Pertanika J. Soc. Sci. & Hum. 12(2): 87-100 (2004) ISSN: 0128-7702i Universiti Putra Malaysia Press

On the Twin Deficits Hypothesis: Is Malaysia Different?

EVAN IAU & AHMAD ZUBAIDI BAHARUMSHAHDepartment of EconomicsUniversiti Putra Malaysia

43400 UPMf Serdang, Selangor, Malaysia

Keywords: Current account deficit, budget deficit, Granger non-causality, JEL classification: F30,H60, H62

ABSTRAK

Kajian ini membincangkan isu semasa berkaitan defisit berkembar di Malaysia. Teknik statistikyang dipelopori oleh Toda dan Yamamoto (1995) digunakan di dalam kajian ini untuk menganalisishubungan penyebab jangka panjang di antara defisit akaun semasa dan belanjawan kerajaan.Keputusan empirikal jelas menunjukkan bahawa terdapat hubungan penyebab dua hala di antarakedua-dua defisit di Malaysia. Keputusan ini menunjukkan dengan jelas bahawa Malaysia berbezasekali daripada kebanyakan negara-negara industri. Dari satu perspektif, keputusan inimencadangkan bahawa hubungan penyebab daripada defisit belanjawan kerajaan kepada defisitakaun semasa yang membuktikan fenomena defisit berkembar. Dari sudut lain, hubunganpenyebab daripada defisit akaun semasa kepada defisit belanjawan kerajaan mencadangkanelemen-elemen polisi pengukuhan akaun semasa. Oleh itu, masalah menguruskan akaun semasadefisit dalam akaun semasa tidak hanya boleh bergantung sepenuhnya pada mengurangkandefisit di dalam belanjawan kerajaan.

ABSTRACT

This paper discusses the on-going debates surrounding the issue of twin deficits in Malaysia. Thestatistical technique advocated by Toda and Yamamoto (1995) for handling economic variablesthat might spuriously move together is utilized to examine the long run causal relationshipsbetween budget and current account deficits. We examine more than the three decades of timeseries data to answer the question of whether the budget deficit had led to current accountdeficit. The empirical result reveals the presence of bi-directional causality between the twodeficits in Malaysia. It is this finding that makes Malaysia different from the major industrializedcountries. On the one hand, we find that the causal relationship is from budget to currentaccount deficits providing evidence of twin deficits phenomena. On the other hand, the reversecausation as detected in this study tends to suggest some evidence of current account targeting.Therefore, policy to curb 'chronic' current account deficit cannot be achieved if the policymarkers simply rely on curtailing budget deficit.

INTRODUCTION

The past two decades have witnessed large swingsin budget as well as large fluctuations inemployment, output, interest rates, exchangerate and the trade balance in the majorindustrialized countries. Economists view these

events as harmful to the economy. The bestknown events took place during the "Reaganfiscal experiment" in the 1980s which marked aperiod of strong appreciation of the dollar andan unusual shift in external balance not in favorof the United States.1 In Europe, both Germany

In the period 1981-1985, budgetary deficits in US rose from almost zero to a total of USD140 billion in 1985. In thesame period, there was simultaneous depreciation of US dollar in real as well as nominal terms as well asdeterioration in current account balance from a current account surplus of USD6.0 billion in 1981 to a deficit ofUSD 120 billion by the year 1985. The two deficits were called twin deficits because they mpve in the same direction(amount) and they derived from the same economic fundamentals.

Evan Lau & Ahmad Zubaidi Baharumshah

and Sweden faced similar problems in the earlypart of the 1990s where the rise in the budgetdeficits was accompanied by a real appreciationof their national currencies. This in turn adverselyaffects the current account balances. Developingcountries have also experienced severe problemswith external debts in the early 1980s. The hugebudget deficits during these periods widencurrent account deficits in these debt-crisiscountries. The emergence of current accountdeficit and budget deficit phenomena in manycountries in recent years has drawn increasingattention to the problem of "twin deficits".

The issues relating to the two deficits haveimportant policy implications on the economicperformance of a country. Large and persistentcurrent account deficits are troublesome due tothe transfer of a nation's wealth to foreigners.More importantly, countries with large deficitsface difficult economic problems that necessitatedsome kind of policy response if such tendenciesare expected to continue for a long period oftime. Suppose that the basic reason for therising of current account imbalances is primarilydue to the escalating of government budgetdeficit, then the deficit in current account cannotbe remedied unless policies that addressgovernment budget deficit are put into place.The success of such policy measures of coursedepends upon whether budget deficit causescurrent account deficit or the other way round.If the causal link between the two variables isincorrect, then reduction in the governmentbudget deficit may not solve the dilemma ofcurrent account imbalances. In other words, todesign an appropriate policy stance, the essenceof the problems has to be examined thoroughly.2

It is worth noting that the experiences of adeveloping country can sometimes be verydifferent from large industrialized nations. Forinstance, the developing nations have poorinfrastructure, trade impediments and tightregulations in the financial sector, not to mentionpolitical uncertainty that usually follows theseproblems. We can expect some differences inthe macroeconomic dynamics governing budgetand current account deficits between developing

and developed economies. Therefore, lessonsfrom the industrialized countries may not applyto the emerging economies because thecircumstances may differ. In addition, thediscussion is also especially relevant given thebackdrop of the financial crisis that engulfedMalaysia. Malaysia and most of the crisis-affectedAsian countries recorded large current (andbudget) deficits. Indeed, due to the size of theexternal deficit, some economists havequestioned the sustainability of the deficit inperiods prior to the 1997 crisis (Lau andBaharumshah 2003).

Malaysia now belongs to the upper middle-income developing country with per capita GNPof USD 3,640 in 2001 (World Bank 2003).Following the recent Asian financial crisis, theringgit was pegged to the US dollar in September1, 1998: Prior to the financial crisis, the economyrecorded persistent current account deficits goingas far back as 1989. The current account deficitsgrew from 5% of GDP in 1993 to 8% in 1994and increased to 10.5% in 1995. Although thecurrent account deficits have alternated in thepast two decades or so with some years ofsurpluses it had, on average, a larger deficit(5%) compared to its neighboring countrieslike Thailand (2%) and Indonesia (2.5%) overthe same period.

Malaysia's current account deficits in thelast decade reflected the movements of foreigncapital inflow, mainly foreign direct investment(FDI) from the US, Japan and the NewlyIndustrialized Countries (NIEs). FDI accountedmore than 60% of the capital inflows in the1990s. The FDI boom provides the needed capitalfor investment, employment, managerial skills aswell as technology and therefore, acceleratesgrowth and development (DeMello 1997). Thenation's experience with budget deficits in the1990s differs somewhat from that of the previousdecade. Budget deficits, which exceeded 10% ofGDP in the early 1980s, were closely related tothe current account deficits. The current accountdeficits during these periods were closelyconnected to the imbalances in fiscal budgetlargely due to investments in large infrastructure

Some authors like Edwards (2001) and Megarbane (2002) address the twin deficits issue from the point of view ofmacroeconomic stability of the country. They underlined that the negative implications of a combination of adversefactors (e.g. twin deficits, high interest rates and exchange rate depreciation) would increase the vulnerability of acountry and that the fiscal instruments are crucial for sound macroeconomic policy for transition and developingcountries. Therefore, twin deficits should be avoided.

88 PertanikaJ. Soc. Sci. & Hum. Vol. 12 No. 2 2004

On the Twin Deficits Hypothesis: Is Malaysia Different?

projects. On the other hand, in the late 1980sand 1990s fiscal deficit shrank (sometimepositive) but external deficit was large suggestingthat the external imbalances in the recent yearswere mainly due to private saving-investmentdecisions. It is worth noting that Malaysia's savingrates are one of the highest in the world butthey were insufficient to close-up the saving-investment gap because of the increase inmarginal propensity to invest. In other words,the gap between the national savings andinvestments were filled by the foreign savings.

Figs, 1 and 2 plot Malaysia's current accountand budgetary positions from 1975 to 2000. Inthe early 1980s and the most part of the 1990sthe current account balance is in deficit and sois the budgetary position. Visual inspection ofthe plot suggests that fiscal deficits areaccompanied by wide current account deficits,reflecting the twin deficit phenomena asexperienced in the industrialized countries. Thisobservation is also supported by the highcorrelation (r=0.801) between the two deficitsfor the sample period under investigation.3 Thetwo variables appear to move closely togetherovertime, but the budget deficits appear to bemore volatile than the current accountimbalances, especially as one moves to the recentyears. In spite of the importance of the effect of

fiscal (budget) policy on current account deficit,the subject on twin deficits is under research inMalaysia. The reason is partly because Malaysiahas not experienced any difficulty in managingthe two deficits in the past except in the early1980s due to the collapse of the commodityprices and the recent 1997/98 Asian financialcrisis.

The aim of the paper is to investigate thecausal link between the two deficits. To this end,the Toda and Yamamoto (1995) Granger non-causality test is utilized to examine the long runrelationship of the two deficits. Based on theexperience of Malaysia, this present work aimsto seek and contribute to the debate on the twindeficit phenomena in emerging economies,which we find is still lacking in the literature.Specifically, the purpose is to identify the causaldirection of the relationship between the twovariables. This in our view is important as it willprovide the right policy option (or mix) tocombat the above-mentioned issue.

The present paper differs from all previousstudies in the following ways: First, we utilizedan alternative testing methodology, endorsed byToda and Yamamoto (1995) which has verygood power properties against the causality testbased on vector error correction model (VECM).Importantly, the Toda-Yamamoto overcome the

-1076 78 80 82 84 86 88 90 92 94 96 98 00

Timc(Ycar)

Fig. 1: Malaysian current account

76 78 80 82 84 86 88 90 92 94 96 98 00Time(Year)

Fig. 2; Malaysian budgetary position

The correlation coefficient analysis measures the strength or the degree of linear association between two variables.In this study, we are interested in finding the correlation between current account and budget deficits. The twovariables are treated in a symmetrically fashion where there is no distinction between the dependent and theexplanatory variable. After all, the correlation between current account and budget deficits is the same as thatbetween budget and current account deficits. We like to express gratitude for the anonymous referee for providingthis insight.

PertanikaJ. Soc. Sci. & Hum. Vol. 12 No. 2 2004

Evan Lau & Ahmad Zubaidi Baharumshah

pretest bias associated with unit root andcointegration tests.4 Second, the empiricalevidence on the link between the two deficits isdrawn from the experience of an emergingmarket economy - Malaysia. Few studiesinvestigated the twin deficit hypothesis based onthe data from emerging market economies(exceptions are Anoruo and Ramchander 1998and Khalid and Teo 1999). In this way, we hopeto add to existing literature on the host subject.

The remainder of this paper is structured asfollows. Section II provides the relevant literaturein the research area. A simple theoreticalframework for analyzing the causal relationshipbetween the budgetary and current accountdeficits is also given in Section III. In section IV,we briefly outlined the methodology and dataused in the analysis. Section V presents ourempirical results. We also included the furtheranalysis of the multivariate setting in this section.The concluding remarks and policy implicationsare contained in the final section.

PREVIOUS LITERATURE

The connection between budget deficit andcurrent account deficit has sparked aconsiderable amount of interest amongeconomists in the past few decades. Thediscussion has mainly centered on two majortheoretical models. The first view is based on thepopular Keynes proposition. By using the well-known Mundell-Fleming framework, Keynesshowed that an increase in budget deficit wouldinduce upward pressure on interest rates, causingcapital inflows and what follows the appreciationof exchange rates. According to this absorptiontheory, an increase in budget deficit wouldinduce domestic absorption and hence importexpansion, causing current account moves intodeficit. Therefore, Keynes suggests aunidirectional Granger causality that runs frombudget deficit to current account deficit.Research that used modern statistical technique

includes authors like Vamvoukas (1999), Piersanti(2000) and Leachman and Francis (2002) whohave all found convincing evidence to supportthe Keynesian view that the budget deficits causethe current account deficits.5

Second, the more controversial and probablyleast accepted view is the Ricardian EquivalenceHypothesis (REH)6, initially developed by Ricardo(Buchanan 1976). According to this hypothesis,an intertemporal shift between taxes and budgetdeficits does not matter for the real interest rate,the quantity of investment or the current accountbalance. In fact, neither a crowding-out effect ofdomestic investment nor a trade deficitnecessarily emerges from a budget deficit. Hence,non-Granger causality relationship between thetwo deficits would be in accordance with theREH. Meanwhile, the empirical evidence foundin Seater and Mariano (1985), Enders and Lee(1990), Evans and Hasan (1994), among othersare supportive of REH. Moreover, the validity ofthe REH is questionable for an emergingeconomy like Malaysia.

Khalid (1996) examined the effectiveness ofthe policy applied on the developing countries.He argued that if the REH is a validapproximation for developing economies, thenthe International Monetary Fund (IMF) shouldrevise their policies to curtail problems like fiscaldeficits and the misalignment of exchange rate.The empirical results support for the validity ofREH is rejected for most LDCs. For Malaysia,the findings suggest the presence of largeproportion of income subject to liquidity-constrained individuals is the main source ofdeviation from Ricardian neutrality. Ghatak andGhatak (1996) examine the validity of REH forIndia. They found that imperfect credit marketsin India are inconsistent with the assumption ofREH.

Third, a unidirectional causality that runsfrom current account to budgetary variable ispossible. This outcome may occur when the

In most application as in ours, it is not known a prior of which order of integration the variables are and whetherthey are cointegrated or not. Consequently, unit root and cointegration are normally required before estimating theVAR model and the hypothesis therefore conditional on these pretests. As the power of unit root test are known tobe low and test of cointegration are known not to be very reliable for small sample, these pretest biased might besevere (see Toda 1995).Some earlier work that attempted to resolve the issue includes Hutchison and Pigott (1984) and Bachman (1992).These studies also identify a causal relationship running from budget to current account deficits.For a comprehensive understanding on the REH, interested reader could refer to Barro (1974), Barro (1989) andSeater (1993).

90 PertanikaJ. Soc. Sci. & Hum. Vol. 12 No. 2 2004

On the Twin Deficits Hypothesis: Is Malaysia Different?

deterioration in current account leads to slowerpace of economic growth and hence increasesthe budget deficits. This outcome is possible fora small open developing economy (e.g. Malaysia)that depends largely on foreign capital inflowsto finance economic developments. Thebudgetary position of a country is usually affectedby large capital inflows or through debtaccumulations from a donor country and withthat the host country will eventually run intobudget deficits. The experience of LatinAmerican and to some extent East Asiancountries illustrate this point (Reisen 1998). Forinstance, in the 1980s most of the Latin Americandomestic investment is growing more than thedomestic savings that have adverse effects oncurrent account. The fiscal position hadexacerbated the private sector imbalances. Thisreverse causality usually observed in LDCs istermed as * current account targeting' bySummers (1988), where he argued that externaladjustment may be sought via fiscal policy.

Motivated by the large and unprecedentedcurrent account deficits as well as massive federalbudget deficits in the developing countries,Anoruo and Ramchander (1998) examine thetwin deficits issue in five developing Asiancountries includes India, Indonesia, Korea,Malaysia and Philippines. They found aunidirectional Granger causal link running fromcurrent account to budget deficits for all thesample countries investigated, except for Malaysiawhere a bi-directional causality is documented.Recently, Khalid and Teo (1999) documentedthe reverse causality for Indonesia and Pakistanwhile Alkswani (2000) reported the reversecausation between the two deficits for SaudiArabia. According to them, this will occur if thegovernment of a country utilized their budget(fiscal) stance to target the current accountbalance.

Lastly, a bi-directional causality between thetwo deficits is also possible. The results obtainedby Darrat (1988), Islam (1998) and Normandin(1999) are supportive of this outcome. Islam(1998), for example, analyzed the relevance oftwin deficit hypothesis in Brazil for the period1973-1991. He found a bi-directional link betweenbudget and trade imbalances. This finding is inaccordance with the result reported by Darrat(1988). These authors went on to argue that inthe case of a bidirectional relationship, budgetcut will not be effective to overcome the problemwith current account deficit. In fact,complementary options such as interest ratepolicy, exchange rate policy, trade policy with abudget cut are a better option. The abovediscussion identified four direct possible linksbetween budget and current account deficits.7

The body of evidence, however, does notyield a consensus on the causal relationshipbetween the two deficits. The results were foundto be affected by the sampling period as well asthe method used in the investigation. To sumup, the role of fiscal deficit in correct currentaccount imbalances is not without controversy.Henceforth, the issue has become very importantin developing nations and we are motivated toreexamine the relationship between the two, ifany, for Malaysia.

CURRENT ACCOUNT ANDFISCAL BALANCE IN NATIONAL

ACCOUNTS FRAMEWORKThe national account identity provides the basisof the relationship between budget deficit andcurrent account deficit.8 The model starts withthe national income identity for an openeconomy that can be represented as:

Y=C+I+G+X-M (1)

Studies by Haug (1996) and Cardia (1997) found contradict perspective of the REH when they nested Ricardianequivalence within a non-Ricardian equivalence. Their simulation results also show that the lack of a strongrelationship between the current account deficit ratio and budget deficit ratio has been found for the G-7 countries.A low correlation exists between the two series in the nested and non-nested hypothesis. Moreover, they did notsupports any testable hypothesis presented in this study.We adopt a simple bivariate model discussed in Khalid and Teo (1999), Vamvoukas (1999) and Akbostanci and Tune(2001) to identify a casual relationship between the two deficits in developing countries. Similarly, Piersanti (2000),Hatemi and Shukur (2002) and Leachman and Francis (2002) also use the same framework to identify the causalitybetween current account and budget deficits for developed nations. As such, the bivariate analysis adopted in thisstudy well is accepted in the previous literature on the subject matter.

Pertanika J. Soc. Sci. 8c Hum. Vol. 12 No. 2 2004 91

Evan Lau & Ahmad Zubaidi Baharumshah

where Y= gross domestic product (GDP), C =consumption, / = investment, G = governmentexpenditure, X = export and M = import.Defining current account (CA) as the differencebetween export (X) and import (M), equation(1) becomes:

CA = Y - (C + I + G) (2)

where (C + I + G) are the spending of domesticresidents (domestic absorption). In a closedeconomy saving (S) equals investment (I) andgiven that Y - C = S, we have:

S = / + CA (3)

Equation (3) states that unlike a closedeconomy, an open economy can sourcedomestically and internationally for the necessaryfunds for investments to enhance its income. Inother words, external borrowing allowsinvestment at levels beyond those that could befinanced through domestic savings. The nationalsaving (s) can be further decomposed into private(SJ and government savings (SJ :

Y - T- C

and

Sg - T - G

(4)

(5)

where T is the government revenue. Usingequations (4) and (5) and substituting intoequation (3) yield:

Sp = / + CA + (G-T) (6)

or

CA Sp-I-(G-T)

Equation (7) states that a rise in thegovernment (budget) deficits will increase thecurrent account deficit if and only if, the rise ingovernment deficit decreases total nationalsaving. Supposing that current tax revenues areheld constant and (S - I) remains the same, anincrease in temporary government spending will

cause government deficit to rise (G - T)9 whichwill affect the current account positively. In thisway the government deficit resulting fromincreased purchase reduces the nation's currentaccount surplus, which, in other words, suggeststhe worsening of current account balances.

METHODOLOGY AND DATAThe standard procedure used for testing non-causality such as the error correction model(ECM) and vector error correction model(VECM) when the variables are cointegrated arecumbersome and sensitive to the values ofnuisance parameters in finite samples andtherefore 'the virtues of simplicity and ease ofapplication have been largely lost' (Rambaldiand Doran 1996:3). Alternatively, Toda andYamamoto (1995) have proposed a procedurefor testing Granger non-causality which allowscausal inference to be conducted in the levelVARs that may contain integrated processes butdoes not rely upon integration and cointegrationproperties of any or all the variables. Theprocedure offers a simple formulation and isconvenient to apply which permits linear as wellas non-linear tests of restrictions. Moreover, ithas a normal standard limiting chi-squaredistribution and the usual lag selectionprocedures can be applied even if there is nocointegration and /o r the stability and rankconditions are not satisfied ' so long as the orderof integration of the process does not exceedthe true lag length of the model' (Toda andYamamoto 1995:225). The MWALD statistics isvalid regardless the series is 1(0), 1(1) or 1(2), noncointegrated or cointegrated of a arbitrary order.Rambaldi and Doran (1996) had shown that theMWALD could be easily computed by SeeminglyUnrelated Regression (SUR).

They showed that in the integrated and(non-) cointegrated system, the MWALD test forrestrictions on the parameters of a VAR(fc) hasan asymptotic %2 distribution when a VAR (p=k+ d ) is estimated, where d is the maximum

max' max

order of integration suspected to occur in thesystem and k is the lag length selected for theestimation. These restrictions themselves wouldthen imply long run causal inference since unlikeordinary difference VARs, the formulation

The difference between (G - T) is the government budget position. If G > 7'the government will experiencing budgetdeficit while analogously if T > G the country would experiencing budget surplus.

92 PertanikaJ. Soc. Sci. & Hum. Vol. 12 No. 2 2004

On the Twin Deficits Hypothesis: Is Malaysia Different?

involves only variables appearing in their levels.Moreover, the twin deficit nexus is a long runbehavioral relationship that requires metho-dologies for estimating long run equilibria. Thus,the application of the Toda and Yamamotomethod is suitable in permitting the long runequilibrium states.

We now turn to the causal inference betweenthe two deficits. We note that the negativerelationship between the accumulation of foreignassets and budget deficit can be causally linkedin a two-dimensional VAR system (assuming p^3):

[CAD,] CADi! I

(8)

where Ao is an identity matrix. To test whetherBD does not Granger cause movement in CAD(if h=2 and d^l), the null hypothesis HQ:fig = fig = o where #£ are the coefficients ofBDt., i=l,2,...,Kin the first equation of the system.The existence of the causality from BD to CADcan be established through rejecting the abovenull hypothesis, which requires finding thesignificance of the MWALD statistics for BI)t]

and BDt2 identified above while BDt^ is leftunrestricted as a long run correction mechanism.

Similar analogous restrictions and testingprocedures can be applied in testing thehypothesis that CAD does not Granger causemovement in BD, i.e. to test Ho: fig = /IJf = 0where $> are the coefficients of CADtA> i=l,2,...,Kof the second equation of the system (Equation8). Equation (8) has been used as the basis toassess the causality between current account andbudget deficits in the literature. This procedure

can be easily generalized for a larger number oflags in the VAR system.

Quarterly data that spans from the firstquarter of 1976 to the third quarter of 2000 isutilized in the analysis, providing a total of 99observations for the analysis.10 The data are takenfrom International Financial Statistics, publishedby the International Monetary Eund (IMF). Thevariables employed in the study are the currentaccount (CAD) and the budgetary variables (BD)where the variables are expressed as ratio of theGDP in order to account for the economy'sgrowth.11 The IFS provided CAD denominatedin US dollars while the BD and the nominalGDP are measured in domestic currency (RinggitMalaysia). In this study, we express all thevariables in Ringgit Malaysia.12 The data for GDPare available on the annual basis and hence, thequarterly GDP data for this study wereinterpolated from the annual series by employingthe approach suggested by Gandolfo(1981).13

All the data used here are seasonally unadjusteddata. We are aware of the seasonality problem intesting high frequency data. To overcome thisproblem, seasonal dummies are included in theestimation.

EMPIRICAL RESULTS

Prior to testing for causality relationship betweenthe time series, it is necessary to determine theorder of integration and establish that they areintegrated of the same order. To this end, theAugmented Dickey-Fuller test is carried out onthe series, first on level and then on firstdifference forms. Overwhelmingly, the resultspresented in Table 1 suggest that all the seriesare 1(1) variables based on conventionalsignificance levels. We also adopt the Phillips-Perron tests and the results obtained are similarto the ADF tests. To conserve space these resultsare not reported here. Our unit root test results

1(1 The power of unit root and cointegration tests depends on the data's span. It does not help by using data of higherfrequency (e.g. monthly) as the power of most of the test used in the analysis depends on the data's span rather thanits frequencies.

11 We expressed both variables in the straightforward manner of CAD and BD per output following Vamvoukas (1999),Piersanti (2000) and Hatemi and Shukur (2002) rather than the decomposed components of exports and importsfor CAD and revenues and expenditures for BD. The authors are grateful to an anonymous referee for pointing thispoint to clear the uncertainties of data descriptions for this study.

12 We convert the CAD into Ringgit Malaysia by the using the exchange rate between Malaysia and US dollar reportedin IFS.

18 This method provides the convenient and straightforward manner to interpolate the annual basis of GDP intoquarterly observations compared to other interpolation method. The interpolation technique of the continuos flowvariables from the annual basis of GDP to the quarterly observations is documented in the Appendix A of this paper.

Pertanika J. Soc. Sci. & Hum. Vol. 12 No. 2 2004 93

Evan Lau 8c Ahmad Zubaidi Baharumshah

indicate among other things, shocks in thecurrent account and budget deficits tend to bepermanent implying that they are on anunsustainable path. Nevertheless, we caution thereader that the results of these standard unitroot tests may be biased due to the low power ofthe test.

Next, we proceeded with the estimation ofthe VAR model after determining that thed =1. In order to check the robustness andconsistency of the results, we also estimated thesame model with d=2. The lag selection iscommonly done by setting arbitrary lags or byapplying an order estimation criterion such asAkaike Information Criterion and SchwartzCriterion. The choice of the lag length is testedfrom one up to twelve. The optimal lag is basedon minimum Akaike Information Criterion(AIC). Following this criterion we obtain k=6.Therefore, VAR (p = dnax+ k =7) and VAR (p =d + k =8) models were then estimated and the

max

MWALD statistic were was computed. Table 2presented the result of the Granger non-causality.The null hypotheses that the CAD (currentaccount deficits) does not cause the BD (budgetdeficits) and BD (budget deficits) does not causethe CAD (current account deficits) are easilyrejected at the usual significance levels, hence,suggesting the existence of a bi-directionalcausality between the two deficits. It is worthnoting that both Panels A (d=l) and B (d=2)yield similar quantitative conclusions on thebudget-current account deficits nexus.

The finding implies that two deficits movesimultaneously and are connected to each otherin the long run. It turns out that the resultsreplicate the picture drawn in a number ofstudies conducted for the developing economiese.g. (Anoruo and Ramchander 1998; Khalid andTeo 1999 ; Chinn and Prasad 2000). Specifically,Khalid and Teo (1999) and Chinn and Prasad(2000) pointed out that a high correspondencebetween the two deficits is more likely to occurin the developing rather than the developedeconomies. We note that the results contradictwith those studies conducted in the US and EU.It is this finding that makes Malaysia differentfrom the major industrialized countries wherelarge budget deficits normally lead to currentaccount deficit - a one way causal relationship.It also worth noting that the evidence refutedthe Rfcardian hypothesis that suggests that abudget deficit does not matter.

Further Evidence

Further evidence on the interplay betweenbudget deficit and various economic variablesthat affect the real economic activities areexamined to validate the Keynesian propositions.The literature has identified several potentialvariables that may serve as the mediating variables(exchange rate, interest14 rate, investment,inflation, to name a few). In what follows, weextended the 2-vector variable model to includetwo additional variables namely exchange rateand interest rates.15 The results of the causalitytests, which are available from the authors, donot change the causal inference reported earlier.

TABLE 1Augmented Dickey Fuller (ADF) unit root test resultsa

Series

CADBD

-1-1

.879

.844

Series in

(4)(4)

tT

Level

-1.966-2.526

Test

(4)(4)

Statistics

Series in

-4.655b(4)-5.705b (4)

first

tT

Difference

-4.766b

-5.673b(4)(4)

aThe t statistics is the ADF unit root test. tfare the model with constant term and t are the modelwith constant and a time trend. The value in the parentheses represents the number of lag order.The ADF unit root test examine the null hypothesis of a unit root against the stationary alternative.The notation applied in all tables followed: CAD and BD denotes current account deficits andbudget deficits respectively.b denotes significant at 5% level.

We are grateful to the anonymous referee for pointing out this suggestion.For more discussion on this issue, refer to Law (2004).

94 Pertanika J. Soc. Sci. 8c Hum. Vol. 12 No. 2 2004

On the Twin Deficits Hypothesis: Is Malaysia Different?

TABLE 2Long run Granger non-causality test for Malaysia based on Toda and Yamamoto (1995)

Null Hypothesis

Panel A: (k =6, dmax=l)Budget deficits does not Granger cause currentCurrent account deficits does not Granger cause

Panel B: (k =6, <*_ = 2)Budget deficits does not Granger cause currentCurrent account deficits does not Granger cause

accountbudget

accountbudget

deficitsdeficits

deficitsdeficits

TestMWALD

15.25218.865

12.85613.795

Statisticsp-value

0.018*0.004a

0.045 a

0.032 a

Conclusion

Reject HoReject Ho

Reject HoReject Ho

a denotes statistically significant at 5 percent level, k • optimum lag and dmax = maximal order of integration.

To analyze the dynamics as well as the importanceof the variables in the system, we conduct boththe generalized variance decomposition (GVDCs)and the generalized impulse response function(GIRFs). Briefly, the GVDCs provide the strengthof each variable in the system while GIRFs providethe dynamic response of each variable to its ownshocks as well as of other variables included inthe system. Intuitively, the GIRFs trace theresponse over time of a variable say x due to aunit shock given to another variable, say y. Themapping of the GIRFs is represented by agraphical illustrative.

Table 3 provides the decomposition of theforecast error variances of each variable up to24-quarter horizon. The results from this tablemay be summarized as follows.1. BD seems to be the most interactive variable

among the four variables. The GVDCs showthat almost 46% of the forecast errorvariance can be explained by current account(24%), interest rate (13%) and exchangerate (9%) at the end of 24-quarter horizon.

2. Exchange rate is the most exogenous variablein the system with only about 10% of itsforecast variance being explained by theremaining variables in the entire forecasthorizon.

3. Changes in interest rates (budget) are largelydue to the movement in budget deficit(interest rate). The relationship becomesstronger as the horizon increases. Thepicture that emerged seems to suggest a bi-direction relationship between the twovariables. Nevertheless the influence ofbudget on interest rate is much stronger.The variance in BD explains 23% of theerror variance in IR but only 13% of errorvariance of BD are explained by IR.

Together, all these results support theKeynesian proposition i.e. higher budget deficitwill lead to higher interest rates. These resultsclearly do not support the REH that suggestinterest rate movements are independent of thedeficit in the government's budget. A majorpolicy implication of this finding is that budgetdeficit can cause crowding out of privateinvestment and the extent depends on how muchinterest rate is forced up. This in turn canadversely affect capital formation and eventuallyretard long-term growth.

Finally, we examine the pattern of dynamicresponses of each variable using the GIRFs. Givena system of four-dimensional variables, we mayconstruct up to 12 possible scenario of GIRFs foreach of the variables taken separately (ignoringtheir own shocks). The results of GIRFs aretraced out up to 50 quarters in Fig. 3. In allcases, the responses of shock to the subsequentvariable under investigation seems to settle backto its post-shock level after a period of about 6-9 quarters. This suggests that most of theadjustments are rapid and relatively settle as thetransitory impact. Additionally, we observed thatthe impact patterns of CAD and BD to eachother are rather similar. Both adjust back to thelong run equilibrium after 8 quarters. We alsodetect that CA reverts rapidly back to theequilibrium, after the shock from EXC, implyingthat devaluation is taking place in the subsequentquarters. This of course does not follow the J-curve pattern observed in Baharumshah (2001)where the rapid adjustment in trade balancefrom one-time shock of real exchange ratenegates any theoretical J-curve pattern.

PertanikaJ. Soc. Sci. & Hum. Vol. 12 No. 2 2004 95

Evan Lau & Ahmad Zubaidi Baharumshah

Generalized

Percentage of Horizonvariations in (Quarters) ACAD

Quarters Relative Variance in: ACAD1 85.6744 81.7108 80.72124 80.218

Quarters Relative Variance in: ABD1 4,6534 11.5568 16.599

24 24.048

Quarters Relative Variance in: AIR1 6.5664 5.9088 6.02024 6.190

Quarters Relative Variance in: AEXC1 2.3334 1.0088 0.708

24 0.498

TABLEvariance

ABD

8.76010.18610.85811.196

94.65882.32670.77654.231

7.07017,51820.97022.782

0.1730.0800.0760.070

3decomposition11

due to innovation in:

AIR

3,2514.7554.4554.363

0.6473.6887,21312.725

76.13463.18859.93858.064

7.7058.2178.6278.960

AEXC

2.3153,3503.9654.223

0.0422.4305.4128.996

10.23013.38613.07212.964

89.78990.69490.58990.471

ACU

14.32618.29019.27919.782

5.34217.67429.22445.769

23.86636.81240.06241.936

10.2119.3069.4119.529

The last column provides the percentage of forecast error variances of each variable explained collectively by theother variables. The column in bold represent their own shock.

CONCLUSION AND POLICYIMPLICATIONS

Economists have long argued that for developingcountries to reduce 'chronic' current accountdeficits, national savings must rise by reducingthe budget deficits and/or increasing the rate ofprivate savings. The results of this study point tosuggest bi-directional causality between budgetdeficits and current account deficits. This is nota surprising result for an emerging economylike Malaysia. On the one hand, governmentscan have large budget deficits by heavilyborrowing in international markets. Furthermore,even the deficits financed by excessive moneycreation, these are more likely to affect thecurrent account. Excessive monetary expansionin an economy with fixed exchange rate willcause disequilibrium in the money market andwill in turn lead to increase in import demandand a larger current account deficit, other thingsbeing equal. Therefore, we would expect toobserve causality running from budget deficitsto current account deficits.

, On the other hand, higher export prices(or export volumes) generated by increase in

world demand will not only raise export earningsand improve the current account but also reducethe budget deficit (since taxes on export earningsare a significant portion of governments' revenuefor a small economy that depends on exportsectors like Malaysia). Also, an increase in exportprices (or volume) will raise domestic incomefor expansionary or countercycle fiscal policy. Inboth cases, the improvement in the currentaccount could be reflected in an improvementin fiscal balance suggesting the causal relationshipfrom current account deficits to budget deficits(reverse causation). Since both mechanisms areat work in the case of Malaysia, this explainedthe main results recorded in this study.

It is evident from the finding of this paperthat the decision to curb the problem withcurrent account imbalances cannot be achievedby simply relying on fiscal cuts. Policy measuresfocusing on monetary and productivityenchantment may have to be complementedwith the budget cut policy. Monetarists claimthat fiscal policy cannot correct thedisequilibrium in external account. The findingsof this study reject this claim but our research

96 Pertanika J. Soc. Sci. & Hum. Vol. 12 No. 2 2004

On the Twin Deficits Hypothesis: Is Malaysia Different?

« of BO to shock in CA

0.02O:

0.01 £

0 01^

ooos-

oooo;

-0.005-0 5 10 15 20 25 30 35 40 45 50

HofSon(Quattm)

t of EXC to shock fri CA

-0003

-0004-

-0.006

-0 008

•0 010

-0 0170 5 10 15 20 26 30 35 40 45 50

Horizon (Quartets)

G m n l N d fen*** FtespoMM of f t to shack In BO

0.35-

0 30-

025-

02O

015-

0100 5 10 15 20 25 30 35 40 45

tCAlo shocking

knautaa RMpoiiMs or f t to mac* In CA

0 5 10 IS 20 25 30 35 40 45 50

Homon(Qu*rt*r«)

sofCAtottoclcln8D0 0 0 ^

0007

OQ0S-

0 003-

0 0010 5 10 1$ 20 25 30 35 40 45 50

• of EXC to shock In BD

0 5 fO 15 20 25 30 35 40 45 50Hortzon(Qu«rt«rs)

* BO to shock m K

/ CA

0 5 10 15 20 ?5 30 35 40 «5 50

Horizon (Quarter?)

0 5 10 15 20 25 » 35 40 45 $0

s of EXC to shock In R * of CA to shock in EXC

0 5 10 15 20 25 30 35 40 45 50

Hortzon(Qu«rt»rs)

0 5 10 15 20 25 30 35 40 45 50

Honzon(Quart*r«)

/ CA

Ganeraflzed tmpi*M ftesponses of BO to shoe* tn EXC • or«to shock tn EXC

/ BO

0 5 10 15 20 25 30 35 40 45 50

Hortzon<CHj«rttr«)

0 5 10 15 20 25 30 35 40 45 50

Hortzon(Quiritfs)

Fig. 3: Generalized impulse response function

Pertanika J. Soc. Sci. & Hum. Vol. 12 No. 2 2004 97

Evan Lau 8c Ahmad Zubaidi Baharumshah

also suggests fiscal policy (government spendingand taxes) by itself cannot be used to correctcurrent account imbalances. This is because theresults suggest budget deficit is not an exogenousvariable in the policy equation. Given the degreeof openness of Malaysia and how sensitive thecurrent account to foreign interest rate is, wemay expect the current account to be affectedby the budget deficit in the long run. In otherwords, even with budget cuts, external imbalancesin current accounts may lead to deterioration inbudget deficits.

From the dynamic analysis, we foundsufficient evidence to show that the causalrelationships between budget and currentaccount deficits are transmitted through interestrate and exchange rate (BD->IR->EX->CAD).These results strengthen the causality chain inthe bivariate model and lends further support tothe body of literature that suggests that budgetdeficit does indeed have a causal relationshipwith current account.

Finally, our study focuses on Malaysia andhence the results may not be generalized to theother developing countries. Further examinationusing data from other countries may be requiredto understand the twin deficit phenomena indeveloping economies particularly the AsianDeveloping Economies (ADE). We realize theneed for more empirical work in this area ofacademic interest and it is in our next researchagenda.

ACKNOWLEDGEMENT

The authors would like to thank the twoanonymous referees of this journal for providinguseful comments on the earlier draft of thispaper. The findings of this research have beenpresented in APEB (2002) Conference. Thesecond author would like to acknowledge thefinancial supports from MOSTE [IRPA grantsNo: 05-02-04-0532].

REFERENCESAKBOSTANCI, E. and G.I. TUN<;. 2001. Turkish twin

deficits: an error correction model of tradebalance. Economic Research Center (ERC)Working Papers in Economics No. 6.

ALKSWANI, M.A. 2000. The twin deficits phenomenonin petroleum economy: Evidence from SaudiArabia. Presented in Seventh Annual Conference,Economic Research Forum (ERF), 26-29

' October, Amman, Jordan.

ANORUO, E. and S. RAMCHANDER. 1998. Currentaccount and fiscal deficits: evidence from fivedeveloping economies of Asia. Journal of AsianEconomics 9: 487-501.

BACHMAN, D.D. 1992. Why is the US current accountdeficit so large? Evidence from vectorautoregressions. Southern Economic Journal 59:232-240.

BAHARUMSHAH, A.Z. 2001. The effects of exchangerate on bilateral trade balance: new evidencefrom Malaysia and Thailand. Asian EconomicJournal 15: 291-312.

BARRO, RJ. 1974. Are government bonds net wealth?Journal of Political Economy 82: 1095-1117.

BARRO, RJ. 1989. The Ricardian approach to budgetdeficits. Journal of Economic Perspectives 3: 37-54.

BUCHANAN, J.M. 1976. Barro on the Ricardianequivalence theorem. Journal of Political Economy84: 337-342.

CARDIA, E. 1997. Replicating Ricardian equivalencetests with simulated series. American EconomicReview 87: 65-79.

CHINN, M. and E.S. PRASAD. 2000. Medium-termdeterminants of current accounts in industrialand developing countries: An empiricalexploration. NBER Working Paper No. 7581.

DARRAT, A.F. 1988. Have large budget deficits causedrising trade deficits? Southern Economic Journal54: 879-886.

DEMELLO, L.R. 1997. Foreign direct investment indeveloping countries and growth: a selectivesurvey. Journal of Development Studies 34: 1-34.

EDWARDS, S. 2001. Does current account matter?National Bureau of Economic Research(NBER), Working Paper No. 8275.

ENDERS, W. and B-S. LEE. 1990. Current accountand budget deficits: twins or distant cousins?The Review of Economics and Statistics 72: 373-381.

EVANS, P. and I. HASAN. 1994. Are consumersRicardian? Evidence for Canada. QuarterlyReview of Economics and Finance 34: 25-40.

GANDOLFO, G. 1981. Quantitative Analysis andEconometric Estimation of Continuous TimeDynamic. North-Holland Publishing CompanyAmsterdam.

PertanikaJ. Soc. Sci. 8c Hum. Vol. 12 No. 2 2004

On the Twin Deficits Hypothesis: Is Malaysia Different?

GHATAK, A. and S. GHATAK. 1996. Budgetary deficitsand Ricardian equivalence: the case of India,1950-1986. Journal of Public Economics 60: 267-282.

HATEMI, A. and G. SHUKUR. 2002. Multivariate-basedcausality tests of twin deficits in the US. Journalof Applied Statistics 29: 817-824.

HAUG, A. 1996. Blanchard's model of consumption:an empirical study. Journal of Business andEconomic Statistics 14: 169-177.

HUTCHISON, M.M. and C. PIGOTT. 1984. Budget

deficits, exchange rates and current account:theory and U.S. evidence. Federal Reserve Bankof San Francisco Economic Review 4: 5-25.

IBRAHIM, S.B. and F.Y. KUMAH. 1996. Comovements

in budget deficits, money, interest rate,exchange rate and the current account balance:some empirical evidence. Applied Economics 28:117-130.

ISIAM, M. F. 1998. Brazil's twin deficits: an empiricalexamination. Atlantic Economic Journal 26: 121-128.

KHALID, A.M. 1996. Ricardian equivalence: empiricalevidence from developing economies. Journalof Development Economies 51: 413-432.

KHALID, A.M. and W.G. TEO. 1999. Causality tests ofbudget and current account deficits: cross-country comparisons. Empirical Economics 24:389-402.

LAU, E. 2004. The dynamics of the current accountbehavior in the ASEAN-5 countr ies .Unpublished Ph.D thesis, Universiti PutraMalaysia.

LAU, E. and A.Z. BAHARUMSHAH. 2003. Sustainabilityof external imbalances: the case of Malaysia.Singapore Economic Review 48: 61-80.

LEACHMAN, L.L. and B.B. FRANCIS. 2002. Twin deficits:apparition or reality? Applied Economics 34: 1121-1132.

MEGARBANE, P. 2002. Slovakia's external currentaccount deficit: Why so large and is itsustainable? Country Report 210, IMF,Washington D.C.

NORMANDIN, M. 1999. Budget deficit persistenceand the twin deficits hypothesis. Journal ofInternational Economics 49: 171-193.

PIERSANTI, G. 2000. Current account dynamics andexpected future budge t deficits: someinternational evidence. Journal of InternationalMoney and Finance 19: 255-171.

RAMBALDI, A.N. and H. DORAN. 1996. Testing for

granger non-causality in cointegrated systemsmade easy. Department of Econometrics.University of New England. Working Paper inEconometrics and Applied Statistics No. 88.

REISEN, H. 1998. Sustainable and excessive currentaccount deficits. OECD Development Centre.Technical Paper No. 132.

SEATER, J. J. and R.S. MARIANO. 1985. New tests of

the life cycle and tax discounting hypothesis.Journal of Monetary Economics 15: 195-215.

SEATER, J. J. 1993. Ricardian equivalence. Journal ofEconomic Literature 31: 142-190.

SUMMERS, L.H. 1988. Tax policy and internationalcompetitiveness. In International Aspects of FiscalPolicies, ed. J. Frenkel. Chicago: Chicago UP.

TODA, H.Y. 1995. Finite sample performance oflikelihood ratio tests for cointegration ranksvector autoregressive. Econometric Theory 11:1015-1052.

TODA, H.Y, and T. YAMAMOTO. 1995. Statistical

inference in vector autoregressive with possiblyintegrated processes. Journal of Econometric 66:225-250.

VAMVOUKAS, G.A. 1999. The twin deficitsphenomenon: evidence from Greece. AppliedEconomics 31: 1093-1100.

WORLD BANK. 2003. World Development Indicators.World Bank.

(Received: 3 April 2002)

Pertanika J. Soc. Scl 8c Hum. Vol. 12 No. 2 2004 99

Evan Lau & Ahmad Zubaidi Baharumshah

APPENDIX A

Quarterly interpolation of GDP from annual observations

Let us assume that yf y^v yl+] be three consecutive annual observations of continuos flow variables ofyear y(t). In deriving the interpolation formulae, the observed values are actually integrals. Thus, therule of thumb is to integrate the quadratic function in order to obtain the quarterly formulae. Thequarterly formulae after satisfying each of the conditions in any year t are as follows:

j<" = 0.0546875j>M + 0.234375^, - 0.0390625j>,+1 (1)j « = 0.0078125>M + 0.265625j>, - 0.0234375\M (2)<̂3) = 0.0234375^, + 0.265625^ - 0.0078125^ (3)

yj4) = 0.0390625^, + 0.234375j>, - 0.0546875^ (4)

where yt, y^v yt+l are the current, lag and lead values of the variables in question at time t(annual). In other words, three continuous annual observations of variable y(t) are adopted in eachof the equations. In order to calculate the value for the first quarter, we apply the formulae for thefirst quarter and subsequently for the remaining quarters. For example, one may substitute the GDPvalues for ye j M , yt+l in Equation 1 to obtain the calculated, value for the first quarter. One advantageof the interpolation technique is being able to generate the higher frequency data series for the timeseries analysis. Thus, we adopted the Gandolfo (1981) interpolation technique of extracting thequarterly observations based on the annual GDP in this study.

100 Pertanika J. Soc. Sci. & Hum. Vol. 12 No. 2 2004