paricipans in he lunch meeing seminar. 1

2 1 Inroducion In counries ha adop inermediae exchange regimes, policy auhoriies conduc exchange marke inervenion o mainain he nominal exchange rae sabiliy. When a speculaive aack occurs, no only he change in he nominal exchange raes bu also he change in he domesic moneary componen such as he reserve money or he domesic ineres raes should be considered o measure pressure exered on he currency. Giron and Roper(1977) propose an exchange marke pressure index composed of simple sum of he change in he nominal exchange rae and he change in he foreign exchange reserves scaled by he base money. Based on assumpions of a small-counry, hey measured he volume of inervenion necessary o achieve any desired exchange rae arge. On he oher hand, Kaminsky and Reinhar(1999) measures exchange marke pressure as a weighed sum of he change in he nominal exchange rae and he change in he foreign exchange reserves. The weigh is deermined by raio of volailiy o avoid sudden currency depreciaion or large reserve loss dominaing he movemen of index. This measure of exchange marke pressure is widely used in lieraure of nancial crisis and conagion. (Eichengreen, Rose and Wyplosz(1996) e.al) However, neiher of hese speci caions of he exchange marke pressure can capure he magniude of pressure accuraely, due o he fac ha Kaminsky and Reinhar's measure of exchange marke pressure ignores he effec of he change in he domesic credi creaed by he change in he ne foreign asse of moneary auhoriy and Giron and Roper's measure of exchange marke pressure are based on he special assumpion ha he money muliplier is consan over ime. Weymark(1995) proposed a mehod of measuring exchange marke pressure consisen wih simple model of a small open economy. I is shown ha he weigh of he change in he nominal exchange raes and he change in he foreign exchange reserves as proporion of inheried money sock is deermined in he model. The objecive of his paper is o esimae exchange marke pressure for crisis affeced Asian counries which conain Indonesia, Japan, Korea, Malaysia, Philippines, Singapore, and Thailand from 1990 o 2004 based on monhly daa. In he period of Asian nancial crisis occurred in 1997, hese counries excep Japan were suffered from he remendous pressure of depreciaion creaed by massive speculaive aacks. Especially, Indonesia, Korea and Thailand were forced o abandon heir de faco xed exchange rae regimes and shif o he Λoaing exchange regimes. Malaysia adoped an unprecedened capial conrol policy in Sepember 1998 o insulae domesic money marke from he volailiy of inernaional capial Λow. To measure he exen of pressure exered on hese counries precisely, we should consider no only he change in he nominal exchange raes bu also he change in he foreign exchange reserves or he change in he domesic money sock creaed by he change in foreign exchange reserves. Following he discussion above, we use Weymark's small open economy framework o calculae he exchange marke pressure for hese counries. Calculaing he exchange marke pressure, we analyze coinegraing relaionships in he money demand and he price equaion. 2

3 We also calculae he exchange marke pressure indices for Asian counries measured by Giron and Roper's mehod and Kaminsky and Reinhar's mehod, and compare hese hree indices. This paper is organized as follows. In secion 2, we brieλy review he lieraure of exchange marke pressure: Giron and Roper model, Kaminsky and Reinhar speci caion and Weymark's model. To esimae he exchange marke pressure proposed by Weymark, we analyze coinegraing relaionship in he money demand and he price equaion. Before analyzing he coinegraing relaionship, es of uni roo is conduced in secion 3 and we con rm ha mos variables have uni roo. Secion 3 also provides he resuls of c es of coinegraion proposed by Johansen. We nd ha coinegraing relaionship in real money demand, domesic oupu and domesic ineres rae in several Asian counries and calculae measure of exchange marke pressure consisen wih Weymark's model by applying dynamic OLS esimaion mehod. In secion 4, we compare hese hree indices of exchange marke pressure. A brief summary and concluding remark is found in secion 5. 2 Measuring exchange marke pressure 2.1 Giron and Roper model Giron and Roper(1977) consider simple wo counry model. They specify exchange marke pressure as a sum of he change in he nominal exchange rae and change in he foreign exchange reserves. Moneary equilibrium condiion can be wrien as H = R + D = P Y fi exp( ff i ) (1) where H denoes he supply of base money issued by moneary auhoriy in period, R denoes he base money creaed agains he ne purchase of foreign asses, D denoes he base money creaed by domesic credi expansion, P denoes he price level, Y denoes he real income and i denoes he domesic ineres rae. Transforming his equaion ino logarihms and differeniaing i wih respec o ime, we have h = r GR + d GR = p + fi y ff i (2) where h denoes he log of base money, r denoes he log of base money creaed by ne foreign asses, d denoes he log of base money creaedby domesic credi expansion, p denoes he log of price level, y denoes he log of real income, r GR = R =H r and d GR = D =H d * 1.Toexamine he moneary ineracion beween counries, we subrac moneary equilibrium condiion (equaion (2)) in foreign counry from moneary equilibrium condiion (equaion (2)) in home counry. We obain r GR r ΛGR = d GR + d ΛGR + fi y fi Λ y Λ + p p Λ ff( i i Λ ) (3) * 1 Noaions are slighly changed from Giron and Roper's original paper in order o compare wih oher exchange marke pressure indices. 3

4 where aserisk ( Λ ) denoes he foreign values. ff and ff Λ are assumed o be equal. We inroduce he furher noaion, e ha denoes he rae of depreciaion of he home currency in erms of he foreign currency, ha denoes differenial inλaion rae adjused for exchange rae changes ( p p Λ e ), and ffi ha denoes change in uncovered ineres rae differenial ( i i Λ ). Equaion (3) can be rewrien as r GR r ΛGR e = d GR + d ΛGR + fi y fi Λ y Λ + ffffi : (4) This equaion shows he volume of disequilibrium in foreign exchange marke ha mus be removed hrough marke inervenion or change in he nominal exchange rae. The lef hand side of equaion (4) is referred o exchange marke pressure. We also assume ha foreign counry is a keycurrency counry and has abiliy o force all adjusmen burdens on home counry ha has made effors o sabilize he exchange rae. Then equaion (4) can be rewrien as EMP GR e r GR = d GR + h Λ + fi y fi Λ y Λ + ffffi : (5) Giron and Roper de ne exchange marke pressure as r GR e. We muliply i by -1 in accordance wih oher EMP indices, so ha he posiive EMP implies he currency is under he pressure of depreciaion or loss of he foreign exchange reserves. 2.2 Speci caion of Kaminsky and Reinhar Kaminsky and Reinhar(1999) specify exchange marke pressure as a weighed average of he rae of change in he exchange rae, e and in he foreign exchange reserves, r KR, wih weigh such ha he wo componens of he index have equal sample volailiies. EMP KR ff e e r KR (6) ff r where ff e is he sandard deviaion of he rae of change in he nominal exchange rae and ff r is he sandard deviaion of he rae of change in he foreign exchange reserves. Since changes in he nominal exchange rae ener wih a posiive weigh, changes in he foreign exchange reserves have negaive weigh. Posiive exchange marke pressure index indicaes ha he currency is under he pressure of depreciaion. 2.3 Weymark's Model Weymark(1995) generalize he measure of exchange marke pressure proposed by Giron and Roper. The model employedbyweymark is one of a small open economy in which he domesic price level is inλuenced by boh he level of foreign counry and he exchange rae. The purchasing power pariy does no necessarily hold. The gross domesic produc, he foreign price level, he foreign ineres rae and he domesic credi are assumed o be exogenous. I is also assumed ha he domesic nancial 4

5 marke is well developed and he domesic asses and he foreign asses are compleely subsiuable. The foreign and he domesic ineres raes are linked hrough an uncovered ineres rae pariy condiion. The model is characerized by four equaions below. m d = p + b 1 y b 2 i + ν (7) p = a 0 + a 1 p Λ + a 2 e (8) i = i Λ + E[e +1j] e (9) m s = m s 1 + dw + r W (10) where m denoes he log of he money sock in period (superscrip s denoes supply, and d denoes demand), p denoes he log of he domesic price level, y denoes he log of he gross domesic oupu, i denoes he domesic ineres rae, ν denoes a sochasic shock agains he money sock, e denoes he log of he nominal exchange rae measured by home currency. An aserisk( Λ ) denoes foreign value and E[e +1j] denoes expecaion of exchange rae in period +1 which agen expec condiional on he informaion obained in period. Supply of he money is deermined by he foreign exchange reserves and he domesic credi creaed by moneary auhoriy. d W = [μ D μ 1 D 1 ]=M 1 (11) where μ denoes money muliplier in period, D denoes domesic credi supplied by moneary auhoriy in period, and M 1 denoes money sock in period -1. Similarly, he de niion of r W is given by he equaion below. r W = [μ R μ 1 R 1 ]=M 1 (12) where μ denoes he money muliplier in period, R denoes he ne foreign exchange reserves of moneary auhoriy in period, and M 1 denoes he money sock in period -1. Subsiue equaion (8) and equaion (9) o equaion (7), we obain m d = a 0 + a 1 p Λ + (a 2 + b 2 )e + b 1 y b 2 i Λ b 2 E[e +1j] + ν : (13) Under he assumpion ha he money marke clears coninuously, m d = m s = m for all : (14) Subsiuing equaion (10) and equaion (13) ino equaion (14), we obain equilibrium condiion of money marke 5

6 a 0 + a 1 p Λ 1 + (a 2 + b 2 )e + b 1 y b 2 i Λ b 2 E[e +1j] + v = m 1 + d W + r W (15) Taking he rs order differences of he lef hand side of equaion (15) and divide by 1=(a 2 + b 2 )o normalize, we have EMP W e 1 a 2 + b 2 r W = 1 a 2 + b 2 ( d W a 1 p Λ b 1 y + b 2 i Λ + b 2 E[e +1j] + u ): (16) where u = v =(a 2 + b 2 ). Equaion (16) shows he magniude of disequilibrium in he money marke ha mus be removed eiher by he change in he nominal exchange rae or he change in he money sock creaed by foreign exchange reserves. Exogenous variables in his equaion are change in he foreign price level ( p Λ ), change in he domesic produc ( y ), change in he foreign ineres rae ( i Λ ), change in he domesic credi ( dw ) and sochasic disurbance in money demand (u ). Following he discussion of Giron and Roper, we de ne a lef hand side of equaion (16) as an exchange marke pressure. There are four independen equaions and ve endogenous variables; m, i, p, e and r W. Therefore we canno solve he model explicily. However, weareineresed in exchange marke pressure as an aggregae volume of disequilibrium in he money marke which expressed by exogenous variables. The composiion of he change in he nominal exchange rae and he change in he money sock creaed by he foreign exchange reserve scaled by inheried money sock in exchange marke pressure depends on he policy auhoriy's reacion funcion which saes he relaionship beween e and r W. In oher words, if we can specify moneary auhoriy's reacion funcion, we can solve he model explicily and measure he volume of marke inervenion necessary o achieve nominal exchange rae sabiliy or he change in he nominal exchange rae when no marke inervenion is conduced. In his paper, we do no specify he reacion funcion of moneary auhoriy and esimae exchange marke pressure as a disequilibrium in money marke. 3 Empirical invesigaion In his secion, we esimae exchange marke pressure for crisis affeced Asian counries; Indonesia, Japan, Korea, Malaysia, Philippines, Singapore and Thailand. The daa are monhly series from January 1990 o July 2004 excep Indonesia and Korea. Sample period for Indonesia is from January 1991 o June 2004 and sample period for Korea is from January 1990 o June The principal daa source is IFS and deailed daa descripion is discussed in Appendix. To esimae exchange marke pressure speci ed by Weymark, we have o derive he coefficien of he domesic ineres rae in he money demand equaion (b 2 in equaion (7)) and coefficien of he exchange rae in he price equaion (a 2 in equaion (8)) properly. Weymark esimaed he rs differenced equaion by 2SLS esimaion mehod. This paper esimaes he equaion using he coinegraing mehod. Before analyzing coinegraing relaionship in money demand and he price equaion, we rs es he exisence of uni roo in 6

7 ime series we analyze and wha he order of inegraion for each variable is. 3.1 Uni roo es We apply he uni-roo es developed by Dickey and Fuller. The resuls of Augmened Dickey Fuller ess *2 are presened in Table 1 and Table 2. The null hypohesis is ha here exiss uni roo, herefore, he rejecion of null hypohesis indicae ha he variable is saionary. An aserisk (*) denoes ha he ADF saisic is signi can a he ve percen level. The es is conduced wih 0 o 11 lags and number of opimal augmening lags is deermined o minimize AIC. When we es he exisence of uni roo in level of ineres rae, rend componen is excluded. The resuls presened in Table 1 and Table 2 sugges ha virually all variables excep p have uni roo. To avoid modeling I(2) variable, we focus on (m p) and nd ha (m p) had desired I(1) propery. Inhenex subsecion, we analyze coinegraing relaionship in he money demand equaion; (m p), y and i, and he price equaion; p, p Λ,ande and ry o obain consisen coinegraing vecor. * 2 ADF ess is carried ou in he conex of he model: y = μ + fly 1 + P p 1 j=1 ffi j y j + ffl 7

8 (a) Indonesia (num. of obs.=162) variable e e m m y y speci caion cons cons cons cons cons cons rend rend rend num.of lags adf- value :86 Λ :89 Λ :39 Λ variable p p i i (m-p) (m-p) speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :91 Λ 2:83 6:95 Λ :67 Λ (b) Japan (num. of obs.=175) variable e e m m y y speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :28 Λ :91 Λ :33 Λ variable p p i i (m-p) (m-p) speci caion cons cons cons cons cons cons rend rend num.of lags adf- value 3: :29 Λ :14 Λ (c)korea (num. of obs.=174) variable e e m m y y speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :14 Λ :09 Λ variable p p i i (m-p) (m-p) speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :33 Λ :07 Λ :10 Λ (d)malaysia (num. of obs.=175) variable e e m m y y speci caion cons cons cons cons cons cons rend rend rend num.of lags adf- value :19 Λ :34 Λ variable p p i i (m-p) (m-p) speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :23 Λ :83 Λ Table. 1 Augmened Dickey-Fuller saisics 8

9 (e)philippines (num. of obs.=175) variable e e m m y y speci caion cons cons cons cons cons cons rend rend rend num.of lags adf- value :96 Λ :51 Λ variable p p i i (m-p) (m-p) speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :89 Λ :50 Λ :44 Λ (f)singapore (num. of obs.=175) variable e e m m y y speci caion cons cons cons cons cons cons rend rend rend num.of lags adf- value :27 Λ :88 Λ :23 Λ variable p p i i (m-p) (m-p) speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :61 Λ :82 Λ (g)thailand (num. of obs.=175) variable e e m m y y speci caion cons cons cons cons cons cons rend rend rend num.of lags adf- value :02 Λ :33 Λ variable p p i i (m-p) (m-p) speci caion cons cons cons cons cons cons rend rend num.of lags adf- value :10 Λ :36 Λ (h)unied Saes (num. of obs.=175) variable p p i i speci caion cons cons cons cons num.of lags adf- value :22 Λ :77 Λ Table. 2 Augmened Dickey-Fuller saisics: coninued 9

10 3.2 Coinegraion analysis In his secion, we conduc Johansen's race es *3 o deermine he number of coinegraing vecors for money demand and he price equaion. In words, Johansen's race es esimaes he coinegraing vecor in error correcion represenaion of I(1) sysem direcly by applying maximum likelihood esimae o residuals obained by canonical regressions. The null hypohesis is ha here are r or fewer coingraing vecors agains he alernaive hypohesis ha here are more han r coinegraing vecors. Johansen's race es saisics for money demand of Asian counries are summarized in able 3. An aserisk (*) denoes ha he race saisic is signi can a he ve percen level. The es is conduced wih 0 o 11 lags and number of opimal augmening lags is deermined o minimize AIC. The lag lengh adoped is 5 for Indonesia, is 6 for Japan, is 6 for Korea, is 1 for Malaysia, (a)indonesia Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 26:14 Λ (b)japan Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 43:56 Λ (c)korea Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 32:77 Λ 11: (d)malaysia Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 43:04 Λ :63 (e)philippines Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 29:09 Λ (f)singapore Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 34:51 Λ (g)thailand Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 38:18 Λ 11: Table. 3 Johansen's race es saisics for money demand * 3 For he deails of he es procedure, see Maddala and Kim(1998) or Kawai and Ohara(1997) 10

11 Indonesia Japan Korea Malaysia Philippines Singapore Thailand m-p y i Table. 4 Esimaed coinegraing vecors for money demand is 7 for Philippines, is 2 for Singapore, and is 6 for Thailand. Resuls repored in able 3 indicae ha null hypohesis of no coinegraion is rejeced and null hypohesis ha indicaes he exisence of one coinegraing vecor is acceped in all counries a he sandard level of signi cance. Therefore we conclude ha here is one coinegraing relaionship in money demand for Asian counries. These resuls are consisen wih convenional sudies ha con rm he exisence of coinegraing relaionship in money demand. Esimaed coinegraing vecors are presened in able 4. We normalize he vecor using real money sock coefficien. The convenional sudies of money demand show hacoefficien on he domesic produc is posiive and coefficien on he domesic ineres rae is negaive. However, he coefficiens on he ineres rae of Indonesia, Singapore and Thailand are negaive and he coefficien on he ineres rae of Singapore is exremely large. The coefficien on he ineres rae of Korea is posiive, bu oo large o consider sable long-run relaionship beween money demand and ineres rae. The resuls of Johansen's race es on he price equaion for Asian counries are also repored in able 5. Similarly, he es is conduced wih 0 o 10 lags and number of opimal augmening lags is deermined o minimize AIC and an aserisk (*) denoes ha he saisic is signi can a ve percen level. The lags adoped is 6 for Indonesia, is 5 for Japan, is 2 for Korea, is 1 for Malaysia, is 3 for Philippines, is 2 for Singapore, and is 1 for Thailand. Unlike he case of money demand, we canno obain clear evidence ha implies exisence of one coinegraing relaionship. Null hypohesis of no coinegraing relaionship is rejeced and null hypohesis of one coinegraing relaionship is acceped for all counries a he ve percen signi cance level. However, null hypohesis of wo coinegraing relaionship is rejeced for Korea, Philippines and Singapore. This resul may sugges ha no coinegraing relaionship in hese counries. The coinegraing vecors correspond o maximum eigenvalue is repored in able 6. The coefficien on he foreign price level and he nominal exchange rae are negaive in all counires excep Indonesia. This is consisen wih convenional heory of he price equaion. However, he coefficiens are oo large o recognize long-run relaionship. 11

12 a Indonesia Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 40:54 Λ b Japanese Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 45:27 Λ c Korea Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 42:63 Λ :61 Λ d Malaysia Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 37:22 Λ e Philippine Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 58:64 Λ :44 Λ f Singapore Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 35:45 Λ :06 Λ g Thailand Eigenvalue Null hypohesis r 0 r 1 r 2 Trace saisic 35:78 Λ Table. 5 Johansen's race es saisics for he price equaion Indonesia Japan Korea Malaysia Philippine Singapore Thailand p p Λ e Table. 6 Esimaed coinegraing vecors for he price equaion 3.3 Esimaion by Dynamic OLS In he previous subsecion, we examined wheher an I(1) sysem is coinegraed. We can nd sable coinegraing relaionships in money demand of several Asian counries. In his subsecion, we esimae riangular represenaion of coinegraing sysems. We canno nd valid evidence of coinegraion for he price equaions, consisen esimae can be obained by applying DOLS esimae mehod because almos all he variables are assumed o be I(1). The lengh of he lead and lag are arbirarily se 12

13 o 3, so he augmening coinegraing regression for he money demand equaion is represened as m p = μ 1 + fl y y + fl i i + fi y0 y + fi y; 1 y +1 + fi y; 2 y +2 + fi y; 3 y +3 + fi y1 y +1 + fi y2 y 2 + fi y3 y 3 +fi i0 i + fi i; 1 i +1 + fi i; 2 i +2 + fi i; 3 i +3 + fi i1 i +1 + fi i2 i 2 + fi i3 i 3 and he price equaion is similarly represened as p = μ 2 +fl p Λ p Λ +fl ee +fi p Λ 0 p Λ +fi p Λ ; 1 p Λ +1 +fi p Λ ; 2 p Λ +2 +fi p Λ ; 3 p Λ +3 +fi p Λ 1 p Λ +1 +fi p Λ 2 p Λ 2 +fi p Λ 3 p Λ 3 +fi e0 e + fi e; 1 e +1 + fi e; 2 e +2 + fi e; 3 e +3 + fi e1 e +1 + fi e2 e 2 + fi e3 e 3 : The esimaed coefficiens on he domesic oupu and he domesic ineres rae in money demand are repored in able 7. The esimaed coefficiens on he foreign price level and he nominal exchange rae in money demand are repored in able 8. Using hese coefficiens obained by DOLS esimae, we calculae he exchange marke pressure index proposed by Weymark. The movemens of calculaed exchange marke pressure for each counries are presened in gure 1 o gure 7 in appendix. Indonesia Japan Korea Malaysia fl y fl i fl y fl i fl y fl i fl y fl i SOLS esimae DOLS esimae Philippines Singapore Thailand fl y fl i fl y fl i fl y fl i SOLS esimae DOLS esimae Table. 7 Esimaed coefficiens for he price equaion by SOLS and DOLS Indonesia Japan Korea Malaysia fl p Λ fl e fl p Λ fl e fl p Λ fl e fl p Λ fl e SOLS esimae DOLS esimae Philippines Singapore Thailand fl p Λ fl e fl p Λ fl e fl p Λ fl e SOLS esimae DOLS esimae Table. 8 Esimaed coefficiens for money demand by SOLS and DOLS 4 Comparison of exchange marke pressure indices Kaminsky and Reinhar (1999) de ne a currency crisis when he index of exchange marke pressure is hree sandard deviaions above he mean. In his secion, we idenify a currency crisis using each marke pressure index and compare he resuls. The change in nominal exchange rae and he change 13

14 Indonesia Japan Korea Malaysia Philippines Singapore Thailand ff e ff r ff e =ff r Num. of Obs Table. 9 Sandard deviaion of he rae of change in he nominal exchange rae and reserves in foreign exchange reserves are calculaed in able 9. Using he raio of sample volailiy of he change in he nominal exchange raes and he foreign exchange reserves, we calculae he exchange marke pressure speci ed by Kaminsky and Reinhar. The periods ha classi ed as a currency crisisbyeach exchange marker pressure indices are repored in able 10. The periods in he column EMP GR, EMP KR and EMP W are periods ideni ed as a currency crisis by Giron and Roper's mehod, by Kaminsky and Reinhar's mehod, by Weymark's mehod respecively. We can con rm ha enormous pressures of depreciaion were pu on Indonesia, Korea, Malaysia, Philippines and Thailand hrough 1997 o 1998 by all indices. In he case of Indonesia, all indices idenify January 1998 as a currency crisis period and indices speci ed by Kaminsky and Reinhar and Weymark idenify May 1998 as a currency crisis period. Considering he fac ha he Indonesia rupiah depreciae dramaically hrough December 1997 o January 1998 and he Presiden Suharo announced his resignaion in May 1998, i is quie reasonable EMP GR EMP KR EMP W Indonesia Mar Dec Jan Nov Jan Jun Jan Jun Japan Mar Mar Korea Nov Nov Nov Dec Dec Dec Malaysia Jan Jul Dec Mar Jan Jan Philippines Jul Jan May Dec Dec Jul Sep Singapore May Dec Feb May Thailand Jul Jul Jul Aug Aug Aug Nov Nov Dec Dec Jan Jan Jan Table. 10 Periods classi ed as a crisis by each EMP measure 14

15 o idenify January and May 1998 as crisis periods for Indonesia. The index speci ed by Kaminsky and Reinhar ideni es March 1990 and March 1991 as crisis periods for Japan. However, i is no appropriae o regard hese periods as crisis periods because he marke inervenion conduced by he moneary auhoriy is serilized and he change in he foreign exchange reserves does no inλuenced base money. In Japanese case, we can infer ha exchange marke pressure speci ed by Giron and Roper and Weymark are more appropriae measure han by Kaminsky and Reinhar. In Korean case, all indices correspondingly indicae ha November and December 1997 are he currency crisis periods and in he case of Thailand, all indices similarly indicae ha he periods from July 1997 o January 1998 are he currency crisis periods. These ideni caions of he crisis periods are accepable because Korean won experienced severe depreciaion in December 1997 and Thailand exhaused heir foreign exchange reserves and abandon he de faco xed exchange rae regime in July Therefore, we can infer ha exchange marke pressure speci ed by Weymark can idenify a currency crisis period properly han indices speci ed by Giron and Roper and Kaminsky and Reinhar. 5 Consluding remarks Wehave analyzed exchange marke pressure for Asian counries along wih he model proposed by Weymark using coinegraing esimaion. The idea behind he measure of exchange marke pressure is ha he imbalance beween he demand and supply of he domesic currency in he inernaional foreign exchange marke is removed by a change in he exchange rae or in he supply of money. Measuring exchange marke pressure considering he effec ha he increase in foreign exchange reserve inλuences he volume of domesic money sock is quie reasonable. We have conduced Johansen's race ess of coinegraion in he money demand and he price equaion and found ha here is coinegraing relaionship in money demand in several Asian counries. However, he some of hese relaionships canno consider o be sable because he coefficien on he ineres rare is no consisen wih convenional heory of money demand. We calculae he exchange marke pressure using coefficiens obained in dynamic OLS esimaion. By comparing calculaed exchange marke pressure index wih oher exchange marke pressure indices de ned by Giron and Roper and Kaminsky and Reinhar, we have showed ha we can idenify a currency crisisbyweymark's mehod more properly han by Giron and Roper's mehod and Kaminsky and Reinhar's mehod. 15

16 Appendix A Daa descripion The principal daa source is Inernaional Financial Saisics CD-ROM (IFS). When daa is missing from hese sources, saisic from he homepage of cenral-bank is uilized. Price level and ineres rae of Unied Saes are used as foreign price level and foreign ineres rae. Domesic oupu is subsiued by indusrial producion or manufacuring producion. The change in money sock creaed by he change in foreign exchange reserves (r W ) is calculaed as M =M 1 R =(R + D ) R 1 =(R 1 + D 1 )by subsiuing μ = M =(R + D ). In words, he change in money sock creaed by he change in foreign exchange reserves is composed of he rae of moneary expansion and he proporion of reserve money o base money. Because of he insufficiency of he daa, he proporion of reserve money o base money is subsiued by he proporion of ne foreign asses o oal asses of moneary auhoriy for Malaysia, Singapore and Thailand. Daa used in individual counry is described as follows. 1. Indonesia ffl m: M2 (IFS line 35L) ffl p: Consumer prices (IFS line 64) ffl y: obained from Quarerly GDP (IFS line 99b) by spline daa ing ffl e: End-of-period marke rae of naional currency per U.S. dollar (IFS line ae) ffl i: Call money rae (IFS line 60B) ffl r: Toal reserves minus gold (IFS line 1L) ffl R=(R + D): Moneary auhoriy's ne foreign asse (IFS line11-ifs line16c) divided by sum of moneary auhoriy's ne foreign asse, ne governmen claims (IFS line 12a-IFS line 16d) and claims on deposi money bank (IFS line 12e) 2. Japan ffl m: M2 (IFS line 35L) ffl p: Consumer prices (IFS line 64) ffl y: Indusrial producion (Seasonally adjused) (IFS line 66C) ffl e: End-of-period marke rae of naional currency per U.S. dollar (IFS line ae) ffl i: Money marke rae (IFS line 60B) ffl r: Toal reserves minus gold (IFS line 1L) ffl R=(R + D): Moneary auhoriy's ne foreign asse (IFS line11-ifs line16c) divided by Reserve money(ifs line 14) 3. Korea ffl m: M2 (IFS line 35L) ffl p: Consumer prices (IFS line 64) ffl y: Indusrial producion (Seasonally adjused) (IFS line 66C) ffl e: End-of-period marke rae of naional currency per U.S. dollar (IFS line ae) 16

17 ffl i: Money marke rae (IFS line 60B) ffl r: Toal reserves minus gold (IFS line 1L) ffl R=(R + D): Moneary auhoriy's ne foreign asse (IFS line11-ifs line16c) divided by sum of moneary auhoriy's ne foreign asse, ne governmen claims (IFS line 12a-IFS line 16d) and claims on deposi money bank (IFS line 12e) 4. Malaysia ffl m: M2 (IFS line 35L) ffl p: Consumer prices (IFS line 64) ffl y: Indusrial producion (IFS line 66) (Seasonally adjused by X-12a) ffl e: End-of-period marke rae of naional currency per U.S. dollar (IFS line ae) ffl i: Inerbank overnigh money rae (IFS line 60B) ffl r: Toal reserves minus gold (IFS line 1L) ffl R=(R + D): Moneary auhoriy's ne foreign asse (IFS line11-ifs line16c) minus capial accoun (IFS line 17a) divided by sum of moneary auhoriy's ne foreign asse, ne governmen claims (IFS line 12a-IFS line 16d) and claims on deposi money bank(ifs line 12e), nonbank nancial insiuions (IFS line 12g), and privae secor(ifs line 12b) 5. Philippines ffl m: M2 (IFS line 35L) ffl p: Consumer prices (IFS line 64) ffl y: Manufacuring producion (IFS line 66) (Seasonally adjused by X-12a) ffl e: End-of-period marke rae of naional currency per U.S. dollar (IFS line ae) ffl i: Money marke rae (IFS line 60B) ffl r: Toal reserves minus gold (IFS line 1L) ffl R=(R + D): Moneary auhoriy's ne foreign asse (IFS line11-ifs line16c) minus capial accoun (IFS line 17a) divided by sum of moneary auhoriy's ne foreign asse, ne governmen claims (IFS line 12a-IFS line 16d) and claims on deposi money bank(ifs line 12e), nonbank nancial insiuions (IFS line 12g), and privae secor (IFS line 12b) 6. Singapore ffl m: M2 (IFS line 35L) ffl p: Consumer prices (IFS line 64) ffl y: Manufacuring producion (IFS line 66) (Seasonally adjused by X-12a) ffl e: End-of-period marke rae of naional currency per U.S. dollar (IFS line ae) ffl i: Inerbank rae (IFS line 60B) ffl r: Toal reserves minus gold (IFS line 1L) ffl R=(R + D): Moneary auhoriy's ne foreign asse (IFS line11-ifs line16c) divided by oal asses of he moneary auhoriy (IFS line11+ifs line12) 7. Thailand 17

18

19

20

21