The Journal of Development Studies

ISSN: 0022-0388 (Print) 1743-9140 (Online) Journal homepage: http://www.tandfonline.com/loi/fjds20

Foreign direct investment, financial development

and economic growth

Niels Hermes & Robert Lensink

To cite this article: Niels Hermes & Robert Lensink (2003) Foreign direct investment, financial
development and economic growth, The Journal of Development Studies, 40:1, 142-163, DOI:
10.1080/00220380412331293707

To link to this article: http://dx.doi.org/10.1080/00220380412331293707

Published online: 04 Jun 2010.

Submit your article to this journal

Article views: 3400

View related articles

Citing articles: 141 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=fjds20

Download by: [The UC San Diego Library] Date: 30 April 2017, At: 09:09
401jds07.qxd 21/10/2003 15:42 Page 142

Foreign Direct Investment, Financial

Development and Economic Growth

N I E L S H E R M E S and R O B E RT L E N S I N K

This article argues that the development of the financial system of

the recipient country is an important precondition for FDI to have
a positive impact on economic growth. A more developed financial
system positively contributes to the process of technological
diffusion associated with FDI. The article empirically investigates
the role the development of the financial system plays in enhancing
the positive relationship between FDI and economic growth. The
empirical investigation presented in the article strongly suggests
that this is the case. Of the 67 countries in data set, 37 have a
sufficiently developed financial system in order to let FDI
contribute positively to economic growth. Most of these countries
are in Latin America and Asia.

I. INTRODUCTION

The contribution of foreign direct investment (FDI) to economic growth has

been debated quite extensively in the literature. This debate has focused on the
channels through which FDI may help to raise growth in recipient countries.
In particular, it has been discussed to what extent FDI may enhance
technological change through spillover effects of knowledge and new capital
goods, that is, the process of technological diffusion. In this discussion, some
have argued that the contribution FDI can make is strongly dependent on the
circumstances in the recipient countries. However, empirical studies
investigating the relationship between FDI and economic growth on the one
hand, and the role played by the circumstances FDI is confronted with
whenever it enters a recipient country on the other hand, are scarce.1
This article argues that the development of the financial system of the
recipient country is an important precondition for FDI to have a positive

Niels Hermes, Faculty of Management and Organisation; Robert Lensink, Faculty of Economics
University of Groningen, The Netherlands. The authors thank two anonymous referees and
Gerard Kuper for their comments on an earlier version of the article. Please send comments to
Niels Hermes: c.l.m.hermes@bdk.rug.nl.
The Journal of Development Studies, Vol.40, No.1, October 2003, pp.142–163
PUBLISHED BY FRANK CASS, LONDON
401jds07.qxd 21/10/2003 15:42 Page 143

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 143

impact on economic growth. The financial system enhances the efficient
allocation of resources and in this sense it improves the absorptive capacity
of a country with respect to FDI inflows. In particular, a more developed
system may contribute to the process of technological diffusion associated
with FDI. The main contribution of this article is to investigate empirically
the role the development of the financial system plays in enhancing the
positive relationship between FDI and economic growth.
The article is structured as follows. Section II provides a description of
the discussion of the contribution FDI can make to increased economic
growth. The section emphasises the importance of technological diffusion
and the role of FDI. In particular, it focuses on the contribution the financial
system can make in this respect, using a simple theoretical model. Section
III discusses the data and the empirical methodology. Sections IV–VI
discuss the outcomes of the empirical investigation. Finally, section VII
provides a summary and concluding remarks.

II. FDI, THE FINANCIAL SYSTEM, AND ECONOMIC GROWTH:

A THEORETICAL FRAMEWORK

Review of the Literature on FDI and Growth

There is a huge literature emphasising the positive impact FDI may have on
economic growth.2 Next to the direct increase of capital formation of the
recipient economy, FDI may also help increasing growth by introducing
new technologies, such as new production processes and techniques,
managerial skills, ideas, and new varieties of capital goods. In the new
growth literature the importance of technological change for economic
growth has been emphasised [Grossman and Helpman, 1991; Barro and
Sala-i-Martin, 1995]. The growth rate of less developed countries (LDCs)
is perceived to be highly dependent on the extent to which these countries
can adopt and implement new technologies available in developed countries
(DCs). By adapting new technologies and ideas (that is, technological
diffusion) they may catch up to the levels of technology in DCs.
One important channel through which adoption and implementation of
new technologies and ideas by LDCs may take place is FDI. The new
technologies they introduce in these countries may spillover from
subsidiaries of multinationals to domestic firms [Findlay, 1978]. The use of
new technologies may be important in contributing to higher productivity of
capital and labour in the host country. The spillover may take place through
demonstration and/or imitation (domestic firms imitate new technologies of
foreign firms), competition (entrance of foreign firms leads to pressure on
domestic firms to adjust their activities and to introduce new technologies),
linkages (spillovers through transactions between multinationals and
401jds07.qxd 21/10/2003 15:42 Page 144

144 THE JOURNAL OF DEVELOPMENT STUDIES

domestic firms), and/or training (domestic firms upgrade the skills of their
employees to enable them to work with the new technologies) [Kinoshita,
1998: 2–4; Sjöholm, 1999a: 560].
The next question is about the conditions in the host country that are
important to maximise the technology spillovers discussed above. In the
literature it has been emphasised by some that the spillover effect can only
be successful given certain characteristics of the environment in the host
country. These characteristics together determine the absorption capacity of
technology spillovers of the host country. Thus, FDI can only contribute to
economic growth through spillovers when there is a sufficient absorptive
capacity in the host country.
Several country studies have been carried out, providing diverging results
on the role of FDI spillovers with respect to stimulating economic growth.
These studies deal with the productivity effects of FDI spillovers on firms or
plants using micro level data. Whereas positive effects from spillovers have
been found for, for example, Mexico [Blomström and Persson, 1983;
Blomström and Wolff, 1994; Kokko, 1994], Uruguay [Kokko, Tansini, and
Zejan, 1996] and Indonesia [Sjöholm, 1999b], no spillovers were traced in
studies for Morocco [Haddad and Harrison, 1993] and Venezuela [Aitken
and Harrison, 1999]. These diverging results may underline the crucial role
of certain host country characteristics necessary to let FDI contribute
positively to economic growth through spillovers. They emphasise the
difference in absorptive capacity between countries to adopt FDI.3
Some authors argue that the adoption of new technologies and
management skills requires inputs from the labour force. High-level capital
goods need to be combined with labour that is able to understand and work
with the new technology. Therefore, technological spillover is possible only
when there is a certain minimum, or ‘threshold’ level of human capital
available in the host country [Borensztein, et al., 1998]. This suggests that
FDI and human capital are complementary in the process of technological
diffusion. Other authors argue that the process of technological spillovers
may be more efficient in the presence of well-functioning markets. Under
these circumstances, the environment in which FDI operates ensures
competition and reduces market distortions, enhancing the exchange of
knowledge among firms [Bhagwati, 1978; Ozawa, 1992;
Balasubramanyam et al., 1996]. Some authors stress that the establishment
of property rights – in particular intellectual property rights – is crucial to
attract high technology FDI [Smarzynska, 1999]. If intellectual property
rights are only weakly protected in a country, foreign firms will undertake
low technology investments, which reduces the opportunities for spillover
effects and improvements of productivity of domestic firms.
401jds07.qxd 21/10/2003 15:42 Page 145

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 145

FDI and the Domestic Financial System
The previous discussion shows there may be several characteristics that
may indeed be important to promote the use of absorptive capacity of a
country with respect to maximising technology spillovers from foreign
firms. Yet, this article argues that one crucial characteristic of the
environment in the host country has not been mentioned in the literature,
that is, the development of the domestic financial system. The importance
of the domestic financial system as a precondition for the positive growth
effects of FDI can be illustrated with a simple model of technological
change.4
The model assumes that technical progress is represented through the
variety of capital goods available. There are three types of agents in the
model: final goods producers, innovators and consumers. Every producer of
final goods rents N varieties of capital good from specialised firms that
produce a type of capital good (the innovators). The producer has monopoly
rights over the production and sale of the capital goods. The purchase price
Pj of the capital good is set by optimising the present value of the returns
from inventing (and producing in several periods), V(t). This leads to a fixed
mark-up over production costs. Barro and Sala-i-Martin [1995: 218],
assuming free entry of inventors, show that in equilibrium with positive
R&D (at cost price) and increasing N, the (constant) rate of return (interest
rate, r) is given by:
1 − α 2 /(1−α )
r = (1 / η ) LA1 /(1−α ) ( )α (1)
α
where α measures capital’s share of income (that is, the coefficient in Cobb-
Douglas production function) and L is labour input. FDI is introduced in the
model by assuming that there are fixed maintenance costs, equal to 1, and
fixed set up costs (R&D costs, η ). The costs of discovering a new variety
of a good (costs of innovation) are assumed to be the same for all goods. In
line with Borensztein et al. [1998] the costs of R&D depend on FDI: more
FDI leads to a decline in the costs of innovation. This reflects the idea that
it is cheaper to imitate than to innovate [Borensztein et al. 1998], and that
the possibility to imitate increases if more goods are produced in other
countries (that is, when FDI is higher). So, the costs of discovering a new
good can be modelled as (using FDI = F): η=f(F), where ∂η/∂F < 0.
The impact of the financial sector enters the model via A, the level of
technology. It is well known that the financial sector may improve
economic growth by enhancing the average level of technology (see below).
So, A is a function of the development of the financial sector (H), A=h(H),
where ∂A/∂H > 0. This implies that:
401jds07.qxd 21/10/2003 15:42 Page 146

146 THE JOURNAL OF DEVELOPMENT STUDIES

L 1 − α 2 /(1−α )
r=( )h( H )1 /(1−α ) ( )α (2)
f (F ) α

To introduce the link to economic growth we close the model by

considering behaviour of households. Households maximise a standard
inter-temporal utility function, subject to the budget constraint. This gives
the well-known Euler condition for the growth rate of consumption, gC =
(1/θ )(r – ρ), where -θ is the elasticity of marginal utility and ρ is the
discount rate. In the steady state the growth rate of consumption equals the
growth rate of output, g.
Using the expression for r from (2) we finally get:
L 1 − α 2 /(1−α )
g = (1 / θ )[( )h( H )1 /(1−α ) ( )α − ρ] (3)
f (F ) α
It is now easy to see that an increase in FDI leads to an increase in the
growth rate of output (g) and that the effect of FDI depends on the
development of the financial sector. An increase in FDI lowers set-up costs
(for technology adaptation) and raises the return on assets (r). This leads to
an increase in saving and so a higher growth rate in consumption and
output. This effect will be greater the higher the level of technology in a
country, that is, the better the financial system is developed.
A crucial assumption in the above model is that the domestic financial
system influences growth through the level of technology. We need to
specify further this link, however. First of all, the financial system
influences the allocative efficiency of financial resources over investment
projects. Thus, the financial system may contribute to economic growth
through two main channels (next to providing and maintaining a generally
accepted means of exchange). On the one hand, it mobilises savings; this
increases the volume of resources available to finance investment. On the
other hand, it screens and monitors investment projects (that is, lowering
information acquisition costs); this contributes to increasing the efficiency
of the projects carried out (see, for example, Greenwood and Jovanovic
[1990]; Levine [1991]; Saint-Paul [1992]).5 The more developed the
domestic financial system, the better it will be able to mobilise savings, and
screen and monitor investment projects, which will contribute to higher
economic growth.
Second, investment related to upgrade existing or adopt new
technologies is more risky than other investment projects. The financial
system in general, and specific financial institutions in particular, may help
to reduce these risks, thereby stimulating domestic entrepreneurs to actually
401jds07.qxd 21/10/2003 15:42 Page 147

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 147

undertake the upgrading of existing technology or to adopt new
technologies introduced by foreign firms. Thus, financial institutions
positively affect the speed of technological innovation, thereby enhancing
economic growth [Huang and Xu, 1999]. This argument also holds for
technological innovation that results from one or more of the channels of
technology spillovers from FDI as described above. The more developed
the domestic financial system, the better it will be able to reduce risks
associated with investment in upgrading old and/or new technologies.
Third, when we reconsider the different channels through which
technology spillover may take place, it becomes clear that in many cases
domestic firms will need to invest when upgrading their own technology or
adopting new technologies, based either on a demonstration effect, a
competition effect, and/or a linkage effect. The same holds in case they aim
at upgrading the skills of their employees (the training effect). These
investments should be financed, however. The development of the domestic
financial system at least partly determines to what extent domestic firms
may be able to realise their investment plans in case external finance from
banks or stock markets is needed.
Finally, the development of the domestic financial system may also
determine to what extent foreign firms will be able to borrow in order to
extend their innovative activities in the host country, which would further
increase the scope for technological spillovers to domestic firms. FDI as
measured by the financial flow data may be only part of the FDI to
developing countries, as some of the investment is financed through debt
and/or equity raised in financial markets in the host countries [Borensztein
et al., 1998: 134]. Thus, the availability and quality of domestic financial
markets also may influence FDI and its impact on the diffusion of
technology in the host country. This diffusion process may be more efficient
once financial markets in the host country are better developed, since this
allows the subsidiary of a MNC to elaborate on the investment once it has
entered the host country.
Thus, in conclusion, FDI and domestic financial markets are
complementary with respect to enhancing the process of technological
diffusion, thereby increasing the rate of economic growth. This hypothesis
can be tested empirically, which will be the subject of the next two sections.

I I I . D ATA A N D M E T H O D O L O G Y O F E M P I R I C A L I N V E S T I G AT I O N

The data set used in this article applies to the 1970–95 period and contains
67 LDCs (see Appendix II for a complete list of the countries). For this set
of countries data is available for all variables used in this study, which
means that the estimations have been carried out with a balanced data set.
401jds07.qxd 21/10/2003 15:42 Page 148

148 THE JOURNAL OF DEVELOPMENT STUDIES

Table 1 provides basic descriptive statistics for the dependent variable, that
is, the per capita growth rate (PCGROWTH) and the crucial variable in this
study, that is, gross FDI inflows as a percentage of GDP (FDI). Both
variables (and all other variables in this study) are average values for the
1970–95 period.
The table shows that PCGROWTH and FDI are not normally distributed.
The distribution of these variables is skewed. With respect to PCGROWTH,
30 countries have an average growth rate varying between zero and two per
cent, for 13 countries the growth rate is between two and four, for five the
growth rate is above four per cent, and for 19 countries the growth rate is
negative. For 42 countries FDI as a percentage of GDP is between zero and
one, for 17 countries it is between one and two, for five countries between
two and three, and for three countries between four and five. The largest
recipients of FDI as a percentage of GDP are Swaziland, Trinidad and
Tobago and Malaysia.
The methodology of the empirical investigation follows the voluminous
growth regression literature, which was stimulated by the seminal paper of
Barro [1991]. Unfortunately, theory does not provide clear guidance
concerning the set of variables that should be included in the growth
equation. Depending on the aim of the study and the insights and beliefs of
the author(s), different explanatory variables have been included and found
to be significant in the literature. Recently, some studies have shown that
only a few variables have a robust effect on economic growth (see, for
example, Levine and Renelt [1992] and King and Levine [1993a]),
implying the importance of stability tests. Sala-i-Martin [1997a; 1997b]
provides a useful method to test for the robustness of different variables in
explaining economic growth. The empirical analysis in this article closely
follows his approach. In particular, the regression analysis for the cross-
section of 67 countries is specified as follows:
PCGROWTH = α i + β i ,j I + βmj M + β z,jZ + e (4)

TA B L E 1
D E S C R I P T I V E S TAT I S T I C S F O R P E R C A P I TA G R O W T H A N D F D I

PCGROWTH FDI

Mean 0.938 0.998

Median 0.529 0.693
Maximum 6.832 4.698
Minimum –3.134 0.003
Standard Deviation 1.923 0.989
Skewness 0.719 1.824
Kurtosis 4.120 6.686
401jds07.qxd 21/10/2003 15:42 Page 149

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 149

where I, M and Z are vectors of variables and e is an error term. I is a vector
of variables that are ‘generally accepted’ to be important to explain
economic growth. M is a vector of variables containing the variables of
interest in this study. In this study these variables are the log of the FDI to
GDP ratio (LFDI), and LFDI interacted with the log of the private sector
bank loans to GDP ratio (LCREDP). LCREDP is chosen here as a measure
of financial development (see below). The vector of Z variables contains a
limited number of variables from a large set of variables that have been used
in the literature to explain per capita economic growth. These variables are
used as control variables in the estimations.6
The vector of I variables contains variables that, according to Levine and
Renelt [1992], and King and Levine [1993a], have a robust effect on
economic growth. These variables are: the log of the initial level of the
secondary enrolment rate (LSECENR), the log of the initial level of GDP
per capita (LGDPPC), the variable proxying for financial market
development over the 1970–95 period (LCREDP) and the log of the
investment share in GDP (LINVGDP). Table 2 presents the correlation
matrix for the I variables, PCGROWTH and LFDI.
The choice of the I variables needs some further explanation:

(1) LSECENR measures human development;

(2) The introduction of LGDPPC reflects the process of catch up;
(3) With respect to the choice of the financial development variable, we
note that several variables have been suggested in the literature to
measure financial development, depending on the specific
characteristics of the financial system of interest. These variables focus
on the size, the efficiency and/or the relative importance of different
financial intermediaries in the total financial system. The problem is
that for several of these variables data are only available for a limited
number of countries. Therefore in the analysis the log of credit to the
private sector as a percentage of GDP (LCREDP) is used to measure
financial development, since for this variable data are available for all

TA B L E 2
C O R R E L AT I O N M AT R I X

PCGROWTH LFDI LSECENR LINVGDP LCREDP LGDPPC

PCGROWTH 1
LFDI 0.21 1
LSECENR 0.43 0.30 1
LINVGDP 0.58 0.29 0.27 1
LCREDP 0.48 0.38 0.53 0.52 1
LGDPPC –0.10 0.37 0.52 0.18 0.45 1
401jds07.qxd 21/10/2003 15:42 Page 150

150 THE JOURNAL OF DEVELOPMENT STUDIES

countries in the data set. Moreover, this variable is used in several other
studies (see, for example, Demirgüç-Kunt and Levine [1996]);7
(4) With respect to LINVGDP, regression models are estimated including
and excluding this variable in the vector of I variables. The reason for
this is that the interpretation of a significant coefficient for a certain
variable x depends on whether or not LINVGDP is included in the
regression model. If LINVGDP is included and the coefficient of
variable x is significant, this is interpreted as x affecting growth via the
‘level of efficiency’. If LINVGDP is not included, it is unclear whether
variable x affects growth via investment or via efficiency. This
distinction is of importance to obtain more information with respect to
how exactly FDI is related to economic growth.

I V. R E S U LT S O F T H E E M P I R I C A L I N V E S T I G AT I O N

The analysis starts by estimating a number of base equations, that is, the Z
variables are not yet included in the regression models. The results of these
estimations are presented in Table 3 (without LINVGDP) and Table 4 (with
LINVGDP). Column [1] in both tables shows the relevance of including the
different I variables as determinants of GDP per capita growth. The tables
show that LGDPPC, LSECENR, LCREDP and LINVGDP have a significant
impact on economic growth. In column [2] LFDI is added to this equation.
This variable does not have a significantly positive direct effect on
economic growth. This may be interpreted as a confirmation of the view
that without additional requirements FDI does not enhance economic
growth of a country.
As explained above, the aim of this article is to empirically investigate
the hypothesis that FDI and domestic financial markets are complementary
with respect to enhancing the process of technological diffusion, thereby
increasing the rate of economic growth. Therefore, the empirical analysis
focuses on the variables LFDI and the interactive term LFDI*LCREDP,
which represent the vector of M variables as specified in equation (1). The
model presented in column [3] of Tables 3 and 4 directly tests the central
hypothesis of this article. The outcomes in the tables show that the
interactive term LFDI*LCREDP is positive and significantly related to the
dependent variable PCGROWTH, whereas LFDI alone is significantly
negative.8 This supports the view that FDI only has a positive effect on
economic growth if the development of the domestic financial system has
reached a certain minimum level. Thus, we find preliminary support for the
central hypothesis of this article.
It may be argued that the results presented in column [3] of Tables 3 and
4 are due to high multi-collinearity between LSECENR and LCREDP (see
401jds07.qxd 21/10/2003 15:42 Page 151

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 151

TA B L E 3
FDI AND ECONOMIC GROWTH

[1] [2] [3] [4]

LGDPPC –1.182*** –1.238*** –1.180*** –1.247***

(–5.15) (–5.29) (–5.20) (–4.98)
LSECENR 0.891*** 0.882*** 0.740*** 0.964***
(4.47) (4.45) (3.75) (4.09)
LCREDP 1.562*** 1.471*** 1.827*** 1.620***
(4.12) (3.97) (4.59) (4.02)
LFDI 0.156 –1.587** –1.574***
(1.11) (–2.60) (–2.68)
LFDI*LSECENR 0.215*
(1.69)
LFDI*LCREDP 0.621*** 0.429*
(2.85) (1.86)
C 1.376 2.127 0.943 1.386
(1.01) (1.47) (0.60) (0.85)
R2 0.46 0.46 0.51 0.51
F 19.94 15.32 14.53 12.46

Note: See Appendix A for abbreviations used. Dependent variable: PCGROWTH. Amount of
observations in all regressions: 67. Values in parentheses are White heteroskedastic
adjusted t-values. * denotes significance at the ten per cent level; ** denotes significance
at the five per cent level; *** denotes significance at the one per cent level. R2 is the
adjusted R2. F is the F-statistic.

TA B L E 4
FDI AND ECONOMIC GROWTH: EFFECTS VIA EFFICIENCY

[1] [2] [3] [4]

LGDPPC –1.117*** –1.148*** –1.081*** –1.113***

(–6.83) (–6.33) (–7.14) (–6.86)
LSECENR 0.868*** 0.863*** 0.706*** 0.805***
(5.40) (5.43) (4.64) (4.03)
LINVGDP 2.539*** 2.488*** 2.594*** 2.536***
(5.56) (5.55) (5.60) (5.25)
LCREDP 0.843** 0.807** 1.171*** 1.095***
(2.46) (2.40) (3.23) (2.86)
LFDI 0.085 –1.839*** –1.828***
(0.65) (–3.23) (–3.28)
LFDI*LSECENR 0.095
(0.807)
LFDI*LCREDP 0.685*** 0.599***
(3.45) (2.91)
C –4.437*** –3.910*** –5.473*** –5.125***
(–3.07) (–2.48) (–3.44) (–3.04)
R2 0.59 0.58 0.64 0.63
F 24.43 19.47 20.40 17.36

Note: See note to Table 3.

401jds07.qxd 21/10/2003 15:42 Page 152

152 THE JOURNAL OF DEVELOPMENT STUDIES

Table 2). This would mean that the results found are in fact due to the level
of human development in a country (that is, the hypothesis forwarded by
Borensztein et al. [1998]),9 rather than due to the level of financial
development. To further investigate this issue we estimate a model
incorporating LFDI, LFDI*LCREDP, and LFDI*LSECENR. This model is
presented in column [4] of both tables. If we concentrate on the results for
the model including LINVGDP (Table 4), the results of the estimation show
that LFDI*LCREDP remains significant; however, LFDI*LSECENR
becomes insignificant. These results can be interpreted as follows.
First, it again confirms the hypothesis that a certain level of financial
market development is an important prerequisite for FDI to have a positive
effect on economic growth. Second, it suggests that the importance of a
certain level of human capital as a prerequisite for the growth effects of FDI
(the argument made by Borensztein et al. [1998]) is at least partly explained
by the existence of a well-developed financial sector. Moreover, the fact that
the variable LFDI*LCREDP remains significant in the models where
LINVGDP is included suggests that FDI affects economic growth mainly
via the level of efficiency.10
What do the results of the analysis presented in Tables 3 and 4 imply for
the countries in the data set? Based on the results of our empirical analysis
we are able to determine the threshold value of LCREDP above which LFDI
starts to have a positive effect on growth. In order to be able to this, we
differentiate the model presented in column [3] of both tables with respect
to LFDI. We get:
∆(PCGROWTH)/∆ LFDI = -1.587+0.621*LCREDP
(model without LINVGDP); and
∆(PCGROWTH)/∆ LFDI = -1.839+0.685*LCREDP
(model with LINVGDP).
The threshold level of LCREDP above which LFDI has a positive effect on
economic growth can be calculated by setting the first derivative of the
above equations equal to zero. The threshold levels then equal:
(1.587/0.621) = 2.56 and (1.839/0.685) = 2.68. Since LCREDP is the
logarithm of credit to the private sector as a percentage of GDP, the results
imply that LFDI (and hence also FDI) will have a positive effect on growth
in countries where credit to the private sector as a percentage of GDP is
above 12.9 (when LINVGDP is excluded from the basic model) and 14.6
(when LINVGDP is included). In other words, CREDP should be larger than
12 per cent in order for FDI to have a positive effect on growth. In our data
set 37 out 67 countries (or 55 per cent) satisfy this threshold value for
CREDP. Table 5 presents the countries for which the domestic financial
401jds07.qxd 21/10/2003 15:42 Page 153

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 153

TA B L E 5
R E L AT I O N S H I P B E T W E E N F D I A N D G R O W T H A N D T H E R O L E O F T H E L E V E L O F
DEVELOPMENT OF THE DOMESTIC FINANCIAL SYSTEM

No positive effect of LCREDP on relationship between LFDI and PCGROWTH

AFRICA:
Algeria; Benin; Burkina Faso; Burundi; Cameroon; Cape Verde; Central African Rep.; Chad;
Gabon; Gambia; Guinea-Bissau; Cote d’Ivoire; Kenya; Lesotho; Madagascar; Mali;
Mauritania; Niger; Nigeria; Rwanda; Senegal; Sierra Leone; Somalia; Sudan; Togo; Zimbabwe

LATIN AMERICA:
Guatemala; Haiti

ASIA AND OTHER COUNTRIES:

Nepal; Papua New Guinea

Positive effect of LCREDP on relationship between LFDI and PCGROWTH

AFRICA:
Egypt; Ghana; Morocco; Swaziland; Tunisia; Zambia

LATIN AMERICA:
Barbados; Costa Rica; Dominican Rep.; El Salvador; Honduras; Jamaica; Mexico; Nicaragua;
Panama; Trinidad and Tobago; Argentina; Bolivia; Chile; Colombia; Ecuador; Paraguay; Peru;
Uruguay; Venezuela

ASIA AND OTHER COUNTRIES:

Bangladesh; China; India; Malaysia; Pakistan; Philippines; Sri Lanka; Syria; Thailand;
Hungary; Malta; Fiji

system has reached a sufficient level of development, that is, for these
countries FDI contributes positively to economic growth. The table shows
that for most sub-Saharan African countries it appears to be the case that the
level of development of their domestic financial system is insufficient, so
that FDI probably will not have a positive impact on their economic growth.

V. F U RT H E R A N A LY S I S O F T H E R E L AT I O N S H I P B E T W E E N F D I ,
FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH

The relationship between FDI, financial development and growth may be

further investigated in several directions. We will discuss a number of these
directions below.11
First, we investigate whether the results found in column [3] of Tables 3
and 4 also hold when using alternative indicators of financial development.
As was already discussed several variables have been used in the literature
to measure financial development. One alternative used extensively is the
log of the average money and quasi money to GDP ratio (LMGDP).12 We
401jds07.qxd 21/10/2003 15:42 Page 154

154 THE JOURNAL OF DEVELOPMENT STUDIES

use this variable and present the results of the analysis in Table 6, columns
[1] and [2] (with and without INVGDP). The estimation results first of all
show that LMGDP is positively and significantly related to growth.
Moreover, the interactive term LFDI*LMGDP has a positive and significant
coefficient, but only when INVGDP is included. These results seem to
support the view that FDI only has a positive effect on economic growth if
the development of the domestic financial system has reached a certain
minimum level. More precisely, it suggests that FDI affects economic
growth mainly via the level of efficiency.13
We also investigate whether the results we find in Tables 3 and 4 are
different for specific geographic regions. To analyse this we repeat the
regression analysis of column [3] in Tables 3 and 4 and include region
dummies for Africa (DUMAFR), Latin America (DUMLA) and Asia and
other countries (DUMAS), and present the results in columns [3] and [4] of
Table 6. The outcomes of the estimation show that the main results of Tables

TA B L E 6
F D I A N D E C O N O M I C G R O W T H : U S I N G A LT E R N AT I V E M E A S U R E S O F F I N A N C I A L
D E V E L O P M E N T A N D C O U N T RY R E G I O N D U M M I E S

[1] [2] [3] [4]

LGDPPC –0.9014*** –0.929*** –0.979*** –0.999***

(–4.05) (–4.98) (–3.46) (–5.52)
LSECENR 0.610*** 0.676*** 0.522* 0.424*
(3.32) (4.32) (1.98) (1.84)
LINVGDP 2.236*** 2.587***
(4.45) (5.36)
LCREDP 1.648*** 1.053***
(4.79) (3.17)
LMGDP 2.409*** 1.615***
(5.75) (3.72)
LFDI –0.721 –1.532* –1.236*** –1.583***
(–0.80) (–1.74) (–2.27) (–2.90)
LFDI*LMGDP 0.277 0.498*
(1.04) (1.94)
LFDI*LCREDP 0.503*** 0.596***
(2.75) (3.35)
C –2.738* –6.957*** 2.046 –3.611**
(–1.76) (–3.93) (1.05) (–2.15)
DUMAFR –1.463 –1.631
(–1.38) (–1.59)
DUMLA –1.470 –1.159
(–1.54) (–1.22)
DUMAS –0.491 –0.689
(–0.45) (–0.66)
R2 0.55 0.64 0.52 0.65
F 17.30 20.39 10.07 14.80

Note: See note to Table 3.

401jds07.qxd 21/10/2003 15:42 Page 155

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 155

3 and 4 remain unchanged. At the same time, none of the region dummies
appears to have statistically significant coefficient. These results suggest
that the findings as presented in column [3] of Tables 3 and 4 are not
different for specific country regions.
Finally, we analyse whether our results remain the same when we
average the variables over five-year periods, instead of over 25 years as we
did in the analyses presented in column [3] of Tables 3 and 4. For this
analysis we create a panel data set of four five-year periods (1975–79,
1980–84, 1985–89 and 1990–95). Because of missing data for some of the
variables, we do the estimations with an unbalanced panel data set. We use
three different estimation techniques: estimations with a common constant,
with fixed effects and with random effects. The results of the estimations are
shown in Table 7. The results in columns [1] and [4] refer to the estimations
with a common constant; columns [2] and [5] refer the estimations with
fixed effects; and columns [3] and [6] show the outcomes for the
estimations with random effects.
In general, the results we find using the panel data set are similar to our
results shown previously. This is especially true for estimations with a
common constant and with random effects. The estimations with fixed
effects are less satisfactory, since some of the control variables appear not
to be significant; yet, even these estimations show a statistically significant
positive relation between LFDI*LCREDP and growth. Thus, also when we
use a panel data set, the analysis appears to confirm the central hypothesis
of this article on the relationship between FDI, financial development and
growth.14

V I . F U RT H E R S TA B I L I T Y A N A LY S I S

In this section we further investigate the robustness of the results, by

conducting a stability analysis in line with Sala-i-Martin [1997a, 1997b].
This stability analysis tests whether the coefficients for LFDI and the
interactive term LFDI*LCREDP remain robust after adding a vector Z of a
limited number of control variables to the models presented in Tables 3 and
4. We define a group of 14 variables from which the additional control
variables are taken. These variables are shown to be important for
explaining economic growth in several other studies. Since we aim at using
a fully balanced data set in our analysis, other possibly relevant variables
were not taken into account due to lack of observations.
The additional variables we take into account in our analysis are
AIDGDP (development aid as a percentage of GDP), BANKL (bank and
trade related lending as a percentage of GDP), BMP (black market
premium), CIVLIB (index of civil liberties), DEBTGDP (the external debt
401jds07.qxd 21/10/2003 15:42 Page 156

156 THE JOURNAL OF DEVELOPMENT STUDIES

TA B L E 7
F D I A N D E C O N O M I C G R O W T H : PA N E L D ATA E S T I M AT I O N S

[1] [2] [3] [4] [5] [6]

LGDPPC –0.828** –8.877*** –0.909** –1.007*** –9.600*** –1.135***

(–2.50) (–7.44) (–2.34) (–3.20) (–8.23) (–3.11)
LSECENR 1.028*** –0.771 0.779* 0.919*** –0.280 0.840**
(2.84) (–1.39) (1.90) (2.71) (–0.54) (2.19)
LINVGDP 2.275*** 1.676* 2.262***
(3.85) (1.79) (3.93)
LCREDP 1.581*** 0.656* 1.146*** 1.253*** 0.605 0.935**
(4.68) (1.67) (2.84) (3.71) (1.46) (2.37)
LFDI –1.443*** 0.0419 –0.745 –1.112** 0.018 –0.597
(–3.09) (0.069) (–1.23) (–2.37) (0.03) (–1.02)
LFDI*LCREDP 0.566*** 0.333* 0.399** 0.450*** 0.342* 0.325*
(3.47) (1.67) (1.99) (2.77) (1.65) (1.67)
C –1.696 1.195 –5.924*** –3.637
(–0.99) (0.55) (–2.77) (–1.60)
N 226 226 226 224 224 224
R2 0.14 0.47 0.39 0.21 0.49 0.33
F 8.51 66.43 11.04 55.79

Note: Dependent variable: PCGROWTH. Values in parentheses are White heteroskedastic

adjusted t-values. * denotes significance at the ten per cent level; ** denotes significance
at the five per cent level; *** denotes significance at the one per cent level. The results in
columns [1] and [4] refer to the estimations with a common constant; columns [2] and [5]
refer the estimations with fixed effects; and columns [3] and [6] show the outcomes for the
estimations with random effects. N is the number of observations. R2 is the adjusted R2. In
case of the estimations with random effects R2 refers to unweighted statistics including
random effects. F is the F-statistic. For the estimations with random effects the F-statistic
is not given.

to GDP ratio), DEBTS (total external debt service as a percentage of GDP),

EINFL (uncertainty with respect to inflation), EGOVC (uncertainty with
respect to government expenditures), EXPGDP (exports of goods and
services as a percentage of GDP), GOVCGDP (government consumption as
a percentage of GDP), INFL (the annual inflation rate), PRIGHTS (index of
political rights), STDINFL (the standard deviation of the annual inflation
rate), and TRADE (exports plus imports to GDP).15 In all estimates discussed
below, these variables have been transformed into logarithmic form.
The stability test starts by determining all possible combinations of a
limited number of the above-presented set of 14 variables. We have chosen
to perform the stability test by adding combinations of three, respectively
four control variables to the models discussed above. Next, we carry out
regression analysis including all variables presented in column [3] from
Table 3, respectively Table 4, as well as all possible combinations of three
(respectively four) control variables. This means that in case of three
additional variables we estimate 14!/(11! 3!) = 364 different specifications
401jds07.qxd 21/10/2003 15:42 Page 157

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 157

of the model presented in column [3] of Tables 3 and 4 (that is, with and
without LINVGDP). In case we use four additional variables the amount of
different specifications equals 14!/(10!4!) = 1,001.
After having estimated all different equation specifications, the next step
of the stability test is to look at the distribution of the coefficients of the
individual equations, and calculate the fraction of the cumulative
distribution function lying on each side of zero. By assuming that the
distribution of the estimates of the coefficients is normal and calculating the
mean and the standard deviation of this distribution, the cumulative
distribution function (CDF) can be computed.
More precisely, if βj is the coefficient for a variable in the specification
j of the estimated model and σj is the standard error of the coefficient βj , we
proxy the mean and the standard deviation of the distribution by:

Σβj
β =
n
Σσ j
σ =
n .
The number of estimated equations is 364 (in case we add combinations of
three Z variables), respectively 1,001 (when we add combinations of four Z
variables). In Table 8 the mean estimate is presented in the column entitled
COEF and the mean standard deviation is given in the column entitled
STERR.
Next, we calculate the fraction of the cumulative distribution function
lying on the right or left-hand side of zero, using a table for the (cumulative)
normal distribution. The test statistic we use is defined as the mean over the
standard deviation of the distribution. The column entitled CDF in Table 8
denotes the larger of the two areas. Finally, as an additional stability test, the
last column of the table presents the percentage of all regressions for which
the variable of interest (that is, LFDI or LFDI*LCREDP) is significant at
the 95 per cent level.
The results presented in Table 8 show that the coefficients for LFDI and
the interactive term LFDI*LCREDP are very robust. In the models
including LINVGDP as an additional I variable t-values for LFDI and
LFDI*LCREDP are significant at the 95 per cent level in all cases. These
results strongly suggest that FDI enhances economic growth only if
domestic financial markets are well-developed, thus supporting the main
hypothesis investigated in this article.
401jds07.qxd 21/10/2003 15:42 Page 158

158 THE JOURNAL OF DEVELOPMENT STUDIES

TA B L E 8
S TA B I L I T Y T E S T

NUMBER R2 COEF STERR CDF PERC

Without LINVGDP in the base model

LFDI 364 0.57 –1.348 0.544 0.993 1.000
LFDI*LCREDP 364 0.57 0.534 0.194 0.997 0.997
LFDI 1,001 0.58 –1.310 0.535 0.993 0.921
LFDI*LCREDP 1,001 0.58 0.519 0.192 0.997 0.983

With LINVGDP in the base model

LFDI 364 0.72 –1.620 0.505 1.000 1.000
LFDI*LCREDP 364 0.72 0.615 0.173 1.000 1.000
LFDI 1,001 0.73 –1.528 0.500 1.000 1.000
LFDI*LCREDP 1,001 0.73 0.603 0.172 1.000 1.000

Note: NUMBER denotes the number of equations tested. R2 is the adjusted R2. CDF is the
cumulative distribution function. COEF is the mean estimate of the coefficient of the
variable of interest (i.e. LFDI or LFDI*LCREDP). STERR is the mean standard deviation
of the variable of interest. PERC is the percentage of all regressions for which the variable
of interest is significant at the 95 per cent level.

VII. CONCLUSIONS

FDI may help to raise economic growth in recipient countries. Yet, the
contribution FDI can make may strongly depend on the circumstances in the
recipient countries. Few empirical studies have investigated the relationship
between FDI and economic growth and the role played by the circumstances
FDI is confronted with whenever it enters a recipient country. These studies
focused on the role of human capital available in and the export-
orientedness of the recipient country. The original contribution this article
makes is that it argues that the development of the financial system of the
recipient country is an important precondition for FDI to have a positive
impact on economic growth. A more developed financial system positively
contributes to the process of technological diffusion associated with FDI.
The article empirically investigates the role the development of the
financial system plays in enhancing the positive relationship between FDI
and economic growth. The empirical investigation presented in the article
strongly suggests that this is the case. Of the 67 countries in data set, 37
have a sufficiently developed financial system in order to let FDI contribute
positively to economic growth. Most of these countries are in Latin America
and Asia. Almost all other countries in our data set are in sub-Saharan
Africa. These countries have very weak financial systems and consequently
FDI does not contribute positively to growth.
The results of the empirical investigation in this article provide a number
of policy-relevant conclusions. First, the results contradict the widely
401jds07.qxd 21/10/2003 15:42 Page 159

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 159

accepted view that an increase in FDI may important to enhance economic
growth of LDCs. This is only true after these countries have improved their
domestic financial systems. Second, the analysis in this article may
contribute to the discussion on the order of economic liberalisation in
LDCs. The outcomes of the empirical investigation suggest that these
countries should first reform their domestic financial system before
liberalising the capital account to allow for enlarged FDI inflows.

final revision accepted September 2002

NOTES

1. Exceptions are Balasubramanyam et al. [1996], Borensztein et al. [1998] and Lichtenberg
and van Pottelsberghe de la Potterie [1998].
2. For an overview of the literature on the relationship between FDI and economic growth, see
De Mello [1997] and World Bank [2001].
3. Görg and Strobl [2001] suggest that the econometric methodology used also has an important
impact on the results found. In particular, they report that studies using cross-section data
may overstate the spillover effects of FDI, since these studies do not take into account other
time-invariant or firm-specific effects. Yet, these effects may have an impact on the spillover
effects of FDI on productivity.
4. This model is based on Chapters 6 and 7 in Barro and Sala-i-Martin [1995].
5. See Levine [1997] or Berthelémy and Varoudakis [1996] for good surveys on the role of the
domestic financial system and its relationship to economic growth.
6. The list of Z variables used in this study will be discussed in the next section. See also
Appendix I.
7. Below, we will also discuss the results of an analysis in which we have used the log of the
average money and quasi money to GDP ratio (LMGDP).
8. Borensztein et al. [1998] also find a statistically significant negative relationship between the
linear term of FDI and GDP per capita growth. Lensink and Morrissey [2001] find a positive
relationship between both variables, but they use a data set for developed and less developed
countries. When we redo their analysis by using only data for LDCs, we find that there is no
statistically significant relationship between FDI and growth (see also our results in column
[2] of Tables 3 and 4).
9. Lensink and Morrissey [2001] show that the results presented by Borensztein et al. [1998]
are not statistically robust.
10. We have also explored the relationship between LFDI and LFDI interacted with a financial
development variable as exogenous variables and total investment as a share of GDP as the
endogenous variable. In line with Borensztein et al. [1998] it appears that LFDI and LFDI
interacted with financial market development do not have a robust effect on investment
levels. This confirms that FDI mainly affects growth via the level of efficiency.
11. Two anonymous referees suggested us to investigate the relationship between FDI, financial
development and growth in the directions discussed below. We thank them for these suggestions.
12. See, for example, the seminal studies by King and Levine on the relationship between
financial development and economic growth [King and Levine, 1993a, 1993b]. For an
overview of the (empirical) literature on financial development, see Levine [1997].
13. We acknowledge that there are more variables to measure financial development. In
particular, the two variables we have used here (LCREDP and LMGDP) focus only on the
banking sector. Ideally, we would have liked to also use variables that focus on other
financial markets, for example, stock market variables. Yet, using such variables would have
led to a substantial reduction of the number of countries in our data set due to lack of long-
term time series data.
401jds07.qxd 21/10/2003 15:42 Page 160

160 THE JOURNAL OF DEVELOPMENT STUDIES

14. We also investigated whether the relationship between FDI and financial development is
non-linear. In particular, we included two new interactive terms LFDI*LCREDP2 and
LFDI*LCREDP3 separately into the regression models presented in column [3] of Tables 3
and 4 (that is, with and without including LINVGDP). The results from this analysis suggest
that, using a specification in which the linear interactive term is also included, none of the
interactive terms have statistically significant coefficients. The results of this analysis are
available upon request from the authors.
15. See Appendix I for the exact specification and data sources of these variables.

REFERENCES

Aitken, B.J. and A.E. Harrison, 1999, ‘Do Domestic Firms Benefit from Direct Foreign
Investment? Evidence from Venezuela’, American Economic Review, Vol.89, No.3,
pp.605–18.
Balasubramanyam, V.N., Salisu, M. and D. Sapsford, 1996, ‘Foreign Direct Investment and
Growth in EP and IS Countries’, The Economic Journal, Vol.106, No.1, pp.92–105.
Barro, R.J., 1991, ‘Economic Growth in a Cross-Section of Countries’, Quarterly Journal of
Economics, Vol.106, No.2, pp.407–43.
Barro, R.J. and J.W. Lee 1994, Data Set for a Panel of 138 Countries, Cambridge, MA: NBER.
Barro, R.J. and X. Sala-I-Martin, 1995, Economic Growth, Cambridge, MA: McGraw-Hill.
Berthelemy, J.C. and A. Varoudakis, 1996, ‘Models of Financial Development and Growth: A
Survey of Recent Literature’, in N. Hermes and R. Lensink (eds.), Financial Development
and Economic Growth: Theory and Experiences from Developing Countries, London:
Routledge, pp.7–34.
Bhagwati, J.N., 1978, ‘Anatomy and Consequences of Exchange Rate Regimes’, Studies in
International Economic Relations, 10, New York: NBER.
Blomström, M. and H. Persson 1983, ‘Foreign Investment and Spillover Efficiency in an
Underdeveloped Economy: Evidence from the Mexican Manufacturing Industry’, World
Development, Vol.11, No.6, pp.493–501.
Blomström, M. and E.N. Wolff, 1994, ‘Multinational Corporations and Productivity Convergence
in Mexico’, in W. Baumol, R. Nelson and E.N. Wolff (eds.), Convergence of Productivity:
Cross-National Studies and Historical Evidence, Oxford: Oxford University Press.
Bo, H., 1999, ‘Empirics of the Investment-Uncertainty Relationship’, unpublished manuscript,
Groningen: University of Groningen.
Borensztein, E., De Gregorio, J. and J.W. Lee, 1998, ‘How Does Foreign Direct Investment
Affect Economic Growth?’ Journal of International Economics, Vol.45, No.1, pp.115–35.
De Mello, Jr., L.R., 1997, ‘Foreign Direct Investment in Developing Countries and Growth: A
Selective Survey’, The Journal of Development Studies, Vol.34, No.1, pp.1–34.
Demirgüç-Kunt, A. and R. Levine, 1996, ‘Stock Market Development and Financial
Intermediaries: Stylized Facts’, The World Bank Economic Review, Vol.10, No.2,
pp.291–321.
Findlay, R., 1978, ‘Relative Backwardness, Direct Foreign Investment and the Transfer of
Technology: A Simple Dynamic Model’, Quarterly Journal of Economics, Vol.92, No.1,
pp.1–16.
Görg, H. and E., Strobl 2001, ‘Multinational Companies and Productivity Spillovers: A Meta-
Analysis’, The Economic Journal, Vol.111, No.475, pp.F723–39.
Greenwood, J. and B. Jovanovic, 1990, ‘Financial Development, Growth, and the Distribution of
Income’, Journal of Political Economy, Vol.98, No.5, pp.1076–107.
Grossman, G.M. and E. Helpman, 1991, Innovation and Growth in the Global Economy,
Cambridge, MA: MIT Press.
Haddad, M. and A. Harrison, 1993, ‘Are There Positive Spillovers from Direct Foreign
Investment? Evidence from Panel Data for Morocco’, Journal of Development Economics,
Vol.42, No.1, pp.51–74.
Huang, H. and C. Xu, 1999, ‘Institutions, Innovations and Growth’, American Economic Review,
Vol.89, No.2, pp.438–43.
401jds07.qxd 21/10/2003 15:42 Page 161

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 161

King, R.G. and R. Levine, 1993a, ‘Finance and Growth: Schumpeter Might be Right’, Quarterly
Journal of Economics, Vol.108, No.3, pp.717–37.
King, R.G. and R. Levine, 1993b, ‘Finance, Entrepreneurship, and Growth: Theory and
Evidence’, Journal of Monetary Economics, Vol.32, No.3, pp.513–42.
Kinoshita, Y., 1998, ‘Technology Spillovers Through Foreign Direct Investment’, unpublished
working paper, Prague: CERGE-EI.
Kokko, A., 1994, ‘Technology, Market Characteristics, and Spillovers’, Journal of Development
Economics, Vol.43, No.2, pp.279–93.
Kokko, A., Tansini, R. and M.C. Zejan, 1996, ‘Local Technological Capability and Productivity
Spillovers from FDI in the Uruguayan Manufacturing Sector’, The Journal of Development
Studies, Vol.32, No.4, pp.602–11.
Lensink, R. and O. Morrissey, 2001, ‘Foreign Direct Investment: Flows, Volatility and Growth in
Developing Countries’, SOM research report 01E16, Groningen: SOM Research School,
University of Groningen.
Levine, R., 1997, ‘Financial Development and Economic Growth: Views and Agenda’, Journal
of Economic Literature, Vol.35, No.2, pp.688–726.
Levine, R., 1991, ‘Stock Markets, Growth, and Tax Policy’, Journal of Finance, Vol.46, No.4,
pp.1445–65.
Levine, R. and D. Renelt 1992, ‘A Sensitivity Analysis of Cross-Country Growth Regressions’,
American Economic Review, Vol.82, No.4, pp.942–63.
Lichtenberg, F.R. and B. van Pottelsberghe de la Potterie, 1998, ‘International R&D Spillovers’,
European Economic Review, Vol.42, No.8, pp.1483–91.
Ozawa, T., 1992, ‘Cross-Investments between Japan and the EC: Income Similarity,
Technological Congruity and Economies of Scope’, in J. Cantwell (ed.), Multinational
Investment in Modern Europe: Strategic Interaction in the Integrated Community, Aldershot:
Edward Elgar, pp.13–45.
Saint-Paul, G., 1992, ‘Technological Choice, Financial Markets and Economic Development’,
European Economic Review, Vol.36, No.4, pp.763–81.
Sala-i-Martin, X., 1997a, ‘I Just Ran Two Million Regressions’, American Economic Review,
Vol.87, No.2, pp.178–83.
Sala-i-Martin, X., 1997b, ‘I Just Ran Four Million Regressions’, unpublished manuscript,
Colombia University and Universitat Pompeu Fabra.
Sjöholm, F., 1999a, ‘Productivity Growth in Indonesia: The Role of Regional Characteristics and
Direct Foreign Investment’, Economic Development and Cultural Change, Vol.47, No.3,
pp.559–84.
Sjöholm, F., 1999b, ‘Technology Gap, Competition and Spillovers from Direct Foreign
Investment: Evidence from Establishment Data’, The Journal of Development Studies,
Vol.36, No.1, pp.53–73.
Smarzynska, B.K., 1999, ‘Composition of Foreign Direct Investment and Protection of
Intellectual Property Rights in Transition Economies’, unpublished working paper, New
Haven, CT: Yale University.
World Bank, 1997, World Development Indicators 1997, Washington, DC: World Bank.
World Bank, 2001, Global Development Finance 2001, Washington, DC: World Bank.
401jds07.qxd 21/10/2003 15:42 Page 162

162 THE JOURNAL OF DEVELOPMENT STUDIES

APPENDIX I
L I S T O F VA R I A B L E S U S E D I N T H E A N A LY S I S

AIDGDP = development aid as a percentage of GDP

BANKL = bank and trade related lending as a percentage of GDP
BMP = black market premium, calculated as (black market rate/official rate)-1
CIVLIB = index of civil liberties
CREDP = credit to the private sector as a percentage of GDP
DEBTGDP = the external debt to GDP ratio
DEBTS = total external debt service as a percentage of GDP
DUMAFR = dummy for African countries
DUMAS = dummy for Asian (and other) countries
DUMLA = dummy for Latin American countries
EINFL = uncertainty with respect to inflation
EGOVC = uncertainty with respect to government expenditures
EXPGDP = exports of goods and services as a percentage of GDP
FDI = foreign direct investment as a percentage of GDP
GDPPC = GDP per capita in 1970
GOVCGDP = government consumption as a percentage of GDP
INFL = the annual inflation rate
INVGDP = average investment to GDP ratio over 1970–95 period
MGDP = average money and quasi money to GDP ratio over the 1970–95 period
PCGROWTH = average real per capita growth rate over 1970–95 period.
PRIGHTS = index of political rights
SECENR = secondary school enrolment rate in 1970
STDINFL = the standard deviation of the annual inflation rate, calculated from the
inflation figures
TRADE = exports plus imports to GDP; measure of the degree of openness

The source for all variables is World Bank [1997], which is available on CD-ROM, except for
BMP, CIVLIB and PRIGHTS. These variables are obtained from the data set created by Barro and
Lee [1994]. Moreover, EINFL and EXPGDP have been calculated by the authors (see below).
The variables from Barro and Lee [1994] refer to averages for the 1970–90 period. Unless
otherwise stated, all other variables refer to averages over 1970–95 period. For all variables
logarithmic transformations are used.
We need to explain how the uncertainty variables EINFL and EGOVC have been constructed.
Both variables are constructed by using the standard deviation of the unpredictable part of INFL
and GOVC; see Bo [1999] for a survey of different methods to measure uncertainty. We first
specify and estimate a forecasting equation to determine the expected part of INFL and GOVC.
The standard deviation of the unexpected part of INFL and GOVC (that is, the residuals from the
forecasting equation) is used as a measure of uncertainty. We have used a second-order
autoregressive process, extended with a time trend, as the forecasting equation:
Pt = a1 + a2T + a3FDIt-1 + a4FDIt-2 + et ,
where Pt is the variable under consideration, T is a time trend, a1 is an intercept, a3 and a4 are
the autoregressive parameters and et is an error term. We estimate the above equation for all
countries in the data set. By calculating the standard deviation of the residuals for the entire
sample period for each individual country, we obtain the variables EINFL and EGOVC.
401jds07.qxd 21/10/2003 15:42 Page 163

FDI, FINANCIAL DEVELOPMENT AND ECONOMIC GROWTH 163

APPENDIX II
C O U N T R I E S I N T H E D ATA S E T

Africa:
Algeria, Benin; Burkina Faso; Burundi; Cameroon; Cape Verde; Central African Rep.; Chad;
Egypt; Gabon; Gambia; Ghana; Guinea-Bissau; Côte d’Ivoire; Kenya; Lesotho; Madagascar;
Mali; Mauritania; Morocco; Niger; Nigeria; Rwanda; Senegal; Sierra Leone; Somalia; Sudan;
Swaziland; Togo; Tunisia; Zambia; Zimbabwe.

Latin America:
Barbados; Costa Rica; Dominican Rep.; El Salvador; Guatemala; Haiti; Honduras; Jamaica;
Mexico; Nicaragua; Panama; Trinidad and Tobago; Argentina; Bolivia; Chile; Colombia;
Ecuador; Paraguay; Peru; Uruguay; Venezuela.

Asia and others

Bangladesh; China; India; Malaysia; Nepal; Pakistan; Philippines; Sri Lanka; Syria; Thailand;
Hungary; Malta; Fiji; and Papua New Guinea.