broad money demand and financial liberalization in ... papers/jssh vol... · pembetulan ralat telah...

Pertanika J. Soc. Sci. & Hum. 9(2): 131 - 141 (2001) ISSN: 0128-7702© Universiti Putra Malaysia Press

Broad Money Demand and Financial Liberalization in Malaysia:An Application of the Nonlinear Learning Function

and Error-Correction Models

'M AZALI & 2KENT MATTHEWS'Department of EconomicsUniversiti Putra Malaysia

43400 UPM Serdang Selangor2 Cardiff Business School

University of Wales Cardiff

Keywords: Broad money demand, financial liberalization and innovation, nonlinear learningfunction, and error-correction model

ABSTRAK

Kajian ini bertujuan melakukan pemo.delan terhadap kesan liberalisasi dan inovasi kewangan keatas fungsi permintaan wang meluas di Malaysia. Fungsi pembelajaran tak linear dan modelpembetulan ralat telah digunakan. Keputusan ujian keharmonian dan simulasi dinamik telah jugadigunakan sebagai spesifikasi penerimaan. Kajian mendapati pengolahan model pembetulanralat yang merangkumi angkubah tindak balas tidak berupaya mengesan kejadian liberalisasikewangan bagi fungsi wang meluas di Malaysia.

ABSTRACT

This paper attempts to model the effects of financial liberalization and innovations on thedemand for broad money in Malaysia, The nonlinear learning function and error-correctionmechanism model were utilized. The results of encompassing tests and dynamic (ex post)simulation confirm the error-correction as the parsimonious specification. The augmented error-correction model with nonlinear interacted variables is unable to detect the effects of financialliberalization on Malaysian broad money demand.

INTRODUCTION

Relatively little research work has beenundertaken on the underlying effects of thefinancial liberalization and innovations on thedemand for money in Malaysia. The view thatfinancial liberalization and innovations areimportant for money demand and thus haveimportant consequences for the conduct ofmonetary policy, has a pedigree that goes backto the mid-1950s. The precursor to recentstudies was the work of Gurley and Shaw (1955,1960); Cagan and Schwartz (1975); and Friedman(1984).

An important feature in the conduct ofmonetary policy and hence the monetarytransmission mechanism, is the existence of astable and predictable relationship betweenmonetary aggregates, output, prices and interest

rates. However, financial liberalization andfinancial innovations, have been the primesuspects for the breakdown in the money demandrelationship, leading the authority to focus itscontrol on other indicators of monetary policysuch as bank credit and interest rate (Ismail andSmith 1993).

Moreover, one must take notice that thereare various types of financial liberalization takenplace in the context of developing countries.For example, Tseng and Corker (1991) hademphasized on interest rates liberalization. As aresult they argued that 'it is important to questionwhich interest rates have been liberalized: if, forexample, interest rates on time deposits increaseafter liberalization, the demand for broad moneymight rise at any given level of income, but thedemand for narrow money might decline. This

M. Azali & Kent Matthews

means that financial deregulation and innovationmight cause instability on both monetaryaggregate demand functions - cause a one-timeor a gradual shift in money holdings and hencealter the sensitivity of money demand to changesin both income and interest rates' (p. II)1.

Having mentioned the above issues,certainly, additional work is needed before theimpact of financial deregulation and financialinnovation on the money demand in Malaysiacan be fully understood. The purpose of thispaper is to clarify the situation. In so doing, thisstudy adopts the nonlinear learning functionmodel (NLM) recently proposed by Hendry andEricsson (1991) and Baba, Hendry and Starr(1992) to capture the impact of financialderegulation and financial innovation on incomeand interest rates, which technique to my bestknowledge, has never been tested on theMalaysian money demand function2. In addition,the error-correction mechanism model (ECM)is utilized for comparison purposes. Theprocedures are elaborated in the immediatesection. Specifically, this study emphasizes onthe demand for real broad money (m2) due tothe following reasons:

First: As a result of financial liberalizationand financial innovation, the Malaysian monetaryauthority currently monitors the growth of broadmonetary aggregate (m2). It is argued that thegrowth of m2 reflects the growth of privatesector liquidity (BNM 1994). Thus, focussing onm2 is justified in order to parallel to the presentpractice of the Malaysian monetary authority(Mohamed 1996).

Second: A study by Mohamed (1998) hasdiscovered that the short-run narrow moneydemand function has suffered from one seriousproblem such as temporal structural instabilityand thus reducing the robustness of policyimplications.

Having mentioned the introduction above,the remainder of the paper is organized asfollows. Section II presents the methodologyused in estimating the nonlinear model. SectionIII deals with the series properties. The empiricalresults are presented and discussed in SectionIV. Finally, Section V presents the paper'sconcluding remark.

METHODOLOGYA Model of the Nonlinear learning Process (NLM)In the literature on financial deregulation andfinancial innovation it is often suggested thatderegulation and innovation take time to exerttheir full impact on money demand as economicagents need time to learn about newly innovatedfinancial assets and to change their behaviour ofmoney holdings. To model the transition phaseof financial adaptation; Hester (1981), Hendryet al (1991), and Baba et al (1992); amongstothers, have used the NLM, They haveconstructed the nonlinear interacted variablesusing the following nonlinear weighting function\w) to represent agents learning process aboutthe newly introduced financial assets. Thenonlinear weighting function is constructed as:

wg= (1 + e x p [ a - P ( * - J*for t > /* and 0 otherwise. (1)

where t is time, £* is the date of introduction ofthe assets, and a and /3 correspond to initialknowledge and the rate of learning. FollowingHendry et al (1991), Baba et al (1992), theyhave set a = 7 and j3 = 0.8 for the US, implyingthe learning adjustment w{ - 0.50 after two yearsand wt = 0.999 after 3.5 years; while for the UKthey set a = 5 and (3 = 1.2, implying w( - 0.50after one year and wt = 0.999 after two years,indicating higher initial knowledge and morerapid learning for the UK counterpart. In hiswork on the Australian data, Hossain (1994) setthe a and (3 values exactly as Hendry et al(1991) had done in the US case. Hendry et al(1991) showed that the inclusion of the learning-adjusted ml retail sight-deposit interest rate inthe error-correction model explained sufficientlythe rapid growth of ml in the UK and in the US.Baba et al (1992) also found a similar result onthe demand for ml. On the other hand, thestudy by Hossain (1994) indicated that the resultswere qualitatively similar, either using thenonlinear learning adjustment function or usingthe linear time trend3.

1. See, Hossain (1994); Fry (1995), Hoque and Al-Mutairi (1996), amongst others.2. Hendry and Ericsson (1991) adopted this technique in investigating the demand for narrow money in the United Kingdom and the United States. Hossain

(1994) utilised this technique on the Australian short-run narrow money demand function.3. Hossain (1994) employed partial adjustment as his base line model.

132 PertanikaJ. Soc. Sci. & Hum. Vol. 9 No, 2 2001

An Application of the Nonlinear Learning Function and Error-Correction Models

Following those studies, we have adjustedthe initial values for a and )8 in order to be wellfitted to the Malaysian case. We set the a = 10and /3 = 0.5, implying wt = 0.50 after five yearsand wt = 0.999 after 8.5 years to indicate a lowerinitial knowledge and slower learning processamong the economic agents in Malaysia4. Theseries (wt) were constructed beginning in 1979and reached the peak at the end of the 1980s, sothat the generated weighting series would capturemost of the financial deregulation and financialinnovation during the 1980s. Next, the generatedweighting series were used to define twononlinear interacted variables for income (y)and interest rate (ir), respectively: u/*A y andto*A ir, and finally these variables were includedin the augmented error-correction model forestimation purposes. The equation can bewritten as follows:

A(m2-p)t = aQ + ax A?, + a2 Airt

+ Q, AA p + a, ecm2+ a xu^Ay + CL zu*Air (2)

where m2-p is real m2; Apt is an annualizedinflation rate, and ecm2vX is an error-correctionterm. From equation (2) there are three testablenull hypotheses: a6 = 0, a7 = 0 and a6 = a7 = 0.The first two hypotheses indicate that for eachinteracted variable has no effect on moneydemand, whilst the latter shows the combiningeffects are insignificant on money demand. Inthis paper, the final estimated ECM is estimatedusing Hendry-type general-to-specificmethodology (Hendry 1979; Hendry et al. 1984).

The Time Series Properties

Detailed sources and data description used inthis study are presented in Data Appendix. Thequarterly data used spanning 1972:ql to 1993:q4.The Dickey-Fuller (DF)5, the Augmented Dickey-Fuller (ADF)6 and the Phillip-Perron (1988)procedures were utilised in order to check theorder of integration or the data stationarity. Forall tests, the results show that all variables areintegrated of order one or stationary after first

differences7. All variables are specified in naturallogarithm. The cost of holding money ismeasured by the ir, which is the margin betweenthe 3-month Treasury bill rate (tbr$) and the 3-month banks' fixed deposit rate (dr$). Inmodeling the long-run broad money demandfunction, we have tried various measurements ofopportunity cost of holding broad money ue. 3-month, 6-month and 12-month Treasury billrates. However, the results were inconsistentwith the standard theoretical sign (negative) inthe money demand study8. This problem isactually well recognized in the literature of moneydemand in the developing countries (Aghevli etal. (1979); Tseng and Corker (1991); and Tariqand Matthews (1996)). Following Tariq andMatthews (1996) and Johansen and Juselius(1990), the irwas constructed as the margin ratebetween the 3-month Treasure bill rate and the3-month commercial banks fixed deposit rate9.Moreover, the ir could be considered as atightness of holding money. The fixed depositrate acts as the own rate of interest and the tbr3acts as the rate on the alternative asset.

Cointegration test - The Johansen-Juselius's(1990, 1992) maximum likelihood procedurewas utilised to test for long-run cointegratingrelationships among real broad money (m2-p),real income (y), interest rate margin (ir) andannualised inflation rate (Ap). The estimates ofthe cointegrating vector was based on a VARmodel with two lags10, a constant, 'centered'seasonal dummies (SI, S2 and S3) to capturethe presence of the linear time trend in thenonstationary part of the data generating process,and a shift variable with value zero to 1978:4inclusive and non-zero thereafter (DUM1) toallow for the effects of financial liberalization.The results of the rank tests are shown in Table1 below.

The null hypothesis of non-stationary withno cointegration (r= 0) was rejected at the 5%level on the basis of the trace and the maximal-eigenvalue (k-max) test statistics, whereas the nullhypothesis of r to be, at most, one was notrejected. This led to the conclusion that there

4. This is an ad hoc procedure. The weight (w) was chosen by plotting equation {1) using the irial-and error technique on various combinations of ( and( values.

5. See, Dickey and Fuller (1979, 1981).6. See, Engel and Granger (1987).7. Full results are available upon request from the author.8. A similar problem exists when the short-run broad money demand function is estimated.9. The 3-month Treasury bill rate was chosen because it is well established, while the 3-month fixed deposit rate is the main part of broad money, m2.10. The Akaike Information Criteria (A1C) test statistics are as follows: 0 lag = 351.4; 1 lag • 555.9; 2 lags = 564.9; 3 lags = 559.7; and 4 lags = 563.0.

Pertanika J. Soc. Sci. 8c Hum. Vol. 9 No. 2 2001 133


TABLE 1Cointegration with unrestricted intercepts and restricted trends in the VAR

Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix

80 observations from 1974Q1 to 1993Q4. Order of VAR - 2.List of variables included in the cointegrating vector:m2 y ir ApList of 1(0) variables included in the VAR:DUM1 SI S2 S3List of eigenvalues in descending order:.32567 .19297 .065559 .0082307

Nullr = 0r<= 1r<= 2r<= 3

Alternativer = 1r« 2r = 3r • 4

Cointegration LR Test

Nullr = 0r<= 1r<= 2r<= 3

Alternativer>= 1r>= 2r>= 3r = 4

Statistic31.5229*17.15115.42450.6611

Based on the

Statistic54.7597*23.23686.08570.6611

95%

Trace of

95%

Critical Value27.420021.120014.88008.0700

the Stochastic

Critical Value48.880031.540017.86008.0700

90%

Matrix

90%

Critical Value24.990019.020012.98006.5000

Critical Value45.700028.780015.75006.5000

was at most one cointegrating vector (3 forminga cointegrating relationship (3'xwhere x*t = [(m2-p)lt y%i irt, A/>J\ Normalised for {m2-p)v theunrestricted cointegrating vector can be written as:

(m%p) = 1.1726 y - 3.0412 ir - 0.24394 Ap(0.3679)* (2.7624) (0.1285)** (3)

where standard errors in bracket; and **'[**]denotes significant at the 1%[5%] level.

The null hypothesis of a unit incomeelasticity (0^=1) also was accepted by the data asindicated by the log likelihood ratio test statistics,asymptotically chi-squared distribution with onedegree of freedom; %2(1) = 0.15452 [0.694];yielding the long-run restricted cointegratingvector:

= ! . « > - 3.2482 ir - 0.29385 Ap (4)

or the so-called an error-correction term (ECM2)and can be written as

ECM2 = 1.0 (m2-p) - 1.0 y + 3.2482 ir+ 0.29385 Ap (4.1)

In summary, the results of cointegration testssuggest that it seems reasonable to proceed witha single-equation analysis for {m2-p)c However,

before doing so, the explanatory variables weretested for exogeneity by applying the Wu-Hausman test. The computed Wu-Hausman T2

statistic F(3,71) = 0.595 [0.554] supported thehypothesis that the explanatory variables in thecointegrating vector were exogenous.

There are two main results that can bederived from the estimates of long-run imposedrestrictions money demand function:

First: Although the imposed unit long-runincome elasticity was not rejected, however, theunrestricted income elasticity was found to be1.1726 which is consistent with the results foundin most previous studies in the developingcountries, Chowdhury (1997), Tariq andMatthews (1996); Tseng and Corker (1991); andAghevli et al (1979). The results indicated thatthe velocity of broad money in Malaysia haddeclined over time. A likely explanation is thatas a result of financial deregulation andinnovation, the banking institutions can providemore services, i.e. transaction services, andelectronic facilities, and thus have a greaterimpact on the growth of broad money, implyingthe growing degree of monetisation in theeconomy.

Second: The estimation result also indicatedthat a strong exogenity assumption on theexplanatory variables was not rejected by the

134 PertanikaJ. Soc. Sci. & Hum. Vol. 9 No. 2 2001


Wu-Hausman test statistics, In addition, theacceptance of one cointegrating vector in thedata allowed an efficient estimation of the short-run dynamic of the demand for broad money inMalaysia.

RESULTS

IV The Estimates of NLM and ECMThe final estimated base line ECM model issummarised in Table 2. However for comparisonpurposes, we have also estimated the augmentedECM model which includes two nonlinearinteracted variables namely, io*Ay and uf*Air (seeTable 3). As mentioned in Section II, the finalestimated models were derived using the Hendry-type general-to-specific methodology. Byinspection on both tables, the former estimationresults indicate better outcomes particularly inthe acceptance of the income variable. Thatmeans the inclusion of two nonlinear interactedvariables to capture for the effects of financialliberalization on the demand for broad money

are unable to improve the final specification.The results are supported by the nonrejection ofthe null hypotheses: a6 = 0, a7 = 0 and a6 = a7 =0 using the standard F-test (see, Table 4).

Encompassing Alternative Models

Two different models have been estimated inthe last section: [1] the base line ECM model orthe error-correction model without the nonlinearinteracted variables, and [2] the augmented ECMmodel - the base line ECM model with thenonlinear interacted variables. The next questionto be asked is which equation represents thebest approximation of the data generatingprocess. Following an approach developed byMizon and Richard (1986), this section conductsan F-test, the so-called encompassing F-test, whichis a test for 'restricted' against 'unrestricted'models. The encompassing F-test is conductedon the 'restricted' model (base line ECM model)against the 'unrestricted* models (augmentedECM model) as follows.

TABLE 2Ordinary Least Squares estimation of the base line error-correction model

Dependent variable is (Am2-p)80 observations used for estimation from 1974Q1 to 1993Q4

Regressor Coefficientconstant .03891Ay .18102Air -.90363Air(-2) -.90451AAp -.00586ecm2(-l) -.01551

R-SquaredS.E. of Regression

Standard Error T-RatiofProb].00917.09660.30209,29742.00236.00854

0.31310.0194

Mean of Dependent Variable 0.0265Residual Sum of SquaresDW-statistic

Diagnostic TestsTest StatisticsA: Serial CorrelationB: Functional FormC: NormalityD: Heteroscedasticity

0.02791.4413

LM Version

4.2426[.000]1.8739[.065]-2.9913[.004]-3.0412[.003]-2.4777[.016]-1.8167[.O73]

R-Bar-SquaredF-Stat, F(5, 74)S.D. of Dependent VariableEquation Log-likelihood

F Version

0.26676,7469 [.000]0.0227204.8309

CHSQ (4)= 7.4404[.114] F(4, 70)= 1.7945[.14O]CHSQ (1)= 0.4149[,519] F(l, 73)= 0,3805[CHSQ (2)= 3.2765[.194] Not applicableCHSQ (1)= 0.3002[.584] F(l, 78)= 0.2938[

.539]

.589]

Notes:A:Lagrange multiplier test of residual serial correlationB:Ramsey's RESET test using the square of the fitted valuesGBased on a test of skewness and kurtosis of residualsD:Based on the regression of squared residuals on squared fitted values

where ecm2 = 1.0*w2 -1.0*)? + 3.2482*ir + 0.2938b*Ap

Pertanika J. Soc. Sci. 8c Hum. Vol. 9 No. 2 2001 135


TABLE 3The augmented error-correction model with nonlinear interacted variables

Dependent variable is (Am2-p)80 observations used for estimation from 1974Q1 to 1993Q4

RegressorconstantAyAirAiri-2)AApecm2(-\)nf*Ay

Coefficient0.041240.17657-0.64338-0.87181-0.00628-0.017630.02294-0.79335

Standard Error.00995.11625.38157.30094.00242.00941.18266.68145

T-Ratio[Prob]4.1441 [.000]1.5189[.133]

-1.6861 [.096]-2.8970[.005]-2.5972[.O11]-1.8738[.065]0.1256[,900]-L1642[.248]

R-SquaredS.E. of Regression

0.3260800.019523

Mean of Dependent Variable 0.026557Residual Sum of SquaresDW-statistic

Diagnostic TestsTest StatisticsA:Serial CorrelationB:Functional FormCrNormalityD:Heteroscedasticity

0.0274431.4615

LM VersionCHSQ (4)= 6.7641 [CHSQ (1)= 0.0860[CHSQ (2)= 2.5753[CHSQ (1)= 0.2079[

R-Bar-SquaredF-Stat.S.D. of

.149]

.769]

.276]

.648]

F(7, 72)0.26054.9768[.000]

Dependent Variable 0,0227

F VersionF(4, 68)=F(l, 71)=Not applicableF(l, 78)=

1.5701 [.192]0,0764[.783]

0.2033[.653]

Notes:A.Lagrange multiplier test of residual serial correlationBrRamsey's RESET test using the square of the fitted valuesC:Based on a test of skewness and kurtosis of residualsD:Based on the regression of squared residuals on squared fitted values

TABLE 4The OLS regression results: The error-correction model with nonlinear interacted variables

A(m2-p)

sub-period

F-test forHypothesis:

a6 = 0

a7 = 0

a6 = a7 = 0

1974:1 - 1987:4k=56

F(l,48)= 0.709F(l,48)= 0.184F(2,48)= 0,486

Air( + a3 Airi2 + a4A^<Ayt + a7 Zi/*Airt

1974:1 - 1992:4k=72

F(l,68)= .0200F(l,68)= 0.516F(2,68)= 0.269

* A + a5 ecmKx

1974:1 - 1993:4k=80

F(l,72)= 0.150F(l,72)= 1.350F(2,72)= 0.690

Notes : '*' denotes significant at a 10% level.

Restricted base line error-correction model:

A(m2-p)t = a0 + a, A^ + a2 Airt + a3 Ai+ a4AA p( + a5

Unrestricted augmented error-correction model withnonlinear interacted variables:

(5) A(m2-p)( = a0 + aj Ayt + a2 Mrt

+ a4AA pt + a5

+ a6 w*Ayt + a7 (6)



The test results are reported in Table 5 below.The tests in Table 5 indicate that all the nullhypotheses cannot be rejected, implying theoriginal restricted specification is not mis-specified. However, whether the respecified'restricted' error-correction model (equation 5)can be regarded as a satisfactory short-run broadmoney demand function, depends on its abilityto provide adequate out-of-sample (ex post)forecasts. The results are reported in the nextsubsection.

TABLE 5Encompassing F-test - Restricted vs.

unrestricted models

sample period1974ql-1991q41974ql-1992q41974ql-1993q4

Eqn 5 vs. Eqn 6F(2, 64) = 0.213[.808]F(2, 68) = 0.269[.764]F(2, 72) = 0.692[.504]

Ex post Dynamic Forecast ResultsA summary of the forecast statistics is detailed inTable 6. The results show that the root-mean-square prediction errors (rmse) of the originalrestricted model are quantitatively similar to theunrestricted models. In addition, all of theequations passed the predictive failure tests. AnF-test was also conducted to test the nullhypothesis that there is no difference in theaccuracy of these forecasts (see, Holden, Peeland Thompson 1990; pp. 33-37 for the details ofthe procedure). However, the latter test alsofailed to reject the null hypothesis. As analternative criterion to select the bestspecification, all equations were re-estimated andthen their properties were compared. Theestimation results on equation (5) were reported

earlier in Table 5, while results on equation (6)were reported in Table 4. Based on theseestimation results, it is clear that the originalmodel ('restricted* version) provided the bestoutcomes. Although they retained the desirablediagnostic residual properties, the estimates ofequation (6) indicate that the income variablewas insignificant. Therefore, the restrictedspecification (equation 5) was considered to bethe parsimonious model for broad moneydemand in this study.

CONCLUSION

This study has shown that equation (5) is thepreferred error-correction specification for theshort-run broad money demand function, whileequation (4) for the long-run function. Theresults of encompassing tests and dynamic (expost) simulations confirm equation (5) as theparsimonious error-correction specification. Theinclusion of the nonlinear interacted variableswas unable to detect the underlying effects offinancial liberalization and innovations on thedemand for broad money in Malaysia.

The estimated short-run elasticities ofincome, cost of holding money, and inflation,are all reasonably estimated by the restrictedmodel as reported in Table 2. Overall, theparsimonious model is rather simple, and itincludes basic explanatory variables in thecontemporary literature of money demandfunction.

Given the present results, the following twopolicy implications can be drawn from this study.First, the evidence supports the present strategytaken by the Malaysian monetary authority inmonitoring the growth of the broad monetaryaggregate as one of several strategies to curb

TABLE 6Summary statistics for ex post forecasts

Estimation period

Equation (5)1974:1 - 1992:4

Equation (6)1974:1 - 1992:4

Significance testamong the RMSE

Forecasting period

1993:1 - 1993:4

1993:1 - 1993:4

Hull Hypothesis, Ho:Equation (5) has asmaller RMSE thanequation (6)F,«, = °-27

Predictive failure test

F(4, 70)- 1.597[.185]

F(4, 68)= 1.346[.262]

Conclusion:No difference in the

accuracy of theseforecasts

RMSE

0.025

0.025

PertanikaJ. Soc. Sci. 8c Hum. Vol. 9 No, 2 2001 137

M, Azali & Kent Matthews

inflationary pressure. This study shows thatinflation has a significant negative impact on thedemand for real m2. For example, a 1% rise inthe price level could bring about a 0.58% declinein the demand for real m2, ceteris paribus. Hence,monetary policy has an important role incontrolling the price stability in Malaysia.

Second, the impact of the financialderegulation and innovation on the models'coefficients had been tested using two formalprocedures. The result showed that the recursiveestimates of the short-run interest elasticity hadgradually decreased over time". (Figs. 1 and 2)Thus, the evidence suggested that the interestsensitivity of money demand had increased,implying a small slope of the LM curve (a flatterLM curve). The flatter the LM curve, the greaterwould be the rise in real income and the smallerwould be the rise in the interest rate (Parkinand Bade 1988; p.256). This result should provevaluable for the effectiveness of monetary policy12.

Finally, recent development in the moneydemand literature suggests to estimate Divisiaaggregates (Barnet et al 1984). For example,Cuthbertson (1988, p. 97) argues that 'theapproach is probably most useful where interestrates are market determined rather than subjectto regulation and hence may prove useful in anincreasing competitive financial environment/Hence, the deregulation of interest rates opensup the possibility that future innovations in m2may need to be remodelled within the frameworkof the 'divisia' approach13. I leave this issue forfuture research studies.

REFERENCES

AGHEVLI, B.B., M.S. KHAN, P.R. NARVEKAR and B.K.

SHORT. 1979. Monetary policy in selected Asiancountries. IMF Satff Papers 26(4): 775-824.

BABA, Y., D.F. HENDRY and R.M. STARR. 1992. The

demand for Ml in the U.S.A., 1960- 1988.Review of Economic Studies 59(1): 25-61.

Coef. of DIR and its 2 S.E. bands on recursive OLS

.33761

-.58161

-1.5008

-2.4200

_

-

, , , , , , I . . .

\

Ii

/ A - - '

^ . _

r

;

x . — • • •* . y '

\/ X

1974Q1 1979Q1 1984Q1 1989Q1

Fig. 1. Recursive estimates of coefficient Air( and its 2SE (Equation 5)

11. The estimated slow speed of adjustment is. however, a cause for concern but could indicate that monetary disequilibrium in Malaysia works through widerchannel i.e. credit channel.

12. Kandil (1991) estimated the slope of LM curve for Malaysia was about 0.20. Thus, the present finding was consistent with a higher sensitivity of moneydemand to changes in the interest rate.

13. Interested readers can refer to Habibullah (1999)

PertanikaJ. Soc. Sci. & Hum. Vol. 9 No. 2 2001


Coef. of DIR and its 2 S.E. bands based on recursive OLS

-.0064007

-.83014

-1.6539 - . -

•2.47761 .. .i i i1973Q1 1978Q2 ^ 983Q3 1988Q4 1993Q4

Fig. 2. Recursive estimates of the coefficient of Air( (Equation 6)

BANK NEGARA MALAYSIA. 1994. Money and Banking in

Malaysia. Kuala Lumpur.

BANK NEGARA MALAYSIA. Quarterly Bulletin. Various

issues.

BARNET, W.A., E.K. OFFENBACHER and P.A. SPINDT.

1984. The new divisia monetary aggregates.fournal of Political Economy 92(6): 11-48.

CAGAN, P. and A. SCHWARTZ. 1975. Has the growth of

Mmoney substitutes hindered monetary policy.fournal of Money, Credit, and Banking7: 137-159.

CHOWDHURY, A.R. 1997. The financial structure and

the demand for money in Thailand. AppliedEconomics 29(3): 401-409.

CUTHBERTSON, K. 1988. The Supply and Demand for

Money. New York: Basil Blackwell Inc.

DICKEY, D.A. and W. FULLER. 1979. Distribution of

the estimators for autoregressive time serieswith a unit root, fournal of the American StatisticalAssociation 74: 427-431.

DICKEY, DA. and W. FULLER. 1981. The likelihoodratio statistics for autoregressive time serieswith a unit root. Econometrica 49(4): 1057-1072.

ENGLE, R.F. and C.WJ. GRANGER. 1987. Cointegrationand error correction representation,estimation, and testing. Econometrica. 55(2): 251-276.

FRIEDMAN, M. 1984. Lesson from the 1979-1982monetary policy experiment. The AmericanEconomic Review Papers and Proceedings 74: 339-443.

FRY, MJ. 1995. Money, Interest, and Banking inEconomic Development Second Edition. London:The Johns Hopkins University Press.

GINSBURGH, V.A. 1973. A further note on the

derivation of quarterly figures consistent withannual data. Applied Statistics. 22: 368-374.

GURLEY,J.G. and E.S. SHAW. 1955. Financial aspects

of economic development. American EconomicReview. 45(4): 515-538.

GURLEY, J.G. and E.S. SHAW, 1960. Money in a Theory

of Finance. Washington D.C: BrookingsInstitution.

HABIBULLAH, M.S. 1999. Divisia Monetary Aggregatesand Economic Activities in Asian DevelopingEconomies. Aldershot: Ashgate.

HAQUE, A. and N. AL-MUTAIRI. 1996. Financial

deregulation, demand for narrow money andmonetary policy in Australia. Applied FinancialEconomics 6(4): 301-306.

PertanikaJ. Soc. Sci. & Hum. Vol. 9 No. 2 2001 139


HENDRY, D.F. 1979. Predictive Failure andEconometric Modelling in Macroeconomics:The Transaction Demand for Money, Chapter9. In Economic Modelling, ed. P. Ormerod. pp.217-242. London: Heinemann EducationBooks.

HENDRY, D.F. and N.R. ERICSSON. 199L Modellingthe demand for narrow money in the UnitedKingdom and the United States. EuropeanEconomic Review 35: 833-886.

HENDRY, D.F., A.R. PAGAN and J.D. SARGAN. 1984.Dynamic Specifications. In Handbook ofEconometrics, eds. Z. Griliches and M.D.Intriligator, Vol.2. Amsterdam: North-Holland.

HESTER, D.D. 1981. Innovations and MonetaryControl, Brookings Papers in EconomicActivity, pp. 141-189.

HOLDEN, K., D.A. PEEL and J.L. THOMPSON. 1990.Economic Forecasting: An Introduction. New York:Cambridge University Press.

HOSSAIN, A. 1994. Financial deregulation, financialinnovation and the stability of the Australianshort run narrow money demand function.Economics Notes 23(3): 410-437.

HYLLEBERC, S., R.F. ENGLE, C.W.J. GRANGER and B.S.Yoo. 1990. Seasonal integration andcointegration. Journal of Econometrics 44: 215-38.

INTERNATIONAL MONETARY FUND. InternationalFinancial Statistics, [monthly], Washington.Various issues.

ISMAIL, A.G. and P. SMITH. 1993. Monetary policyand commercial banks: an overview. JurnalEkonomi Malaysia 27: 29-55.

JOHANSEN, S. 1988. Statistical analysis of cointegrationvectors. Journal of Economic Dynamics and Control12: 231-254.

JOHANSEN, S. and K. JUSELIUS. 1990. Maximumlikelihood estimation and inference oncointegration -with application to the demandfor money. Oxford Bulletin of Economics andStatistic 52(2): 169-210.

JOHANSEN, S. and K. JUSELIUS. 1992. Testing structuralhypotheses in a multivariate cointegrationanalysis of the PPP and the UIP for UK Journalof Econometrics 53: 211-244.

KANDIL, M. 1991. Structural differences betweendeveloping and developed countries: someevidence and implications. Economic Notes 20(2):254-278.

MOHAMED, A, 1996. Velocity and the variability ofanticipated and unanticipated money growthin Malaysia. Applied Economics Letters 3: 697-700.

MOHAMED, A. 1998. The monetary policytransmission mechanism: the Malaysianexperience during the pre-liberalization andpost-liberalization periods. Unpublished Ph.D.Thesis, University of Wales Cardiff, the UK.

MIZON, G.E. and J.F. RICHARD. 1986. Theencompassing principle and its application totesting non-nested hypotheses. Econometrica 54:657-678.

PARKIN, M. and R. BADE. 1988. Modern Macroeconomics.Hertfordshire: Simon and Schuster InternationalGroup.

PESARAN, M.H. and B. PESARAN. 1996. Working xvithMicrofit 4,0: Interactive Econometric Analysis,Oxford University Press.

PHILLIPS, P.C.B. and P. PERRON. 1988. Testing for aunit root in time series regression. Biometrika75(2): 335-346.

TARIQ, S.M. and K.G.P. MATTHEWS. 1996. Demandfor simple-sum and divisia monetary aggregatesfor Pakistan. Cardiff Business School WorkingPaper, University of Wales Cardiff, the UK.

TSENG, W. and R. CORKER. 1991. Financialliberalization, money demand and monetarypolicy in Asian countries, internationalmonetary fund. Occasional Paper No. 84,Washington D.C.

Received: 16 August 2001



DATA APPENDIX method (see Ginsburgh 1973)). ObservationsData used in the paper come from various issues spanning 1972:ql to 1993:q4. The variablesof the International Financial Statistics published w h i c h h a v e b e e n u s e d i n t h e P a P e r a r e a s follows:by the International Monetary Fund and the GDP-Gross Domestic Product; m2 - broad moneyQuarterly Bulletin Bank Negara Malaysia. The supply; p- consumer price index (CPI, 1990=100);quarterly data of the Industrial Production Index d& ' 3-month deposit rate; and tinS - 3-month(IPI, 1990=100) were utilised to interpolate the Treasury bill rate,quarterly GDP data (y) using the Vangrevelinghe's

PertanikaJ. Soc. Sci. & Hum. Vol. 9 No. 2 2001 141

broad money demand and financial liberalization in ... papers/jssh vol... · pembetulan ralat telah...

Documents