This article was downloaded by: [University of Haifa Library]On: 05 October 2013, At: 10:16Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

Applied EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/raec20

India's trade balance in the 1980s an econometricanalysisD. M. Nachane & Prasad P. RanadePublished online: 04 Oct 2010.

To cite this article: D. M. Nachane & Prasad P. Ranade (1998) India's trade balance in the 1980s an econometricanalysis, Applied Economics, 30:6, 761-774, DOI: 10.1080/000368498325462

To link to this article: http://dx.doi.org/10.1080/000368498325462

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”)contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensorsmake no representations or warranties whatsoever as to the accuracy, completeness, or suitability for anypurpose of the Content. Any opinions and views expressed in this publication are the opinions and viewsof the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sources of information.Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs,expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly inconnection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Applied Economics, 1998, 30, 761 ± 774

India’s trade balance in the 1980s ±an econometric analysis

D. M. NACHANE* and PRASAD P. RANADE ³

* Department of Economics, University of Bombay, Justice M.G. Ranade Bhavan,V idyanagari Marg, Mumbai 400098, India and ³ Infrastructure L easing and FinancialServices L td, Mahindra T owers, 4th Floor, Dr. G. M. Bhosale Marg, W orli, Mumbai400018, India

The paper is an examination of India’s trade balance in the 1980s. The approachattempted is of examining the bilateral trade balances of India with nine tradingpartners from the non-communist bloc. The long-run equilibrium relations arestudied via two VAR models in Johansen’s multivariate cointegration framework.Five hypotheses of interest are singled out for attention. The nominal and realexchange rates consistently emerge as important in¯ uences on the trade balance.However, as the exchange rates fail weak exogeneity tests, policy implications are notclear cut.

I . INTRODUCTION

India’s travails on the Balance of Payments front com-menced from the beginning of its Second Five Year Plan(1956) and have persisted over the past four decades withvarying degrees of intensity. Following previous studies(Ahluwalia, 1986; Rangarajan, 1990; Jalan 1993), one maydistinguish three chronological epochs.

(a) Period I (1956 ± 76)In this phase, India’s current account de® cit averaged about1.8% of its GDP and this high level of the de® cit wasoccasioned mainly by sluggish export growth in relation toimport requirements. Other aggravating factors were thedroughts of 1965 ± 66 and 1966 ± 67, the external wars withChina and Pakistan and the Oil Shock of 1973. However,® scal conservatism and generous external assistance pre-vented the current account imbalance from spilling overinto a general balance-of-payments crisis.

(b) Period II (1976 ± 79)This brief period was a halcyon period for India’s externalsector, with the current account posting a small surplus of0.6% of GDP. The major contributing factor to this happystate of a� airs was a sharp rise in the in¯ ow on invisibles’account due to remittances from abroad. The trade de® citwas also somewhat lower during this period. However, the

second Oil Shock of 1979 once again put the balance ofpayments situation in jeopardy.

(c) Period III (1979 ± 91)This period was witness to several structural changes in theexternal sector scenario. The trade de® cit burgeoned, espe-cially after 1985; simultaneously, there was a gradual declinein the receipts from invisibles’. As a result, the currentaccount de® cit once again reverted back to the 1.8% (ofGDP) level characterizing Period I. But, in contrast to thesituation prevailing in Period I, external assistance was notforthcoming on facile terms and increasing reliance had tobe placed on external commercial borrowings with a conse-quent steep build-up of foreign debt. The situation wasfurther aggravated by indiscriminate ® scal pro¯ igacy, andmatters had assumed crisis proportions by September 1990,when the IMF stepped in with its bail-out package.

To the foregoing periods, one should logically appenda fourth, beginning July 1991, which saw a newly installedgovernment embark on several bold liberalization measureson the external and domestic fronts. While some analyststend to see in these structural reforms nothing but anextrapolation of policies initiated earlier on, in the mid-1980s, the bulk of opinion seems to favour a view of these asmarking a signi® cant structural break from the past. Thelatter view seems particularly appropriate as regards the

0003 ± 6846 Ó 1998 Routledge 761

Dow

nloa

ded

by [

Uni

vers

ity o

f H

aifa

Lib

rary

] at

10:

16 0

5 O

ctob

er 2

013

1 The other component, invisibles’, is notoriously di� cult to model both because of its volatility and its being subject to a scattered host ofin¯ uences.

Fig. 1. Share of trade in India’s GDPNote: Exports, imports and GDP are in real terms

external sector where reforms of far reaching signi® cancewere instituted (such as convertibility of the Indian Rupeeon the Current Account, liberalizing foreign investment,accelerated scaling down of import duties on capital goodsand intermediates, etc). From an analytical point of view,therefore, it is desirable to study this period separately fromthe others. However, a formal econometric study of this(current) period is ruled out because of the dearth of obser-vation points ± the impact of the policies initiated in 1991have begun to be felt only since 1993 and, additionally, dataon the trade ¯ ows is subject to fairly signi® cant lags.

While from a policy perspective, interest attaches to thebalance of payments as a whole, the state of the currentaccount is usually regarded as being of critical importancefor adjudging the overall health of the external sector. With-in the current account, the trade balance is of predominantimportance in LDCs like India, apart from being analyti-cally the more amenable component.1

In this paper we propose to examine the behaviour ofIndia’s trade balance over the period 1979 to 1991 (PeriodIII above). Whereas, most available studies in this area focuson a country’s total trade balance, our analysis focuses onthe bilateral trade balances between India and its ninemajor non-communist trading partners, namely Australia,Canada, France, Germany, Japan, Netherlands, Sweden,the UK, and the USA. This approach has the advantage ofhighlighting the major in¯ uences on the trade balance in eachtrading partner’s case separately. Since the focus is on longterm relations, we resorted to the multivariate cointegrationframework of Johansen and Juselius (1990). In most cases,data on the relevant variables were available on a monthlybasis and, in these cases, we operated with monthly models.For three countries (UK, France and Australia) data on someof the variables were not forthcoming on a monthly basis,and hence we adopted quarterly models. It would also beuseful to have an indication of the foreign trade stance of theIndian Government over the period of our analyses 1979 ± 91.To this end, we fall back upon the standard measure ofopenness generally invoked in this context, namely

[(exports + imports)/(GDP)] (all variables in real terms)

Fig. 1 graphs the dynamic behaviour of this measure andit clearly shows that the Indian economy opened up sub-stantially during the 1980s. Interest attaches also to thedynamics of the bilateral trade with the nine trading part-ners that we have selected for our analysis. Fig. 2 depicts theratio of the bilateral trade with each partner to India’sGDP. Once again a clear upward trend is evident. Thus, notonly has trade in the aggregate increased (as a proportion ofGDP), but the bilateral trade with each individual partnerhas also registered improvements.

II . THE THEORETICAL FRAMEWORK

The relationship between the exchange rate and the tradebalance is a key issue in international trade theory.

Four distinct schools of thought may be identi® ed.

(a) First, there is the traditional Mundell ± Flemingapproach (Mundell, 1962; Fleming, 1962 but alsoRobinson, 1947; Meade, 1951; Metzler, 1948) whichpostulates that (subject to the ful® lment of theMarshall ± Lerner conditions on export and import elas-ticities) a nominal devaluation will improve the tradebalance in the long run, although the short run e� ectsmay be adverse. This is the famous J-curve hypothesis.

(b) The Absorption Approach (Alexander, 1952; Johnson,1963) also predicts a J-curve type of e� ect of nominaldevaluations on the trade balance but based on ananalytical framework quite distinct from the Mundell ±Fleming approach. The Marshall ± Lerner conditions arereplaced by the condition that the propensity to absorb’is less than unity (see SoÈ dersten, 1990, for a de® nition ofthis propensity). The absorption e� ect may operate viaseveral factors including the terms-of-trade, productione� ects, expenditure switching etc.

(c) In direct contrast to the above two approaches, themonetary approach to the balance of payments’(Dornbusch, 1973; Grubel, 1976; Frenkel and Rod-riguez, 1975) rules out any impact of nominal devalu-ations (whether in the short or the long run) on the tradebalance. Any possible impact of nominal devaluationson the balance of payments is essentially via the capitalaccount, involving only portfolio stock adjustments.

762 D. M. Nachane and P. P. RanadeD

ownl

oade

d by

[U

nive

rsity

of

Hai

fa L

ibra

ry]

at 1

0:16

05

Oct

ober

201

3

Fig. 2. Share of bilateral trade in India’s GDPNotes:1. X-axis denotes time in years.2. The Y -axis plots, in each plate, the ratio of India’s trade with that country to India’s total GDP.

2 A systematic exposition of this viewpoint may be found in Edwards (1995), where empirical and policy implications of the approach arealso examined in detail.

(d) In recent years, neoclassical general equilibrium modelshave shifted the focus from nominal to real exchangerates as the decisive in¯ uence on the trade balance (seeKrueger, 1978; Bautista, 1981; Edwards, 1985, 1995).2

The real exchange rate is de® ned as the relative price oftradables to non-tradables , but since this entity is di� -cult to estimate in practice, it is usually proxied by the

quantity r de® ned as

r = e(p*/p) (1)

where e is the nominal e� ective exchange rate, p the domes-tic consumer price index (intended to represent thedomestic non-tradable prices) and p* the foreign whole-sale price index (representing world prices of tradables).

India’s trade balance in the 1980s 763

Dow

nloa

ded

by [

Uni

vers

ity o

f H

aifa

Lib

rary

] at

10:

16 0

5 O

ctob

er 2

013

3 The goods may be non-traded either due to high transport costs or because of government prohibitions on imports/exports of thesegoods.

The con¯ icting views about the role of exchange rateadjustments in in¯ uencing trade ¯ ows are far from being ofacademic interest only. As noted by Himarios (1985) amongothers, such views impinge in a direct fashion on policyissues. `Getting the exchange rate right’ has proved to bea major item on the agenda of IMF-led stabilization pack-ages. Other related policy issues are, of course, the use ofnominal exchange rates as an anchor for policy credibilityand real exchange-rate targeting.

However, the exchange rate (whether real or nominal) isnot the sole in¯ uence on the trade balance. Several otherfactors enter the picture as well. Of these, the three mostimportant ones are (1) income variables (2) monetary vari-ables and (3) price variables.

(1) Income variables. The impact of a rise in domestic orforeign income levels on a country’s exports and importshas received a great deal of attention in the literature. TheHouthakker ± Magee (1969) hypothesis asserted that theelasticities of a country’s demand for imports (w.r.t. a rise indomestic incomes) and of the demand for the country’sexports (w.r.t. a rise in foreign incomes) are likely to beunequal. The hypothesis has been extensively investigatedby Hooper (1980), Goldstein and Khan (1988) and Marquez(1989). As a consequence of the hypothesis, the trade bal-ance will be a� ected whenever there is an equiproportionaterise in domestic and world incomes ± whether the e� ect willbe favourable or adverse will depend on the relative magni-tudes of the respective elasticities.

A rise in domestic incomes, of itself, will usually increasethe demand for the country’s imports but it is also likely toincrease the country’s exports if the latter have been supplyconstrained. This latter e� ect is likely to be particularlyimportant in LDCs. Thus, the net impact on the tradebalance of a rise in domestic incomes is not certain a priori.A rise in foreign incomes, ceteris paribus, is expected how-ever to lead to an increase in the demand for the country’sexports and thereby to an improvement in the tradebalance.

(2) Monetary variables. Several analysts have noted thepossibility of a real balance e� ect associated with monetaryvariables on the trade balance (Johnson, 1973; Dornbusch,1973; Miles, 1979). A rise in money supply would be per-ceived as an increase in domestic wealth leading to anincrease in imports via a spurt in expenditure. Similarly,a rise in foreign money supply is expected to act as a ® llip forexports. However, the strength of the real balance e� ectwould depend on the ratio of nominal money balances tothe total wealth of the economy.

(3) Price variables. One of the earliest hypotheses relating tothe in¯ uence of price variables on exports and imports is theMarshall ± Lerner condition. However, in this paper, ourfocus being on the trade-balance rather than on exports andimports separately, we do not enter into the issue of testingthis hypothesis. Rather, of more interest to us is the issueoriginally raised by Orcutt (1950), which raises the possibi-lity of trade ¯ ows responding di� erentially to the two com-ponents of the real exchange rate r, namely the nominalexchange rate e and relative price ratio (p*/p). Krueger (1983)attributes this di� erential response to the existence of a classof domestically produced non-traded goods.3 The issue hasattracted considerable attention and con¯ icting empiricalevidence has been posted by Junz and Rhomberg (1973),Wilson and Takacs (1979) and Bahmani-Oskooee (1986).

The foregoing discussion indicates that one may identifythe following six distinct hypotheses as being of policyrelevance and theoretical interest:

H1 : (Neoclassical general equilibrium hypothesis) ± Tradebalance responds favourably in the long run to a de-valuation in the real exchange rate.

H2 : (J-curve hypothesis) ± Trade balance responds favour-ably to a nominal devaluation in the long run (aspredicted both by the Mundell ± Fleming approach andthe Absorption approach). The rejection of this hy-pothesis would favour the monetary approach to thebalance of payments.

H3 : (Houthakker ± Magee e� ect) ± An equiproportionaterise in domestic and foreign incomes does a� ect thetrade balance (although the e� ect does not specifywhether favourably or adversely).

H4 : (Income e� ects) ± Trade balance responds (a) favour-ably to a rise in foreign incomes and (b) adversely toa rise in domestic incomes.

H5 : (Monetary e� ects) ± Trade balance responds (a) favour-ably to a rise in foreign money supply and (b) adverselyto a rise in domestic money supply.

H6 : (Orcutt hypothesis) ± Trade balance responds di� eren-tially to the nominal exchange rate and the relativeprice ratio (of foreign to domestic prices).

III . DATA DESCRIPTION

As indicated earlier, we propose to examine the six hypothe-ses delineated above within a multivariate cointegrationframework. For this purpose, we develop two VAR systems

764 D. M. Nachane and P. P. RanadeD

ownl

oade

d by

[U

nive

rsity

of

Hai

fa L

ibra

ry]

at 1

0:16

05

Oct

ober

201

3

4 For the de® nition of an I (d) process (d a non-negative integer), reference may be made to Engle and Granger (1987).

for each bilateral trade balance. Typically, for India’s tradewith the jth country, we would have the following twosystems

VAR I (T B)j , ln Y , ln Y *j , ln HM, ln HM*j , ln rj

VAR II (T B)j , ln Y , ln Y *j , ln HM, ln HM*j , ln ej , ln (p*j /p)

where(T B)j India’s real trade balance with the jth trading

partnerY India’s industrial production indexY *j the jth country’s industrial production indexHM India’s real reserve moneyHM*j the jth country’s real reserve moneyej nominal bilateral exchange rate between the In-

dian Rupee and the jth country’s currency (ex-pressed as Rupees per unit of jth currency)

rj real bilateral exchange rate between the IndianRupee and the jth currency

(p*j /p) ratio of the jth country’s wholesale price index tothe Indian wholesale price index

Note that all variables, except (T B)j are in log form.

The trade-balance data have been gathered from variousissues of the Direction of Trade Statistics (IMF). These dataon exports and imports are denominated in terms of USdollars. We convert these data into the real trade balance(T B)j by a process of double de¯ ation ± exports are ® rstde¯ ated by an index of the jth country’s nominal exchangerate vis-aÁ -vis the US dollar and then by the jth country’swholesale price index, whereas imports are de¯ ated by anindex of the nominal Rupee rate (vis-aÁ -vis the US dollar) andIndia’s wholesale price index. The use of wholesale priceindices is resorted to, owing to the non-availability of separ-ate unit values indices for exports and imports for eachtrading partner.

As far as the income measures are concerned, the idealones would, of course, be the GDP or GNP. However, ouranalysis is intended to be on a monthly or quarterly basisand since GDP or GNP ® gures for India are available onlyon an annual basis, we resort to the standard practice ofusing the industrial production index as a proxy variable forincome levels. Data on Y and Y *j are obtained from success-ive issues of the International Financial Statistics (IMF)database.

In choosing the monetary base as a variable to measuremonetary in¯ uences, we avoid the problems associated withdi� ering de® nitions of money supply across countries.Nominal reserve money data are once again sourced fromthe International Financial Statistics (IFS) Database andare converted to real terms by de¯ ating via the wholesaleprice index of the country concerned.

The nominal exchange rate is the end-of-period valuegiven by line ae’ of the IFS. The real bilateral exchange rateis given by Equation (1) above.

One ® nal clari® cation pertains to the ratio (p*/p). Eventhough on theoretical grounds, as we had indicated in thediscussion following equation (1), a domestic consumer priceindex is suggested as an appropriate proxy for the nominalprice level, we have chosen the Indian wholesale price indexfor this role. This is so because, in India, there are severaldi� erent consumer price indexes (for industrial workers, foragricultural labourers, for urban non-manual workers etc)and, purely on a priori grounds, it would have been di� cultto choose a particular one as the most appropriate. The IFSdatabase once again provides data on the wholesale priceindices of the various countries.

As indicated in Section I, a monthly analysis has beenundertaken for six of India’s trading partners and a quarter-ly analysis for three. For the latter, the period of analysis hasbeen taken from 1979QI to 1991QIII, whereas for the for-mer it is from January 1979 to September 1991. VAR I hasbeen designed speci® cally for testing H1 and VAR II fortesting H2 and H6 . The remaining hypotheses H3 , H4 andH5 can be tested in either system and we test them in themore general system VAR II.

IV. ECONOM ETRIC FRAMEWORK

The multivariate cointegration framework that we proposeto use here has now come to be established as a standardone for VAR systems. The procedure may be summarized asfollows (see for example, Johansen, 1988; Johansen andJuselius, 1990).

Let Xt be an I(1) vector representing the n-series ofinterest.4 A VAR of length p for Xt would then be of theform

Xt =p

+j = 1

P j Xt ± j + m + e t t = 1, 2, 3, ¼ T (2)

where the P j are matrices of constant coe� cients, m is anintercept (see discussion below), e t is a Gaussian error termand T the total number of observations.

The ECM (Error Correction Model) corresponding toEquation (2) is

D Xt =p

+j = 1

G j D Xt ± j + P Xt ± p + m + e t (3)

where D is the ® rst-di� erence operator and the expres-sions for G j and P are as given in Johansen and Juselius(1990).

India’s trade balance in the 1980s 765

Dow

nloa

ded

by [

Uni

vers

ity o

f H

aifa

Lib

rary

] at

10:

16 0

5 O

ctob

er 2

013

5 The exact form of the restriction when the intercept enters only the ECM may be found in Johansen and Juselius (1990).

If Rank (P ) = r (r < n) then cointegration is indicated(with r cointegrating vectors present) and further, in thiscase P may be factored as P = a b 9 , with the matrix b com-prising the r cointegrating vectors and a can be interpretedas the matrix of corresponding ECM weights.

We can summarize the main steps in our analytical frame-work as follows.

Step 1. First, note that models (2) and (3) can be easilyextended to include a vector Zt of I(0) variables as regressors.Zt could include seasonal dummies as well as other station-ary variables. These I(0) variables pose no additional es-timation problems, since they can be simply concentratedout of the likelihood function (Banerjee et al., 1993).

Step 2. The lag length p of the VAR system in Equation (2)can be decided upon by the likelihood-ratio (LR) test sug-gested by Sims (1980).

De® ne

S = (T - c) [ln| S r| - ln| S u|] (4)

where (i) S r and S u are the determinants of the restricted andunrestricted covariance matrices respectively; (ii) T is thenumber of observations; and (iii) c is the Sims’ correctionfactor de® ned as the total number of variables in the un-restricted system.

Step 3. The next step is to check whether all variables® guring in the VARs are I(1). For this purpose, we use thestandard Dickey ± Fuller, Augmented Dickey ± Fuller andPhillips ± Perron tests. The sequential procedure suggestedby Dolado et al. (1990) and elaborated by Holden andPerman (1994) is followed. Brie¯ y, this procedure comprisesthe following stages. If Xt is the series of interest, we run theregression

D Xt = r Xt ± 1 + a + b t +p

+i = 1

D Xt ± i + e t (5)

with the lag p chosen to render the residuals free of serialcorrelation. We then test the composite null hypothesisH0 : b = 0, r = 0 using the Dickey ± Fuller (1981) statistic/ 3 and its Phillips ± Perron (1988) transform Z (/ 3 ) . If H0 isrejected, there is no unit root and the presence of drift andtrend can be ascertained by conventional t-tests on a andb respectively. (ii) If H90 is not rejected, we re-estimateEquation (5) setting b = 0 and then use the Dickey ± Fullerstatistic t m and its Phillips ± Perron transform, Z (t m ) to testthe hypothesis H 90 : r = 0. (iii) If H 9 0 is favoured, we getadditional con® rmation about the presence of a unit root.We may then resort to the statistics / 2 and Z (/ 2 ) to test thenull hypothesis H 9 90 : a = 0, r = 0. (iv) Rejection of H 9 90

argues for the presence of a unit root with drift, and itsnon-rejection is interpreted as having a unit root withoutdrift.

Step 4. The selection of r (the number of cointegratingvectors) is the next item on the agenda. This decision pro-ceeds via an examination of the maximal eigenvalue (l max)and the trace (JT) statistics of Johansen (1988).

Step 5. The treatment of the intercept term l is quite impor-tant. Two possibilities are present ± namely (i) that theintercept term is restricted to enter the ECM only and (ii)that the intercept term is unrestricted. Johansen andJuselius (1990) furnish the following statistic for discriminat-ing between the two possibilities

L = - Tn

+i = r+ 1

ln[(1 - l à *i )/(1 - l à i)] (6)

where l à i and l à *i denote the eigenvalues (arranged in de-scending order) of the unrestricted and restricted systemsrespectively5 and L is distributed as a x 2 with d.f. (n - r).

It is important to realize that the critical value of thestatistics (l max) and (JT) would depend upon the treatment ofthe intercept term m ± a fact explicitly recognized byOsterwald ± Lenum (1992) who furnishes the requisite criticalvalues for each treatment of m .

Step 6. Once the value of r and the treatment of m have beendecided upon, the Maximum Likelihood Estimates (MLEs)of the cointegrating vectors, a and b , can be derived alongthe lines detailed by Johansen ± Juselius (1990). We adoptthe convention of denoting the variables in VAR I byXi (i = 1, ¼ , 6) in the order in which they have been writtenat the beginning of Section III. (Thus, X1 denotes thebilateral trade balance, X2 India’s industrial productionindex etc.) Similarly, the variables ® guring in VAR II aredenoted by Yi (i = 1, ¼ , 7). The ith cointegrating vector inVAR I is written as b i (i = 1, ¼ , r) with a i denoting thecorresponding error correction weights. b 9i and a 9 I will standfor the corresponding vectors in VAR II. Further, b ik willrepresent the kth element of b i and so on. The cointegratingvector estimates are normalized in each VAR on the ® rstvariable (i.e. the trade balance) so that b à i1 = - b = 9i1 = - 1.

Step 7. As decided above, we propose to test H1 in VARI with VAR II being used for testing H2 to H6 . For a speci® ctrading partner, these hypotheses translate themselves intohypotheses on the various components of the r cointegrat-ing vectors. Thus:

H1 : b i6 > 0 (i = 1, ¼ r) (7)

H2 : b 9i6 > 0 (i = 1, ¼ r) (8)

766 D. M. Nachane and P. P. RanadeD

ownl

oade

d by

[U

nive

rsity

of

Hai

fa L

ibra

ry]

at 1

0:16

05

Oct

ober

201

3

H3 : b 9i2 + b 9i3 ¹ 0 (i = 1, ¼ r) (9)

H4 : (a) b 9i3 > 0 (b) b 9i2 < 0 (i = 1, ¼ r) (10)

H5 : (a) b 9i5 > 0 (b) b 9i4 < 0 (i = 1, ¼ r) (11)

H6 : b 9i6 ¹ b 9i7 (i = 1, ¼ r) (12)

The hypothesis H6 has to be tested by ® rst postulating thenull b 91 6 - b 91 7 = 0, with a rejection of the null being viewedas evidence favouring H6 .

Step 8. The ECM weights a i (and a 9 i) determine the short-term error correction responses of the variables to devi-ations from long-run equilibrium values. A low value of a ik,for example, indicates that the corresponding variable Xk (inVAR I) responds slowly to the disequilibrium term andnon-rejection of the hypothesis a ik = 0 (i = 1, ¼ , r) may beinterpreted as evidence in favour of weak exogeneity’ ofXk to VAR I. (`Weak Exogeneity’ is being used here in itsformal sense as de® ned by Engle et al., 1983).

The crucial question of policy interest is whether the tradebalance can be managed to some extent by an appropriatepolicy of nominal or real exchange rate targeting. This issueis really tied down to four di� erent sets of hypotheses: (a) itrequires the acceptance of H1 in VAR I (favourable impacton trade balance of real exchange depreciation); (b) theacceptance of H2 in VAR II (favourable impact on the tradebalance of a nominal exchange devaluation; (c) weakexogeneity of nominal and real exchange rates implying thefeasibility of monitoring them without interference fromother variables in the system; and (d) absence of weakexogeneity of the trade balance, since if the trade balancewere weakly exogenous, it would be outside the ambit ofin¯ uence of any variables whatsoever in the system.

The above discussion implies that, apart from verifyingH1 and H2 , we should also verify that

a i6 = 0 (i = 1, ¼ , r)

(exogeneity of the real exchange rate in VAR I) (13)

a 9 i6 = 0 (i = 1, ¼ , r)

(exogeneity of the nominal exchange rate in VAR II) (14)

a i1 ¹ 0 and a 9 i1 ¹ 0 (i = 1, ¼ , r)

(non-exogeneity of trade balance in VAR I and VAR II)

(15)

The exogeneity tests have to be based on the LR statistic ofJohansen and Juselius (1990). These tests are based on thelikelihood-ratio (LR). We adopt the convention of denotingthe LR statistic corresponding to a ik = 0 by lk (and similarlythat corresponding to a 9ik = 0 by l9k).

Step 9. One ® nal consequence of the Johansen ± Juseliusmethod is especially noteworthy. The matrix P = a b T

furnishes the long-run impact in VAR I (and a similarmatrix for VAR II may also be written down ± we will call itP *). The (i, j)th entry of P shows the long-run impact onXi of a unit change in Xj .

V . EMPIRICAL RESULTS

As outlined above, this paper attempts to model thebilateral trade balance between India and nine of its majortrading partners (accounting for 65% of its total trade withthe non-communist bloc during the 1980s). A monthly anal-ysis was conducted for six out of the nine countries, andfor the remaining three countries (owing to a paucity ofmonthly data) a quarterly analysis was attempted.

It would be cumbersome to present the details of all thestatistics emanating from our exercise. We have thereforedecided to present detailed statistics for one illustrative case,namely India ± Australia and summarize the qualitative con-clusions of the rest of our calculations in Tables 6 to 10.

T he India± Australia case

Using the Sims’ statistic, a value of lag length p = 3 wasdecided upon for both VAR I and VAR II. The unit rootstatistics are displayed in Table 1, and they clearly indicatethat all the variables ® guring in the two VARs are I (1), sothat the Johansen ± Juselius procedure may be legitimatelyinvoked. The l max and JT statistics for both VARs arepresented in Table 2 and indicate a single cointegratingvector in each case. Our next decision pertains to the treat-ment of the intercept term. The ® rst row of Table 3 showsthat L (as given by Equation (6)) is highly signi® cant in bothVAR I and VAR II. Because of this, the intercept term willbe unrestricted in both ECMs.

We are now in a position to estimate the cointegratingvectors for VAR I and VAR II using the ML method. In thenotation introduced in Section III, we may write

b à 1 = ( - 1, + 0.2846, - 1.4144, + 0.1871, + 0.1932,

+ 0.0743) (16)

b à 91 = ( - 1, - 0.8874, - 6.5038, + 3.8314, - 1.5658,

+ 0.8107,+ 3.5685) (17)

The detailed statistics needed for testing Hi (i = 1, 2, ¼ ,6)are given in Table 3. The coe� cients b à 1 6 , b = 91 6 are positive inEquations (16) and (17) and the associated LR statistics inRows 2 and 3 of Table 3 are signi® cant, lending credence toH1 and H2 . From the same table, the LR statistic for the nullof (b 9i2 + b 9i3 = 0) is signi® cant, so that H3 stands its groundtoo. As far as H4 is concerned, the ® rst part of the hypothesisis rejected since b à 9i3 is insigni® cant (and in addition has thewrong sign in Equation (17)). The other part of the hypo-thesis is accepted however, H5 may be rejected outright,

India’s trade balance in the 1980s 767

Dow

nloa

ded

by [

Uni

vers

ity o

f H

aifa

Lib

rary

] at

10:

16 0

5 O

ctob

er 2

013

Table 1. Unit root tests (India± Australia case)

Variable / 3 Z (/ 3 ) t m Z (t m ) / 2 Z(/ 2 ) Inference

X1 4.81 5.97 - 2.13 - 1.93 5.72* 5.98* Unit root with driftD X1 8.05* 9.37** Ð Ð Ð Ð StationaryX2 5.11 6.04 - 1.43 - 1.67 11.18** 13.27** Unit root with driftD X2 5.97 6.51 - 3.41* - 3.66** Ð Ð StationaryX3 5.88 6.23 - 1.97 - 2.38 4.59 5.27* Unit root with

possible driftD X3 6.16 6.82* - 3.12* - 3.37* Ð Ð StationaryX4 3.34 4.12 - 2.66 - 2.31 6.78* 7.13* Unit root with driftD X4 7.83* 7.99* Ð Ð Ð Ð StationaryX5 4.63 5.90 - 1.87 - 2.18 4.43 4.92 Unit root without drift

D X5 9.07* 9.59** Ð Ð Ð Ð StationaryX6 6.17 6.69 - 2.49 - 2.07 5.72* 5.81* Unit root with driftD X6 12.53** 13.06** Ð Ð Ð Ð StationaryY 6 2.08 3.19 - 1.06 - 1.67 4.39 4.86 Unit root without

driftD X6 5.37 6.10 - 2.98* - 3.20* Ð Ð StationaryY 7 4.43 4.87 - 2.17 - 2.55 5.73* 6.11* Unit root with driftD Y 7 7.79* 8.41* Ð Ð Ð Ð Stationary

Notes:(i) Our terminology de® nes the variables in VAR I as Xi, i = 1, 2, ¼ , 6 and the variables in VAR II as Y i, i = 1, ¼ , 7.

However, Xi = Y i, i = 1, 2, ¼ , 5 (i.e. the ® rst ® ve variables in both VARs are identical). Hence, the variables we considerare Xi, i = 1, 2, ¼ , 5 and Y 6 , Y 7 .

(ii) Our sample size is 55 and hence the critical values that we employ are those given for a sample size of 50. The values fort m are from Fuller (1976) Table 8.5.2 and for / 2 and / 3 from Dickey and Fuller (1981) Tables V and VI respectively.

(iii) (*) and (**) denote 5% and 1% levels of signi® cance respectively.

Table 2. Determination of r in VAR I and VAR II (India± Australia)

VAR I VAR II VAR I VAR II

95% 95% 95% 95%Null Alternate critical critical critical criticalhypothesis hypothesis l max value l max value JT value JT value

r = 0 r = 1 36.6197 39.3720 50.4091 45.2770 Ð Ð Ð Ðr < 1 r = 2 21.7245 33.4610 36.5258 39.3720 Ð Ð Ð Ðr = 0 r > 1 Ð Ð Ð Ð 100.3914 94.1550 155.3551 124.2430r < 1 r > 2 Ð Ð Ð Ð 61.3717 68.5240 92.9460 94.1550

Note: The table is divided into two parts, since the hypotheses for l max and JT are slightly di� erent. The 95% critical values for VAR I andVAR II di� er since the number of variables ® guring in each are di� erent.

6 Note that the exchange rates are measured in logs, whereas the trade balance is in original units.

whereas H6 is supported by the data. Table 3 also throwslight on the weak exogeneity issues of policy interest. Thetrade balance is not exogenous in either of the VARs and thenominal exchange rate is not exogenous in VAR II whereasthe real exchange rate fails to be exogenous in VAR I.

Our ® nal point of interest pertains to the long-run impactmultipliers displayed in Tables 4 and 5. Two impact multi-pliers are especially relevant. From Table 4 we infer thata 1% change in the real exchange rate produces an absolutechange of 2.20 in the real trade balance (VAR I), and a 1%

change in the nominal exchange rate leads to a change of0.23 absolute units in the real trade balance (VAR II).6

Other results

It is not necessary to present results for the other cases in thesame detail as was done in the India ± Australia case, sincethe general pattern would now be familiar. Accordingly,Table 6 presents only the basic features of the VAR modelsin all nine cases. Two features are noteworthy in the table.

768 D. M. Nachane and P. P. RanadeD

ownl

oade

d by

[U

nive

rsity

of

Hai

fa L

ibra

ry]

at 1

0:16

05

Oct

ober

201

3

Table 3. Hypotheses tests (India± Australia)

Observed Distribution Observed Distributionvalue and d.f. value and d.f.

Purpose Statistic VAR I VAR I VAR II VAR II

Treatment of intercept L as given by (6) 18.1239* x 2 (5) 20.2945* x 2 (6)Testing H1 b i6 = 0 16.21** x 2 (1) Ð ÐTesting H2 b 9i6 = 0 Ð Ð 7.74** x 2 (1)Testing H3 b 9i2 + b 9i3 = 0 Ð Ð 9.81** x 2 (1)Testing H4 b 9i3 = 0 Ð Ð 2.09 x 2 (1)

b 9i2 = 0 Ð Ð 5.86* x 2 (1)Testing H5 b 9i5 = 0 Ð Ð 0.93 x 2 (1)

b 9i4 = 0 Ð Ð 1.66 x 2 (1)Testing H6 b 9i6 - b 9i7 = 0 Ð Ð 7.82* x 2 (1)Testing exogeneity of

Trade Balance a i1 = 0 37.12** x 2 (1) Ð Ða 9 i1 = 0 Ð Ð 82.10** x 2 (1)

Testing exogeneity of realrate exchange a i6 = 0 32.80** x 2 (1) Ð Ð

Testing exogeneity ofnominal exchange rate a 9 i6 = 0 Ð Ð 16.44 x 2 (1)

Notes:(i) The d.fs. in the ® rst row of the table are determined as (n - r), where n is the number of variables in the VAR and r the number of

cointegrating vectors.(ii) (*) and (**) denote signi® cance at 5% and 1% levels respectively.

Table 4. L ong run impact matrix P for VAR I (India± Australia)

X1 X2 X3 X4 X5 X6

X1 - 1.2464 - 1.7629 0.3547 0.2408 0.0926 2.2026X2 0.0849 0.1202 - 0.0242 - 0.0164 - 0.0063 - 0.0159X3 - 0.0178 - 0.0253 0.0051 0.0035 0.0013 0.0033X4 0.0161 0.0228 - 0.0046 - 0.0031 - 0.0012 - 0.0030X5 0.0019 0.0028 - 0.0005 - 0.0004 - 0.0001 - 0.0003X6 0.0319 0.0451 - 0.0091 - 0.0062 - 0.0024 - 0.0059

Note: In the above matrix the (i, j)th entry shows the impact on Xi of a 1% change in Xj . The impact isabsolute if Xi is measured in original units and in % terms if Xi is in logs.

Table 5. L ong run impact matrix P for VAR II (India± Australia)

Y 1 Y 2 Y 3 Y 4 Y 5 Y 6 Y 7

Y 1 - 0.6183 - 4.0216 - 0.5487 - 0.9682 0.5013 0.2332 2.3691Y 2 0.0535 0.3482 0.0475 0.0838 - 0.0433 - 0.1910 - 0.2051Y 3 - 0.0907 - 0.5898 - 0.0805 - 0.1420 0.0735 0.3236 0.3474Y 4 - 0.0207 - 0.1344 - 0.0183 - 0.0324 0.0166 0.0737 0.0792Y 5 0.0375 0.2439 0.0332 0.0587 - 0.0304 - 0.1339 - 0.1437Y 6 - 0.0255 - 0.1658 - 0.0226 - 0.0399 0.0207 0.0909 0.0977Y 7 0.0393 0.2558 0.0349 0.0616 - 0.0319 - 0.1404 - 0.1508

Note: In the above matrix the (i, j)th entry shows the impact on Y i of a 1% change in Y j . The impact is absolute if Y i ismeasured in original units and in % terms if Y i is in logs.

India’s trade balance in the 1980s 769

Dow

nloa

ded

by [

Uni

vers

ity o

f H

aifa

Lib

rary

] at

10:

16 0

5 O

ctob

er 2

013

Table 6. Summary features of VAR models

Lag Treatment Number ofModel Data length of intercept cointegrating

Country-pair type frequency p vectors r

India± Australia Var I Quarterly 3 Unrestricted r = 1Var II Quarterly 3 Unrestricted r = 1

India± Canada Var I Monthly 4 Restricted r = 1Var II Monthly 3 Unrestricted r = 1

India± France Var I Quarterly 3 Unrestricted r = 1Var II Quarterly 2 Unrestricted r = 1

India± Germany Var I Monthly 3 Unrestricted r = 1Var II Monthly 3 Unrestricted r = 2

India± Japan Var I Monthly 2 Unrestricted r = 1Var II Monthly 2 Unrestricted r = 1

India± The Netherlands Var I Monthly 4 Restricted r = 1Var II Monthly 3 Unrestricted r = 1

India± Sweden Var I Monthly 3 Unrestricted r = 0Var II Monthly 3 Unrestricted r = 0

India± UK Var I Quarterly 2 Unrestricted r = 1Var II Quarterly 1 Unrestricted r = 1

India± US Var I Monthly 4 Unrestricted r = 1Var II Monthly 3 Unrestricted r = 1

Table 7. Cointegrating vectors for VAR I and VAR II

Number ofModel cointegrating

Country-pair type vectors Cointegrating vectors (normalized on the trade balance)

India± Australia Var I 1 b à 1 = [ - 1, 0.2846, - 1.4144, 0.1871, 0.1932, 0.0743]Var II 1 b à 91 = [ - 1, - 0.8874, - 6.5038, 3.8314, - 1.5658, 0.8107, 3.5685]

India± Canada Var I 1 b à 1 = [ - 1, 0.8770, - 0.5876, - 1.0656, 3.2152, 0.2587]; intercept + 4.1759Var II 1 b à 91 = [ - 1, 0.3125, - 0.9645, 0.8505, 1.4772, 0.4617, 0.5861]

India± France Var I 1 b à 1 = [ - 1, - 15.3853, 196.0921, - 2.2837, - 162.6952, 21.9763]Var II 1 b à 91 = [ - 1, 36.8434, 5.7687, - 35.3357, - 20.5732, 6.6256, 10.9939]

India± Germany Var I 1 b à 1 = [ - 1, 3.2477, - 12.7713, - 2.5966, 8.9064, 4.3176]Var II 2 b à 91 = [ - 1, 1.8357, 17.8628, - 6.1417, 14.1771, 18.7442, 13.2675]

b à 92 = [ - 1, 3.2617, 1.9975, - 4.5476, 0.4183, 3.5647, 1.3795]

India± Japan Var I 1 b à 91 = [ - 1, 326.2957, 795.5333, - 2027.1, 1431.2, 101.3936]Var II 1 b à 91 = [ - 1, 333.3189, 804.1608, - 2177.2, 1627.1, 104.7768, 147.5062]

India± Var I 1 b à 1 = [ - 1, - 2.2815, - 7.1322, 3.2590, - 0.1852, 0.7815]; intercept + 41.0744The Netherlands Var II 1 b à 91 = [ - 1, 0.8728, - 2.7899, - 1.8867, 1.4814, 2.0636, 1.6117]

India± UK Var 1 1 b à 1 = [ - 1, 8.3331, 4.3390, - 8.9502, 3.2043, 0.3215]Var II 1 b à 91 = [ - 1, 2.5110, 2.1796, - 3.6694, 3.9578, 1.2630, 4.4975]

India± US Var I 1 b à 1 = [ - 1, 3.6155, 1.5638, - 7.2095, 7.8805, 3.0774]Var II 1 b à 91 = [ - 1, 1.4662, 3.5810, - 0.5576, 2.8099, 2.8220, 6.7020]

Note: The India ± Sweden case does not ® gure above, since there are no cointegrating vectors

(i) First, the number of cointegrating vectors in most casesturns out to be one, which facilitates interpretationconsiderably. In the India ± Sweden case, there are nocointegrating vectors, pointing to an absence of any

long-term equilibrium relations in this case. TheIndia ± Sweden case is then dropped from the subsequentanalysis. Secondly, except in two cases, the interceptterm emerges as unrestricted. Separate tables of signi® cance

770 D. M. Nachane and P. P. RanadeD

ownl

oade

d by

[U

nive

rsity

of

Hai

fa L

ibra

ry]

at 1

0:16

05

Oct

ober

201

3

Table 8. L R statistics for null hypotheses on elements on b i and b 9i

Country pair Null hypothesis

b i6 = 0 b 9i6 = 0 b 9i2 + b 9i3 = 0 b 9i3 = 0 b 9i2 = 0 b 9i5 = 0 b 9i4 = 0 b 9i6 = b 9 i7

India± Australia 16.21** 7.74** 9.81** 2.09 5.86* 0.93 1.66 7.82*India± Canada 0.13 7.031** 8.63** 18.64** 5.16* 28.61** 25.03** 1.08India± France 5.78* 3.16 18.04** 0.07 19.53** 4.48* 19.52** 8.26**India± Germany 18.76** 11.42** 16.83** 6.24* 7.42* 34.08** 7.74* 2.13India± Japan 4.69* 4.82* 34.73** 7.93** 7.69** 14.83** 18.12** 0.88India± The Netherlands 21.55** 8.24** 14.91** 2.85 4.04* 3.72 6.94** 1.47India± UK 20.78** 3.63 16.29** 3.13 2.25 12.94** 4.21* 2.65India± US 31.79** 15.89** 25.43** 2.74 0.94 0.57 0.01 14.73**

Notes:1) (*) and (**) denote signi® cance at 5% and 1% levels respectively.2) All LR statistics are distributed as x 2 with 1 d.f. except the statistics on VAR II in the India ± Germany case which are distributed as

x 2 with 2 d.f. These latter statistics have been italicized for convenience.

Table 9. Qualitative summary of main results

India’s trading partners for which:

Hypothesis Hypothesis is not rejected Hypothesis is rejected

H1 Australia, France, Germany, Japan, CanadaNetherlands, UK, US

H2 Australia, Canada, Germany, Japan, FranceNetherlands, UK, US

H3 Adverse: Australia, Canada and Netherlands NilFavourable : France, Germany, Japan, UK, US

H4 (a) Germany, Japan Australia, France, Netherlands, UK, US, Canada

(b) Australia Canada, France, Germany, UK, US, Japan, Netherlands

H5 (a) Canada, Germany, Japan, UK Australia, Netherlands, US, France

(b) France, Germany, Japan, Netherlands, UK Australia, US, Canada

H6 Australia, France, US Canada, Germany, Japan, Netherlands, UK

Notes:(1) For a description of the hypotheses, see text (Section II).(2) For H3 we distinguish between the cases where an equiproportionate rise in domestic and foreign incomes a� ects the trade balance

adversely and favourably.

are used for the restricted and unrestricted cases (Oster-wald-Lenum, 1992).

The actual cointegrating vectors are presented in Table 7and the likelihood-ratio statistics for tests for various nullhypotheses are presented in Table 8. These null hypotheseshave been selected with a view to fu rnishing evidence Hi:(i = 1, 2, ¼ , 6) ± our economic hypotheses of interest. (Theeconomic hypotheses themselves may not be directly test-able in the Johansen cointegrating framework but have tobe adjudged by combining the evidence on the null hypothe-ses in Table 8 with that on the signs of the cointegratingvectors in Table 7.)

Table 9 presents the qualitative evidence on the economichypotheses of interest to us, and the long term impacts of

the exchange rate variables on the trade balance are present-ed in Table 10.

One ® nal point deserves to be noted regarding theweak exogeneity question. We tested the weak exogeneity ofthe trade-balance in both VAR I and VAR II (for alleight pairs of countries) and found that weak exogeneitywas rejected in all cases. Similarly, there was outrightrejection of weak exogeneity of the real exchange rate(in VAR I) and of the nominal exchange rate in (VAR II)for all pairs of countries studied. Since weak exogeneitywas rejected across the board, we thought there was nopoint in presenting all the LR statistics for the weakexogeneity hypotheses.

India’s trade balance in the 1980s 771

Dow

nloa

ded

by [

Uni

vers

ity o

f H

aifa

Lib

rary

] at

10:

16 0

5 O

ctob

er 2

013

Table 10. L ong run impact of exchange rate variables on the realtrade balance

Real NominalCountry-pair exchange rate exchange rate

India± Australia 2.20 0.23India± Canada 0.29 0.01India± France 2.16 0.35India± Germany 2.64 0.28India± Japan 0.0056 0.000027India± The Netherlands 1.24 0.06India± UK 1.22 0.36India± US 1.02 0.45

Notes:(i) The ® gures show the absolute changes in the real bilateral

trade balance corresponding to a 1% change in the (bilateral)exchange rate variable.

(ii) The table is based on the long-run impact matrices P (of thetype presented in Tables 4 and 5 for the India ± Australia case).

VI . CONCLUSIONS

We are now in a position to gather our main conclusions.But before that, it may be well to emphasize what we believeare the major contributions of this study. Several of thehypotheses that we have examined have been investigatedempirically before (although not for India). But they havetaken into account the aggregate trade balance, and theanalysis is usually with single-equation econometric modelsbased on annual data. Our study, in contrast, breaks newground by focusing on bilateral trade balances betweenIndia and its major trading partners. This shift of emphasisis important because the factors in¯ uencing trade ¯ owsdi� er substantially across India’s trading partners. Thetari� s and quota restrictions for example (in the 1980sbefore regional trading blocs arrived on the scene) varyconsiderably between the European Countries and NorthAmerica because of di� erences in the Multi-® bre Agree-ment (MFA) provisions, Generalized Scheme of Preferences(GSP) etc. Another major improvement e� ected by ourstudy over its predecessors is the use of the multivariatecointegration framework (Johansen, 1988; Johansen andJuselius, 1990). This framework permits the highlighting oflong-term equilibrium relationships and avoids several es-timation and inferential problems which usually plaguenon-stationary data in single-equation models. Addition-ally, we have introduced greater degrees of freedomin our analysis by resorting to monthly (and occasionallyquarterly) data rather than annual data as used in mostearlier studies. It is hoped that these features imbueour conclusions with a greater degree of reliability as well as¯ exibility.

One of the major and robust conclusions of our studypertains to the e� ective role played by the exchange ratein¯ uencing the trade balance. Both the real and nominal

exchange rates enter all the cointegrating vectors with thecorrect sign (see Table 7). Further, the real exchange rate issigni® cant in all cases except Canada (see Table 8, ® rstcolumn) and the nominal exchange rate in all cases exceptFrance (see Table 8, second column). Thus, both the neo-classical viewpoint about the role played by the real ex-change rate and the J-curve hypothesis about nominaldepreciations derive unstinted support from our exercise.The Houthakker ± Magee hypothesis (H3 ) is accepted for allthe countries studied (except of course Sweden, for which nolong-term analysis was possible due to the absence of coin-tegrating relations). The hypothesis however merely predi-cates that an equiproportionate rise in domestic and foreignincomes does not leave the trade balance unaltered, butdoes not specify whether such a change a� ects the tradebalance favourably or adversely. We ® nd that for three ofIndia’s trading partners the net impact of such a change isadverse, whereas for the remaining ® ve it is favourable.

Hypothesis H4 is concerned with capturing the impact onthe trade balance of the income variables. A rise in foreignincomes provides a stimulus to exports and thereby contri-butes to an improved trade balance. This is the ® rst part ofour hypothesis, namely H4 (a). This hypothesis is rejected ina majority of cases (the exceptions being Germany andJapan). The rejection is probably attributable to the quotalimitations faced by India’s major exports (textiles andclothing) in these countries. However, it could also arisebecause of the low value addition of some of India’s tradi-tional exports with purchasers switching over to highervalue added goods from India’s competitors as their (pur-chasers’) incomes rose. The second part of H4 (namelyH4 (b)) considers the impact of a rise in domestic incomes.This rise increases imports via the conventional expendituree� ect, but in the case of an LDC like India, exports facevarious supply bottlenecks, which may also be eased by therise in domestic incomes. Economic theory does not predictwhich e� ect is likely to predominate but, for the sake ofspeci® city, we set up the hypothesis as a rise in domesticincomes adversely a� ecting the trade balance (i.e. the im-port e� ect swamping the export e� ect). This hypothesis isaccepted only for the Australian case. In the case of the UKand the US the corresponding coe� cient is insigni® cant, butfor the remaining ® ve countries (i.e. Canada, France, Ger-many, Japan and the Netherlands) a rise in India’s incomesstimulates exports more than imports ± a phenomenonunderlining both the presence of import controls (high tar-i� s and quota restrictions on several categories of imports,especially consumer goods), as well as supply bottlenecks inthe export sector.

The in¯ uence of monetary variables in the trade balanceis sought to be encapsulated in hypothesis H5 . On theore-tical grounds, an expansionary monetary stance in a foreigncountry is re¯ ected in an improvement in India’s tradebalance with that country (H5 (a)) whereas a domesticmoney supply expansion is expected to produce the

772 D. M. Nachane and P. P. RanadeD

ownl

oade

d by

[U

nive

rsity

of

Hai

fa L

ibra

ry]

at 1

0:16

05

Oct

ober

201

3

contrary e� ect (H6 (b)). Both components of the hypothesisget strong support from our exercise (see Table 8). Of thefour countries which fail H5 (a), only France does so witha signi® cant and wrong sign (the remaining three fail thehypothesis by virtue of insigni® cant coe� cients) and, in thecase of H5 (b), only Canada, among the three rejections,exhibits a signi® cant perverse e� ect. Thus, by and large, ourresults do indicate an important role for monetary variablesin in¯ uencing the trade balance.

The Orcutt hypothesis (H6 ) is not strongly favoured byour analysis. For ® ve out of the eight countries consideredthere seems to be no substantial di� erence between theimpact of a nominal devaluation and of a relative pricechange on the trade balance.

A ® nal word ought to be said about the results of ourweak exogeneity tests. As mentioned earlier, weak exo-geneity is rejected throughout for the three variables ± tradebalance, nominal exchange rate and the real exchange rate.Taken in conjunction with the results of our hypothesesH1 and H2 , this may be taken to imply that even thoughexchange rates (both real and nominal) can in¯ uence thetrade balance signi® cantly, there may be constraints inmanipulating them freely as policy variables. The latterobservation accords well with the experience of the ReserveBank of India, which has found it a daunting task to e� ectdeadbeat targeting of either the real or the nominal ex-change rate of the Rupee.

Our study is certainly not without its limitations. First,a signi® cant portion of India’s trade during the early 1980s(about 23%) was with the Soviet and East European bloc.However, this trade was considerably out of alignment withmarket signals and would have been di� cult to analyse withmodels of the type discussed here. Secondly, even within thegroup of non-communist trading partners, our selection ofcountries is far from complete. A few important tradingpartners in Asia and Europe had to be omitted because ofa lack of availability of data for the relevant variables ona consistent time-series basis. (The included group of coun-tries accounted for 55% of India’s total trade and for 65% oftrade with the non-communist world.) Thirdly, our analysisdoes not permit reliable inferences about the trade possi-bilities in the post-1991 liberalization scenario, owing toa paucity of data points.

Subject to the above limitations, it is hoped that ouranalysis has satisfactorily worked out the testable implica-tions of several interesting hypotheses and thereby contri-buted to some understanding of the web of complex issuesenmeshing India’s foreign trade sector.

ACKNOWLEDGEMENTS

The authors wish to record their appreciation of KarlLopez, Neeraj Hatekar and Ms. R. Lakshmi who helpedin the formidable tasks of data collection and software

development. Errors and omissions are the authors’ soleresponsibility. An anonymous referee’s suggestions provedmost valuable and are gratefully acknowledged.

REFERENCES

Ahluwalia, M. S. (1986) Balance of payments adjustments in India,1970 ± 71 to 1983 ± 84, W orld Development, 14, 937 ± 62.

Alexander, S. S. (1952) E� ects of devaluation on a trade balance,IMF Sta� Papers, 2, 263 ± 78.

Bahmani-Oskooee, M. (1986) Determinants of international trade¯ ows: the case of developing countries, Journal of Develop-ment Economics, 20 (1), 107 ± 23.

Banerjee, A., Dolado, J. J., Galbraith, J. W. and Hendry, D. F.(1993) Cointegration, Error Correction and the EconometricAnalysis of Non-Stationary Data, Oxford University Press,Oxford.

Bautista, R. M. (1981) Exchange rate changes and LDC exportpreferences under generalised currency ¯ oating, W elt-wirtscha¯ iches Archiv, 117, 443 ± 68.

Dickey, D. A. and Fuller, W. A. (1981) Likelihood ratio statisticsfor autoregressive time series with a unit root, Econometrica ,49, 1057 ± 72.

Dolado, J. J., Jenkinson, T. and Sosvilla-Rivero, S. (1990)Cointegration and unit roots, Journal of Economic Surveys, 4,249 ± 73.

Dornbusch, R. (1973) Devaluation, money and non-traded goods,American Economic Review, 63, 871 ± 80.

Edwards, S. (1985) Stabilisation with liberalisation and evaluationof ten years of experiment with free market policies± 1973 ± 1983, Economic Development and Cultural Change, 32(1), 223 ± 54.

Edwards, S. (1995) Capital Controls, Exchange Rates and Mone-tary Policy in the W orld, Cambridge University Press,Cambridge.

Engle, R. F. and Granger, C. W. J. (1987) Cointegration and errorcorrection: representation, estimation and testing, Econo-metrica, 55, 251 ± 76.

Engle, R. F., Hendry, D. F. and Richard, J. F. (1983) Exogeneity,Econometrica , 51, 227 ± 304.

Fleming, M. J. (1962) Domestic ® nancial policies under ® xed andunder ¯ oating exchange rates, IMF Sta� Papers, 9, 369 ± 80.

Frenkel, J. A. and Rodriguez, C. A. (1975) Portfolio equilibriumand the balance of payments: a monetary approach, AmericanEconomic Review, 65, 674 ± 88.

Fuller, W. A. (1976) Introduction to Statistical T ime Series, Wiley,New York.

Goldstein, M. and Khan, M. S. (1988) Income and price e� ects inforeign trade, in P. B. Kenen (Ed.) Handbook of InternationalEconomics, Vol. 2, North-Holland, Amsterdam.

Grubel, H. G. (1976) Domestic Origins of the Monetary Approach tothe Balance of Payments, Essays in International Finance, No.117, Princeton University.

Himarios, D. (1985) The e� ects of devaluation on the tradebalance: a critical view and reexamination of Miles’ newresults’, Journal of International Money and Finance, 4,553 ± 63.

Holden, D. and Perman, R. (1994) Unit roots and cointegration forthe economist, in B. B. Rao (Ed.), Cointegration for the AppliedEconomist, St. Martin’s Press, New York.

Hooper, P. (1980) The stability of income and price elasticities inU.S. Trade 1957 ± 1977, International Finance Division PaperNo. 119, Board of Governors of the Federal Reserve System.

India’s trade balance in the 1980s 773

Dow

nloa

ded

by [

Uni

vers

ity o

f H

aifa

Lib

rary

] at

10:

16 0

5 O

ctob

er 2

013

Houthakker, H. S. and Magee, P. (1969) Income and price elastici-ties in world trade, T he Review of Economics and Statistics, 51,111 ± 28.

International Monetary Fund (various issues) Direction of T radeStatistics.

International Monetary Fund (various issues) International Finan-cial Statistics.

Jalan, B. (Ed) (1993) T he Indian Economy: Problems and Prospects,Penguin Books, New Delhi, India.

Johansen, S. (1988) Statistical analysis of cointegrating vectors,Journal of Economic Dynamics and Control, 12, 231 ± 54.

Johansen, S. and Juselius, K. (1990) Maximum likelihood estima-tion and inference on cointegration ± with applications to thedemand for money, Oxford Bulletin of Economics and Statis-tics, 52, 169 ± 210.

Johnson, H. G. (1963) Recent developments in monetary theory,Indian Economic Review, 6, 29 ± 69.

Johnson, H. G. (1973) The monetary approach to the balanceof payments, in M. Connoly and A. Swoboda (Eds) Interna-tional T rade and Money, University of Toronto Press,Toronto.

Junz, H. and Rhomberg, R. (1973) Price competitiveness in exporttrade among industrial countries, American Economic Review,Papers and Proceedings, 63, 412 ± 18.

Krueger, A. O. (1978) Foreign T rade Regimes and Economic Devel-opment: L iberalization Attempts and Consequences, Ballinger,Cambridge.

Krueger, A. O. (1983) T rade and Employment in Developing Coun-tries, Vol. 3, University of Chicago Press, Chicago.

Marquez, J. (1989) Income and price elasticities of foreign trade¯ ows: econometric estimation and analysis of the U.S. tradede® cit, in L. R. Klein and J. Marquez (Eds), Economics inT heory and Practice: An Eclectic Approach, Kluwer AcademicPublishers, Amsterdam.

Meade, J. E. (1951) T he Balance of Payments, Oxford UniversityPress, Oxford.

Metzler, L. (1948) The theory of international trade, in H. S. Ellis(Ed.) A Survey of Contemporary Economics, Vol. 1, Blakiston,Philadelphia.

Miles, M. A. (1979) The e� ects of devaluation on the trade balanceand the balance of payments; some new results, Journal ofPolitical Economy, 87, 600 ± 20.

Mundell, R. A. (1962) The appropriate use of monetary and ® scalpolicy for internal and external stability, IMF Sta� Papers, 9,70 ± 79.

Orcutt, G. H. (1950) Measurement of price elasticities in interna-tional trade, Review of Economics and Statistics, 32, 117 ± 32.

Osterwald-Lenum, M. (1992) A note with fractiles of the asymp-totic distribution of the maximum likelihood cointegrationrank test statistics: four cases, Oxford Bulletin of Economicsand Statistics, 54, 461 ± 72.

Perron, P. (1988) Trends and random walks in macro-economictime series: further evidence from a new approach, Journal ofEconomic Dynamics and Control, 12, 297 ± 332.

Pesaran, M. H. and Pesaran B (1991) Micro® t 3.0: An InteractiveEconometric Software Package, Oxford University Press, Ox-ford.

Phillips, P. C. B. and Perron, P. (1988) Testing for a unit root intime series regression, Biometrika, 75, 335 ± 46.

Rangarajan, C. (1990) The balance of payments scenario, SeventhG. L. Mehta Memorial Lecture, The Indian Institute of Tech-nology, Mumbai.

Robinson, J. (1947) The foreign exchanges, in H. S. Ellis and L. A.Metzler (Eds) Readings in the T heory of International T rade,Allen and Unwin, Philadelphia.

Sims, C. A. (1980) Macroeconomics and reality, Econometrica , 48(1), 1 ± 48.

SoÈ dersten, B. (1990) International Economics, Macmillan, London.Wilson, J. F. and Tackacs, W. E. (1979) Di� erential responses to

price and exchange rate in¯ uences in the foreign trade ofselected industrial countries, Review of Economics And Statis-tics, 61 (2), 267 ± 79.

774 D. M. Nachane and P. P. RanadeD

ownl

oade

d by

[U

nive

rsity

of

Hai

fa L

ibra

ry]

at 1

0:16

05

Oct

ober

201

3