-
EXPORTS AND ECONOMIC GROWTH: AN ERROR CORRECTION
MODEL
Emmanuel Anoruo
Department of Management Science and Economics
Coppin State College 2500 W. North Avenue
Baltimore, MD 21216
U.S.A.
Ph: (410) 383-5582
Email: [email protected]
Sanjay Ramchander*
Department of Finance and Real Estate
College of Business
Colorado State University Fort Collins, CO 80523
Ph: (970) 491-6681
Email: [email protected]
________________ * Corresponding author
-
EXPORTS AND ECONOMIC GROWTH: AN ERROR CORRECTION MODEL
Abstract
The relationship between exports and economic growth has been a
popular subject of debate among development
economists. This paper uses a theoretically consistent method to
examine the export-led growth (ELG) hypothesis
for five emerging economies of Asia namely — India, Indonesia,
Korea, Malaysia, and the Philippines.
Specifically, the paper employs a cointegration estimation
procedure to examine the export-economic growth nexus,
and employs a vector error correction model to abstract
simultaneously the short- and long-run information in the
modeling process. Results from the study provide evidence in
support of the ELG hypothesis in that export growth
has a causal influence on economic growth for all countries with
the exception of Indonesia. From a policy
perspective, the acceptance of the ELG hypothesis lends credence
to the view of ‘outward orientation’ as an
effective policy for economic growth, especially for countries
with nascent economies.
-
3
EXPORTS AND ECONOMIC GROWTH: AN ERROR CORRECTION MODEL
I. Introduction
The purpose of this paper is to test the probity of the
export-led growth (ELG) hypothesis for
five emerging economies of Asia — namely India, Indonesia,
Korea, Malaysia, and the
Philippines. The issue of the links between export performance
and economic growth in a
trading world economy are a perennial source of concern and
controversy, more so with the
emergence of a significant body of empirical work in the
development economics literature since
the late 1960s. While classical trade theory provides important
insights into the static gains of
trade (i.e., the impact of trade on national economic
well-being), it fails to fully account for the
dynamic relationship between trade policies and economic growth.
The rapid economic growth
witnessed by the so-called newly industrialized countries has
revived the debate on optimal
growth strategies for emerging market economies.
The current debate centers on whether a developing country would
be better served by
trade policies oriented toward import substitution or export
promotion. Import substitution
strategies seek to promote rapid industrialization and therefore
development by erecting high
barriers to foreign goods such as tariffs and quotas to
encourage local production. This approach
to development thus applies the ‘infant industry’ argument for
protection to one or more targeted
industries in the developing country. As the industrialization
process takes hold, the government
lowers the trade barriers. On the other hand, outward-looking
development (or ELG) strategies
involve government support for manufacturing sectors in which a
country has a potential
comparative advantage. This framework argues that international
trade promotes specialization
in production of export products, which in turn boosts the
productivity level and causes the
general level of skills to rise in the export sector. This then
leads to a re-allocation of resources
from the inefficient non-trade sector to the trade sector. Thus,
the entire economy would benefit
due to the dynamic spillover benefit from the export sector’s
growth. Empirical and anecdotal
evidence tends to support the notion that those economies which
actively pursue export-
promotion policy have been more successful than those that have
pursued import substitution
-
4
policies (see, for example, Feder, 1982 and Krueger, 1990)1.
This paper incorporates the recent advances made in time series
analysis, and proposes a
theoretically consistent method to examine the ELG hypothesis
for several emerging economies.
Specifically, unit root tests, cointegration analysis and
error-correction techniques are employed
in a multi-variate framework that directly addresses the problem
of omitted variables (an issue
that is often overlooked in past studies).2 The estimation
technique places minimal restrictions
on the explicit structure of the relationship between exports
and economic growth, and abstracts
simultaneously the short- and long-run information in the
modeling process. Additionally, this
study by using an extensive sample period and large information
set proposes to obtain more
robust results than those of the earlier studies.
Apart from its important policy implications, the present
discussion is topical considering
that many economists attribute the recent Asian economic crisis
to the unsustainable level of
current account deficits that were maintained by these
countries. Furthermore, emerging
economies may be characterized by potentially unique monetary
policy and macroeconomic
transmission mechanisms that are arguably very different from
those of industrialized nations.
Developing economies also experience numerous other drawbacks,
such as an inefficient public
enterprise, deficient infrastructure, tight trade controls,
restrictive regulations in the financial
sector, pro-cyclical macroeconomic policy responses to large
capital inflows, poor corporate
governance, and political uncertainty. Under such conditions,
there may be wide disparities in
the macroeconomic dynamics governing policy transmission between
developing and developed
economies.
The outline of the remainder of this study is as follows. The
next section conducts a brief
1 Import-substituting industrialization has come under
increasingly harsh criticism, since many countries that
pursued such strategies have not shown any signs of catching up
with the advanced countries. India is an excellent
example. After 40 years of ambitious economic plans between the
1950s and late 1980s, India found itself with per
capita income only a few percent higher than before. But after
adopting market friendly reforms beginning in the
early 1990s, India has shown tremendous strides in both export
revenue and economic growth.
2 The deployment of a multi-variate estimation procedure is
especially important since causality findings from bi-
variate VARS can easily be overturned by the addition of a third
(or more) variable (see Lutkepohl, 1989). We
thank the anonymous referee for this suggestion.
-
5
review of the existing literature, their methodological
drawbacks and our approach to redress this
issue. Section III provides a discussion on the methodological
issues. The data employed and
results of the study are presented in Section IV. The final
section summarizes the findings of the
study and makes several policy implications.
II. Literature Review
The empirical investigation into the relationship between export
growth and economic expansion
has primarily taken three different, but related, forms. The
context of these studies has ranged
from individual-country analyses to multi-country
investigations. Early studies have undertaken
correlation-type analysis between an economic growth variable
and some variant of export
growth (example, Michaely, 1977, Balassa, 1978, Heller and
Porter, 1978, Tyler, 1981 and
Kavoussi, 1984). The evidence of a highly significant positive
correlation between the two
variables was interpreted as support of the hypothesis that
export-promoting measures have
fueled economic growth. The second type of investigation, which
derives its basis from
neoclassical growth accounting technique of production function,
specifies and estimates a
production function of labor, capital and export levels
regressed on real gross domestic product
(example, Michalopoulos and Jay, 1973, Feder, 1982, Balassa,
1985, Rana, 1988 and Ram,
1987). A highly significant positive value of the coefficient of
the export growth variable in the
growth accounting equation was treated as evidence supporting
the export-oriented growth
hypothesis. Recent studies examine the issue by employing
Granger causality tests based on
vector autoregressive (VAR) models to determine the direction of
the causality in this
relationship3. The evidence from the causality investigations
has been conflicting. Marin’s
(1992) and Serletis’ (1992) test results, for instance, support
the ELG hypothesis. Giles et al.
(1992), on the other hand, using New Zealand data, finds support
in only specific commodity
groups. Moreover, others such as Jung and Marshall (1985), Chow
(1987), Ahmad and Kwan
3 The ‘technology theory of trade’ posits that causality runs
from output growth to exports. For instance, if a certain
sector of the economy achieves technological innovation, it is
possible that the output from this sector will far
exceed the increase in domestic demand. Thus, the producers are
likely to sell this surplus in the foreign market.
-
6
(1991) and Sharma and Dhakal (1994) find only marginal support
for uni-directional causality
from exports to economic growth.
Although the existing literature has helped provide numerous
insights and raised the
general awareness of policy makers toward this issue, the
conceptual and methodological
approach undertaken in these studies raises a number of serious
concerns. First, the single-
equation studies using OLS regression may suffer from a
simultaneous-equation bias which can
lead to invalid inferences. Second, most early studies make the
a priori assumption that export
growth causes output growth, thus ignoring the potential of a
feed back effect (see Michaely,
1977, Kavoussi, 1984 and Kunst and Marin, 1989). Third, the few
studies that do accommodate
the concepts of causality and exogeneity suffer from an
additional methodological constraint, in
that the ELG nexus, inherently, is a long run behavioral
relationship whose analysis requires
methodologies for estimating a long run equilibria (see Ahmad
and Harnihurun, 1995).
Furthermore, VAR/Granger type analyses (which are essentially
autoregressive distributed lag
models) are strictly appropriate only when all the variables in
the model are stationary (see
Charemza and Deadman, 1992, pg. 194). If stochastic trends
exist, detrended values of the time-
series with appropriate differencing should be used in order to
make the regression analysis
meaningful.4 Finally, the mixed and conflicting evidence amassed
by previous studies is
possibly a result of omitted variables that serve to mediate the
linkages between export growth
and economic development. Modeling the ELG hypothesis in a
bi-variate framework entails the
risk of inaccurate inferences being drawn, since it is clear
that economic growth depends on
many other factors besides exports (see for example, Glasure and
Lee, 1999). By not accounting
for these variables in the model, the results may mask or
overstate the causal relationship
between exports and economic growth. This study attempts to
overcome these methodological
deficiencies by examining the export led growth hypothesis in a
multi-variate framework that is
consistent with the theoretical inferences posited by the ELG
hypothesis.
4 In fact, Toda and Phillips (1993) argue that in the presence
of stochastic trends, the empirical use of the
asymptotic Granger causality tests in first difference vector
error correction models is superior to Granger tests in
level VAR models.
-
7
III. Methodological Issues
This paper employs a methodology that attempts to address the
shortcomings in the earlier
literature. The empirical process comprises three parts: (1)
testing for a unit root, I(1), in each
series; (2) testing for the number of cointegrating vectors in
the system, given that we cannot
reject the null hypothesis of a unit root in the variables; and
(3) estimating and testing for
causality in the framework of a multi-variate vector
error-correction model (VECM). If the
variables for a particular country are found to be stationary in
their level representation, then the
standard vector auto regression (VAR) model is appropriate in
detecting the direction of
causality (in the Granger sense) between exports and economic
growth.
Unit Root Test
To test for a unit root in each series, we employ the augmented
Dickey-Fuller (ADF)
methodology (see Dickey-Fuller, 1981). The ADF test is estimated
by the following regression:
t
p
1i
1ti1t10t YaYaztaY ε+∑ ∆++++=∆=
−− (1)
where a0 is a constant, t is a deterministic trend, and enough
lagged differences are included to
ensure that the error term becomes white noise. If the
autoregressive representation of Yt
contains a unit root, the t-ratio for a1 should be consistent
with the hypothesis a1=0.
Cointegration Test
Engle and Granger (1987) observe that even though economic time
series may wander
through time, that is, may have the characteristic of
nonstationarity in their level, there may exist
some linear combination of these variables that converges to a
long run relationship over time. If
the series individually are stationary only after differencing
but one finds that a linear
combination of their levels is stationary, then the series are
said to be cointegrated. In the
context of the present analysis, the existence of a common trend
between the export and
economic development variables means that in the long run the
behavior of the common trend
will drive the behavior of the two variables, and that there
exists some convergence of policies.
-
8
In other words, a finding of cointegration would simply mean
that the transmission mechanism
underlying the export led growth hypothesis is stable, and thus
more predictable over long
periods. Furthermore, shocks that are unique to one time series
will quicky dissipate as the
variables adjust back to their common trend.
To investigate the existence of a long run equilibrium
relationship between exports and
economic growth, we employ the maximum-likelihood test procedure
established by Johansen
and Juselius (1990) and Johansen (1991).5 Specifically, Yt is a
vector of n stochastic variables,
then there exists a k-lag vector autoregression with Gaussian
errors of the following form:
tt1t zYY...YaY 1t1k1k1t +Π+∆Γ++∆Γ+=∆ −−−−− (2)
where '1,......, 'k-1 and A are coefficient matrices, zt is a
vector of white noise process and "
contains all deterministic elements.
The focal point of conducting Johansen’s cointegration test is
to determine the rank (r) of
the p x p A matrix. In the present application, there are three
possible ranks. First, it can be of
full rank , which would imply that the variables are given by a
stationary process, which would
contradict the earlier finding that the two variables are
nonstationary. Second, the rank of A can
be zero, in which case it indicates that there is no long run
relationship between export growth
and economic development. In instances when A is of either full
rank or zero rank, it will be
appropriate to estimate the model in either levels or first
differences, respectively. Finally, in the
intermediate case when 0 < r < p (reduced rank), there are
r cointegrating relations among the
elements of Yt and p-r common stochastic trends. The number of
lags used in the vector
autoregression is chosen based on the evidence provided by
Akaike’s Information Criterion
5 This approach is especially appealing since it provides a
unified framework for estimating and testing
cointegrating relations in the context of a VECM model. Thus, by
treating all the variables as endogenous, this
approach avoids the arbitrary choice of the dependent variable
in the cointegrating equations, as in the Engle-
Granger methodology. They have also been shown to have good
large- and finite-sample properties (see Phillips,
1991, Cheung and Lai, 1993, and Gonzala, 1994).
-
9
(AIC) (see Akaike, 1973).6
The cointegration procedure yields two likelihood ratio test
statistics, referred to as the
trace test and the maximum eigenvalue (8-max) test, which will
help determine which of the
three possibilities is supported by the data. 7 The study
employs both tests to examine the
sensitivity of the results to different tests. In the trace
test, the null hypothesis that there are at
most r cointegrating vectors is tested against the general
alternative, whereas in the maximum
eigenvalue test the null hypothesis of r cointegrating vectors
is tested against the alternative of at
least (r+1) cointegrating vectors.8
Causality Test Under the Multi-variate VECM Framework
Causality inferences in the multi-variate framework are made by
estimating the parameters of the
following VECM equations.
tt
p
l
s
k
jtj
n
j
it
m
i
ZRERMEGrowGGrowiGGrow εθζδγβα ++∆+∆+∆+∆+=∆ −==
−
=
−
=∑∑∑∑ 1
1
0
111
(3)
tt
p
l
s
k
jtj
n
j
m
i
i fZREReMdEGrowcGGrowbaEGrow ξ++∆+∆+∆+∆+=∆ −==
−
==∑∑∑∑ 1
1
0
111
(4)
6 The optimal lag length chosen is the one that minimizes AIC,
where
AIC = ln det Skn + (2d
2k)/T
and k = 1, 2,...., n, d is the number of variables in the
system, n is the maximum lag length considered, det denotes
the determinant, and Sk is the estimated residual
variance-covariance matrix for lag k.
7 The trace test statistic is given by:
)1ln(1
∑ −=+=
N
ri
iTTR λ
where 8r+1, ...., 8N are the N-r smallest squared canonical
correlations between Xt-k and ) Xt series, corrected for the effect
of the lagged differences of the Xt. The maximum eigenvalue
statistic is given by
8max = T ln(1-8r+1) Since the asymptotic distributions of the
trace and maximum eigenvalue test statistics follow P2
distributions, a simulation procedure is needed to identify proper
critical values for each test (see Osterwald-Lenum, 1992).
8 In order to mitigate the bias arising from small sample size,
this study utilizes both the Reinsel and Ahn (1988)
and Cheung and Lai (1993) test procedures to check for the
significance of the results. Under the Reinsel and Ahn
(1988) procedure, the trace test statistic is multiplied by a
factor of (T-nK)/T, where T represents the size of the
sample, n stands for the lag length, and K represents the number
of series in the system. Under the Cheung and Lai
procedure, the Osterwald-Lenum (1992) critical values are
multiplied by a factor equal to 0.1+0.9T/(T-nk).
-
10
where GGrow and EGrow denote GDP and export growth rates
respectively, Ms is the M2 real
money supply, RER is the real exchange rate (with respect to the
U.S. dollar) and zt-1 is the error-
correction term which is the lagged residual series of the
cointegrating vector. The error-
correction term measures the deviations of the series from the
long run equilibrium relation. For
example, from equation (3), the null hypothesis that EGrow does
not Granger-cause GGrow is
rejected (in other words, the ELG hypothesis is supported) if
the set of estimated coefficients on
the lagged values of EGrow is jointly significant. Furthermore,
in those instances where EGrow
appears in the cointegrating relationship, the ELG hypothesis is
also supported if the coefficient
of the lagged error-correction term is significant. Changes in
an independent variable may be
interpreted as representing the short run causal impact while
the error-correction term provides
the adjustment of GGrow and EGrow toward their respective long
run equilibrium. Thus, the
VECM representation allows us to differentiate between the
short- and long-run dynamic
relationships.
IV. Data and Empirical Findings
The empirical analysis is conducted using annual observations of
GDP, exports, broad real
money supply (under the M2 definition) and real exchange rate
covering the periods, 1950 to
1998 for India; 1969 to 1998 for Indonesia, 1953 to 1998 for
Korea; 1955 to 1998 for Malaysia;
and 1949 to 1998 for the Philippines. All data were obtained
from the International Financial
Statistics published by the International Monetary Fund (IMF).
Growth rates are calculated by
the transformation, (Yit-Yit-1)/Yit*100, where Y represents GDP,
exports, and broad money
supply. This study employs data on broad money supply (M2) and
real exchange rate to act as
variables mediating the relationship between economic growth and
exports. The choice of the
control variables is motivated by existing theoretical and
empirical work in the growth literature.
For instance, Glasure and Lee (1999), Cheng and Lai (1997),
Piazola (1995), Ahsan, Kwan and
Balbir (1992) and Grier and Tullock (1989) supply evidence that
changes in the real money
-
11
supply are important determinants of GDP growth rate. Other
studies such as Glasure (1998),
Lee and Glasure (1998) and Marin (1992) document the importance
of real exchange rates in
transmitting the effects of external shocks (such as the oil
price shock in the 1970's and 1980's)
on trade balance.
The time series properties of GDP growth rate (GGrow), export
growth rate (EGrow),
real money supply (M2) and real exchange rate (RER) are first
investigated. Table 1 reports
ADF test results for stationarity of all the time series over
the various sample periods. For the
levels of the series, with the exception of the M2 variable for
India and Indonesia, none rejects
the null hypothesis of nonstationarity at the 5 percent level.
In general, the evidence suggests the
presence of I(1) for most of the variables.
Tests for cointegration are performed for those countries whose
variables were found to
be nonstationary in the levels (i.e., Korea, Malaysia and the
Philippines). Table 2 reports the
Johansen test results for cointegration. For the trace test, we
start with r#0 and move upwards.
We stop the first time we are unable to reject the null
hypothesis. For instance, in the case of
Korea, the hypothesis of r=0 is rejected as the computed value
of the test statistic (153.85) is
greater than the critical value (58.93). Similarly, the null
hypothesis of r#1 and r#2 is also
rejected. However, in the next step, the null hypothesis of at
most three cointegrating vectors
(r#3) cannot be rejected at the 5 percent level of significance.
Thus, there is evidence of three or
fewer CV’s in the system. The maximum eigenvalue test provides a
more conclusive evidence
regarding the exact number of CV’s in the system. The results
again confirm that there are three
cointegrating vectors (r=3). Based on these results it can be
said that there are three common
factors (permanent components) driving the entire system in
Korea. The results for Malaysia
and the Philippines suggest that there are two and three
cointegrating equations, respectively.
The existence of more than one cointegrating vector indicates
that the system under examination
is stationary in more than one direction and, hence, more
stable. In sum, the Johansen test results
suggest that there is a long run, steady state relationship
among exports, economic growth,
-
12
money supply and real exchange rates for Korea, Malaysia and the
Philippines.9 We applied
both the Reinsel and Ahn (1988) nor the Cheung and Lai (1993)
procedures to check for small
sample bias. Neither test provided evidence against our
cointegration results.
Given the cointegration results, the next stage in our model
building process requires the
construction of a multi-variate VECM for Korea, Malaysia and the
Philippines where the time
series are found to be cointegrated. Table 3 provides causality
results that are ascertained from
estimating the parameters in the GDP and export growth equations
given in Equations (3) and
(4), and the VAR system of equations. Several important
observations pertaining to the ELG
hypothesis can be made by first examining the results of the GDP
growth equation that is
exhibited in Panel A. First, the error-correction term, which
measures the speed of adjustment to
past shocks in equilibrium, emerges as an important channel of
influence for Korea. This
implies that the variables in the Korean system have a strong
tendency to adjust to their past
disequilibrium by moving toward the trend values of their
counterparts. Second, and perhaps
most important, in terms of the short run dynamics between
exports and GDP growth, it can be
seen that changes in exports have a significant causal influence
(in the Granger-sense) on GDP
growth rates for all the three countries - Korea, Malaysia and
the Philippines. Third, while on
the one hand exchange rate movements play an influential role in
the GDP growth equation for
Korea and Malaysia, on the other hand, money supply changes are
an important channel of
influence on the Philippine economic performance.
Panel B reports the results from the export growth (EGrow)
equation. It is theoretically
plausible for economic growth to cause export growth especially
if innovation and technical
progress in a growing economy help improve export performance.
Such evidence have in fact
been found for the United States (see Ghartey, 1993). Our
results indicate that the error-
correction terms are statistically significant for all countries
examined. This corroborates the
previous finding of a cointegrating relationship. With the
exception of the Philippines, the
9 India and Indonesia did not enter the cointegration system
since their money supply variables were found to be
stationary in the levels.
-
13
hypothesis that output growth does not prima facie causes export
growth in the short run is
rejected for all countries in the system (at the 10 percent
level of significance). Furthermore,
money supply changes in Korea and the Philippines are found to
have an important influence on
their exports.
Table 4 presents the short-run dynamic relationships that are
based on a VAR system, for
India and Indonesia. The paper employs de-trended values of the
time-series with appropriate
differencing in order to make the VAR analysis meaningful.
Specifically, the following VAR is
estimated:
India
t
p
l
s
k
jtj
n
j
it
m
i
RERMEGrowGGrowiGGrow εζδγβα ∑∑∑∑==
−
=
−
=
+∆++∆+∆+=∆1
2
0
111
(5)
t
p
l
s
k
jtj
n
j
m
i
i eRERMdEGrowcGGrowbaEGrow ξ∑∑∑∑==
−
==
+∆++∆+∆+=∆1
2
0
111
(6)
Indonesia
t
p
l
s
k
jtj
n
j
it
m
i
RERMEGrowGGrowiGGrow εζδγβα ∑∑∑∑==
−
=
−
=
+∆++∆+∆+=∆1
0
111
(7)
t
p
l
s
k
jtj
n
j
m
i
i REReMdEGrowcGGrowbaEGrow ξ∑∑∑∑==
−
==
+∆++∆+∆+=∆1
0
111
(8)
In the above equations, ) represents the first difference
operator and )2 is the second
difference operator. It is observed from Table 4, that while
exports lead economic growth for
India, the converse situation where economic growth stimulates
export performance is
documented for Indonesia. The control variables, exchange rates
and money supply, do not
-
14
carry statistically significant coefficients.
In sum, the results from Tables 3 and 4 taken together suggest
that (a) the export-led
growth hypothesis is clearly supported by our results for India,
Korea, Malaysia and the
Philippines, (b) a weak feedback relationship (i.e.,
bi-directional causality) emanating from
economic growth to exports is observed for Indonesia, Korea and
Malaysia; (c) exchange rate
movements have a significant influence on Korean and Malaysian
economic growth; and (d)
change in money supply have a pronounced impact on Korean and
Philippine export growth.
The above results are largely consistent with the development
economics literature in that export
promotion policies engender economic growth by encouraging and
making it feasible for firms
in the trade sector to efficiently and fully utilize their
economic resources. A re-allocation of
resources takes place within the economy from the inefficient
non-trade sector to the efficient
trade sector. The ensuing re-allocation of resources leads to a
more efficient allocation of a
nation’s resources and a higher level of material well-being in
the domestic economy. The
simultaneous short run feedback influence of Indonesian, Korean
and Malaysian economic
growth on their exports may be attributed to the favorable shift
in their country’s production
possibilities frontier (which are primarily driven by expanding
resource supplies and/or
technological progress) that enables its producers to sell their
surplus units to foreign markets.
To obtain additional insights into the short-run transmission
mechanisms between exports
and economic growth, impulse response functions (IRFs) are
computed. The study employs
Choleski decomposition to produce the orthogonal residuals
necessary to compute IRFs.10
The
Choleski decomposition requires that variables in the VAR be
ordered in a particular fashion.
Specifically, in the presence of cross-equation residual
correlation, a change in the higher-
ordered variable will result in a corresponding change in all
lower-ordered variables. The extent
of the response among the lower-ordered variables depends on the
degree of the residual
correlation. The present study employs two different ordering
schemes: (i) GGrow, EGrow, M2,
10
It must be noted that the Choleski decomposition is not without
any shortcomings (see Wheeler, 1999). A major
criticism of the Choleski decomposition is that it places a
recursive structure on contemporaneous relationships.
-
15
RER; and (ii) EGrow, GGrow, M2, RER. In the former ordering
system, GGrow is the higher-
ordered variable, and the corresponding response of EGrow to
changes in GGrow is presented in
Figures 1A-5A. In the second ordering system, EGrow takes
precedence over GGrow as the
higher-ordered variable, and its impact on GGrow is shown in
Figures 1B-5B. Of course, other
such ordering systems could be constructed, but our ordering
systems seem reasonable in light of
the information lags present and the deployment of annual data.
It is also consistent with the
principal purpose of our investigation, i.e., testing the
dynamic relationship between exports and
economic growth.
The IRFs (10 periods) from shocks of each variable are traced by
using the simulated
response of the estimated autoregressive system. An inspection
of the graphs reveals that the IR
analysis are in conformity with the causality tests. Looking at
the individual country impulse
response graphs, it can be observed that both GDP and exports,
on average, fully accommodate
shocks to the other variable within four to five periods. India,
however, stands out as an
exception to this observation. The country’s economic growth is
seen to take an extended period
of time to fully digest innovations in its export sector.
Furthermore, in the cases of Korea and
the Philippines, it is surprising to observe that the immediate
impact of a one-unit shock in
exports on economic performance is negative. However, the sign
is quickly reversed in the
subsequent periods as their economies respond positively to the
stimulus in exports. In
summary, the results from the impulse response functions support
the presence of significant
dynamic relationship between exports and economic growth.
V. Summary and Conclusions
During the past few decades, the export-led growth hypothesis
has been a topic of
sustained interest and controversy in the economic development
literature. This study improves
upon past studies by proposing a theoretically reasonable
approach to reexamine the GDP-export
relationship for five emerging economies of Asia namely — India,
Indonesia, Korea, Malaysia,
and the Philippines. The emerging countries of Asia provide an
excellent avenue to examine the
-
16
issues relevant to our study. Specifically, we utilize the
Johansen’s cointegration process for
testing the rank of the cointegration space spanned by the
stochastic process of exports, GDP
growth, real money supply, and real exchange rate. We then
employ the long run equilibrium
restriction from the cointegration model to examine the temporal
interrelationships between
these variables.
The study makes several important findings. First, we confirm
that export-led growth
nexus is inherently a steady state, long run phenomenon, in that
they are found to be cointegrated
in the cases of Korea, Malaysia and the Philippines. Second,
based on the VECM results, we
surmise that both exports and economic growth are related to
past deviations (error-correction
terms) from the empirical long run relationship. This implies
that all variables in the system
have a tendency to quickly revert back to their equilibrium
relationship. Finally, we find support
in favor of ELG hypothesis in that export growth has a causal
influence on economic growth for
all countries with the notable exception of Indonesia. This
implies that any rise in export growth
would have a positive influence on economic development in both
the long- and short-runs.
Evidence from the impulse response function corroborates this
finding while providing
additional insights into the transmission mechanism. From a
policy perspective, the results from
our study imply that countries having nascent economies should
adopt export-oriented measures
in conjunction with sound fiscal and monetary policies in order
to stimulate economic growth.
ACKNOWLEDGMENTS
The authors would like to thank an anonymous referee whose
helpful comments and suggestions
have been instrumental in improving the paper. The authors are
responsible for any remaining
errors.
-
17
References
Ahmad, J. and S. Harnihurun, 1995. Unit Roots and Cointegration
in Estimating Causality
Between Exports and Economic Growth: Empirical Evidence from the
ASEAN Countries.
Economics Letters 49: 329-334
Ahmad, J. and A.C. Kwan, 1991. Causality Between Exports and
Economic Growth: Empirical
Evidence from Africa. Economics Letters 37: 243-248
Ahsan, S.M., A.C. Kwan, and B.S. Sahni, 1992. Public
Expenditures and National Income
Causality: Further Evidence on the Role of Omitted Variables.
Southern Economic Journal 59:
623-634
Akaike, H, 1973. Information Theory and an Extension of the
Maximum Likelihood Principle. In
2nd International Symposium on Information Theory, edited by
B.N. Petrov and F. Craki,
Budapest: Akademiai Kiado
Balassa, B., 1978. Exports and Economic Growth: Further
Evidence. Journal of Development
Economics 5: 181-189
Balassa, B. 1985. Exports, Policy Choices and Economic Growth in
Developing Countries After
the 1973 Oil Shock. Journal of Development Economics 18:
25-35
Charemza, W.W. and Deadman, D.F. 1992. New Directions in
Economic Practice, (Edward
Elgar Publishing Limited, England)
Cheng, B.S. and T.W. Lai, 1997. Government Expenditures and
Economic Growth in South
Korea: A VAR Approach. Journal of Economic Development. 22:
11-24
Cheung, Y.W. and K.S. Lai, 1993. Finite-sample Sizes of
Johansen’s Likelihood Ratio Tests for
Cointegration. Oxford Bulletin of Economics and Statistics 55:
313-328
Chow, P.C.Y. 1987. Causality Between Export Growth and
Industrial Performance: Empirical
Evidence from the NIC’s. Journal of Development Economics 26:
53-63
Dickey, D.A. and W.A. Fuller, 1981. Likelihood Ratio Statistics
for Autoregressive Time Series.
Econometrica 49: 1057-1072
Engle, R.F. and C.W.J. Granger, 1987. Cointegration and Error
Correction: Representation,
Estimation, and Testing. Econometrica 55: 251-276
Feder, G., 1982. On Exports and Economic Growth. Journal of
Development Economics 12:
59-73
Ghartey, E.E., 1993. Causal Relationship Between Exports and
Economic Growth: Some
Empirical Evidence in Taiwan, Japan and the U.S. Applied
Economics 25: 1145-1153
-
18
Giles, D.E.A., J.A. Giles and E. McCann, 1992. Causality, Unit
Roots and Export-Led Growth:
The New Zealand Experience. Journal of International Trade and
Economic Development 2:
195-218
Glasure, Y., 1998. Trade Conflict Resolutions and Economic and
Export Performances: The
Korean Experience Between 1973-1994. presented at the American
Economic Association
Conference, Chicago, IL
Glasure, Y. U. and A.R. Lee, 1999. The Export-Led Growth
Hypothesis: The Role of the
Exchange Rate, Money and Government Expenditure from Korea.
Atlantic Economic Journal,
27: 260-272
Grier, K.B. and G. Tullock (1989), “An Empirical Analysis of
Cross-National Economic
Growth, 1951-80,” Journal of Monetary Economics, 259-276
Gonzala, J.,1994. Comparison of Five Alternative Methods of
Estimating Long Run Equilibrium
Relationships. Journal of Econometrics, 60: 203-233
Heller, P.S. and R.C. Porter, 1978. Exports and Growth: An
Empirical Investigation. Journal of
Development Economics 3: 191-193
Johansen, S., 1991. Estimation and Hypothesis Testing for
cointegration Vectors in Gaussian
Vector Autoregressive Models. Econometrica 59: 1551-1580
Johansen, S. and K. Juselius, 1990. Maximum Likelihood
Estimation and Inference on
Cointegration - With Applications to the Demand for Money.
Oxford Bulletin of Economics
and Statistics 52: 169-210
Jung, W.S. and P.J. Marshall, 1985. Exports, Growth and
Causality in Developing Countries.
Journal of Development Economics 18: 1-12
Kavoussi, R.M., 1984. Export Expansion and Economic Growth:
Further Empirical Evidence.
Journal of Development Economics 14: 241-250
Krueger, A., 1990. Perspectives on Trade and Development.
Chicago: University of Chicago
Press.
Kunst, R.M. and D. Marin, 1989. On Exports and Productivity: A
Causal Analysis. Review of
Economics and Statistics 71: 699-703
Lee, A.R. and Y.U. Glasure, 1998. The Political Dynamics of
Trade Negotiation: The Korean-
US Experience Between 1960 and 1990,” in Walter Jung, Xiaobing
Li eds, Korea and Regional
Geographics, New York, NY: University Press of America
Lutkepohl, H., 1989. Asymptotic Distributions of Impulse
Response Functions of Estimated
VAR Models with Orthogonal Residuals. Journal of Econometrics.
72: 116-125
-
19
Marin, D., 1992. Is the Export-Led Growth Hypothesis Valid for
Industrialized Countries?.
The Review of Economics and Statistics 74: 678-688
Michaely, M., 1977. Exports and Growth: An Empirical
Investigation. Journal of Development
Economics 4: 49-53
Michalopoulos, D. And K. Jay, 1973. Growth of Exports and Income
in the Developing World: A
Neo-classical View. Discussion Paper No. 28, Agency for
International Development,
Washington, D.C.
Osterwald-Lenum, M., 1992. A Note with Quantiles of the
Asymptotic Distribution of the
Likelihood Cointegration Rank Test Statistics: Four Cases Oxford
Bulletin of Economics and
Statistics 54: 461-472
Phillips, P.C.B., 1991. Optimal Inference in Cointegrated
Systems. Econometrica 59: 283-
306
Piazola, M., 1995. Determinants of South Korean Economic Growth,
1955-1990. International
Economic Journal 9: 109-133
Ram, R., 1987. Exports and Economic Growth in Developing
Countries: Evidence from Time-
Series and Cross-section Data. Economic Development and Cultural
Change 36: 51-72
Rana, P.B., 1988. Exports, Policy Changes and Economic Growth in
Developing Countries
After the 1973 Oil Shock: Comment. Journal of Development
Economics 18: 261-264
Reinsel, G.C. and S. K. Ahn, 1988. Asymptotic Distribution of
the Likelihood Ratio Test for
Cointegration in the Nonstationary Vector AR Model. Technical
Report, University of
Wisconsin, Madison.
Serletis, A., 1992. Export Growth and Canadian Economic
Development. Journal of
Development Economics 38: 133-145
Sharma, C.S. and D. Dhakal, 1994. Causal Analyses Between
Exports and Economic Growth in
Developing Countries. Applied Economics 26: 1145-1157
Toda, H.Y. and P.C.B. Phillips, 1993. Vector Autoregressions and
Causality. Econometrica 61:
1367-1393
Tyler, G.W., 1981. Growth and Export Expansion in Developing
Countries: Some Empirical
Evidence. Journal of Development Economics 9: 121-130
Wheeler, M., 1999. The Macroeconomic Impacts of Government Debt:
An Empirical Analysis
of the 1980s and 1990s. Atlantic Economic Journal 27:
273-284
-
20
Table 1. ADF Unit Root Test
Country/Period SeriesR Level First Difference Second
Difference
India 1950-1998 GGrow Tµ = -2.41 Tµ = -6.70** —
TJ = -5.25 TJ = -6.66** —
EGrow Tµ = -2.65 Tµ = -4.00**
TJ = -3.07 TJ = -4.06** —
RER Tµ = -2.45 Tµ = -1.15 Tµ = -4.84**
TJ = -2.05 TJ = -2.54 TJ = -4.90**
M2 Tµ = -3.82** — —
TJ = -3.94** — — Indonesia 1969-1998 GGrow Tµ = -1.69 Tµ =
-8.07** —
TJ = -3.19 TJ = -8.74** —
EGrow Tµ = -2.31 Tµ = -4.40** —
TJ = -2.99 TJ = -4.42** —
RER Tµ = -0.94 Tµ = -3.67** —
TJ = -1.98 TJ = -4.20** —
M2 Tµ = -16.20** — —
TJ = -15.98** — — Korea 1953-1998 GGrow Tµ = -1.58 Tµ = -4.94**
—
TJ = -2.59 TJ = -5.22** —
EGrow Tµ = -0.33 Tµ = -3.72** —
TJ = -2.60 TJ = -3.79** —
RER Tµ = -1.07 Tµ = -3.56** —
TJ = -2.41 TJ = -3.63** —
M2 Tµ = -1.75 Tµ = -4.37** —
TJ = -2.81 TJ = -4.50** —
* indicates statistical significance at the 5% level. Tµ =
without trend; TJ = with trend. The critical values at the 5%
significance level are –2.97 and –3.58, respectively, for without
trend and with trend. The critical values at the 10%
significance level are –2.60 and –3.18, respectively, for
without trend and with trend.
R GGrow = GDP growth rate; EGrow = export growth rate, RER =
real exchange rate and M2=broad money supply.
-
21
Table 1. ADF Unit Root Test (Continued)
Country/Period SeriesR Level First Difference Second
Difference
Malaysia 1955-1998 GGrow Tµ = -2.38 Tµ = -5.53
** —
TJ = -2.40 TJ = -5.48**
—
EGrow Tµ = -2.01 T µ = -4.80** —
TJ = -2.40 T J = -4.80**
—
RER Tµ = -1.20 Tµ = -2.74* —
TJ = -0.06 TJ = -3.20*
—
M2 Tµ = -1.94 Tµ = -3.90**
—
TJ = -1.94 TJ = -3.88** —
The Philippines 1949-1998 GGrow Tµ = -2.77 Tµ = -6.53
** —
TJ = -2.83 TJ = -6.61**
—
EGrow Tµ = -2.62 Tµ = -5.40** —
TJ = -2.85 TJ = -5.41** —
RER Tµ = -1.82 Tµ = -3.39**
—
TJ = -0.66 TJ = -4.12**
—
M2 Tµ = -2.18 Tµ = -4.28**
—
TJ = -2.47 TJ = -4.28**
—
**
indicates statistical significance at the 5% level. Tµ = without
trend; TJ = with trend. The critical values at the 5%
significance level are –2.97 and –3.58, respectively, for
without trend and with trend. The critical values at the 10%
significance level are –2.60 and –3.18, respectively, for
without trend and with trend. R GGrow = GDP growth rate; EGrow =
export growth rate, RER= real exchange rate and M2=broad money
supply.
-
22
Table 2. Multi-variate Cointegration Tests
Trace Test Maximum Eigenvalue Test Country
Test Critical Null Test Critical
(Null hypothesis) Statistic Value hypothesis Statistic Value
Korea
r=0 153.85** 58.93 r=0 76.55**
31.00
r#1 77.30**
39.33 r#1 46.82**
24.35
r#2 30.48**
23.83 r#2 20.40**
18.33
r#3 10.09 11.54 r#3 10.09 11.54
Malaysia
r=0 94.08** 58.93 r=0 41.28**
31.00
r#1 52.80**
39.33 r#1 36.42**
24.35
r#2 16.38 23.83 r#2 16.13
18.33
r#3 0.25 11.54 r#3 0.25 11.54
Philippines
r=0 123.07** 58.93 r=0 59.72**
31.00
r#1 63.28**
39.33 r#1 35.72**
24.35
r#2 27.56** 23.83 r#2 26.42** 18.33
r#3 1.14 11.54 r#3 1.14 11.54
**
indicates statistical significance at the 5% level. The critical
values are obtained from the Microfit 4.0 program.
-
23
Table 3. Multi-variate Granger-Causality Tests Based on VECM
(F-Statistics)
Panel A: GDP Growth Equation (Dependent Variable: GGrow)S
INDEPENDENT VARIABLES
Country zt-1 'EGrow 'GGrow 'RER 'M2 LagsR
Korea 22.41***
22.56***
1.25 4.44**
0.64 1, 1, 1, 1
Malaysia 0.71 6.56***
0.24 8.44***
1.48 2, 1, 1, 1
Philippines 1.06
3.59**
2.34 0.12 3.57**
3, 1, 1, 1
Panel B: Export Growth Equation (Dependent Variable: EGrow)S
INDEPENDENT VARIABLES
Country zt-1 'EGrow 'GGrow 'RER 'M2 LagsR
Korea 4.51**
0.92 3.11* 1.09 24.86
*** 1, 1, 1, 1
Malaysia 14.56***
2.84* 2.65
* 0.05 0.90 1, 1, 1, 1
Philippines 37.27***
1.55 0.97 0.04 21.71***
1, 1, 1, 1
*,
**,
*** associated with the F-statistics represent statistical
significance at the 10%, 5% and 1% level respectively.
The standard t-test is used to determine the level of marginal
significance for the error correction term (zt-1). S Results for
Panels A and B are obtained from the estimation of Equations (3)
and (4) respectively.
R Lags represent the optimal lag length employed for GGrow and
EGrow as determined by the AIC.
-
24
Table 4. Causality Tests based on VAR (F-Statistics)
Panel A: GDP Growth Equation (Dependent Variable: GGrow)S
INDEPENDENT VARIABLES
Country 'EGrow 'GGrow 'RER 'M2 LagsR
India 5.95***
16.22***
0.41 1.08 2, 2, 2, 2
Indonesia 0.46 1.21 0.09 1.29 2, 2, 2, 2
Panel B: Export Growth Equation (Dependent Variable: EGrow)S
INDEPENDENT VARIABLES
Country 'EGrow 'GGrow 'RER 'M2 LagsR
India 15.38***
0.10 0.07 0.28 2, 2, 2, 2
Indonesia 2.89* 2.79
* 0.55 0.29 2, 2, 2, 2
*,
**,
*** associated with the F-statistics represent statistical
significance at the 10%, 5% and 1% level respectively.
S Results for Panels A and B are obtained from the estimation of
Equations (5), (6), (7) and (8) respectively.
R Lags represent the optimal lag length employed for GGrow and
EGrow as determined by the AIC.
-
25
-6
-4
-2
0
2
4
6
8
1 2 3 4 5 6 7 8 9 10
Figure 1A
India: Response of GDP Growth to Exports
Sta
nd
ard
Dev
iati
on
Periods
-40
-20
0
20
40
60
80
100
1 2 3 4 5 6 7 8 9 10
Figure 2A
Indonesia: Response of GDP Growth to Exports
Periods
Sta
ndar
d D
evia
tion
-4
-2
0
2
4
6
8
10
1 2 3 4 5 6 7 8 9 10
Figure 3A
Korea: Response of GDP Growth to Exports
Periods
Sta
ndar
d D
evia
tion
-4
-2
0
2
4
6
8
10
1 2 3 4 5 6 7 8 9 10
Figure 4A
Malaysia: Response of GDP Growth to Exports
Periods
Sta
nd
ard
Dev
iati
on
-4
-2
0
2
4
6
8
1 2 3 4 5 6 7 8 9 10
Figure 5A
Philippines: Response of GDP Growth to Exports
Periods
Sta
nd
ard
Dev
iati
on
-10
-5
0
5
10
15
1 2 3 4 5 6 7 8 9 10
Figure 1B
India: Response of Exports to GDP Growth
Periods
Sta
ndar
d D
evia
tion
-50
0
50
100
150
200
1 2 3 4 5 6 7 8 9 10
Figure 2B
Indonesia: Response of Exports to GDP Growth
Periods
Sta
nd
ard
Dev
iati
on
-10
-5
0
5
10
15
20
1 2 3 4 5 6 7 8 9 10
Figure 3B
Korea: Response of Exports to GDP Growth
Periods
Sta
nd
ard
Dev
iati
on
-
26
-10
-5
0
5
10
15
20
1 2 3 4 5 6 7 8 9 10
Figure 4B
Malaysia: Response of Exports to GDP Growth
Periods
Sta
ndar
d D
evia
tion
-20
-10
0
10
20
30
40
1 2 3 4 5 6 7 8 9 10
Figure 5B
Philippines: Response of Exports toDGDP Growth
Periods
Sta
ndar
d D
evia
tion