Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

Paper 4 Growth FDI and Exports in Pakistan

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Forman Journal of Economic Studies

Vol. 6, 2010 (January–December) pp. 65-84

Growth, FDI and Exports in Pakistan:


A Co-integration Analysis

Ghulam Shabbir and Amjad Naveed1

Abstract
Most of the empirical work has identified openness with trade to
analyze the impacts of outward-oriented policies in developing
countries. The exports promote both growth and foreign trade,
and foreign direct investment is one of the important factors of
foreign trade. Therefore, this study analyses the long run co-
integrating relationship among foreign direct investment,
exports and GDP. The result shows that long run relationship
exists between growth and exports but not with FDI. Therefore,
FDI is not a significant determinant of growth in the long run.
This study also analyses the causality among these variables in
the short-run, FDI is also not affecting the growth, but growth
affects both FDI and exports. Moreover, export and FDI are
also not significantly influencing each other in the short-run.
I. Introduction
A large numbers of empirical studies have analysed the impacts of
foreign direct investment (FDI) and exports on economic growth. Most of the
studies focussed the export-led-growth hypotheses (ELG) to analyze the
relationship between exports and growth. The ELG hypothesis (ELG)
suggests a positive correlation between export and growth, and also
considered exports as a main determinant of overall economic growth.
Because export sector; generates positive externalities through more efficient
management and improved production techniques (Feder, 1982), increase
productivity by offering potential for scale economies and alleviate foreign
exchange constraints and greater access to international market. Therefore,
export-oriented policies always positively contributed to economic growth.
The Growth rates attained by the South-East Asian tigers since the mid 1960s
(higher than those achieved elsewhere in the world) are normally cited as the

1
The authors are Assistant Professors at Department of Economics, Forman Christian College
(A Chartered University), Lahore. E-mail: ghulamshabbir@fccollege.edu.pk
Shabbir and Amjad

best example for the success of the ELG strategy. These export oriented
policies not only increased the trade but also foreign direct investment.
The flow of FDI to was significantly increased during the decades of
1980s & 1990s, and is considered as an important source of advanced
technologies for the recipient countries (Barrell and Pain 1997, Cuadros 2001
and Henrik et al 2004). The FDI has grown at least twice rapidly as trade
(Meyer, 2003). The very first reason for this increase is the shortage of capital
in developing countries that leads to higher marginal productivity of capital in
these nations. The second reason is capital owner’s desire for higher return on
their capital. According to Borensztein et al, (1998), FDI played a key role in
the technological progress of the recipient countries. Besides this, FDI also
effects economic growth by generating productivity spill over. Blomstrom
(1986) found that FDI has positive significant spillover effects on the labour
productivity of domestic firms and growth of domestic productivity in
Mexico.
The developing countries have changed their strategies from import
substitution to export-orientation. These policies were in line with the body of
literature, created to examine the determinants of exports & FDI and their
impacts on economic growth. The FDI has multi-dimensional issues that
include trade, employment, cost of production etc2. The inward FDI not only
stimulate the local investment but also increase the host country’s export
capacity. Therefore, liberalization policies of developing countries increased
their trade as well as inflow of FDI (Goldberg and Klien, 1999).
Pakistan has also followed the export-oriented policies and reframed
its commercial policy towards fewer and fewer controls. The tariffs rates were
reduced to enhance the degree of openness. These policies have significantly
affected the infow of FDI and it reached to $500.27 million in the decade of
1990s as compared to $88.83 million in the decades of 1980s. These
liberalization policies further enhanced the inflow of FDI that reached to
$3.52 billion in 2005-2006, $5.14 billion in 2006-2007 and $5.41 billion in
2007-2008. But poor economic performance and terrorist’s activities badly
affected the flow of FDI. The FDI was suddenly dropped to the level of $3.72
billion in 2008-09 and $1.72 billion in 2009-10. This reduction in FDI was
53% as compared to previous year and 68% as compared to its highest value
in 2007-08. The main sectors that have shows largely reduction in FDI were

2
See, Freenstra and Hanson (1997).

66
Growth, FDI and Exports in Pakistan

telecommunication and financial services. Only these two sectors contribute


81% of total reduction in FDI3.
This study has tried to analyse the long run relationship among GDP,
Exports and FDI in the present scenario. We used annual data series and co-
integration technique to investigate this relationship. Before proceeding to co-
integration, we checked all data series for stationarity which one is basic
condition for cointegration. For this purpose, we applied two tests;
Augmented Dickey-Fuller (ADF) and Kwiatkowski, et al. (KPSS). Both tests
results show that all data series are non-stationary but their first differences
are stationary. This implies that all data series are having integrating order one
I(1). We used Johansen and Juselius (1990) procedure to examine the long run
relationship among variables. The results show that GDP has long run
relationship with exports but not with FDI. Therefore, FDI is not considered
as a determinant of GDP in the long run. We also tried to capture the short-
tern dynamics by using the error correction model (ECM). The results of
Granger Causality tests predict that FDI is not affecting the GDP whereas
GDP affects both FDI and Exports. The study’s results also argued that
exports and FDI are also not causing each other in the short-run.
The rest of the study is organized as follows: Section II deals with
literature review and theoretical framework, section III is specified to specify
methodology and data sources whereas section IV discusses empirical results.
The last section gives the brief summary of study and conclusion.
II. Literature Review and Theoretical Framework
The existing empirical studies focused a number of factors that
determined the flow of FDI but technology and domestic market size got
special attention. The FDI will be considered as growth enhancing if it
positively influences technology or it increases production through spillover
and technological transfers (Shiva and Somwaru, 2004). Some studies like
Lim (2001), and Moosa (2002) tried to create a link between domestic market
sizes, differences in factor costs and location of FDI. The importance of
market size and its growth will further intensified in case of those foreign
investors who prefer to invest in those industries, which exhibit relatively
large economies of scale. The most of the studies have used GDP,
GDP/Capita or growth in GDP as a proxy to measures the market size. All
these studies predict a positive relation between FDI and market size.

3
Pakistan Economic Survey (various issues).

67
Shabbir and Amjad

Therefore, growth augmenting FDI and its positive relation with market size
created a bi-directional behaviour between two variables FDI and GDP. This
bi-directional behaviour also becomes the cause for simultaneity bias between
these two variables. In contrast, Charkovic and Levine (2005) have shown
either insignificant or negative impact of FDI on growth. This might be due to
crowding out effect of FDI on domestic capital.
A similar two-way causality discussion exists in empirical literature
for exports and GDP. The first way makes export led growth hypothesis
(ELG), while the other way is also equally appealing hypothesis that GDP
growth leads to exports growth also. According to Shiva and Somwaru
(2004), argued that export growth leads to increase in factor productivity. This
might be result of gains acquired from increasing returns to scale. Moreover,
export’s growth enhances the foreign exchange reserves that make possible to
increase the import of capital/technology-intensive intermediate inputs. The
rises in exports also enhance the efficiency of exporters and make them able
to compete in foreign markets. This practice results in domestically
improvement in technology and grooming of domestic entrepreneurs. The
open trade regime supports the inflow of better technologies that lead to better
investment environment (Grossman and Helpman, 1991). This hypothesis is
further supported by the findings of studies like Giles and Williams (2000)
and, Ahmad, Alam and Butt (2004).
On the other hand, Jung and Marshal (1985) supported the second
hypothesis and suggested that in a growing economy, the process of
technological change and learning are not the results of government’s export
promotion policies. This process might be the result of human capital
formation, cumulative productive process, transfer of technology via direct
investment or physical capital accumulation. This increased growth leads to
increase in the production of goods and if domestic market not able to absorb
it. Then, exporters look outward to sell this increased output. This implies that
increased growth leads to increase in exports. This causal relationship
becomes negative if increased output result in decrease in export instead of
increase. This might be due to domestically increased consumers demand for
these exportable goods.
At last, there exists a third bi-directional causality between FDI and
export. According to Hsiao and Hsiao (2006), exports increase the inflow of
FDI. This might be through paving the way for FDI by gathering information
about the host country that helps in reduction of investors’ transaction costs.

68
Growth, FDI and Exports in Pakistan

The FDI also become the source of reduction in exports by serving the foreign
markets through establishing production facilities there. Therefore, Petri and
Plummer (1998) argued that it is not clear whether causality runs from FDI to
exports or exports to FDI. Some studies have also mentioned other aspects of
FDI like Gray (1998) pointed out market seeking FDI or efficiency seeking
FDI, Kjima (1973) mentioned whether FDI is trade oriented or anti trade
oriented and Vernon (1966) explored that FDI can be at the early product life
cycle stage (substitute) or at the mature stage (complement). Moreover,
Johanson & Widershen (1975), Nicholas (1982) and UNCTAD (1996) tried to
analyze the linkages between exports and FDI. These studies suggest that
manufacturing firms first start trade with foreign nation before making the
investment in these economies. They consieder it less risky than FDI. After
getting full informations about these countries economies, political and social
conditions, these firms establish subsidiaries in these nations and then
subsidiary exports. Therefore, FDI-export nexus is also as complicated as the
other bivariate causal discussion.
This study uses a multivariate cointegration instead of bi-variate
causality tests using three variables FDI, exports and GDP. When we consider
two relations like export-GDP and GDP-FDI, then relation between exports
and FDI might be through GDP. Because export’s growth leads to GDP
growth and GDP growth make possible further inflow of FDI. This implies
that exports are the driving force for FDI through GDP. After pinpointing the
channeling effect, it is necessary to explore that this established causality is
either effective in the short run, long run or both.
III. Methodology and Data Sources
This study used Vector Auto Regressive (VAR) model and co-
integration to investigate the short run and long run relationship among
exports, FDI and Growth. For VAR model estimation and co-integration
analysis, it is necessary that all data series must have same cointegration
order. Therefore, first we examined the all data series for the presence of unit
root (stationarity test for data series). The stationarity of data series can be
verified by various techniques like plotting the correlogram of the data series,
applying Dickey and Fuller (1979), Augmented Dickey-Fuller (1981),
Phillips-Perron test (1988) and Kwiatkowski, et al. (KPSS, 1992). Out of
these, we used two main tests KPSS and ADF to check the stationarity
problem of all three data series.

69
Shabbir and Amjad

3.1. Augmented Dickey-Fuller (ADF) Test


The Dickey-Fuller (1979) used first order auto-regressive model to test
the stationarity of data series by including drift and linear time trend in the
model as follows:
Yt is a random walk: ∆ Y t = δ Y t −1 + ε t
Yt is a random walk with drift: ∆ Y t = β 1 + δ Y t −1 + ε t

Yt is a random walk with drift around a stochastic trend

∆ Y t = β 1 + β 2 t + δ Y t −1 + ε t

According to Dickey and Fuller (1979), under the null hypothesis (δ =


0), the estimated t value of the coefficient of Yt-1 follows τ (tau) statistic
instead of following t distribution even in large samples. In empirical
literature, tau statistic or test is called Dickey-Fuller test.4 This test suggests
that if (δ = 0) is rejected against the (δ < 0), then concerned data series is
stationary (no unit root). This test assumes that error term (εt) is uncorrelated.
Dickey and Fuller also developed a new test to resolve this issue. This test is
known as Augmented Dickey-Fuller (1981), ADF test. This test adjusts the
Dickey-Fuller test to take care of serial correlation in the error terms by
including lagged difference terms of dependent variable (∆Yt) in the above
mentioned equations and becomes as follow:
m
∆ Yt = β 1 + β 2 t + δ Y t −1 + α i ∑
i=1
∆ Y t −1 + ε t

This test also faces some weaknesses; Blough (1992) discussed the
trade-off between the size and power of unit root tests, these tests have either
a high probability of falsely rejecting the null of non-stationarity when the
DGP (data generating process) is a nearly stationary process, or low power
against a stationary alternative. Because, in finite samples, it was observed
that some unit root processes display their behaviour closer to stationary
(white noise) than to a non-stationary (random walk), while some trend-
stationary processes behave more likely to random walks (Harris, 1995).
Therefore, considering these issues, we also used KPSS test to examine the
unit root problem of all data series.

4
For detail see, Basic Econometrics (4th edition), page 815.

70
Growth, FDI and Exports in Pakistan

3.2. KPSS Test


The Augmented Dickey-Fuller test, tested the null hypothesis (unit
root exists) against the alternative of no unit root (stationarity). However,
Charemza & Syczewska (1998) and Maddala & Kim (1998) argued that null
hypothesis of no unit root (stationary) must be tested against its alternative,
unit root exists (non-stationary) also. This implies that the null hypothesis of
data series is stationarity (no unit root) is tested against the alternative of a
unit root. For this purpose, Kwiatkowski et al. (1992) developed a test that is
known as KPSS test. This test is based on LM test and formulated as follows:
Yt = δ t + rt + ε t
Where ε t is stationary but rt is random walk component and treated as:
rt = rt −1 + υ t υ t ~ i.i.d (0, σ 2υ )
The initial value of r0 is treated as fixed and it plays as a role of an intercept.
The stationary hypothesis is simply, H 0 :σ 2υ = 0 (Stationary process). Then
KPSS statistic can be calculated as:
T

−2
∑ S t2
KPSS = LM =T t =1

s 2
(l )
l
s 2 (l ) = T −1
∑ e t2 + 2 T −1
∑ w τl ∑ e t e t −τ
τ =1

Where S 2 (l ) indicates long-run


τ
variance, ωτl =1 − is Bartlett weight function
l +1
t
and S = ∑ ei t = 1,2,..., T is partial sum process of the residuals.
i =1
This study used both ADF and KPSS tests to check the stationarity
problem of data series. According to Baillie et al. (1996), combination of both
ADF and KPSS tests statistic make following four possible outcomes:
(i) Rejection of null hypothesis by ADF and failure to reject it by the
KPSS will considered as a strong evidence of covariance stationary
I(0) process.
(ii) Failure to reject null hypothesis by ADF and rejection by the KPSS
statistic will strongly recommend a unit root I(1) process.

71
Shabbir and Amjad

(iii) Failure to reject null hypothesis both ADF and KPSS will indicate that
data is not able to give sufficient information required for the long-run
characteristics of the process.
(iv) Rejection of null hypothesis by both tests ADF and KPSS indicate that
the process is described by neither I(0) nor I(1) processes.
3.3. Multivariate Cointegration Analysis and Error Correction
Modeling
A common method for testing the long run relationship (cointegration)
among the economic series is the Engel-Granger’s two-step bivariate,
residual-based method. According to Banerjee et al. (1990), “this method is
incapable to deal multivariate cases due to; a priori assumption of a single co-
integrating vector in the system, it tends to yield biased parameter estimates in
small samples and is sensitive to the choice of endogenous variables in the co-
integrating regression”. Therefore, Johansen (1988), Johansen and Juselius’
(1990), Maximum Likelihood (ML) procedures are considered the best
alternative to the Engle-Granger technique. The main charm in this method is
its capability to test the possibility of multiple co-integrating relationships
among the variables. Johansen and Juselius (1990), further formulated time
series in the form of reduced rank regression. In this procedure, they
calculated the ML estimates in the multivariate cointegration model with the
help of Gaussian Errors. This model is based on the Error Correction that can
be shown as follows:
p −1
∆X = µ + ∑ Γi ∆X t −i + ∏ X t −1 + ε t (A)
i =1
Where Xt is an (n*1) column vector of k variables, µ is an (n*1) vector
of constant terms, Γ and Π represent coefficient matrices, ∆ is a difference
operator, p denotes the lag length and εt is i.i.d. k-dimensional Gaussian Error,
which has mean zero and variance matrix (white noise disturbance term). The
coefficient matrix Π is called impact matrix that contains information about
the long-run relationships among variables. Equation (A) shows VAR model
in first differences, except the term, lagged level of Xt-1 and an error correction
term. This error correction term provides information about the long run
relationship among variables in the vector Xt. This way of specifying the
equation system is known as VECM model. This model gives information
about short run as well as long run adjustment to changes in Xt through the
estimates of Γ and Π, respectively. The VECM equation allows three
following model specifications.

72
Growth, FDI and Exports in Pakistan

(i) If Π is of full rank, then Xt is stationary in levels and a VAR in levels


is an appropriate model.
(ii) If Π has zero rank, then it contains no long run information, and the
appropriate model is a VAR in first differences.
(iii)If the rank of Π is a positive number, r and is less than k (where k is
the number of variables in the system), there exists matrices α and β, with
dimensions (k x r), such that β α =Π. In this presentation β contains the
coefficients of the r distinct long run co-integrating vectors that render β’
Xt stationary, even though Xt is itself non-stationary, and α contains the
short run speed of adjustment coefficients for the equations in the system.
Johansen’s methodology requires the estimation of the VAR, equation
(A) because its residuals are used to compute two likelihood ratio (LR) test
statistics that can be used in the determination of the unique cointegrating
vectors of Xt. The first test, which considers the hypothesis that the rank of Π
is less than or equal to r cointegrating vectors, is given by the ‘trace test’ as
below:
n
Trace = − T ∑ ln(1 − λi )
i = r +1

The second test statistic is known as the maximal eigen value test or
‘max test’ that computes the null hypothesis that there are exactly r
cointegrating vectors in Xt and is as follows:
λmas = − T ln(1 − λr )
The distributions for these tests are not usual chi-squared distributions.
The asymptotic critical values for these likelihood ratio (LR) tests are
calculated by Johansen & Juselius (1990) and Osterwald-Lenum (1992)
through numerical simulations.
For short run dynamic analysis, we included an Error Correction (EC)
term in the differenced model to capture the equilibrium relationship among
the co-integrating variables in their dynamic behavior, following the Granger
Representation Theorem. This addition of EC in the first differentiated VAR
model, make possible to separate the long-term relationship among economic
variables from their short-run responses. This will also determine the direction
of the Granger causality. Following Johansen and Juselius (1990), the
corresponding ECM can be written as follows:

73
Shabbir and Amjad

∆GDPt = µ1 + δ1∆GDPt −1 + δ 2 ∆FDI t −1 + δ 3∆EXPt −1 + Φ1 ECT 1t −1 + ε1


∆FDI t = µ 2 + δ 4 ∆GDPt −1 + δ 5 ∆FDI t −1 + δ 6 ∆EXPt −1 + Φ 2 ECT 2t −1 + ε 2
∆EXPt = µ3 + δ 7 ∆GDPt −1 + δ 8∆FDI t −1 + δ 9 ∆EXPt −1 + Φ 3 ECT 3t −1 + ε 3
We used annual data series to investigate long run as well as short run
relationships among these three variables; FDI, exports and GDP for Pakistan.
All data series are collected from International Financial Statistic (IFS) for the
period of 1960 to 2010, except FDI that is taken from the annual publications
of State Bank of Pakistan5.

IV. Empirical Results


We used Johansen & Juselius (1990) Cointegration technique to
investigate the long run equilibrium relationship among the variables; FDI,
exports and growth. For this, we need to address the issue of unit root for all
data series. We used two unit root tests ADF and KPSS.
4.1. Unit Root Test Results
Before applying the unit root tests, we plot the graphs for all data
series. These graphs show that all data series have trend6, so we included trend
in the model in case of unit root tests. The results of both tests are presented in
table 1. The statistic estimated with the help of both unit root tests gave the
strong evidence of difference stationary because ADF is failed to reject the
Table: 1. Unit-root Analysis on Nominal Variables
Variables Level First difference Second difference Decision
ADF KPSS ADF KPSS ADF KPSS
GDP -1.30 0.237 -5.88 0.145 -6.23 0.065 Both
(-3.5162) [0.146(2)] (-3.53) [0.463(5)] (-2.94) [0.463(2)] recommend for
I(1)
FDI -2.36 0.251 -7.55 0.152 -7.57 0.0461 Both
(-3.5162) [0.146(2)] (-2.94) [0.463(3)] (-2.94) [0.463(2)] recommend for
I(1)

EXP -2.92 0. .194 -4.92 0.060 -5.60 0.029 Both


(-3.5162) [0.146(2)] (-2.94) [0.463(1)] (-2.94) [0.463(2)] recommend for
I(1)

Values in parenthesis () are MacKinnon critical values while in KPSS tests, we used Automatic
bandwidth selection (maxlag) and Autocovariances weighted by Quadratic Spectral kernel.

5
Annual Publications of State Bank of Pakistan (various issues).
6
For detail see, figure 1 in appendix.

74
Growth, FDI and Exports in Pakistan

null hypothesis while KPSS rejected the null hypotheses in all cases. As all
data series are having same integrating order I(1), so we further proceed for
the investigation of long run equilibrium analysis through co-integration. The
co-integration concept is closely linked with the notion of long-run
equilibrium in case of economic theory. According to this notion, the
variables in a system may deviate from their steady sate value in the short run
but in the long run, it is expected that they ultimately converge to their steady
state. To examine the long run equilibrium relationship among FDI, exports
and GDP, we used Johansen and Juselius (1990) procedure.
4.2. Multivariate Cointegration Results
The results of unit root tests predict that all data series are first
difference stationary I(1). We used VAR to investigate the relationship
between variables. First, we determined the lag length of the VAR model
using AIC and SC criterion. We also applied LR test by estimating VAR
twice, each with different lags (to compute LR statistic). We estimated VAR
with lag order 2 and 37. Since the lower the values of AIC or SC, statistics
give the evidence for better model. On the basis of these values, we can say
that model one with lag order 2 is more parsimonious than the model with lag
order 3. The same is confirmed by the results of LR test.
We estimated VAR (2) to investigate the long run relationship among
Exports, GDP and FDI. We applied Johansen (1988) procedure to estimate the
VAR (2) cointegration analysis because it gives the most efficient estimate of
the long-run relationship between non-stationary variables at level. The Trace
test results indicate that only one co-integrating vector exists. The co-
integrating vector is shown in figure 1. The results of trace test are presented
Figure: 1. Co-integrating Vector
4.8
vector1

4.6

4.4

4.2

4.0

3.8

3.6

1965 1970 1975 1980 1985 1990 1995 2000 2005

7
For results see, table A1 in appendix.

75
Shabbir and Amjad

in table 2. With one cointigrating vector (r = 1) and three variables (k=3),


there are (k-r = 2) common stochastic trend driving the system. As the co
integrating vector is not identified, cannot be interpreted without further
restrictions. Therefore, we assume that the cointegration rank is one and GDP
is normalized to have a unit coefficient in order to identify β. The results are
reported in table 3.
Table: 2. Johansen’s Cointegration Test VAR (2)

Hypothesis Trace Test P Value Decision

H0: r=0, r>0 35.321 0.010* Trace test results


H0: r<1, r>1 8.8833 0.383 indicate that there
H0: r<2, r>2 0.17264 0.678 is only one co-
integrating vector

Table: 3. Identification of Restriction

Variables GDP FDI Export


Coefficient of β 1.000 0.101 -0.8001
(standard errors) (0.000) (0.046) (0.0651)
Coefficient of α -0.0487 -1.6621 0.41754
(standard errors) (0.0223) (0.429) (0.152)

We analysed co-integrating vector further by imposing different


restrictions on β matrix. For example, we impose zero restriction on
coefficient of FDI in β matrix. This restriction is accepted by applying LR
test and indicates that FDI is not co-integrated with GDP and Exports in the
long run. These results differ from previous studies carried on Pakistan8. This
might be due to difference in sample periods or estimation technique. We
excluded the FDI variable from long run co-integrating vector due to its zero
performance and test results are presented in table 4. We also imposed zero
restriction on export variable but it was rejected9. So we concluded from long
run co-integrating vector that GDP and exports are co-integrating in long run
but not with FDI. Beside co integrating vector, we also imposed and tested
different restriction on adjustment (weighting) coefficient (α). The variables

8
See for example, Ahmad et al. 2004.
9
For results see, appendix table A2.

76
Growth, FDI and Exports in Pakistan

GDP and FDI have negative sign in α metrics. This implies that both
variables also respond to correct their own past disequilibrium error.
According to the signs of the adjustment coefficient, co-integrating relation is
error correcting10.
Table: 4. Identification Restrictions on β

Variables GDP FDI Export LR Test Status

Coefficient 1.000 0.000 -0.662 2.89 Accepted


(standard
(0.000) (0.000) (0.029) Prob. (0.09)
errors)
Coefficient 1.000 0.4356 0.000 16.065 Rejected
(standard
(0.000) (0.072) (0.000) Prob. (0.0001)
errors)

Moreover, we also applied a test on constant, whether it should be


included inside the cointegrating vector or not. By imposing restriction on
constant, two models are estimated; one restricted and other is unrestricted.
Then, we used LR test to select more parsimonious model. The null
hypothesis assumes that constant lie inside the cointegrating vector and is
rejected. It means, we cannot restrict the constant inside the cointegratingn
vector. It is due to the fact that series has trend and it was confirmed by the
graph of these series. The test results are presented in table 5. All these results
Table: 5. Restriction on Constant

Log-Likelihood LR Statistics P value


with chi2(1)
Restricted
Constant 78.20 17.56 0.0002
Un restricted 86,98
Constant

confirm that long-run equilibrium relationship exists between GDP and export
but FDI is not co-integrating with these variables. We also carried out short
run dynamic analysis to capture the short-run affects.

10
For restrictions on α metrics, see appendix table A3.

77
Shabbir and Amjad

4.3. Short-run Dynamics Analysis


We used Johansen’s (1988) technique to determine the order of
integration between data series and identify the possible long-term
relationships among the integrated variables. Following the Granger
Representation Theorem, we included ECT in equation of the first difference
VAR model in order to capture the equilibrium relationship among the
cointegrated variables in their dynamic behavior. We estimated the model
after adding ECT and imposed different exclusion restriction on the
coefficients of lag differenced variables and ECT. Then, we apply exclusion
restrictions test of the joint significance of lags of other variables (Wald test),
and the significance of the lagged ECT. The results for these tests are reported
in table 6.
Table: 6. Results of Short-run dynamic analysis on ECM model
(Imposing exclusion restrictions)

Variables ∆GDPt-1 ∆FDI t-1 ∆EXP t-1 ECM t-1 Wald Prob Status
Test
∆GDPt ------ 00000 ------ ------ 2.73 0.11 Accepted
∆GDPt ------ ------ 00000 00000 2.71 0.08 Accepted
∆FDIt 00000 ------ ------ ------ 4.63 0.04 Rejected
∆FDIt ------ ------ 00000 ------ 0.204 0.654 Accepted
∆EXPt 00000 ------ ------ 00000 5.75 0.007 Rejected
∆EXPt ------ 00000 ------ ------ 3.13 0.08 Accepted
0000 indicates the exclusion restrictions on particular variable and ------ indicates no restrictions just
their coefficients.

The results indicate that there is a uni-directional causality that runs


from GDP to FDI but not in reserve order. It implies that FDI play zero roles
in GDP growth in short run. The results also predict uni-directional causality
from GDP to exports but not from exports to GDP. It means the export-led-
growth (ELG) hypothesis does not hold in short-run in case of Pakistan. The
Granger Causality test also explores that FDI lead growth in trade sector also
does not hold. We can conclude that FDI and exports both does not causing
the GDP in short-term in case of Pakistan where as GDP causing both FDI
and exports in short run.

78
Growth, FDI and Exports in Pakistan

V. Summary and Conclusions


This study has tried to investigate the long-run co-integrating
relationship among GDP, Exports and FDI for Pakistan. Annual data series
are used for this analysis. The ADF and KPSS unit root tests are used for the
presence of unit root. The results of these tests indicate that all data series are
stationary at their first difference. After resolving the issue of stationarity, we
used VAR model for long run cointegration analyses. The estimated results
with VAR suggest that long-run relationship exists between GDP and exports
but not with FDI. It means FDI has very little role in determining the GDP
growth in long run for Pakistan. However, VAR results suggest that exports
are having long run co-integrating relation with GDP.
We also tried to capture the short-term dynamics by using the error
correction model. In short-run analysis of Granger causality indicates that FDI
not affecting the GDP even in the short run whereas GDP affects both FDI as
well as Exports. We also tried to investigate the notion that FDI increases the
host country’s exports. The causality results fail to support this hypothesis
also in case of Pakistan. It means, both export and FDI are not causing each
other in the short-run. The results indicate that role of FDI in Pakistan’s
economy is very negligible in long run as well as in short run. Therefore, it is
recommended that government must increase the inflow of FDI along with its
direction towards those sectors that has spill over effects and can increase
Pakistan’s exports.

79
Shabbir and Amjad

Reference
Ahmad, M. H., Alam, S., & Butt, M. S. (2004). Foreign direct investment,
exports and domestic output in Pakistan. Paper presented at the 19th
Annual General Meeting, PIDE, Quaid-e-Azam University, Islamabad.
Alguacil, M., Cuadros, A. & Orts, V. (2011). Inward FDI and growth: The
role of macroeconomic and institutional environment. Journal of
Policy Modeling, Vol. 33(3), pp. 481-496.
Baillie, R. T., Chung C. F. & Tieslau, M. A. (1996). Analysing inflation by
the fractionally integrated ARFIMA-GARCH Model. Journal of
Applied Econometrics, Vol. 11, pp. 23-40.
Barrell, R. & Pain, N. (1997). Foreign direct investment, technological
change, and economic growth within Europe. The Economic Journal
Vol. 107, pp. 1770-1786.
Banerjee, A. R., L. Lumsdaine & Stock, J. H. (1992). Recursive and
sequential tests of the Unit Root and Trend Break Hypothesis: Theory
and international evidence. Journal of Business and Economics
Statistics, Vol. 10 (3).
Blomstrom, M. (1986). Foreign investment and productivity efficiency: The
case of Mexico. Journal of Industrial Economics, Vol. 15, pp. 97-110.
Blough, S. R. (1992). The relationship between power and level for generic
Unit Root tests in finite samples. Journal of Applied Econometrics,
Vol. 7(3).
Borensztein, E., De Gregorio, J. & Lee, J. W. (1998). How does foreign
investment affect economic growth? Journal of International
Economics, Vol. 45, pp. 115-135.
Charemza, W. W. & Syczewska, E. M. (1998). Joint application of the
Dickey-Fuller and KPSS Tests. Economics Letters, Vol. 2 (1).
Cuadros, A, Orts, V. & Alguacil, M. T. (2001). Openness and growth: Re-
examining foreign direct investment, trade and output linkages in Latin
America. CREDIT Research Paper, University of Nottingham.
Dickey, D. A. & Fuller, W. A. (1979). Distribution of the estimators for
autoregressive time series with a unit root. Journal of the American
Statistical Association, Vol. 74, pp. 427-431.
Dickey, D. A. & Fuller, W. A. (1981). Likelihood ratio statistics for
autoregressive time series with a unit root. Econometric Reviews, Vol.
15, pp. 369-386.

80
Growth, FDI and Exports in Pakistan

Engle & Granger, C. (1987). Cointegration and error correction:


Representation, estimation and testing. Econometrica, Vo1. 55, pp.
251-76.
Feder, G. (1982). On exports and economic growth. Journal of Development
Economics, Vol. 12, pp. 59-73.
Freenstra & Hanson (1997). Foreign direct investment and relative wages:
Evidence from Mexico’s Maquiladoras. Journal of International
Economics, Vol. 42, pp. 371-393.
Giles, J. A. & Williams, C. L. (2000). 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.
Golberg, S. & Klien, W. (1999). Foreign direct investment, trade, and real
exchange rate linkages in developing countries. In Reuven Glick (ed.),
Managing Capital Flows and Exchange Rates: Lessons from the
Pacific Basin. Cambridge University Press.
Gray, H. P. (1998). International trade and foreign direct investment: the
Interface, in J. H. Dunning, ed. Globalization, Trade and Foreign
Direct Investment, Oxford: Elesvier. 19-27.
Grossman, G, & Helpman (1991). Innovation and growth in the world
economy, Cambridge: MIT Press.
Harris R. I. D. (1995). Using cointegration analysis in econometric modeling.
Harvester Wheatsheaf, Prentice Hall.
Henrik, H. & Rand, J. (2004). On the causal links between FDI and growth in
developing countries. Development Economics Research Group
(DERG), University of Copenhagen.
Herzer, D., Klasen, S. & Nowak-Lehmann, D. F. (2008). In search of FDI-led
growth in developing countries: The way forward. Economic
Modelling, Vol. 25(5), pp. 793-810.
Hsiao, F. S. T & Hsiao, M. C. W. (2006). FDI, exports, and GDP in east and
southeast Asia-Panel data versus time series causality analyses.
Journal of Asian Economics, Vol. (17), pp. 1082-1106.
Johansen (1988). Statistical analysis of cointegrating vectors. Journal of
Economic Dynamic and Control, Vol. 12, pp. 231-54.
Johansen & Juselius (1990). Maximum Likelihood estimation and inference
on cointegration with applications to the demand for money. Oxford
Bulletin of Economics and Statistics, Vo1. 52(2), pp. 169-210.
Johanson, J. & Wiedersheim-Paul, F. (1975). The internationalisation of the
firm-four Swedish cases. Journal of Management Studies, Vol. 12(3),
pp. 305- 322.

81
Shabbir and Amjad

Jung, W.S. & Marshall, P. J. (1985). Exports, growth and causality in


developing Countries. Journal of Development Economics, Vol. (18),
pp.1-12.
Kojima, K. (1973). A macroeconomic approach to foreign direct
investment. Hitotsubashi Journal of Economics, Vol. 14 (1).
Kwiatkowski, D. et al. (1992). Testing the null hypothesis of stationarity
against the alternative of a Unit Root: How sure are we that economic
time series have Unit Root? Journal of Econometrics, Vol. 54. pp.
159-178.
Lim, E. (2001). Determinants of and the relation between foreign direct
investment and growth: A summary of the recent literature. IMF
Working Paper 01/175
Maddla, G. S. & Kim, M. (1998). Unit Roots, Cointegration and Structural
Change, Cambridge University Press.
Meyer K. E. (2003). FDI spillovers in emerging markets: A literature review
and new perspectives. Mimeo, Copenhagen Business School.
Moosa, I. A. (2002). FDI: theory, evidence and practice. New York: Palgrave.
Nicholas, S. J. (1982). British multinational investment before 1939. Journal
of European Economic History, Vol. 11.
Osterwald-Lenum, M. (1992). A note with quantiles of the asymptotic
distribution of the Maximum Likelihood cointegration rank test
statistics. Oxford Bulletin of Economics and Statistics, Vol. 54, pp.
461-471.
Petri, P. A. & Plummer, M. G. (1998). The determinants of foreign direct
investment: A survey with applications to the United States, Chapter 7
in Hiro Lee and David W. Roland- Holst, eds., Economic Development
and Cooperation in the Pacific Basin: Trade, Investment, and
Environmental Issues, New York: Cambridge University Press.
Phillips, P. C. B. & Perron, P. (1988). Testing for a unit root in time series
regression. Biometrika, Vol. 75(2).
Shiva, M. & Somwaru, A. (2004). Impact of foreign direct investment and
trade on economic growth: Evidence from developing countries.
American Journal of Agricultural Economics, Vol. 86, pp. 795-801.
UNCTAD (1996). United Nations conference on trade and development,
incentives and foreign direct investment, current studies, Series A,
No. 30, United Nations, New York and Geneva.
Vernon, R. (1966). International investment and international trade in the
product Cycle. Quarterly Journal of Economics, Vol. 80, pp. 190-207.
World Bank (1999). World Development Indicators 1999. Washington D.C.

82
Growth, FDI and Exports in Pakistan

Appendix

Figure: A1. Graph of all data series showing trend

GDP FDI
15 10.0
14
7.5
13

12
5.0
11

0 10 20 30 40 0 10 20 30 40
EXP

12

10

0 10 20 30 40

Table: A1. Choice of Lag length of VAR


Lag Interval Lag 1-3 Lag 1-2 LR Statistic
(sample 1960-2010) (sample 1960-2010)
Log 93.927850 89.898203 Chi2(9)=
Likelihood 8.0593
[0.5282]
AIC -3.1184 -3.3609

Table: A2. Imposing different Restrictions on Coefficient β


Dependent Independent β Standard error LR Stat P value
Variable variable Coefficient of β Chi2(1)
GDP Export 0.0000 0.0000 12.667 0.0004
(Export=0)
FDI 0.41912 0.0734
FDI EXP 0.000 0.000 12.667 0.0004
(Export=0) GDP 2.3860 0.38434
Export FDI 0.0000 0.000 2.1316 0.1443
(FDI=0)
GDP 1.5286 0.0625

83
Shabbir and Amjad

Table: A3. Test results of imposing different Restrictions on Coefficient α


Restrictions Variables Coefficient of Standard LR P value
α error of Statistics
α chi2(1)
Alpha coefficient GDP 0.0000 0.0000 2.667 0.0246
of GDP=0
FDI -1.73 0.41

EXPORT 0.46 0.14


Alpha coefficient GDP -0.06 0.03 0.0012
of FDI=0
FDI 0.000 0.000 10.498
EXPORT 0.63 0.19
Alpha coefficient EXPORT 0.000 0.000 6.1599 0.0131
of EXPORT=0
GDP -0.035 0.015
FDI -1.1745 0.28014

84

You might also like