You are on page 1of 26

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/241767836

Exports, Imports and Economic Growth in India

Article

CITATIONS READS

5 6,677

2 authors, including:

Laszlo Konya
Independent
60 PUBLICATIONS 718 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Unemployment Rates Converge in the EU View project

All content following this page was uploaded by Laszlo Konya on 22 May 2014.

The user has requested enhancement of the downloaded file.


Exports, Imports and Economic Growth in India.

László Kónya
and
Jai Pal Singh

Discussion Paper No. A06.06


ISBN 1 92094 8112
ISSN 1441 3213 December 2006
Exports, Imports and Economic Growth in India

László Kónya1 and Jai Pal Singh2


1
Department of Economics and Finance
La Trobe University, Melbourne, Australia
2
Department of Business Management
CCS Haryana Agricultural University, Hisar, India

December 2006

Abstract

With the advent of WTO, India entered into the era of trade reforms in 1991 and is moving
gradually towards an open economy. It is widely believed that export and import growth is
crucial in providing the impetus for economic growth in developing countries and imports
provide the important 'virtuous' link between trade and output growth. Therefore, our aim,
here, is to address the export/import-led growth and growth-driven export/import hypotheses
for India. In spite of some ambiguity, the results clearly show that exports and imports
Granger-cause GDP, both individually and jointly, lending support to the export/import led
growth hypotheses. There is also some indication of GDP and exports jointly Granger-
causing imports, and GDP and imports jointly Granger-causing exports, but the growth
driven export/import hypotheses seem implausible. A possible reason for the results is the
favourable trade environment of India.

JEL Classification Numbers: C22, C32, F14, F43, O11


Key Words: Export, Import, Economic Growth, India, Unit root, Cointegration, Causality
1. Introduction

Since the early 1960s both policy makers and academics have shown great interest in
exploring the possible relationship between international trade and economic growth. The
reason is obvious. Nations are concerned about improving the quality of life of their
countrymen, which mainly comes from overall, i.e. macroeconomic, development in a highly
competitive and globalised world. Thus, creating wealth, increasing GDP is of prime
importance for any economy. There are many different approaches to achieve this goal,
though not a single foolproof. One possibility is to find new export markets for goods and
services, as exports, along with the imports of new technologies, is an important engine of
development. This strategy, however, raises the question: should a country promote exports
and/or imports to speed up economic development and growth, or should it primarily focus on
economic growth to generate international trade?

In the literature there has been considerable debate on the export-led growth (ELG) and
growth-driven exports (GDE) hypotheses, with special regard to their implications on
development policies and international trade. As reported by Giles and Williams (2000a), the
story goes back at least to Nurkse (1961). A large number of empirical studies have focused
on this issue, some of them using Indian data. Giles and Williams (2000a, 2000b) and Ahmad
(2001) offer almost exhaustive and comprehensive reviews of these studies, highlighting their
similarities and differences.

In 1991, with the advent of WTO, India has entered into the era of trade reforms and has been
moving gradually towards an open economy since then. It is widely believed that exports are
crucial in providing the impetus for economic growth in developing countries. Thus, export-
led growth has been put forward as an efficient alternative to inward-oriented strategy of
development. Outward orientation is said to lead to higher total factor productivity growth
(Bhagwati 1978, Krueger 1978, Kavoussi 1984, Ram, 1987) and encourages capital material
investment including foreign direct investment. The pressure to compete with the best in the
world may lead to better products and service quality and force the domestic producers to
reduce inefficiencies. For example, foreign exchange liberalisation, which is an important
component of the export-led growth strategy, is likely to reduce the allocation inefficiencies
of exchange control. MacDonald (1994) argues that the imports of final and intermediate
goods will force domestic producers to innovate and increase their efficiency to compete with
foreign imports.

Anoruo and Ahmad (2000), referring to Esfahani (1991) and Ram (1990), note that imports
have positive influence on economic growth. Imports of capital goods are especially
important for developing countries which depend on foreign capital for their economic
development programmes. However, to be beneficial, imported capital must be productively
engaged in the production of goods and services.

Reviewing the relevant literature, Nidugala (2000) provided the following explanations for
the importance of the ELG strategy: i) Exports enable an economy to specialise in the
production of goods in which it has comparative advantage, resulting in optimal allocation of
resources and enhanced overall productivity; ii) Foreign trade can be beneficial through its
effect on balance of payments; iii) Trade can expand production possibilities through its effect
on competition, access to new technologies and ideas (i.e. through the so-called 'dynamic
gains from trade'); iv) Exports enable imports of essential raw materials and capital goods,
thus increase the investment in the economy and thereby lead to higher output; v) Foreign

2
competition creates strong pressure for innovation and efficiency in both export and import
competing industries, thereby raising the productivity of sectors not directly exposed to
foreign competition; vi) Economic success under outward orientation depends on innovation
and efficiency, rather than on the directly unproductive profit seeking activities observed
under import substitution; vii) Export promotion has increasing marginal returns while import
substitution has decreasing marginal returns over time.

Piana (2001), while discussing exports, advocates that increasing exports raise production,
GDP, and employment. In turn, through the Keynesian multiplier effect, it engenders higher
consumption and production, giving rise to a positive feedback loop. Probably, imports will
also rise as a consequence. On the other hand, Thangavelu and Rajaguru (2004) suggest that
trade has an important impact on productivity and output growth in the economy, however it
is imports that provide the important 'virtuous' link between trade and output growth.

The current study is a modest attempt to further investigate the relationship between Indian
exports, imports and GDP growth, and to re-address the export-led growth (ELG), import-led
growth (ILG), and growth driven export/import (GDE/GDI) hypotheses. Specifically, our aim
is to study the potentially causal relationship between the logarithms of exports, imports, and
GDP (all measured at current prices) in India from 1951/52 to 2003/2004, and to test whether:

• Export and/or import and GDP are cointegrated?


• Export and/or import Granger cause GDP or vice versa?

Since our findings have economic policy implications on the Indian trade policy, this paper is
expected to make a useful contribution to the empirical literature.

The applied methodology is mainly based on the framework of Kónya (2004a, 2004b) and
contributes to the existing literature in the following manner. Firstly, most studies that tested
the ELG, ILG and GDE/GDI hypotheses for India, either did not cover the liberalisation
(post-1991) period at all, or did so but only for four to five years at most. The current study is
based on a reasonably long sample period, spanning from 1950/51 to 2003/2004. This sample
period includes 13 years of the post-reform era, making possible to capture the effects of
liberalisation on exports, imports and output growth, if there is any effect at all. Secondly,
most of the earlier studies have been based on a bivariate framework. An important feature of
this paper is the inclusion of imports as an endogenous variable in a vector autoregressive
framework, making possible to test the ILG and GDI hypotheses as advocated by Thangavelu
and Rajaguru (2004). Thirdly, depending on the time series properties of the data, causality is
usually tested with standard Wald tests within vector error correction (VEC) or autoregressive
(VAR) models in levels and/or in first-differences. The disadvantage of this strategy is that
the final outcome might heavily depend on the preliminary test results which, themselves, are
often uncertain and misleading. In this paper, in order to reduce the impact of pre-testing on
the conclusions regarding causality, the modified Wald (MWald) test of Toda and Yamamoto
(1995) and Dolado and Lütkepohl (1996) is also used in augmented level VAR systems. The
advantage of this procedure is that it is valid even under uncertainty about integration and
cointegration.
In spite of some ambiguity, our results clearly show that in case of India exports and imports
Granger-cause GDP, both individually and jointly, lending support to the ELG/ILG
hypotheses. There is also some indication of GDP and exports jointly Granger-causing
imports, and GDP and imports jointly Granger-causing exports, but the GDE/GDI hypotheses
seem implausible.

3
A possible explanation of these results is the favourable trade environment of India. Sharma
(2000) has noted several factors: real depreciation of exchange rate, liberalisation in
investment policy (especially from the early eighties), and provision of export subsidies to
reduce the anti-export bias created by the import-substitution policy. A sharp devaluation of
rupee since the early nineties has further strengthened export growth. Moreover, in July 1991
India opened up its market by lowering tariff and non-tariff barriers and by liberalizing
investment policy. Piana (2001) has noted that in most countries exports have been growing
much faster than GDP and, as a consequence, the share of exports in GDP is much larger
nowadays than 30 years ago. Piana (2001) also notes that higher exports increase production,
GDP, and employment, provided that exports do not simply replace production previously
directed to domestic demand. This hypothesis seems to be reasonable in the case of India as
the 76% export share in 1997 was due to manufactured goods (Sharma 2000). If exports
require raw materials, semi-manufactured goods, and technology, like in India, then export
growth is expected to increase imports as well, further inducing GDP growth.

The rest of the paper is organized as follows. Section 2 presents a review of studies related to
India. The methodological issues are discussed in Section 3, while the sources and properties
of the data are described in Section 4. The Granger causality test results are reported and
discussed in Section 5. Finally, the summary conclusions and policy implications can be
found in Section 6.

2. Review of Studies Related to India

India remained a protected economy for quite a long time that had been described as an
‘import substituting country par excellence’ (Rodrik 1996). Prior to the 1990s, her import
regime was dominated by quantitative restrictions on imports and a highly protectionist
import tariff structure. The World Bank included India in the list of ‘strongly inward-oriented’
countries, meaning that the overall incentive structure strongly favoured production for the
domestic market (Dutta and Ahmed, 2004). Nevertheless, the Indian economy has been
undergoing substantial changes since 1991 (Dean et al. 1994). Almost all areas of the
economy have been opened to both domestic and foreign private investment, import licensing
restrictions on intermediaries and capital goods have been mostly eliminated, tariffs have been
significantly reduced, and full convertibility of foreign exchange earnings for current account
transactions has been established (Dutta 1998).

Table 1 summarises twenty-nine export-growth time-series studies published between 1978


and 2005 related to India. Using various time-series techniques (unit-root and cointegration
tests, single-equation, VAR, EC modelling) they focus either exclusively on India, or on a
group of countries, including India. Though the results are sometimes controversial and
sensitive to model specification, four studies found support for a significant positive
correlation between exports and economic growth, eight for the export-led growth hypothesis,
seven for the growth-driven export hypothesis, and two for two-way causality between
exports and growth.

4
Table 1: Time–Series Studies of Exports and Growth Related to India

Reference Variables Period Conclusion


Krueger(1978) Real GNP, real exports relative to average 1954-71 PEG
exports
Schenzler (1982) Real GDP, real exports, export share 1950-79 PEG
Jung and Marshall Real GDP, GNP, export 1950-81 NC
(1985)
Ram (1987) Real GDP, exports, % share of changes in 1960-82 No PEG
exports in GDP
Nandi and Biswas real GDP, export growth 1960-85 ELG
(1991)
Salvatore and Hatcher Real GDP, exports 1963-85, PEG
(1991) 1963-73,
1973-85
Hutchison and Singh Real GDP, non-export GDP and exports 1950-85 NC in bivariate,
(1992) GDE in trivariate
Dodaro (1993) Real GDP, real exports of goods and non- 1967-86 NC
factor services
Dutt and Ghosh (1994) Real GDP, GNP, exports 1953-91 cointegration
Sharma and Dhakal Real GDP, exports 1960-88 ELG
(1994)
Bhatt (1995) Export, GDP 1950-93 TWC
Rashid (1995) Real GDP, exports 1960-89 No PEG
Dutt and Ghosh (1996) Real GDP, GNP, exports 1953-91 No-cointegration
Mallick (1996) Real GNP, exports 1951-92 GDE
Riezman et al. (1996) GDP, exports 1950-90 ELG in bivariate,
NC in trivariate
Xu (1996) Real GDP, exports 1951-90 ELG
Dhananjayan and Devi Real GNP, total exports, manufactured 1981-94 PEG
(1997) commodity exports (MCE), MNC as % of
total exports
Ghatak and Price Real GDP, exports, imports 1960-92 GDE
(1997)
Rehman and Mustafa Real GDP, exports 1965-94 GDE, both in
(1997) short-run and
long-run
Islam (1998) Real GDP, proportion of export earnings 1967-91 ELG in
in GDP, change in share of non-export in multivariate
GDP, other variables
Asafu-Adjaye and Exports, real output 1960-94 NC
Chakraborty (1999)
Dhawan and Biswal Real GDP, exports, terms of trade 1961-93 GDE
(1999)
Ekanayake (1999) Real GDP, exports 1960-96 TWC,
GDE in short-run
Anwer and Sampath Real GDP, exports 1960-92 NC
(2000)
Nidugala (2000) Real GDP, exports, imports and other 1960-89 ELG
variables
Kemal et al. (2002) Real GDP, exports 1960-98 ELG in long-run,
GDE in short-run
Dutta and Ahmad Real GDP, imports, import price 1971-95 GDI, liberalisation

5
(2004) had little impact
on import demand
Love and Chandra Real GDP, exports 1950-98 ELG
(2005)
Sharma and Real GDP, exports, imports, gross capital 1971-01 NC
Panagiotidis (2005) formation, employment
Note: PEG = significant positive correlation between exports and economic growth; ELG = export-led growth,
GDE= growth-driven export, NC = no-causality between exports and growth, TWC = two-way causality
between exports and growth.

3. Methodologies

As mentioned in the Introduction, this paper aims to test the ELG/ILG and GDE/GDI
hypotheses for India using the natural logarithms of exports, imports, and GDP measured at
current prices between 1951/52 and 2003/2004.1 In order to re-enforce the causality test
results, two complementary strategies are applied. The first one, referred to as the indirect
approach, assumes that the variables are stationary or can be made so by differencing, and
causality is tested with standard Wald tests within VAR (in levels and/or in first-differences)
or VEC models. The second strategy suggested by Toda and Yamamoto (1995) and Dolado
and Lütkepohl (1996), and referred to as the direct approach, requires less pretesting and is
applied in an appropriately augmented level VAR model.

As regards the first strategy, prior to testing for Granger causality, it is important to establish
the time-series properties of the data. In particular, the order of integration2 and the existence
of common trend(s) are of major importance.

Unit Root/Stationarity

To start with, following Engle and Granger (1987), we have tested all three time series
for unit roots versus stationarity. This issue is of great importance since, while standard
econometric methodologies assume stationarity, many economic time series are non-
stationary rendering the usual statistical tests inappropriate and the inferences erroneous and
misleading. For instance, in the presence of non-stationary variables the ordinary least squares
(OLS) estimation method gives rise to spurious regressions, unless these variables are
cointegrated (Granger and Newbold 1974).

Due to the generally low power of unit root tests, we have performed five different tests,
namely, the augmented Dickey-Fuller (ADF) test, the Dickey-Fuller test with GLS detrending
(DF-GLS), the Elliot-Rothenberg-Stock (ERS) point optimal test, the Dickey-Pantula (DP)
test for at most two unit roots, and the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) test for
stationarity.3 As our analyses demonstrate, these tests often generate contradicting outcomes
and none of them is universally superior.

1
We use logarithms because this transformation is the most commonly used variance stabilizing tool
for variables that have wide range (Weisberg, 1980).
2
Time series Xt is said to be integrated of order d, I(d), if it achieves stationarity after being
differenced d times. For d = 0, Xt is stationary in levels and no differencing is necessary, while for d =
1 Xt is non-stationary but its first difference is stationary.
3
About these tests see e.g. Maddala and Kim (1998), Part II.

6
The most commonly used unit root test is the ADF test. The simple Dickey-Fuller (DF) test is
valid only if the series has been generated by a first-order autoregressive process, while the
ADF test, by adding lagged differences of the dependent variable to the right-hand side of the
test regression, incorporates a parametric correction for higher-order autocorrelation. The DF-
GLS test is a simple modification of ADF test, in which the data are detrended so that all
explanatory variables are taken out of the data prior to running the test regression, while the
ERS point optimal test is based on a local to unity detrending of the time series. The DP test
allows for multiple unit roots. It is a sequential procedure which starts with the supposedly
highest order of integration and reduces it gradually. Assuming that there are at the most two
unit roots, first we tested for two unit roots against one, and then, provided that the possibility
of two unit roots is rejected, we tested for only one unit root against none. Finally, unlike in
the ADF, DF-GLS, ERS and DP tests, in the KPSS test tested series is assumed to be trend-
stationary under the null hypothesis and integrated of order one under the alternative
hypothesis. By exchanging the null and alternative hypotheses, the KPSS test provides a
useful comparison to the other four tests.

Performing these tests, we faced two practical problems. First, there is the choice of
exogenous variables (a constant, a constant and a linear time trend, or neither) in the test
regression. Second, we had to specify the lag length, i.e. the number of lagged differences to
be added to the test regression.4

Cointegration

It is well documented in the literature that most economic variables are non-stationary in
their levels but stationary in their first differences. If individual time series turn out to be non-
stationary in their levels i.e. contain stochastic trends, it is still possible that these stochastic
trends are common across the series, leading to one or more stationary combinations of the
level series. The existence of a long-run equilibrium (stationary) relationship among
economic variables is referred to in the economic literature as cointegration. In the presence
of integrated variables, a necessary pre-condition to test for causality is to check whether the
variables are cointegrated.

Granger (1986), Engle and Granger (1987), and Engle and Yoo (1987) have investigated the
relationship between variables when common trend exists between them. In particular, Engle
and Granger (1987) have shown that if two series, say {yt} and {xt}, are both I(d), then their
linear combination, zt = yt - αxt, will, in general, be I(d). However, if there is a constant α
where {zt} is I(d-b), and b > 0, then yt and xt are said to be cointegrated, denoted as CI(d, b).
Specifically, if yt and xt are CI(1,1), then they share the same stochastic trend and thus have a

4
We decided to include enough number of lags to remove serial correlation in the residuals. In
particular, in the ADF, DF-GLS, ERS and DP regressions the optimum lag lengths were primarily
selected by minimising the modified Akaike and Schwarz information criterion. Subsequently, the
residuals were tested for autocorrelation using the Breusch-Godfrey LM and the Ljung-Box
portmanteau tests, and when autocorrelation was detected, the lag length was increased (up to a
maximum of 8 years) in order to whiten the residuals. In the KPSS test we used the Bartlett spectral
window kernel-based estimator to obtain a consistent estimate of the variance and selected the
bandwidth by using the Newey-West method.

7
long-run equilibrium relationship. Moreover, as Engle and Granger (1987) have shown, there
must be either unidirectional or bi-directional Granger causality between them. 5

To this end, we have applied two maximum likelihood tests, the trace and maximum
eigenvalue tests, advocated by Johansen (1988) and Johansen and Juselius (1990). The first
tests the null hypothesis of at most r cointegrating vectors against the alternative hypothesis of
more than r cointegrating vectors, while the second tests the null hypothesis of exactly r
cointegrating vectors against the alternative hypothesis of r + 1 cointegrating vectors.

Granger Causality

The concept of Granger causality, more precisely precedence, is based on the idea that a
cause cannot come after its effect. More precisely, variable X is said to Granger-cause an
other variable, Y, if the future value of Y (yt+1) is conditional on the past values of X (xt-1, xt-2,
... , x0) and thus the history of X is likely to help predict Y.

The application of the indirect approach to testing for Granger causality requires us to
distinguish three possibilities. Assuming that both variables are stationary, Granger causality
between them can be tested within the following level VAR model:
L L
yt = α1 + ∑ β1, l yi , t − l + ∑ γ 1, l xi , t − l + ε1, t
l =1 l =1
L L
(1)
xt = α 2 + ∑ β 2,l yi , t − l + ∑ γ 2,l xi , t − l + ε 2,t
l =1 l =1

where index t refers to the time period (t = 1, ..., T ), l to the lag, and ε1,i ,t , ε 2, i , t are supposed to
be white-noise errors (i.e. they have zero means, constant variances and are individually
serially uncorrelated) that may be correlated with each other.

With respect to this system, there is one-way Granger causality running from X to Y if in the
first equation not all γ 1,l 's are zero but in the second all β 2,l 's are zero, there is one-way
Granger causality from Y to X if in the first equation all γ 1,l 's are zero but in the second not all
β 2,l 's are zero, there is two-way Granger causality between Y and X if neither all β 2,l 's nor
all γ 1,l 's are zero, and there is no Granger causality between Y and X if all β 2,l 's and γ 1,l 's are
zero.
If {yt} and {xt}, are both I(1) but are not cointegrated, then their first differences are
stationary, hence Granger causality between them can be tested within a VAR model in first
differences:

5
To be cointegrated, Xt and Yt must have the same order of integration (Engle and Granger, 1987;
Granger, 1986). However, in case of more than two variables cointegration might occur even if the
variables are integrated of different orders. For example, two I(2) and a third I(1) variable can be
cointegrated if the former two has an I(1) combination which in turn is cointegrated with the third
variable.

8
L L
∆yt = α1 + ∑ β1, l ∆yi ,t − l + ∑ γ 1, l ∆xi , t − l + ε1, t
l =1 l =1
L L
(2)
∆xt = α 2 + ∑ β 2, l ∆yi , t − l + ∑ γ 2, l ∆xi , t − l + ε 2, t
l =1 l =1

Finally, if {yt} and {xt}, are CI(1,1), then they have an error-correction (EC) representation
L L
∆yt = α1 + λ1 ( yt −1 − δ xt −1 ) + ∑ β1, l ∆yi , t − l + ∑ γ 1, l ∆xi , t − l + ε1, t
l =1 l =1
L L
(3)
∆xt = α 2 + λ2 ( yt −1 − δ xt −1 ) + ∑ β 2, l ∆yi , t − l + ∑ γ 2,l ∆xi , t − l + ε 2, t
l =1 l =1

and the inclusion of the EC term ( yt −1 − δ xt −1 ) provides an additional channel through which
the variables can Granger-cause each other.

Since in each of these cases all the variables included in the system are stationary, the direct
approach to testing for Granger causality is based on standard Wald tests for zero restrictions
on the coefficients of the appropriate VAR or VEC model.

In the presence of I(1) variables, the Wald tests are likely to have non-standard asymptotic
properties. However, as Toda and Yamamoto (1995) and Dolado and Lütkepohl (1996)
pointed out, Wald tests that do not restrict the coefficients of all lagged terms under the null
hypothesis still have their usual chi-square distribution. Consequently, given that all variables
are I(0) or I(1), the direct approach to testing for Granger causality is performed by a so-called
modified Wald (MWald) procedure, where a level VAR model augmented by an extra,
redundant lag (VARAL) is estimated and a Wald test is performed on the first L non-
redundant lags (i.e. t-1, t-2, …, t-L).

4. Data and its Properties

Data Sources

Annual data for fifty-four years ranging from 1950/51 to 2003/04 were used for testing
the various hypotheses. The data on Indian exports, imports and gross domestic product
(GDP) are measured in current prices and in local currency (Indian rupees). The data were
collected from several publications and websites, such as the Directorate General of
Commercial Intelligence and Statistics (Ministry of Commerce, Government of India),
National Accounts Statistics (Central Statistical Organisation, Ministry of Statistics and
Programme implementation, Government of India), Planning Commission of India, Reserve
Bank of India, and various issues of Economic Surveys. The natural logarithms of these time
series, denoted as LNGDP, LNEXP and LNINMP, respectively, are plotted in Figure 1.

9
15
14
13
12
11
10
9
8
7
6
5
1950 1960 1970 1980 1990 2000

LNGDP LNEXPORT LNIMPORT

Figure 1: The Logarithms of Indian GDP, Exports and Imports

Unit-root / Stationarity Test Results

Unit-root / stationarity tests have been performed on the levels and on the first (also on
the second, if necessary) differences of LNGDP, LNEXP and LNINMP. Since each series
exhibits a trend, the level series were tested assuming that the data generating processes have
a linear trend component, while in case of the first differences only drift terms were used. The
results are summarized in Tables 2 trough 4.

Starting with Table 2, the DF-GLS (with lags 1 and 3), the ERS (with lags 0 and 1), the DP
(with lag 1) and the KPSS tests (with lag 5) all suggest that LNGDP has a unit root, implying
non stationarity in levels. However, the ADF test (with lags 0 and 1) suggests that LNGDP is
(trend) stationary, i.e. it is I(0), while the DP test with 2 lags provides no evidence against two
unit roots in LNGDP, i.e. it might be I(2). The results are unambiguous for LNEXP and
LNIMP; all tests suggest that they are non-stationary because they have a unit root.

Unit root and stationarity tests for the first-differences of the logarithms of GDP (∆LNGDP),
export (∆LNEXP) and import (∆LNIMP) are displayed in Table 3. The results are again
mixed. The ADF and ERS tests with zero lag and the DF-GLS test suggest that ∆LNGDP is
stationary, i.e. LNGDP is at most I(1). However, the ADF and ERS tests with 2 lags and the
KPSS test imply that ∆LNGDP is not stationary, i.e. LNGDP is at least I(1). As regards the
other two series, all tests but KPSS reject the presence of a unit root in ∆LNEXP and
∆LNIMP, so LNEXP and LNIMP are likely I(1).

10
Table 2: Unit-Root and Stationarity Tests for the Logarithms of GDP,
Exports and Imports

Variables
Unit-Root /
LNGDP LNEXP LNIMP
Stationarity Tests
Lag Test stat. Lag Test stat. Lag Test stat.
*
ADF 0 -3.250 0 -2.686 1 -2.908
1 -3.323*
DF-GLS 1 -1.032 0 -0.849 1 -1.706
3 -1.310 2 -0.945
ERS 0 160.4 0 144.1 1 45.46
1 158.9
DP step 1 1 -5.328*** 0 -5.215*** 0 -6.153***
2 -2.568 1 -3.540**
step 2 1 1.769 0 2.377 0 1.851
1 1.402
KPSS 5 0.246*** 5 0.230*** 5 0.236***
Notes: a) ADF: Augmented Dickey-Fuller test;
DF-GLS: Dickey-Fuller test with GLS detrending;
ERS: Elliot-Rothenberg-Stock Point Optimal test;
KPSS: Kwiatkowski-Phillips-Schmidt-Shin test;
DP: Dickey-Pantula test for at most two unit roots.
b) In all tests except KPSS, lag refers to lags of the first differences. Allowing for
a maximum lag length of 10 years, the lag lengths are selected by minimizing
the Akaike and Schwarz Information Criteria. In KPSS, lag denotes the
bandwidth selected on the basis of the Newey-West method using Bartlett
kernel.
c) *, **, and *** indicate significance at the 10, 5, and 1% levels.
d) Each test equation has a deterministic linear trend.

Table 3: Unit-Root and Stationarity Tests for the First-Difference of the


Logarithms of GDP, Exports, and Imports

Variables
Unit-Root /
Stationarity Tests ∆LNGDP ∆LNEXP ∆LNIMP
Lag Test stat. Lag Test stat. Lag Test stat.
*** ***
ADF 0 -5.328 0 -5.215 0 -6.153***
2 -2.568 1 -3.540**
DF-GLS 2 -1.950** 0 -5.073*** 0 -3.928***
1 -3.579**
ERS 0 1.490*** 0 1.096*** 0 2.656*
2 4.865 1 2.090**
KPSS 5 0.604** 4 0.633** 0 0.507**
Notes: See notes a-c, Table 2.
d) Each test equation has a constant term.

11
Table 4: Unit-Root and Stationarity Tests for the Second-Difference of the
Logarithm of GDP

Unit-Root / ∆2LNGDP
Stationarity Tests Lag Test stat.
ADF 1 -9.532***
DF-GLS 10 -1.037
ERS 1 0.795
KPSS 0 0.014
Notes: See notes a-c, Table 2.
d) Each test equation has a constant term.

Finally, we also performed the unit-root / stationarity tests on the second difference of the log
of GDP (∆2LNGDP). This time the ADF and KPSS tests indicate stationarity, but the other
three tests fail to reject the unit-root null hypothesis implying non-stationarity in the second
differences (Table 4). However, the fact that the alleged optimal lag length reached the fairly
generous a priori upper limit, i.e. 10 years, casts some doubt on the DF-GLS result.

All things considered we conclude that all three variables can be modelled as I(1), i.e. first-
difference stationary series.

Cointegration Test Results

The next step is to search for cointegration between LNGDP, LNEXP and LNIMP in
order to find out whether they share the same stochastic trend(s). Following the
recommendation of Riezmann et. al (1996) that imports are an important variable while
considering causality between exports and growth, and that the omission of imports could
lead to biased results, we tested for cointegration both within bivariate (LNGDP – LNEXP,
LNGDP – LNIMP) and trivariate (LNGDP – LNEXP – LNIMP) frameworks. We have
applied two cointegration tests, namely Johansen’s trace and maximum eigenvalue tests. The
results are summarised in Table 5.

Both tests performed on the bivariate and trivariate systems strongly reject the null hypothesis
of r = 0 (no-cointegration between the variables) in favour of cointegration between LNGDP
and LNIMP, and between LNGDP, LNEXP and LNIMP, irrespective of the lag structure. In
the trivariate system the null hypothesis of r = 1 can also be firmly rejected, while r = 2 is
maintained by the tests. Consequently, LNGDP, LNEXP and LNIMP share two cointegrating
relations. However, in the LNGDP – LNIMP bivariate system there is some evidence for a
second cointegrating relation at zero lag, suggesting that for LNGDP and LNIMP a level
VAR model might be appropriate. The tests for LNGDP and LNEXP show mixed evidence,
the outcomes depend on the lag structure. Both tests reject null of r = 0 for zero lag, but
neither rejects it for two lags.

In brief, the Johansen cointegration tests imply long-run association between (the logarithms
of) GDP and imports in a bivariate framework and between GDP, export and import in a
trivariate framework, while the results are ambiguous for GDP and export.

12
Table 5: Cointegration Tests for the Logarithms of GDP, Exports and Imports

Cointegration
Variables
Tests
LNGDP
LNGDP LNGDP
LNEXP
LNEXP LNIMP
LNIMP
Lag H0 Test stat. Lag H0 Test stat. Lag H0 Test stat.
*** ***
JT 0 r = 0 28.010 0 r = 0 17.747 0 r=0 43.6(19**
r≤1 0.711 r≤1 2.772* r≤1 )*
r≤2 14.094**
1.438
2 r=0 10.813 6 r = 0 19.698*** 10 r=0 109.243***
r≤1 (0.062) r≤1 0.033 r≤1 37.795***
r≤2 0.695
JME 0 r = 0 27.299*** 0 r = 0 14.974*** 0 r=0 29.525***
r=1 0.711 r=1 2.772* r=1 12.656**
r=2 1.438
2 r=0 10.751 6 r = 0 19.955*** 10 r=0 71.448***
r=1 (0.062) r=1 0.033 r=1 37.146***
r=2 0.649
Notes: a) JT: Johansen’s Trace test;
JME: Johansen’s Maximum Eigenvalue test.
b) Lag refers to lag of the first differences. Allowing for a maximum lag length of 10
years, the lag lengths are selected by minimising the Akaike and Schwarz Information
Criteria.
c) *, **, and *** indicate significance at the 10, 5, and 1% levels.
d) The level data are assumed to have linear trends but the cointegrating equations have
only intercepts (Case 3 in EViews 5).

5. Granger Causality Test Results

As stated earlier, this paper follows a similar approach to Kόnya (2004a, 2004b) to
investigate the ELG, ILG and GDE/GDI hypotheses within bivariate and trivariate
frameworks. In particular, we applied two strategies for testing causality. First, allowing for
the existence and lack of cointegration alike, we tested for Granger causality between LNGDP
and LNEXP; LNGDP and LNIMP; LNGDP, LNEXP and LNIMP by applying Wald tests
within finite-order vector error correction models (VECM) and first-difference vector
autoregressive (VARD) models. Second, we also performed a modified Wald (MWald)
procedure advocated by Dolado and Lütkepohl (1996) and Toda and Yamamota (1995) in
augmented level vector autoregressive (VARAL) models, which is valid for both cointegrated
and non-cointegrated variables as. The Granger causality tests results under these two
strategies are shown in Tables 6 through 9.

The results in Table 6 are based on the first strategy when both series are I(1) but not
cointegrated (VARD) and when they are cointegrated (VECM) (Kόnya 2004a, pp. 79; 2004b,
pp. 84). They imply two-way causality between LNGDP and LNEXP, and between LNGDP
and LNIMP.

13
Table 6: WALD Granger Causality Tests for the Logarithms of GDP, Exports and Imports in
VEC and First-Difference VAR Models of Two Endogenous Variables

Models
Null hypothesis VECM VARD
Rank
Lag t-stat χ -stat
2
Lag χ2-stat
LNGDP →
⁄ LNEXP 0 1 4.137**
2 3.026
1 0 5.203***
2 3.120*** 0.083
LNEXP →
⁄ LNGDP 0 1 8.313***
2 13.323***
1 0 3.731***
2 1.306 12.806***
LNGDP →
⁄ LNIMP 0 1 4.589**
4 4.756
1 0 3.196***
6 3.305*** 14.133**
LNIMP →
⁄ LNGDP 0 1 2.981*
4 13.776***
1 0 3.236***
6 -0.418 17.656***

Notes: a) x → ⁄ y indicates no Granger causality from x to y.


b) Rank refers to the cointegrating rank and lag to lags of the first
differences. The optimal lag length is selected by minimising the Akaike
and Schwarz Information Criteria in level VAR (for VECM) or in first-
difference VAR (for VARD), allowing for a maximum lag length of 10
years. If the residuals generated by the optimal lag length are
autocorrelated, heteroscedastic or non-normal, the lag length is gradually
increased.
c) The models were subjected to three residual tests: Breusch-Godfrey LM
test for general, higher order (up to order 10) ARMA errors, White test
for heteroscedasticity (without cross terms) and Jarque-Bera test for
normality. At the 5% significance level each model passed these tests,
except the VARD(1) models.
d) In VECM Granger causality is tested by the t-statistic (which is
asymptotically normal) on the speed of adjustment coefficient (α) and by
a Wald (χ2) test on the coefficients of the lagged first-differences of the
hypothesized causal variable. In VARD Granger causality is tested by a
Wald (χ2) test on the coefficients of the lagged first-differences of the
hypothesized causal variable.
e) *, **, and *** indicate significance at the 10, 5, and 1% levels.
f) The level data are assumed to have linear trends but the cointegrating equations have
only intercepts (i.e. Case 3 in EViews 5).

Next, we tested causality based on the direct approach with MWald test (Kόnya 2004a, pp.
80, 88), i.e. following the second strategy, which is based on an augmented vector
autoregressive model in levels (VARAL). The results, shown in Table 7 suggest one-way
causality from LNEXP to LNGDP and from LNIMP to LNGDP.
Table 7: MWALD Granger Causality Tests for the Logarithms of GDP, Exports and Imports
in Augmented Level VAR Models of Two Endogenous Variables

14
Model
Null hypothesis VARAL
Lag Order χ2-stat
LNGDP →
⁄ LNEXP 1 1 0.516
2 0.445
3 1 1.899
2 1.234
LNEXP →
⁄ LNGDP 1 1 3.677*
2 4.680**
3 1 9.708**
2 7.123*
LNGDP →
⁄ LNIMP 1 1 1.087
2 1.498
7 1 11.624
2 9.593
LNIMP →
⁄ LNGDP 1 1 3.334*
2 4.591**
7 1 20.840***
2 17.227**

Notes: a) x → ⁄ y indicates no Granger causality from x to y.


b) Lag is the optimal lag order of an augmented level VAR (VARAL)
chosen by minimising the Akaike and Schwarz Information Criteria (p)
and order is the highest degree of integration in the system (d). The order
of VARAL is p+d. If the residuals generated by the optimal lag length
are autocorrelated, heteroscedastic or non-normal, p is gradually
increased.
c) The models were subjected to three residual tests: Breusch-Godfrey LM
test for general, higher order (up to order 10) ARMA errors, White test
for heteroscedasticity (without cross terms) and Jarque-Bera test for
normality. At the 5% significance level each model passed these tests.
d) Granger causality is tested by a Wald (χ2) test on the coefficients of the
first p lags of the hypothesized causal variable.
e) *, **, and *** indicate significance at the 10, 5, and 1% levels.
f) The level data are assumed to have linear trends.

In Table 8 the Wald tests on the coefficients of the lagged first-differences of individual
causal variables indicate one-way causality from LNEXP to LNGDP and from LNIMP to
LNGDP. Interestingly, the Wald tests on the coefficients of the lagged first-differences of two
causal variables at a time also suggest that LNGDP and LNIMP jointly Granger-cause
LNEXP, LNGDP and LNEXP jointly cause LNIMP, and LNEXP, LNIMP jointly Granger-
cause LNGDP.

15
Table 8: WALD Granger Causality Tests for the Logarithms of GDP, Exports and Imports in
VEC Models of Three Endogenous Variables

Model
Null hypothesis VECM
Rank
Lag t-stat χ2-stat
LNGDP → ⁄ LNEXP 2 6 5.118
LNEXP →⁄ LNGDP 2 6 16.676**
LNIMP →⁄ LNGDP 2 6 17.679***
LNGDP → ⁄ LNIMP 2 6 9.269
LNGDP, LNIMP → ⁄ LNEXP 2 0 5.506***
-1.536
6 2.865*** 15.050
-2.013**
LNGDP, LNEXP →
⁄ LNIMP 2 0 3.088***
1.927*
6 1.395 12.101
0.828
LNEXP, LNIMP →
⁄ LNGDP 2 0 3.335***
-1.925*
6 -0.829
-1.256 39.608***

Notes: a) x → ⁄ y indicates no Granger causality from x to y.


b) Rank refers to the cointegrating rank and lag to lags of the first
differences. The optimal lag length is selected by minimising the Akaike
and Schwarz Information Criteria in level VAR (for VECM) or in first-
difference VAR (for VARD), allowing for a maximum lag length of 6
years.
c) The models were subjected to three residual tests: Breusch-Godfrey LM
test for general, higher order (up to order 10) ARMA errors, White test
for heteroscedasticity (without cross terms) and Jarque-Bera test for
normality. At the 5% significance level VECM(0) suffers from
heteroscedasticity and VECM(6) has non-normal errors.
d) In VECM Granger causality is tested by a t-statistic (which is
asymptotically normal) on the speed of adjustment coefficients (α1, α2)
and by a Wald (χ2) test on the coefficients of the lagged first-differences
of the hypothesized causal variable(s).
e) *, **, and *** indicate significance at the 10, 5, and 1% levels.
f) The level data are assumed to have linear trends but the cointegrating equations have
only intercepts (i.e. Case 3 in EViews 5).

Finally, the results in Table 9, based on the modified Wald (MWald) test, also indicate that
LNEXP and LNIMP Granger-cause LNGDP both individually and jointly, while LNGDP
does not cause either LNEXP or LNIMP.

16
Table 9: MWALD Granger Causality Tests for the Logarithms of GDP, Exports and Imports
in Augmented Level VAR Models of Three Endogenous Variables

Model
Null hypothesis VARAL(p+d)
Lag Order χ2-stat
LNGDP →
⁄ LNEXP 1 1 0.864
2 0.775
6 1 3.743
2 2.238
LNEXP →
⁄ LNGDP 1 1 1.671
2 1.877
6 1 29.160***
2 21.859***
LNGDP →
⁄ LNIMP 1 1 0.649
2 2.153
6 1 10.266
2 9.148
LNIMP →
⁄ LNGDP 1 1 1.066
2 2.828*
6 1 24.083***
2 15.013**
LNGDP, LNIMP →
⁄ LNEXP 1 1 0.974
2 1.317
6 1 8.703
2 7.652
LNGDP, LNEXP →
⁄ LNIMP 1 1 1.931
2 4.950*
6 1 16.390
2 14.449
LNEXP, LNIMP →
⁄ LNGDP 1 1 4.887*
2 7.546**
6 1 58.943***
2 49.454***

Notes: a) x → ⁄ y indicates no Granger causality from x to y.


b) Lag is the optimal lag order of a level VAR chosen by minimising the
Akaike and Schwarz Information Criteria (p) and order is the highest
degree of integration in the system (d).
c) The models were subjected to three residual tests: Breusch-Godfrey LM
test for general, higher order (up to order 10) ARMA errors, White test
for heteroscedasticity (without cross terms) and Jarque-Bera test for
normality. At the 5% significance level VARAL(7) and VARAL(8) have
non-normal errors.
d) Granger causality is tested by a Wald (χ2) test on the coefficients of the
first p lags of the hypothesized causal variable.
d) *, **, and *** indicate significance at the 10, 5, and 1% levels.
e) The level data are assumed to have linear trends.

17
Summing up, on the basis of the Wald and MWald tests performed in both bivariate and
trivariate frameworks, there is strong evidence in favour of the ELG and ILG hypotheses, but
not for the GDE/GDI hypotheses.

6. Conclusions and Policy Implications

Summary Conclusions

In this study we aimed at exploring whether India experienced export/import-led


growth, or growth-driven export/import, both or none, during the fifty-four years period
spanning over 1951/52-2003/2004. Our data series are reasonably long and cover the pre- and
post liberalisation periods alike, making it possible to capture the effects of measures to
promote exports and output growth. We studied Granger causality between the logarithms of
exports, imports and GDP. In order to re-enforce the results, we applied two complementary
strategies. The first, indirect, approach assumes that the variables are stationary or can be
made stationary by differencing. It makes use of pre-testing for unit roots and cointegration
and, depending on the outcomes, testing for causality is carried out with Wald tests in VAR
and/or VEC models in levels and/or first differences. The second, direct, approach is based on
a modified Wald test without need to pre-test for unit roots and cointegration.

Based on these strategies, we are of the opinion that exports and imports, both jointly and
individually, Granger-cause GDP (supporting the ELG and ILG hypotheses), as confirmed by
the Wald and MWald tests in the bivariate and trivariate frameworks alike. There is also some
evidence of GDP and exports jointly Granger-causing imports, and GDP and imports jointly
Granger-causing exports, while the growth driven export/import (GDE/GDI) hypotheses seem
implausible in case of India.

Policy Implications for India

This study confirms that export and import growth has been instrumental in accelerating
economic growth in India. The evidence of causality from exports to economic growth
implies that exports can have positive effect on economic growth. Exports, for example, can
boost output growth in the short-run by allowing the use of excess capacity in cases where
domestic demand requires less than full capacity production. Based on outcome similar to
ours, Nidugala (2000) favours the ongoing reforms regarding openness for faster economic
growth and higher GDP in India. By further opening up her market and continuing the
ongoing trade (export/import) promotion policy reforms, India can not only boost its
economic growth further but can also fuel growth in the entire South Asian region.

As suggested by Kemal et al. (2002), in the long-run exports can have beneficial effect on
economic growth in a variety of ways. First, export production allows economies with narrow
domestic markets to overcome size limitations and to reap economies of scale. Second, by
relaxing the foreign exchange constraints, higher exports can permit higher imports of capital
goods thereby strengthening the productive capacity of the economy. Third, exports lead to an
improvement in economic efficiency by enhancing the degree of competition. Fourth, exports
contribute to productivity gains through diffusion of technical knowledge and learning by
doing. Further, export-oriented production and investment tend to take place in the most
efficient sectors of the economy fostering a pattern of production that is consistent with a
country's comparative advantages. Specialisation in these sectors improves productivity in the

18
economy leading to higher output growth, as also advocated by Thangavelu and Rajaguru
(2004). The study of Kemal et al. (2002) lends support to the export-oriented policies that are
hallmark of current trade regimes of the major South Asian economies, and suggest that South
Asian countries ought to continue the strategy of export-led growth to tackle the myriad
development challenges facing their economies.

Finally, according to Thangavelu and Rajaguru (2004), in India and in several other Asian
countries imports tend to have long run 'virtuous cyclical' effect on labour productivity, more
than exports. They suggest that exports and imports are both important for an outer-oriented
economic strategy. Similarly, the empirical evidence reported by Lee (1995) indicates that
imports have a positive effect on long-run output growth. In particular, imports could be an
important vehicle and source to assess foreign technology for developing countries. In an
outer-oriented strategy, countries should allow greater flow of goods and services into the
domestic economy by promoting both exports and imports.

References

Ahmad, J. (2001): Causality Between Exports and Economic Growth: What Do the
Econometric Studies Tell Us?, Pacific Economic Review, vol. 6, pp. 147-167.

Anwer, M.S. and Sampath, R.K. (2000): Exports and Economic Growth, Indian Economic
Journal, vol. 47, pp. 79-88.

Anoruo, E. and Ahmad, Y. (2000): Openness and Economic Growth: Evidence from Selected
ASEAN Countries, Indian Economic Journal, vol. 47, pp. 110-117.

Asafu-Adjaye, J. and Chakraborty, D. (1999): Export-Led Growth and Import Compression:


Further Time Series Evidence from LDCs, Australian Economic Papers, vol. 38, pp.
164-75.

Bhagwati, J. (1978): Foreign Trade Regimes and Economic Development: Anatomy and
Consequences of Exchange Contrast Regimes, Cambridge, MA, Ballinger Publishing
Company.

Bhat, S.K. (1995): Export and Economic Growth in India, Artha Vijana, vol. 37, pp. 350-58.

Dean, J.M., Desai, S. and Riedel, J. (1994): Trade Policy Reforms in Developing Countries
since 1985: A Review of the Evidence’, World Bank Discussion Papers 267.

Dhananjayan, R.S. and Devi, N.S. (1997): Exports and Economic Growth: A Study of Select
Nations in Asia and Europe during 1980-81 to 1993-94, Indian Journal of Applied
Economics, vol. 6, pp. 41-63.
Dhawan, U. and Biswal, B. (1999): Re-Examining Export Led Growth Hypothesis: A
Multivariate Cointegration Analysis for India, Applied Economics, vol. 31, pp. 525-30.

Dickey, D.A. and Pantula, S.G. (1987): Determining the Order of Differencing in
Autoregressive Processes, Journal of Business and Statistics, vol. 5, pp. 455-61.

19
Dodaro, S. (1993): Exports and Growth: A Reconsideration of Causality, Journal of
Developing Areas, vol. 27, pp. 227-44.

Dolado, J.J. and Lütkepohl, H. (1996): Making Wald Tests Work for Cointegrated VAR
Systems, Econometric Reviews, vol. 15, pp. 369-86.

Dutt, S.D. and Ghosh, D. (1994): An Empirical Investigation of the Export Growth-Economic
Growth Relationship, Applied Economics, vol. 1, pp. 44-48.

Dutt, S.D. and Ghosh, D. (1996): The Export Growth-Economic Growth Nexus: A Causality
Analysis, Journal of Developing Areas, vol. 30, pp. 167-82.

Dutta, D. (1998): Fifty Years of Economic Development in India: Rhetoric and Reality,
published in Fifty Years of Indian Development: India in Retrospect, Future Directions
and Investment Outlook, Economic Conference Monograph no. 5, (eds.) P.K. Basu,
N.D. Karunaratne and C.A. Tisdell, The University of Queensland, Australia.

Dutta, D. and Ahmad, N. (2004): An Aggregate Import Demand Function for India: A
Cointegration Analysis, Applied Economics Letters, vol. 11, pp. 607-13.

Ekanayake, E.M. (1999): Exports and Economic Growth in Asian Developing Countries:
Cointegration and Error-Correction Models, Journal of Economic Development, vol. 24,
pp. 43-56.

Elliott, G., Rothenberg, T. and Stock, J. (1996): Efficient Tests for an Autoregressive Unit
Root, Econometrica, vol. 64, pp. 813-36.

Engle, R.F. and Granger, C.W.J. (1987): Cointegration and Error Correction: Representation,
Estimation and Testing, Econometrica, vol. 55, pp. 251-76.

Engle, R.F. and Yoo, B.S. (1987): Forecasting and Testing in Cointegration Systems, Journal
of Econometrics, vol. 35, pp. 143-159.

Esfahani, H.S. (1991): Exports, Imports, and Economic Growth in Semi-Industrialized


Countries, Journal of Development Economics, vol. 35, pp. 93-116.

Ghatak, S. and Price, S.W. (1997): Exports Composition and Economic Growth:
Cointegration and Causality Evidence for India, Weltwirtschaftliches Archive: Review of
World Economics, vol. 133, pp. 538-553.

Giles, J.A. and Williams, C.L. (2000a): Export-Led Growth: A Survey of the Empirical
Literature and Some Non-Causality Results, Part 1, Journal of International Trade and
Economic Development, vol. 9, pp. 261-337.

Giles, J.A. and Williams, C.L. (2000b): Export-Led Growth: A Survey of the Empirical
Literature and Some Non-Causality Results, Part 2, Journal of International Trade and
Economic Development, vol. 9, pp. 445-70.

Granger, C.W.J. and Newbold, P. (1974): Spurious Regressions in Econometrics, Journal of


Econometrics, vol. 2, pp. 111-20.

20
Granger, C.W.J. (1969): Investigating Causal Relations by Econometric Models: Cross
Spectral Methods, Econometrica, vol. 37, pp. 424-38.

Granger, C.W.J. (1986): Development in the Study of Cointegrated Economic Variables,


Oxford Bulletin of Economics and Statistics, vol. 48, pp. 213-228.

Granger, C.W.J. (1988): Some Recent Developments in the Concept of Causality, Journal of
Econometrics, vol. 39, pp. 199-211.

Hutchison, M. and Singh, N. (1992): Exports, Non-Exports, and Externalities: A Granger


Causality Approach, International Economic Journal, vol. 6, pp. 79-94.

Islam, M.N. (1998): Export Expansion and Economic Growth: Testing for Cointegration and
Causality, Applied Economics, vol. 30, pp. 415-25.

Johansen, S. (1988): Statistical Analysis of Cointegration Vectors, Journal of Economic


Dynamics and Control, vol. 12, pp. 231-54.

Johansen, S. and Juselius, K. (1990): Maximum Likelihood Estimation and Inference in


Cointegration – with Application to the Demand for Money, Oxford Bulletin of
Economics and Statistics, vol. 52, pp. 169-210.

Jung, W.S. and Marshall, P.J. (1985): Exports, Growth and Causality in Developing
Countries, Journal of Development Economics, vol. 18, pp. 1-12.

Kavoussi, R. M. (1984): Export Expansion and Economic Growth: Further Empirical


Evidence, Journal of Development Economics, vol. 14, pp. 241-50.

Kemal, A.R., Din, M., Qadir, U., Fernando, L., and Colombage, S.S. (2002): Export and
Economic Growth in South Asia, South Asia Network of Economic Research Institutes,
http://www.saneinetwork.net/.

Kónya, L. (2004a): Export-Led Growth, Growth-Driven Export, Both or None? Granger


Causality Analysis on OECD Countries, Applied Econometrics and International
Development, vol. 4, pp. 73-94.

Kónya, L. (2004b): Unit-Root, Cointegration and Granger Causality Test Results for Export
and Growth in OECD Countries, International Journal of Applied Econometrics and
Quantitative Studies, vol. 1, pp. 67-94.

Krueger, A.O. (1978): Foreign Trade Regimes and Economic Development: Liberalization
Attempts and Consequences, Cambridge (Mass.), Balinger.

Kwiatkowski, D., Phillips, P.C.B., Schmidt, P. and Shin, Y.-C. (1992): Testing the Null
Hypothesis of Stationarity Against the Alternative of a Unit Root, Journal of
Econometrics, vol. 54, pp. 159-78.

Lee, J.W. (1995): Capital Goods Imports and Long-Term Growth, Journal of Developing
Economies, vol. 48, pp. 91-110.

21
Love, L. and Chandra, R. (2005): Testing Export-Led Growth in South Asia, Journal of
Economic Studies, vol. 32, pp. 132-145.

MacDonald, J.M. (1994): Does Import Competition Forces Efficient Production?, Review of
Economics and Statistics, vol. 76, pp. 721-7.

Maddala, G.S. and Kim, I.-M. (1988): Unit Roots, Cointegration, and Structural Change,
Cambridge, Cambridge University Press.

Mallick, S.K. (1996): Causality Between Exports and Economic Growth in India: Evidence
from Cointegration Error Correction Models, Indian Journal of Economics, vol. 76, pp.
307-20.

Nandi, S. and Biswas, B. (1991): Exports and Economic Growth in India: Empirical
Evidence, Indian Economic Journal, vol. 38:, pp. 53-9.

Nidugala, G.K. (2000): Exports and Economic Growth in India: An Empirical Investigation,
Indian Economic Journal, vol. 47, pp. 67-78.

Nurkse, R. (1961): Trade Theory and Development Policy, in Economic Development of


Latin America, ed. H. S. Ellis, New York, St Martin Press.

Piana V. (2001): Exports, www.economicswebinstitute.org.

Rahman, M. and Mustafa, M. (1997): Dynamics of Real Exports and Real Economic Growth
in 13 Selected Asian Countries, Journal of Economic Development, vol. 22, 81-95.

Ram, R. (1987): Exports and Economic Growth in Developing Countries: Evidence from
Time-Series and Cross-Section Data, Economic Development and Cultural Change, vol.
36, pp. 51-72.

Ram, R. (1990): Import and Economic Growth: A Cross Country Study, Economica
Internzionale, vol. 43, pp. 45-66.

Rana, P.B. (1985): Exports and Economic Growth in the Asian Region, ADB Economic Staff
Papers, 25.

Rashid, A.I. (1995): Trade, Growth, and Liberalization: the Indian Experience, 1977-1989,
Journal of Developing Areas, vol. 29, pp. 355-70.

Riezman, R.G., Summers, P.M., and Whiteman, C.H. (1996): The Engine of Growth or Its
Handmaiden? A Time Series Assessment of Export-Led Growth, Empirical Economics,
vol. 21, pp. 77-113.

Rodrik, D. (1996): Understanding Economic Policy Reforms, Journal of Economic


Literature, vol. 34, pp. 9-41.

Salvatore, D. and Hatcher, T. (1991): Inward and Outward Oriented Trade Strategies, Journal
of Developing Studies, vol. 27, pp. 7-25.

22
Schenzler, C. (1982): An Empirical Investigation of the Relationship between Growth of
Gross National Product and Exports in Chile, India, and South Korea, Masters Thesis,
Vanderbilt University, Nashville TN.

Sharma, A. and Panagiotidis, T. (2005): An Analysis of Exports and Growth in India:


Cointegration and Causality evidence (1971-2001), Review of Development Economics,
vol. 9, pp. 232–48.

Sharma, S.C. and Dhakal, D. (1994): Causal Analysis between Exports and Economic Growth
in Developing Countries, Applied Economics, vol. 26, pp. 145-57.

Sharma, K. (2000): Export Growth in India: Has FDI Played a Role? Centre Discussion
Paper no. 816, Economic Growth Centre, Yale University, http://www.econ.yale.edu/.

Thangavelu, S.M. and Rajaguru, G. (2004): Is There an Export or Import-Led Productivity


Growth in Rapidly Developing Asian Countries? A Multivariate VAR Analysis, Applied
Economics, vol. 36, pp. 1083-93.

Toda, H.Y. and Yamamoto, T. (1995): Statistical Inference in Vector Autoregressions with
Possibly Integrated Processes, Journal of Econometrics, vol. 66, pp. 225-50.

Weisberg, S. (1980): Applied Linear Regression, John Wiley and Sons, Inc.

Xu, Z. (1996): On the Causality between Export Growth and GDP Growth: An Empirical
Reinvestigation, Review of International Economics, vol. 4, pp. 172-84.

23
Recent Discussion Papers

School of Business Discussion Papers are available from the Research


Officer, School of Business, La Trobe University VIC 3086, Australia.

04.01 Michael Schneider – Measuring Inequality: The Origins of the


Lorenz Curve and the Gini Coefficient.

04.02 László Kónya – Saving and Growth: Granger Causality


Analysis with Bootstrapping on Panels of Countries

04.03 Buly A. Cardak – Education Vouchers, Growth and Income


Inequality.

05.01 László Kónya and Chris Mouratidis – Income Inequality and


Growth Volatility.

05.02 Shawn Chen-Yu Leu and Jeffrey Sheen – Asymmetric


Monetary Policy in Australia.

06.01 Shawn Chen-Yu Leu – A New Keynesian Perspective of


Monetary Policy in Australia.

06.02 David Prentice – A re-examination of the origins of American


industrial success.

06.03 Elisabetta Magnani and David Prentice – Outsourcing and


Unionization: A tale of misallocated (resistance) resources.

06.04 Buly A. Cardak and Chris Ryan – Why are high ability
individuals from poor backgrounds under-represented at
university?

06.05 Rosaria Burchielli, Donna M. Buttigieg and Annie Delaney –


Why are high ability individuals from poor backgrounds under-
represented at university?

View publication stats

You might also like