THE DETERMINANTS OF FDI A CASE STUDY OF PAKISTAN (FY81

Report
Dr. SALMAN AHMAD
PROF. Department of Management Studies, University of
Central Punjab, LAHORE
ABSTRACT
 This paper analyzes the determinants of Foreign Direct
Investment (FDI) in developing countries like Pakistan and
examines why some countries like Pakistan have been
relatively unsuccessful in attracting FDI despite policy reforms.
This Study applies Auto Regressive distributed Lag (ARDL)
Co-integration technique given by Pesaran. The study found
GDP as a significant variable affecting low FDI inflows. The
reason for small inflows is that Pakistan has small GDP and it
is not growing at a higher rate showing a small size of market
compared to India and China. The paper considers other
variables like corruption, globalization, etc., but found them
statistically insignificant taking into account Pakistan’s data for
about 30 years.
ABSTRACT
 This study investigates the econometrically empirical
evidence of the relationship between FDI and GDP in an
ARDL framework for Pakistan. This study also examines
causal linkages between the variables by applying the
augmented Granger causality test of TodaYamamoto(1995). The results using data on Pakistan’s real
GDP and FDI for the period 1980-81 to 2008-09 show
cointegration between FDI and GDP when FDI is taken as
the dependent variable. Furthermore, unidirectional
Granger causality running from real GDP to real FDI has
been found in bivariate causality framework.
IMPORTANCE
 FDI is crucial to a developing country like Pakistan as it
provides the needed capital for investment. In addition,
FDI brings with it employment, managerial skills and
technology, and thus accelerates growth and development.
The role of FDI as a source of capital has become
increasingly important as IMF has started putting
conditions and forced structural adjustments for loans.
INVESTMENT POLICY 1997
 1.
Previously only manufacturing sector was open to
foreign investors. Now agriculture, services,
infrastructure, and social sectors are also open for foreign
investors on repatriable basis.
 2. Manufacturing sector has been prioritized in four
categories, namely
 a) Value added or export industries, the units which export
80 percent or more of their products in any one year or
have minimum value addition of 40 percent of production
value is treated as value added or export industries
respectively.
Investment Policy
 b) High-tech, this includes information technology, solar
technology, aerospace, defence production, etc.
 c) Priority industries, this includes engineering/capital
goods industries, chemicals, and others.
 d) Agro-based industries, this includes production of
quality/hybrid seeds, edible oil extracting/refining,
livestock/poultry feeds, milk processing, etc.
 3. Revision of labor laws in favor of industrialists.
LITERATURE REVIEW
 The size-of-market hypothesis is based on the assumption that
an inadequate market size has retarded the specialization of
productive factors. The argument holds that the size of the
market has been insufficient to absorb efficiently the
technology which the direct investor desires to introduce.
Based on this and related assumptions that differences exist
among nations in the level of technology; and that some
nations are more able to mobilize financial capital than others,
the size –of-market hypothesis is that foreign investment will
take place as soon as the market is large enough to permit the
capturing of economies of scale.

Literature review
 Bandra and White (1968) found market size to be a
significant determinant of U.S. FDI. Schmitz and Bieri
(1972) found the one-period lagged GNP of the EEC to be
a significant variable in a FDI demand function. Lunn
(1980) also found the one-period-lagged GNP of the EEC
to be a significant explanatory variable for U.S. direct
investment in Europe. For developing countries, Root and
Ahmad(1979), Torrisi(1985), Schneider and Frey (1985),
Petrochilas(1989), and Wheeler and Mody(1992) all find
market size to be significant. For Pakistan, Muhammad
Hanif
Literature Review
 Akhtar(2000) has used data from 1972 to 1996 and
multivariate regression analysis to reveal that market size,
relative interest rates and exchange rates are the major
determinants of FDI in Pakistan. Zahir Shah and Qazi Masood
Ahmad(2003) used cointegration technique to find that in the
error correction model, tariff, per capita GNP and dummy for
democracy were significant variables. Dar, et all (2004) have
used the data for 1970-2002 to check causality and long-term
relationship between FDI, Economic Growth and other sociopolitical determinants. They found economic growth, exchange
rate, interest rates, unemployment, and political instability as
having theoretically expected signs with two-way causality
relationships.
HYPOTHESIS
 Three principal hypotheses have been proposed as to the
motivation of foreign investment: size of market in the
receiving area, economic growth, and tariff
discrimination. These hypotheses are tested using the cointegration technique to determine their relative
importance. The empirical data used in these tests relate to
direct investment in Pakistan for the 1980-2009 period.
ARDL APPROACH
 The prosed ARDL approach to cointegration is developed
by Pesaran and Pesaran(1997), Peasaran and Shin(1995,
1998) and further advanced by Pesaran et al. (2001). It is a
unification of autoregressive models and distributed lag
models. In an ARDL model, a time series is a function of
its lagged values and current and lagged values of one or
more explanatory variables.
 The traditional cointegration technique perform better
only for large sample but this is more appropriate for 30
observations.
MODEL
 The measure of the size of market is the level of GDP
shown by Y. The second category of hypotheses, the
growth hypothesis, are fundamentally based on the
relation between the level of aggregate demand and the
total investment needed to satisfy this demand. Two
variants of growth variables are used. The percentage rate
of growth of Pakistan GDP is designated G1; and dY
represents the absolute change in the Pak GDP.
Model
 The investment demand function incorporating the size-of





market, growth, and tariff-induced burden as arguments may be
specified as:
I = A0 + A1 Y+ A2 T+ A3 G
-------------(1)
Where:
I = the annual book value of Direct foreign investment in
Pakistan (in millions of dollars)
Y = Pak Real GDP (1999-2000 constant factor cost in millions
of rupees)
T= Tariff revenue as a ratio of total tax revenue
G = the general specification for the two variations of the
growth hypotheses: Growth rate Gr, and change in real GDP
dY
DATA
 FDI
 Real inflows of FDI are the annual inflows of FDI for the period 1980-81 to 200809. Real value of the dependent variable is obtained by deflating the nominal values
with GDP deflator at constant prices of 1999-2000.

 GDP
 Real GDP is used as a proxy to estimate the impact of existing market size in
Pakistan on FDI. The data for this variable, being in millions of rupees at constant
prices of 1999-2000, was converted into dollars using each year’s exchange rate of
rupee per dollar.
 Economic Growth
 A high level of economic growth is a strong indication of market opportunities. The
growth of the host market is deemed to be significant for expansionary direct
investment. We have used two types of measures for growth. Real Growth rate of
GDP is used to test the proposition that a growing Pakistani market attracts FDI.


DATA
 Exchange Rate
 An economy with a depreciating currency attracts more
FDI as exporting from abroad becomes expensive, while it
becomes cheaper to produce locally. Hence, exports by the
home country are replaced through local production in the
host country. Real exchange rate is used as a variable. It is
the nominal exchange rate (rupees per US dollar) adjusted
for relative changes in consumer price index (CPI) based
on 2000 prices.
 Globalization
Data
 When one shifts to consider price linked determinants of
foreign investment, direct influences on domestic prices should
be distinguished from direct influences on international prices.
Emphasis is placed on international price patterns to avoid
domestic price change due to change in tariffs. It is said that
foreign investment is undertaken to avoid obstacles to trade. An
implication of this hypothesis is that trade liberalization as a
result of WTO will allow goods to move more freely and
thereby reduce the volume of international investment flow
needed. We use government revenues from taxes on
international trade, mainly import and customs duties. These
revenues are divided by the total tax to compute the relative tax
burden borne by the international sector.
EMPIRICAL FINDINGS
 To investigate the nature of any long-run relationship
between FDI and the variables suggested in our model, we
proceed to examine whether the series are cointegrated.
Unless series are cointegrated, there is no equilibrium
relationship between variables and inference is worthless.
Empirical Findings
 Unit Root Test
 The results of four different unit root tests, augmented
Dickey-fuller(ADF), Phillips-Perron, Dickey-Fuller
Generalised Least Square (DF-GLS), and Ng-Perron tests
are employed for unit roots to find out whether the
variables are integrated of the same order. JohansenJuselius test for cointegration is employed followed by
Error Correction Model to find short-run relationship of
the variables. If not, then we use ARDL approach to test
for long run relationship. Results of the ADF and PhillipsPerron tests are presented in Table 1.
Variables
Lx1
DLx1
Lx2
Augmented Dickey-Fuller Test
(ADF)
Intercept
Intercept &
Trend
-1.5341
-2.6657
(0.5020)
(0.2569)
-5.4522
(0.0001)
-0.5813
-5.7460
(0.8566)
(0.0005)
DLx3
Lx4
DLx4
Lx5
DLx5
-3.6020
(0.0123)
-7.0140
( 0.0000)
-1.3640
(0.5845)
-3.6009
(0.0136)
Intercept
Intercept &
Trend
-2.7374
(0.2304)
-3.8615
(0.0278)
-1.4808
(0.5283)
-5.4533
(0.0001)
-1.2030
(0.6587)
-3.9483
(0.0055)
-3.5532
(0.0138)
-4.4203
(0.0080)
-2.3650
(0.1602)
-4.4033
(0.0083)
-1.5749
(0.4818)
-5.8639
(0.0003)
-1.6416
(0.7499)
DLx2
Lx3
Phillps-Perron Test (PP)
-1.7989
(0.6781)
TABLE 2
Variables
Lx1
DLx1
Lx2
DLx2
Lx3
DLx3
Lx4
DLx4
Lx5
DLx5
Dickey-Fuller Generalized Least Square Test (DF-GLS)
Intercept
Intercept & Trend
-1.4648
-2.7436
-5.4416
0.4390
-2.7736
-2.8076
-3.6663
-1.3228
-2.0992
-4.4877
TABLE 3
Variables
Lx1 with Constant
Lx1 Constant &trend
DLx1 with Constant
DLx1 Constant &trend
Lx2 Constant
Lx2 Constant &trend
DLx2 Constant
DLx2 Constant &trend
Lx3 Constant
Lx3Constant &trend
DLx3 Constant
DLx3 Constant &trend
Lx4 Constant
Lx4 Constant &trend
DLx4 Constant
DLx4 Constant &trend
Lx5 Constant
Lx5 Constant &trend
DLx5 Constant
DLx5 Constant &trend
1% level of significance with constant
5% level of significance with constant
10% level of significance with constant
1% level of significance with constant & Trend
5% level of significance with constant & Trend
10% level of significance with constant & Trend
MZA
-4.2439
-9.4247
-15.8455
MZT
-1.2814
-2.1667
-2.7426
MSB
0.3019
0.2299
0.1730
MPT
5.9959
9.6849
1.8092
-0.1183
-5094.22
-0.0773
-50.4682
0.6531
0.0099
26.9766
0.0182
-12.7850
-2.3322
0.1824
2.6382
-3.4389
-13.5126
-10.2511
-1.1992
-2.5984
-2.2596
0.3487
0.1923
0.2204
7.0644
6.7481
2.4064
-51.2715
-5.0072
0.0976
0.6158
-13.8000
-8.100
-5.7000
-23.8
-17.3
-14.20
-2.5800
-1.9800
-1.6200
-3.42
-2.9
-2.62
0.1740
0.2330
0.2750
0.143
0.168
0.185
1.7800
3.1700
4.4500
4.03
5.48
6.67
TABLE 4
ORDER OF INTEGRATION
variables
LX1
ADF
PP
DF-
NG-
GLS
PERRON
Interc
In
Inte
In
Inter
In
ept
te
rcep
te
cept
te
Intercept
In
te
rc
t
rc
rc
rc
e
e
e
e
pt
pt
pt
pt
&
&
&
&
tr
tr
tr
tr
e
e
e
e
n
n
n
n
d
d
d
d
I(1)
I(1)
I(1)
I(
0)
LX2
I(
I(1)
I(1)
I(
0)
LX3
I(1)
0)
I(0)
I(0)
I(
0)
LX4
LX5
I(0)
I(
I(
I(
I(
0)
0)
0)
0)
I(1)
I(0)
I(
0)
Table 5.
Variable Addition Test (OLS case)
******************************************************************************
Dependent variable is DLX1
List of the variables added to the regression:
LX1
LX2
LX3
LX4
LX5
27 observations used for estimation from 1983 to 2009
******************************************************************************
Regressor
Coefficient
Standard Error T-Ratio[Prob]
INPT
-5.4185
7.5902
-.71388[.486]
DLX1(-1)
-.40578
.20182
-2.0106[.062]
DLX2(-1)
-1.5150
2.6953
-.56209[.582]
DLX3(-1)
.19199
.15940
1.2044[.246]
DLX4(-1)
.061084
.11301
.54054[.596]
DLX5(-1)
.62421
2.5742
.24248[.811]
LX1
.76126
.16750
4.5449[.000]
LX2
.51600
.67575
.76360[.456]
LX3
.41967
.19957
2.1028[.052]
LX4
-.018261
.17693
-.10321[.919]
LX5
3.6916
1.2624
2.9244[.010]
Joint test of zero restrictions on the coefficients of additional variables:
Lagrange Multiplier Statistic
CHSQ( 5)= 18.5838[.002]
Likelihood Ratio Statistic
CHSQ( 5)= 31.4733[.000]
F Statistic
F( 5, 16)=
7.0659[.001]
TABLE 6
Autoregressive Distributed Lag Estimates
ARDL(2,0,2,0,1) selected based on Schwarz Bayesian Criterion
*******************************************************************************
Dependent variable is LX1
27 observations used for estimation from 1983 to 2009
*******************************************************************************
Regressor
Coefficient
Standard Error
T-Ratio[Prob]
LX1(-1)
.60745
.15191
3.9987[.001]
LX1(-2)
-.36779
.19193
-1.9163[.072]
LX2
-1.8355
.61472
-2.9859[.008]
LX3
.18677
.18721
.99770[.332]
LX3(-1)
-.0090175
.19288
-.046752[.963]
LX3(-2)
.57923
.18267
3.1710[.006]
LX4
.16931
.14080
1.2025[.246]
LX5
-7.8992
1.5840
-4.9867[.000]
LX5(-1)
5.7617
1.1193
5.1474[.000]
INPT
19.9557
6.7968
2.9361[.009]
************************************************************************
R-Squared
.90367 R-Bar-Squared
.85267
S.E. of Regression
.31023 F-stat. F( 9, 17) 17.7199[.000]
Mean of Dependent Variable 1.9084 S.D. of Dependent Variable
.80826
Residual Sum of Squares
1.6362 Equation Log-likelihood
-.46442
Akaike Info. Criterion
-10.4644 Schwarz Bayesian Criterion -16.9436
DW-statistic
2.2851
************************************************************************
Diagnostic Tests
**************************************
**********************************
* Test Statistics *
LM Version
*
F Version
*
**************************************
**************************************
***
*
*
*
*
* A:Serial Correlation*CHSQ( 1)=
3.3806[.066]*F( 1, 16)= 2.2901[.150]*
*
*
*
*
* B:Functional Form *CHSQ( 1)=
1.3124[.252]*F( 1, 16)= .81743[.379]*
*
*
*
*
* C:Normality
*CHSQ( 2)=
.72179[.697]*
Not applicable
*
*
*
*
*
* D:Heteroscedasticity*CHSQ( 1)=
3.2248[.073]*F( 1, 25)= 3.3910[.077]*
CAUSALITY TEST
 TODA-YAMAMOTO AUGMENTED GRANGER
CAUSALITY TEST
 The causal linkage among FDI, GDP, Growth rate, Trade
restrictions are being tested by following the Granger causality
procedures adopted by Toda and Yamamoto (1995) and
interpreted and further expanded by Rambaldi and
Doran(1996) and Zapata and Rambaldi(1997). Toda –
Yamamoto Augmented Granger causality Test applied modified
WALD test for restrictions on the parameters of a Seemingly
Unrelated Regression (SUR). The following system of
equations is being estimated to investigate the augmented
Granger causality test.

Causality Test
 FDIt = α1 + ∑ β FDI t-i + ∑ γ GDP t-i + u1
 GDPt = α2 + ∑β FDIt-I + ∑ γ GDP t-i + u2
 The above system of two equations is estimated by SURE
method. To explore that GDP does not Granger cause FDI, the
null hypothesis will be
 Ho : γ = 0. Likewise, the other null hypothesis for second
equation is
 Ho: β =0, that is, the FDI does not Granger cause GDP. This
was carried out by means of a Wald test with the null
hypothesis that the values of the estimated coefficients β and γ
are zero. The results of the Toda-Yamamoto test of augmented
Granger causality are given in the table 7.
Table 7.
TODA-YAMAMOTA GRANGER CAUSALITY TEST
Equation
Null hypothesis
value
d.f.
Prob.
Equation 1
GDP does not
4.006
1
0.045
Granger cause
Reject Ho
FDI
Equation 2
FDI does not
Granger cause
GDP
1.2795
1
0.258
Cannot reject
SUMMARY
 The results of ARDL approach to co-integration show
cointegration between FDI and GDP when GDP, Foreign
exchange rate, Indirect Taxes are taken as Independent
variables.
CONCLUSIONS
 To date, most of the analysis of the determinants of
foreign direct investment has been descriptive in form;
there has been little work directed to statistically
evaluating these determinants. In order to provide a more
comprehensive empirical estimate of the determinants of
foreign direct investment, this article used the ARDL
cointegration technique and data for the 1980-2009 to test
the size of market, growth, and tariff changes as
explanations of why direct investment has been coming
recently in large amounts.
CONCLUSIONS
 The study’s empirical tests lead to the conclusions that
only the size-of-market hypothesis can be supported
statistically. Negative findings were discovered for all
variants of growth and tariff related imports; these
hypotheses were rejected as not statistically significant .
The results suggest that when dealing with policy
problems relating to foreign direct investment, it is
important to focus on the receiving country’s size of
market as a major determinant of foreign direct investment
flows.
THANKS

similar documents