PertanikaJ. Soc. Sci. & Hum. 4(2): 155-164 (1996)ISSN: 0128-7702

© Penerbit Universiti Pertanian Malaysia

Currency Substitution via Expected Exchange Rate and Domestic MoneyDemand: An Empirical Analysis for Japan

T. CHOTIGEAT and SANG H. LEE!

Department of Economics and FinanceNicholls State University

P.O. Box 2015Thibodaux LA 70310, Louisiana, USA

lKorea Security InstituteKorea

Keywords: currency substitution, expected exchange rate, domestic money demand, Japanese

currencyABSTRAK

Kajian ini menguji hipotesis kebolehganti matawang secara empirikal di atas permintaan matawang Jepundengan menggunakan data bulanan dariJanuari 1977 ke Disember 1989. Di bawah sistem kadar pertukaranboleh ubah, ejen ekonomi yang rasional akan membentuk portfolio matawang untuk tujuan permintaanurusniaga dan permintaan spekulasi. Penganggaran tak linear berbilang pembolehubah kebolehjadianmaksimum Jepun dan persamaan kadar pertukaran dijangka: ada kemungkinan terdapat kebolehgantianmatawang. Penemuan kajian ini mencadangkan bahawa (a) kadar pertukaran dijangka bagi matawangJepun (Yen) dengan matawang Dollar Kanada menunjukkan kebolehgantian matawang berdasarkanpermintaan spekulasi dan (b) kadar pertukaran dijangka matawang Jepun (Yen) dan Dollar Amerikamenunjukkan kebolehgantian berdasarkan permintaan urusniaga.

ABSTRACT

In this paper the currency substitution hypothesis is tested empirically on the Japanese money demand, usingmonthly data from January 1977 to December 1989. Under a flexible exchange rate a rational economicagent forms currency portfolios for both transaction and speculative reasons for demanding money. Thenonlinear multivariate maximum likelihood estimation was used to estimate jointly the Japanese moneydemand and the expected exchange rate equation; possible existence of current substitution is found. Theempirical results indicate that (a) the expected exchange rate for the Japanese yen/Canadian dollar showscurrency substitution due to speculative demand, and (b) the expected exchange rate of the yen/US dollarshows currency substitution due to transaction demand.

INTRODUCTION

For more than a decade, the growingintegration of world capital markets hasbrought changes in the environment formonetary policy. The efficacy of monetarypolicy under a flexible exchange rate has beenchallenged by proponents of the currencysubstitution hypothesis [CSHJ (see literaturereview by Bana and Handa 1987). Thecontroversy evolves around the belief that ifcurrencies are substitutable on the demandside, the expectation of future behaviour ofexchange rates under uncertainty may makedomestic money demand unstable and reduce

the autonomy of monetary policy under a fle­xible exchange rate system (Miles 1978a, 1978b;McKinnon 1982; Chen and Tsaur 1983).

One of the main explanations for theCSH is that money demand is influencedthrough a combination of distinctive channelsof currency substitution-transaction demandand speculative demand (King et at. 1980).Specifically, they suggest that with increasingintegration of the goods and services marketaround the world, global traders have moreopportunity to choose from a broader varietyof substitutable goods and services with manydenominations of currencies, i.e., currency

T. Chotigeat and Sang H. Lee

substitution in transaction demand [CSTD].Foreign money demand is thus directlyrelated to foreign goods and services de­mands which, in turn, depend on theirrelative price expectations.

Furthermore, as capital markets aroundthe world have become increasingly inte­grated, risk-averse investors have had moreopportunity to diversify their wealth amongall available currencies, i.e. through currencysubstitution in speculative demand [CSSD].Basically, traders attempting to optimize theircurrency portfolio balances will demand moreof a foreign currency when its value isexpected to appreciate. In particular, acurrency that has a low correlation withtheir holdings is preferable, given the un­certainty associated with the set of exchange

. 1rate expectatlOns.

However, a usual analysis can show thatthe optimizing shifts in demand for currencystocks through these two differen t channelstend to offset each other with a given changein expected exchange rates. Therefore, anattempt to prove empirically the existence ofthese two demand channels, including eachdemand channel's direction and magnitude,is needed. The result would indirectly answerthe question of the efficacy of domesticmonetary policy in the new environment ofworld trade and capi tal liberalization.

In this paper, the Japanese moneydemand function is empirically estimatedand analysed for the possible existence ofsubstitution effects between the yen and USand Canadian dollars. Japan is selected forthis study because it is an open economy withrelatively large trading volume in worldmarkets and a currency widely held amongtraders. The Canadian dollar is included inthe analysis because it has a low correlationwith the yen in their exchange rate move­ments; thus it could provide a channel ofCSSD when traders prefer Canadian dollarsto diversify their currency portfolios.

THEORY AND EVIDENCE

Previous Studies

Standard monetary theory which assumesthat residents of a single country hold only

the national money has been criticized byMiles (1978a), noting that when economicagents hold both domestic and foreigncurrencies, the composition of these cashbalance portfolios will vary with the relevantopportuni ty cost of holding real balances ofthe various types of currencies. Therefore, achange of monetary policy in a foreigncountry will change the relative costs ofholding currencies and thus induce offsettinginflows or outflows of money. An excesssupply of foreign currency flows to adomestic country by the distribution effectand changes in price levels that are notconsistent with the traditional flexible ex­change rate model. McKinnon (1982) arguesthat high substitutability among industrialcountries' currencies destabilizes the demandfor individual national currencies so that onecannot make much sense out of year-to-yearchanges in purely national monetary aggre­gates in explaining cycles in purely nationalrates of inflation. Therefore, the growth in theworld supply is a better predictor of Amer­ican inflation than the US money growthrate. Lapan and Enders (1983) conclude thatin the absence of government control, if theprivate sector views the currencies as perfectsubstitutes for each other, then the exchangerate must be invariant over time andindeterminate. Calvo and Rodriguez (1977)incorporate currency substitution in theirexchange rate determination model andexplain exchange rate overshooting. InDaniel (1981), currency substitution is achannel for in terna tional transmission ofmonetary disturbances under flexible ex­change rates in both the long and shortterm. In the short term, currency substitutioncan be responsible for overshooting theexchange rate and affects the national levelof real wealth and consumption because ofthe unexpected change in money supply.

On the other hand, Thomas (1985)explains that the monetary service model ismore likely than the portfolio balance modelto explain the interactions of domestic andforeign money demand. This implies that thedemand for alternative monies is from thereal rather than the financial channel, andthe transaction services that money produces

156 Pertanika J. Soc. Sci. & Hum. Vol. 4 No.2, 1996

Currency Substitution via Expected Exchange Rate and Domestic Money

where

where

fit = fit + E[Sr] y' + et (3)

(1 )

(2)E[Sr] = Zt-I . (X'

where E[.] is an expectation operator.

The hypothesis for CSTD and CSSDsuggests that the domestic money demand is afunction not only of domestic factors likeincome and the interest rate, but also ofinternational influences through future ex­pectations of exchange rates. Therefore, themoney demand equation can be specified as

An optimal forecast for St then simplyinvolves taking expectations of equation (1)conditional on information available at t-l as

Zt-1 a vector of regressors available attime t-l to forecast St (variableslike interest rate, money supply,and trade balance, etc. known att-l, t-2, t-3, ... , ton),

tJ.' a vector of coefficien ts to beestimated,

U t an error term which is assumed tobe uncorrelated with any informa­tion available at t-l (which in­cludes any variable in Zt-1 or tit-lfor all i > 1).

mt the growth rate of real moneydemand,

mt an equilibrium money demandbased only on the domestic fac-

Empirical Model

Under the rational expectation assumption,an economic agent could optimally expect afuture exchange rate one period ahead of therealized exchange rate, using all availablerelevant information (Mishkin 1983). Theforecasting equation that can be used togenerate the anticipation of the growth rateof a basket of current spot exchange ratesexpressed in domestic currency per unit offoreign currency, So is.

are the main reason for the money demandrather than its expected real return or riskwhen complete bond markets exist.

Both Hamberger (1977) and Britain(1981) support the currency substitutionhypothesis by employing an alternativeempirical procedure which examines thecross-correlations of the residuals from regres­sing a velocity variable on the time variable.Studies on this issue for developing countriesinclude Ramirez-Rojas (1985) for Argentina,Mexico, and Uruguay, Fasano-Filho (1986)for Argentina, Marquez (1987) for Venezue­la, and Arize (1991) for South Korea. Alltheir findings support the evidence of CSH.To a varying degree, it is an importantfeature of these developing countries.

Bordo and Choudhri (1982) note that thedifferential cost of holding foreign instead ofdomestic money is the expected change in theexchange rate, but it does empirically exist inthe Canadian money demand function.Darby et al. (1983) maintain the view thatforeign bonds are generally more importantsubstitutes for domestic money than foreignmoney. They include a foreign interest rateplus expected depreciation term in estimatingthe domestic money demand function andconclude that foreign-asset substitution fordomestic money is not an important channelof economic transmission among countries.Batten and Hafer (1986) raise some empiricalissues about choosing economic variables toexplain the CSH and propose a testingframework that avoids the problems asso­ciated with previous works. By employing thedistributed lag model and the GNP deflatoras the dependent variable, they found thatadding the rest of the world money growth tothe explanatory power of domestic moneygrowth alone did not offer more explanationto US inflation.

The evidence indicates that the role ofcurrency substitution in domestic moneydemand is mixed at best in both empiricaland theoretical studies. Most empiricalstudies, however, do not explicitly differenti­ate the two different channel-of-substitutioneffects and have contributed much to thedebate about the existence and the signifi­cance of the effects.

PertanikaJ. Soc. Sci. & Hum. Vol. 4 No.2, 1996 157

T. Chotigeat and Sang H. Lee

LR = 2(Lu-Lc) which is distributed aschi-square with q degrees of freedom,

tors, i.e. the interest rate andincome variables (however, in thisstudy the interest rate variable wasnot selected as an independentvariable to estimate Japanesemoney demand function),

y' vector of coefficients for forecasted(expected) exchange rates as in­dependent variables, i.e. the ex­pected exchange rates for yenjC$and yenjUS$ from previous twoequations,

et a serially uncorrelated disturb­ance-term.

The CSTD hypothesis proposes that theparameter y' is positive and that the expecteddepreciation of the exchange rate will inducethe substitution of foreign goods and servicesfor domestic ones in the world market andforce both domestic and foreign traders todemand more of domestic than foreigncurrency. On the other hand, the CSSDhypothesis suggests a negative parameter; theexpected depreciation of exchange ratesreduces domestic money demand for theutility-maximizing currency traders. Thesesubstitution effects will create a problem ofmonetary dependency if the coefficients aresignificantly different from zero, although thesubstitution effects tend to offset each other.

Since equations(l) and (3) share infor­mation by using common variables in thesystem, it is likely that the disturbances fromthe two are correlated. The technique ofjointnonlinear estimation of (1) and (3) will beused for the efficient estimation ofparameters.2

To test the hypotheses of CSTD andCSSD, a constrained system of equationsrestricting the parameters for the expectedexchange rates to zero is estimated. Then, alikelihood ratio test is performed to comparethe constrained and unconstrained systems.The likelihood ratio is computed as

whereLu = log likelihood value of uncon­

strained system,

Lc log likelihood value of constrainedsystem,

q number of coefficients being re­stricted.

Furthermore, the Aaike InformationCriterion (AI C) (Amemiya 1980) is used toselect the lag length of relevant variable in theexchange rate and money demand equa­tions. 3

Data

From Citibase, the monthly data for themoney supply, the short-term interest rate,and the industrial production and consumerprice indices are collected for Japan, USAand Canada from January 1977 to December1989 (total 156 months). In addition, themonthly data for the exchange rates andimport and export volume vis-a-vis USA arecollected for Japan and Canada. The Cana­dian exchange rates are then converted into aseries expressed as the number of Japaneseyen per Canadian dollar. The real industrialproduction is used for the real incomemeasure as a proxy. For the stationarity ofeach of these data, the first difference ofinterest rate and the first difference of growthrate for exchange rate, real income, moneysupply, import, and export are used. The testsof stationarity for these series using regressionsof each of these series against time trend andthe square of time trend show no evidence ofthe rejections against stationarity. Note that,in this study, no attempt is made to testwhether systems of equations are cointegratedsince expected values of exchange rates are usedrather than actual exchange rates in the J apa­nese money demand function (equation 3).

EMPIRICAL RESULTS

Table 1 shows the contemporaneous correla­tion coefficients of five major currencies ontheir growth rates of the spot exchange rateper US dollar. It also shows the averagemerchandise trading volume of these coun­tries with USA. The lowest correlationbetween the Canadian dollar and Japaneseyen suggests that the yen per Canadian dollarexchange rate would be a good candidate to

158 Pertanika J. Soc. Sci. & Hum. Vol. 4- No.2, 1996

Currency Substitution via Expected Exchange Rate and Domestic Money

TABLE ICorrelation matrix of five major currencies*

(January 1977 - December 1989)

£. C$ FR GM ¥

£. 1.00C$ 0.25 1.00FR 0.66 0.23 1.00GM 0.69 0.19 0.94 1.00¥ 0.56 0.09 0.68 0.66 1.00

* Growth rates of spot exchange rate (each currency per US dollar)

£. British poundC$ Canadian dollarFR French francGM: German mark¥ Japanese yen

Average Monthly Volume and Growth Rate of Merchandise Trade with USA **(Million US dollars, January 1977 - December 1989)

K Canada France Germany JapanExports to USA: 1,067 4,691 633 1,393 4,442

(0.9%) (0.8%) (1.0%) (0.9%) (1.1 %)

Imports from USA: 1,011 3,740 576 848 1,987(0.9%) (0.7%) (0.8%) (0.7%) (0.9%)

test the CSSD hypothesis on the Japan moneydemand function. Although it is not shown inthe table, these two countries have lowestcorrelation coefficient of monthly growth rateof trade volume with USA. On the otherhand, the highest growth rate and therelatively large volume of trade betweenUSA and Japan suggest a high degree ofCSTD between yen and the US dollar.

For these reasons, the unconstrained sys­tem of two different exchange rates (¥/C$and ¥/US$) and Japanese money demandequations were estimated with the nonlinearmaximum likelihood estimation technique.However, the final specifications of eachequation were first determined by the AlCtechnique for their appropriate variables andlag lengths and reported in Table 2; theequation is repeatedly estimated with variouslag lengths, and then the appropriate order oflag length is determined by selecting theorder with the lowest AlC value. Some usual

tests are run for robustness of these specifica­tions. Since each equation contains laggeddependent variables, Breusch and Pagan's(1980) Lagrange multiplier test was appliedto analyse the possibility of serial correlationsof a higher order. 4 No evidence of significantserial correlations was found in error terms upto the 6th lag at the 5% significance level. Forthe test of heteroscedasticity, Breusch andPagan's (1979) chi-square test method wasused, and no significant evidence of violationsof homoscedasticity at the 5% significancelevel was found. 5 Based on the homoscedasticerror terms, it was possible to perform Chowtests for the exchange rate equations bydividing the samples into two equally sizedsub-groups because each equation is esti­mated first with the OLS procedure beforesubstituting in the system of equations. Theresulting F-statistics show that the stability ofthe parameters was not rejected at 5%significance level for all equations.

PertanikaJ. Soc. Sci. & Hum. Vol. 4 No.2, 1996 159

T. Chotigeat and Sang H. Lee

TABLE 2Joint estimation for exchange rates and Japanese money demand equations

(maximum likelihood estimation from January 1980 to December 1989)

A. The Unconstrained System

2 6 2

¥/C$: SC,t ~ exO + L ex1 i . SC,t-l + L ex2 j • lj,t-l + L ex3 j • Mj,t-l + L ex4i . XI ,t-l + lit1=1 1=1 1=1 1=1

Initial AIC = -4.336 Final AIC = -4.419

3 2

¥jU$: SU,t 1~1 ~O + L ~li' SC,t-1 + L ~2i' lj,t-l + L ~3i' Xj,t-I + Vt- 1=1 1=1

Initial AlC = -4.175 Final AlC = -4.222

2

Md: mI,t 1:1 yO + Ly1i .mI,t-l + Ly2i .Yt-I + y3 . E[Sc,tl + y4· E[Su,tl + et1=1

Initial AlC = -3.964 Final AIC = -3.967(excluding expected exchange rates)

SK,t: Exchange rate as K currency per one unit ofJapanese yen at time t,IK t: Short-term interest rate for the country K at time t,M~,t: Nominal money supply for the country K at time t,mK,t: Real money demand for country K at time t,XK t: Volume of exports to USA for country K at time t,YK',: Real income measure of country K at time t.(All variables except interest rate are the first differences of growth rate of these variables).

B. Parameter Estimation

¥jC$: CoefficientexOexlex2ex3ex4

R 2 = 0.38

¥jU$: ~O

~I

~2

~3

R 2= 0.37

Md: yOy1y2y3y4

R 2 = 0.41

Sum-0.0005-2.4398-0.0248-0.0428-0.0229

-0.0007--2.7150

-0.0284-0.0035

0.0009-3.79940.2975

-0.90610.7468

t-statistics-0.2257-7.7301-3.6345-0.3535-0.8943

-0.2754-7.9466-3.8437-0.1297

0.3423-5.2829

1.1735-2.46172.0808

F-statistic*

19.21056.69620.62415.4687

16.94925.15823.7949

7.63192.5867

* Joint hypotheses test when all individual coefficients for lags are restricted to zero.

C. The log likelihood test for constrained and unconstrained modelsLog likelihood Ratio = 2 (906.24 - 903.159) = 6.162 (restricting y3 and y4 to zeros for the constrainedmodel), which is larger than chi-square with q degrees of freedom (number of restricted coefficients); wecannot reject the null hypothesis that the restrictions do not apply.

160 PertanikaJ. Soc. Sci. & Hum. Vol. 4 No.2, 1996

Currency Substitution via Expected Exchange Rate and Domestic Money

The estimated parameters of the uncon­strained system are presented in part B ofTable 2; both exchange rate equations havesignificant negative effects ofJapanese interestrate on the exchange rate in the long term. 6

When the Japanese interest rate increases,both the ¥jC$ and ¥jUS$ exchange ratesdecrease, implying the appreciation of the yenvis-a-vis the Canadian and US dollars. Thisresult follows the balance of paymentsapproach to exchange rate determinationwhere a rising domestic interest rate shouldlower the exchange rate either by attractingcapital from abroad or by reducing domesticexpenditure for imports (and thereby im­prove the trade balance). Although the sumcoefficients are not significant, the Japaneseexport variables have significant F-statisticsfor all lags of zero. Investigating individualcoefficients of the export variable showspartial support of the fact that an increasein Japanese exports causes the appreciation ofthe yen in the short term. Based on theseobservations, both CSTD and CSSD can beexpected in the Japanese money demandfunction. ote that interest rate variable wasomitted in the Japanese money demandequation because of a technical reason (theAIC value of the equation is higher if it isincluded); no attempt is made here to explainwhy it does not have any effect on theJapanese money demand.

In fact, the coefficients of the expectedexchange rates in the Japanese moneydemand equation show negative and positivesigns in their parameters for ¥ jC$ and¥jUS$ exchange rates, respectively (seeTable 2, part B). The significant andnegative coefficient for the ¥jC$ rate (y3 =

-0.9061) can be interpreted as the expecteddepreciation of the yen inducing currencytraders to reduce their holdings of Japanesemoney in an attempt to avoid the associatedcapital loss, while the expected appreciationof the yen encourages them to demand moreJapanese money. This result, combined withthe negative effect of the Japanese interestrate or exchange rate, is consistent with theCSSD hypothesis that a change in eitherinterest rate or exchange rate expectationexerts an influence on the composition of

optimal money holdings for a currency traderand provides corresponding capital flowswhich affect the domestic money holdings.

On the other hand, the coefficient for the¥jUS$ rate (y4 = 0.7468) has a significantlypositive sign in the money demand equation.This result combined with the negative effectof the Japanese in terest rate on the exchangerate shows that CSTD dominates CSSD inthe Japanese money demand with respect tothe ¥jUS$ exchange rate, given the fact thatUSA is Japan's largest trading partner, andthe growth rate of trading volume is thehighest among the major US trading part­ners. Therefore, when the yen is expected todepreciate, US importers demand moreJapanese goods, and thus demand more yenbalance to finance the increased trade.

A constrained system. of equations (withthe restrictions of the parameters for theexpected exchange rates at zero in the moneydemand equation) is also estimated (seeTable 2, part C). The log likelihood ratiotest on these constrained and unconstrainedsystems shows the rejection of the nullhypothesis that the constrained system ispresent at the 5% significance level. Thisresult and the offsetting influences of differentforeign exchange rates on money demandmay provide a possible explanation for theconflicting results of previous studies on thesignificance of currency substitution effects inother countries. However, by differentiatingthe two different channels of substitutioneffects using the ¥jC$ and ¥jUS$ exchangerates, the results identify the direction andmagnitude of the currency substitution effectof the two different channels. Withoutincluding these two channels of substitution,the domestic money demand function maymot be stable over time as shown in the loglikelihood test.

CONCLUSION

This paper analyses the Japanese moneydemand function and tests the currencysubstitution hypothesis (CSH) in which theexpected exchange rates for yen per Cana­dian dollar (¥jC$) and yen per US dollar (¥jUS$) affect the Japanese money demand in

PertanikaJ. Soc. Sci. & Hum. Vol. 4No. 2,1996 161

T. Chotigeat and Sang H. Lee

two different ways, depending on the magni­tude of two different channels of substi tu tions:the currency substitution in transactiondemand and the currency substitution inspeculative demand for money. The ration­ale for including these two exchange rates inthe Japanese money demand arises from thefollowing: (I) this approach may reconcilethe previous controversies over the signifi­cance of the CSH between the portfolio andmoney-service approaches of money demand;(2) the transaction demand could be thedominant channel with ¥/US$ because of thehigh trade volume between USA and Japan;and finally (3) speculative demand could bethe dominant channel with ¥/C$ because oftheir low correlation in the exchange ratechanges.

The results of maximum likelihoodestimations on the system of equations showthat the current expectation of ¥/C$ has asignificant negative effect on the Japanesemoney demand, implying currency substitu­tion in the speculative demand for money. Onthe contrary, the currency expectation of¥/US$ shows a significant positive effect onthe Japanese money demand function, im­plying currency substitution in the transac­tion demand for money. The offsetting effectsof these two different substitutions could bethe reason for the previous controversies overthe existence and significance of currencysubstitution effects. By identifying these twodifferen t channels explicitly, the Japanesemoney demand function is found to bedependent on foreign exchange rate expecta­tions, thus providing support for the CSH, atleast in the period analysed here. However,although these expected exchange rates aresignificant in determining equilibrium moneydemand for Japan, the effects tend to offseteach other and the net effect of foreignexchange rates may be negligible.

In comparison with previous studies, testsbased on combined rather than separatechannels of currency of substitution showsome doubt about the existence of the CSH,bu t in this paper findings confirm theexistence of the CSH with the opposite signsfor the coefficients of expected exchange ratevariables in the money demand equations

and give more information that may explainthe previous controversy over the existenceand significance of currency substitution effect.An individual country's monetary authoritymust now recognize that its policy can affectforeign exchange rates differently dependingon what money demand channel is dominant inthat currency. In essence, the rapidly integ­rating capital market seems to diminish anindividual country's ability to conduct themonetary policy, e.g. the disaster in Mexico'semerging capital market in 1994; hence a co­ordinating monetary policies or a newly de­signed exchange rate system may be needed.

Footnotes

I. The portfolio optimization decision here assumesthe constant covariances of returns on all other assetsand considers only the behaviour of the exchangerate returns of one currency vis-a-vis other currencies.2. The estimation procedures can be described stepby step as follows: (I) Estimate exchange rateequations with OLS. (2) Get fitted values (expectedvalues) of Yen/C$ and Yen/US$ exchange rates. (3)Use these values as independent variables in thememory demand equation and estimate the threeequations (two exchange rate and one moneydemand equations) jointly with maximum like­lihood method. (4) Use coefficients estimated fromthe procedure (3). (5) Again re-estimate the threeequations jointly, and get estimated coefficients to getfitted values of exchange rates. (6) Re-estimate thethree equations jointly with newly estimated values ofexchange rates to replace the old independentvariables in the money demand equation. (7)Continue these procedures until the highest possibleR 2 value for the joint estimation is obtained.3. In the forecasting of exchange rates, the Granger(1969) causality concept was used for a specificationof the model. A variable Z is said to Granger-causeanother variable S if S can be predicted better fromthe past values ofS alone. The forecasting equation ofS should include lagged values of S to eliminate anyserial correlation in the residuals. IfZ Granger-causesS, then it should also be used in an optimal forecast ofS. To get the relevant variables Z and their optimumlag length, AICs are calculated up to the 10th lag ofboth domestic and foreign variables including thefirst differences of interest rate, first difference of thereal growth rate of income, the first difference of thegrowth rate of nominal money supply, and the firstdifferences of the growth rate of trade volume figures.The variable and its lag with minimum AIC areselected.

162 PertanikaJ. Soc. Sci. & Hum. Vol. 4 No.2, 1996

Currency Substitution via Expected Exchange Rate and Domestic Money

4. Breusch and Pagan's Lagrange multiplier test forassumption of no serial correlations of higher order inerror terms was performed as follows. First, theresidual were obtained from OLS estimation, then aseries of regressions with current residual as depen­dent variable was run with lags of different order, p,of residual and independent variables of originalequation as regressors. The resulting R 2 of eachregression multiplied by its number of observations,NR2, is distributed as chi-square with p degrees offreedom when the null hypothesis of no pth orderserial correlation is true.5. Breusch and Pagan's (1979) test for heterosce­dasticty can be performed as follows. Assume errorterms are normally distributed. First, get theresiduals (E) and variance(cr2) from original equa­tion. Then make a new series, gt = (E

2Jcr2), and run

another regression with g as a dependent variable. Itcan be shown that half the difference between thetotal sum of squares and the residual sum of squarefrom the second regression is distributed asymptoti­cally as chi-square distribution with k degrees offreedom when the null hypothesis ofhomoscedasticityis true.6. The sum coefficients of Japanese interest ratechanges have significant long-term negative effects onthe exchange rates, and the negative coefficients areconsistent for all lags considered in the eq uations.

REFERENCES

AMEMIYA, T 1980. Selection of regressors. Inter­national Economic Review 21: 331-354.

ARIZE, A.C. 1991. Currency substitution III

Korea. American Economist 35(2): 67-72.

BA A, I.M. and J. HANDA. 1987. Currency sub­stitution: A multicurrency study of Canada.International Economic Journal AutuDln: 71-86.

BATTEN, D.S. and R.W. HAFER. 1986. Theimpact of international factors on U.S.inflation: An empirical test of the currencysubstitution hypothesis. Southern EconomicJournal 53(2): 400-412.

BORDO, M.D. and E.U. CHOUDHRI. 1982.Currency substitution and the demand formoney: Some evidence for Canada. Journal ofMoney, Credit, and Banking 14: 48-57.

BREUSCH, TS. and A.R. PAGAN. 1980. TheLagrange multiplier test and its applicationsof model specifi~ation in economics. Review ofEconomic Studies 47: 239-253.

BRITAIN, B. 1981. International currency sub­stitution and the apparent instability ofvelocity in some Western European econo­mies and the United States. Journal oj Money,Credit, and Banking 13(2): 135-155.

CALVO, G.A. and C.A. RODRIQUEZ. 1977. Amodel of exchange rate determination undercurrency substitution and rational expecta­tions. Journal oj Political Economy 85(3): 617­625.

CHEN, CHAU-NA and TW. TSAUR. 1983.Currency substitution and foreign inflation.Quarterly Journal oj Economics Feb: 177-184.

DANIEL, BETTY C. 1984. Monetary autonomyand exchange rate dynamics under currencysubstitution. Journal of International Economics19: 119-139.

DARBY, M.R., JR. LOTHIAN, A.E. GANDOLFI,A.J. SCHWARTZ and A.C. STOCKMAN.1983. The International Transmission oj Inflation.Chicago: University of Chicago Press.

FASANO-FILHO, U. 1986. Currency substitutionand the demand for money: the Argentinecase 1960-1976. Weltwirtschaftliches Archiv122(2): 327-339.

GRANGER, C.W.J 1969. Investigating causalrelations by econometric models and cross­spectral methods. Econometrica 37: 424-438.

HAMBERGER, M.J. 1977. The demand for moneyin an open economy: German and United King­dom. Journal oj Monetary Economics 3: 25-40.

KING, D.T., B.H. PUTNAM and D.S. WILFORD.1980. A currency portfolio approach toexchange rate determination: Exchange ratestability and the independence of monetarypolicy. In The Monetary Approach to InternationalAdjustment, ed. B.H. Putnam and D.S.Wilford, p. 183-184. ew York: Praeger.

LAPAN, H.E. and W. ENDERS. 1983. Rationalexpectations, endogenous currency substitu­tions, and exchange rate determination.Quarterly Journal oj Economics August: 427­439.

MARQUEZ, ]. 1987. Money demand in openeconomies; a currency substitution model for

PertanikaJ. Soc. Sci. & Hum. Vol. 4 No.2, 1996 163

T. Chotigeat and Sang H. Lee

Venezuela. Journal if International Money andFinance 6: 167-178.

McKINNO ,RONALD 1. 1982. Currency substitu­tion and instability in the world dollarstandard. American Economic Review June:320-333.

MILES, MARC A. 1978a. Currency substitution,flexible exchange rates and monetary inde­pendence. American Economic Review 68: 428­436.

MILES, MARC A. 1978b. Currency substitution:perspective, implication, and empirical evi­dence. In The Monetary Approach to InternationalAdjustment, ed. B.H. Putnam and D.S. Wil­ford, p. 170-198. New York: Praeger.

MISHKIN, FREDERIC S. 1983. A Rational Expecta­tions Approach to Macroeconometrics. Chicago:University of Chicago Press.

RAMIREZ-ROJAS, C.L. 1985. Currency substitu­tion in Argentina, Mexico, and Uruguay.IMF Staff Paper 32: 629-667.

THOMAS, LEE R. 1985. Portfolio theory andcurrency substitution. Journal of Money, Credit,and Banking 17(3): 347-357.

(Received 13 March 1995)

164 Pertanika J. Soc. Sci. & Hum. Vol. 4- No.2, 1996