You are on page 1of 65

Determinants of FDI in CIS countries with Transition

Economy
By
Abdulla Azizov

Submitted to
Aarhus School of Business

In partial fulfillment of the requirements for the degree of Master of Science

Academic Advisor: Associate Professor, PhD Philipp J.H. Schrder

Aarhus, Denmark
2007

Abstract
While there has been voluminous research on the determinants of FDI for developed and
developing countries, very little has been done on this issue for transition economies,
especially Commonwealth of Independent States (CIS) countries. The present paper examines
the determinants of Foreign Direct Investment (FDI) in the Commonwealth of Independent
States between 1992 and 2005. Prior empirical and theoretical contributions are discussed
carefully in order to find out potential determinants. Following the description of the variables
under investigation, it is presented a partial adjustment model of FDI to show theoretically the
agglomeration effect of FDI. Fixed effect and random effect are tested and a random effects
panel data model is employed in the econometric analysis. It was found that control of
corruption to have no significant effects on FDI and some other variables have statistically
significant effects on the level of FDI. Additionally, EBRD transition indicators are selected
as explanatory variables. The EBRD large-scale privatization indicator found to be
statistically significant but negative effect on FDI. Test for structural change confirmed
positive shift in the model after 1999. Eventually, following the partial adjustment theory, it
was used Dynamic panel data model (System and Difference GMM) to examine the
agglomeration effect of FDI in CIS economies. Evidence of significant agglomeration effects
of FDI in CIS economies is found.

Keywords: CIS, FDI, dynamic panel

ACKNOWLEDGMENTS

First of all, I would like to express my appreciation to Associate Professor, PhD Philipp
J.H. Schrder, who kindly agreed to supervise the thesis and for his valuable advises and
support. My deepest gratitude is to my parents and Brother Abdulatif Azizov for their moral
support, patience and understanding. I also feel gratefulness to my closest friends Bakhtiyor
Khodjaev, Sayfulla Shaismatov and Kristian Anderssen for their invaluable comments and
care. My special acknowledgement also goes to Prof. R. Shadiev for technical assistance.

Correspondence: azizabdulla@yahoo.com

Table of Contents
Introduction .............................................................................................................................. 5
1. FDI Theory and FDI flow to Transition Economies ......................................................... 9
1.1. FDI Theory ...................................................................................................................... 9
1.2. Foreign Direct Investment and Spillovers..................................................................... 11
1.3. Foreign Direct Investment flow to Transition Economies ............................................ 13
1.4. Literature overview ....................................................................................................... 18
2. DATA and Estimation........................................................................................................ 20
2.1. Description of Variables................................................................................................ 21
2.2. Estimation Methodology and Partial adjustment model of the FDI.............................. 24
3. Panel Data Fixed Effects vs. Panel Data Random Effects .............................................. 27
3.1. The estimation of the Model.......................................................................................... 30
3.2. Estimation Results ......................................................................................................... 32
4. Dynamic Panel Data ........................................................................................................... 36
5. Conclusion ........................................................................................................................... 40
References ............................................................................................................................... 44
APPENDIX A.......................................................................................................................... 47
APPENDIX B.......................................................................................................................... 51
APPENDIX C.......................................................................................................................... 53
APPENDIX D.......................................................................................................................... 55
APPENDIX E.......................................................................................................................... 57
APPENDIX F .......................................................................................................................... 64

Introduction
Over the last couple of decades the approach towards the inflow of Foreign Direct
Investment (FDI) has changed significantly. It gained much attention and most of the
governments in the world have liberalized and promoted generous policies to attract foreign
investment from multinational companies. It is well known and obviously expected from
governments that Multinational Corporations (MNC) are the good source to fight against
unemployment, to raise the tax revenue, exports, job creation and higher wages, technology
and knowledge transfer. Lots of the governments across the world lowered the barriers for
investment entry and provide different forms of investment incentive programs just to
motivate foreign investors to invest in their country. It can be fiscal incentives such as tax
holidays and low taxes for foreign investors, or financial incentives such as grants and
preferential loans to MNCs, as well as measures like market preferences, infrastructure, and
sometimes even monopoly rights.1
Foreign Direct Investment is one of the driving forces of the process of globalization and is
a defining element of the modern-day world economy. Foreign Direct Investment promotes
the restructuring of industry at the regional and global levels and thus ensures the integration
of a national economy into the world economy more effectively than trade.2 Thus it is
important to make an analysis of the determinants of FDI.
What are the factors to attract FDI? What are the benefits for foreign investors from
investing in host country? Dunning (1993) explained by three motivations for FDI:

See UNCTAD (1996) and Brewer and Young (1997) for definitions of various FDI incentives.

U.T. Isayev, General Director of Foreign Investment Agency, Kyrgyz Republic, Session VII: Foreign
Investment During the Transition: How to Attract It, and How to Make the Best Use of It.

1.

Market-seeking with the aim of establishing business that serve a local


market because of its size or to overcome the trade barriers.

2.

Resource-seeking it is when investors are triggered by the availability of


natural resources in the host country.

3.

Efficiency-seeking - in order to achieve efficiency gains investors are taking an


advantage of local assets such as less expensive and skilled workers, cheaper
assets and infrastructure.

At the end of 80s and beginning of 90s most of the countries in Central and Eastern
Europe and former Soviet countries started the transition from communist states to market
economies. One of the most important aspects of transition to those countries was
liberalization of markets, opening the markets for foreign goods and services through trade
and direct investments. FDI has played important role to cover the gap between high levels of
investment and domestic savings required to support economic growth in those countries with
transition economy. Besides, FDI has been providing external financing in the equity form
rather than debt, mostly in export and import sector which supported the improvement of
external position.
The purpose of this paper is to analyze the determinants of FDI inflow to Commonwealth of
Independent States (CIS) countries.3 For the estimation of the model an unbalanced panel data
set covering all 12 CIS economies between 1992 and 2005 is used. Random effects estimation
will be used in the model. Dummy is included in the model to detect structural change after
1999. Additionally, System and Difference Generalized-Method-of Moments (GMM) will be
used in order to estimate agglomeration effect of FDI.

CIS countries Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan, Kyrgyz Republic, Moldova, The Russian
Federation, Tajikistan, Turkmenistan, Ukraine, and Uzbekistan.

In this research several questions can be stated and approached. First of all, it is to reveal the
nature of FDI into CIS countries with transition economy. Second, is to determine whether
corruption has an important effect on the size of FDI inflows an economy receives. There have been
lots of empirical studies on the effect of corruption on FDI inflows to Central and Eastern
Europe countries and to some other developed and developing countries. In overall corruption
has a significant negative effect on FDI inflow. Does this effect of corruption applicable for
CIS countries as well? Third, is to determine the significance of resources in CIS countries.
Do CIS countries are only source of natural resources in foreign investors eye? Fourth is
testing the influence of macroeconomic performance such as privatization, GDP, inflation,
external liberalization, fiscal balance, reforms in banking sector and etc. in attracting foreign
capital into economy.
My main hypothesis is corruption has a significant negative effect on FDI inflow to CIS
countries. Corruption discourages foreign investors; this is because of high uncertainty in the
perspective of their business. The level of investors sovereignty falls, and not everything will
be based on law. Corruption, bureaucratic issues, political and macroeconomic instability
regarded as most decisive barrier in foreign investors decision making (Aristidis Bitzenis,
2006). As stated above previous researches provided proofs for CEEB countries and some
other developed countries. Further, I hypothesize CIS countries has been resource seeking,
even though FDI inflow to more advanced countries with transition economy has been more
often efficiency seeking, i.e. focused on product exports based on low labor cost.
In this paper quantitative analysis is applied in the estimation. The qualitative analysis
findings cant be extended to wider populations with the same degree of certainty that
quantitative analyses can. It is because the findings of the study are not tested to find out
whether they are statistically significant or due to chance. And quantitative research classifies
features, count them, and also construct more complex statistical models in order to explain
7

what is observed. Findings of the research can be generalized to a larger population, and direct
comparisons can be made with previous studies4, so long as valid sampling and significance
techniques have been used.
The paper is organized as follows: Section I describes the theory of FDI, its importance and
effects on host countries economy. This section touches the spillover of FDI and FDI inflow
to CIS countries and slightly country specific facts. Previous literature overview is also
presented in Section I. Description of the data and construction of the sample, partial
adjustment model, also definition of the variables are provided in Section II. Section III
provides the methodology and the reason for the choice of random vs. fixed effects in the
panel data model and its estimation. The estimation results are also given in this section. The
results are described appropriately. Dynamic panel datas estimation and the results are
presented in Section IV. Conclusion of this paper is given in Section V. Later follow
references and appendices.

Most of the previous researches in FDI inward and its determinants are based on quantitative analysis.

1. FDI Theory and FDI flow to Transition Economies


1.1. FDI Theory
Foreign Direct Investment takes place when a company in one country creates or expands
a subsidiary in another country. The importance of this type of capital flow is FDI involves
not only a resource transfer from one location to another, but also acquisition of partial or full
control. FDI is highly valued from receiving countries not only because it doesnt create debt
obligation, but also it plays an important tool in transferring managerial and technical skills
from other countries. For some countries the technology transfer is valued more than capital
inflow. Foreign Direct Investments take place mainly through Multinational Corporations
(MNCs).
If we refer to theory, it suggests that in order to be able to compete in a foreign market an
investing company must have some ownership specific assets in technology, knowledge,
management, marketing, organization and etc. A company who possess such assets has
different ways to claim the rents in foreign markets, including joint ventures, franchising,
subsidiary production, licensing and marketing contracts.
Why multinational companies invest in specific locations? What are the receiving
countries main characteristics that attract FDI? In order to determine the host-countries
characteristics which attract FDI we need to analyze the motives for foreign investor, which of
the three types of motives is relevant to investor. The eclectic theories of FDI view
international production through a simultaneous presence of three types of advantages, or the
OLI (ownership, location and internalization) paradigm.5 The first type of FDI is called
market-seeking FDI, where investors purpose it to serve local markets. It is also referred as

Dunning, J. H. (1993). Multinational Enterprises and the Global Economy. New York: Addison-Wesley.

horizontal FDI, because investor replicates the production facilities in host country. The
reason for horizontal FDI is market size and the market growth. Investor may take an
advantage of market growth and serve better the local market, and can produce locally the
import substitute products. Imposed tariffs for importing goods and services, transportation
costs, various difficulties in gaining the access to local markets are also driving forces for
horizontal FDI.
The second type of FDI is asset-seeking or resource-seeking FDI. It takes place when
companys purpose is to gain access or acquire the resources in the host country which are not
available in home country, such as raw materials, natural resources or low-cost labor. It is
especially related with manufacturing sector. Multinational companies invest directly in
export oriented manufacturing sector and the cost of factors is significantly important. This
type of export oriented FDI or so called vertical FDI, unlike horizontal FDI, relocates one part
of production to host country. Because low cost production factors are the main drivers for
vertical FDI. In addition to that fact, FDI in the resource sector, such as oil and natural gas, is
attracted to countries with abundant natural endowments.6
The third type of FDI is efficiency-seeking FDI. It takes place when the company can gain
when there is a common governance of geographically dispersed activities and presence of
economies of scope and scale. Bevan and Estrin (2000) found this efficiency-seeking FDI to
be true for the first EU accession countries: Czech Republic, Estonia, Hungary, Slovenia and
Poland. It seems that those first wave of countries attracted more efficiency seeking FDI after
the initial accession progress announcement.
If we sum up all three different types of FDI, then the attracting factors of above described
FDI are low-labor cost, large domestic market, richness in natural resources and close

Esanov and others (2001) reported that most FDI in resource-rich countries of the CIS is of this type.

10

proximity to western developed countries. The countries with favorable condition in those
factors would attract more FDI.
However, there exists lots of other factors that plays role in attracting FDI. One of the
important factors is agglomeration economies. When agglomeration economies are present,
new investors mimic past investment decisions by other investors in choosing where to
invest.7 In this situation, new investors locate their investment next to previous investors who
has been there for longer time and they benefit from positive spillover from previous
investors.

1.2. Foreign Direct Investment and Spillovers


The theoretical discussions of investment inflow mostly refer to the transmission of new
and superior technology to host country. In general, the theories of the effect of direct
investment on host countries presented that foreign firms in the host country possessed
superior technology and technological knowledge spillovers. According to Findlay (1978)
the rate of change of technological efficiency in the backward is an increasing function of
the relative extent to which the activities of foreign firms with their superior technology
pervade the local economy.
If we look more specifically into the effects of Multinational companies in host economy,
foreign companies may introduce new know-how by providing the new technology and
training employees, improve their knowledge who can be employed by other local firms later;
contribute to efficiency by breaking supply bottlenecks, however this effect can be less
important as the technology of the host country advances; push local firms to improve their
managerial efforts, or to use the techniques of foreign firms in marketing on the local market
or internationally; depending on the strength and responsiveness of local firms foreign MNCs
7

F. Campos and Yuko Kinoshita (2003), IMF working paper

11

may break down monopolies and contribute to competition and efficiency or against to that
create monopolistic structure in the host country; transfer techniques used by MNCs in
inventory management, quality control and standardization to local distribution channels and
suppliers.8
The growth of developed countries has been due to mostly increases in the production
scale and capital intensity. Consequently, partial contribution of foreign companies is
introducing larger scale or more capital intensive production to host country. One of the
widely accepted effects of foreign entry is introducing new products or industries to hosteconomy and tighter linkage between host country and world trading system. MNCs main
contribution is knowledge, special knowledge of world markets demand, and how host
economy may integrate to world market by finding place in the worldwide allocation. FDI
inflow is associated with faster economic growth through the effects of developing new
products and productivity.
It has been abundantly shown that in host country foreign firms pay higher wages than
domestic firms. Foreign firms pay higher wages for many reasons, such as foreign firms hire
better qualified and better educated workers than domestic firms, foreign firms tend to be
larger, more capital intensive and more intensive in the use of intermediate products. Foreign
firms are mostly in the sector of higher wages. There are many evidences that whatever the
extent and direction of spillovers to domestic firms, foreign firms affect in the increase of
average level of wages. The effect may come from the effects of increasing the demand for
labor even if there is no difference in the wages of workers of foreign and domestic firms, or
from positive spillovers to domestic firms or from higher wages in foreign firms even if it
doesnt influence on domestic firms.
8 Magnus Blomstrm, Ari Kokko, The Economics of Foreign Direct Investment Incentives, 2003

12

1.3. Foreign Direct Investment flow to Transition Economies


FDI plays considerable role in filling the gap between domestic savings and the highly
needed investment to support economic growth in CIS countries over the medium term. FDI
provides external financing in equity form rather than debt, frequently in export and importcompeting sector which improves external position of the country.
Total investment in CIS countries declined dramatically after the breakup of the Soviet
Union. Although progress achieved in macroeconomic stability and relatively high growth
rates, net FDI inflows per capita has remained low in CIS countries in comparison with
Central and Eastern Europe and the Baltic states.

Figure 1. Inward FDI flows to transition economies, by Region and Economy


Inward FDI Flows, by Host Region and Economy, 1992-2005

35 000

Central Europe (EU)

30 000

Southeast Europe

25 000
CIS (without Russia)

20 000
Russian Federation

15 000
10 000

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

5 000

Sources: IMF, International Financial Statistics; IMF, World Economic Outlook Database; and authors' calculations.9

Central Europe (EU) Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia
Southeast Europe Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Macedonia, Romania, Serbia and
Montenegro

13

Prior to Russian crisis in 1998 the level of FDI was low due to Asian crisis in 1997, but it
fell down further in 1998 following the Russian crisis. Economic activity started to recover in
2000 and it was expected that the level of FDI inflow would increase. But statistics show that
there has been upturn in FDI inflow since 2002 (Figure 1). FDI inflow to Central Europe
countries significantly decreased in 2003 and then drastic increase in 2004. The possible
answer to this can be Boom and Bust in Information Technology Investment in 2002 and
economic crisis in South American Countries (Brazil, Argentine and Uruguay). Those crises
could threaten ambitious investors who were investing in intense to Central Europe that newly
joined EU. They didnt influence much on inward FDI to CIS countries as capital flow to CIS
countries were not significant like to Central Europe.
FDI inflow into energy exporting countries such as Russia, Kazakhstan, Azerbaijan and
Turkmenistan has generally been limited to energy sector, except Russia. FDI inflow to
energy sectors of CIS countries are focused on long-term, however the developments in those
countries also depend on investment climate such as corporate governance, rule of low,
transparency and etc.
In Moldova, Georgia and Armenia oil pipeline construction projects or energy sector
privatization accounted for the major inflow of FDI. In Ukraine FDI inflow has been more
diversified reflecting its industrial structure. In Tajikistan and Kyrgyz Republic10 FDI has
been confined mainly to one large gold mine project in each country with limited inward FDI
flows elsewhere in the economy.11

10

Inward FDI flows into the Kyrgyz Republic were mainly related to a few large projects and do not show a
significant impact of the Russian crisis. However, it is possible that inward FDI flows might have been higher
absent the general economic uncertainty in the region which followed in the wake of the Russian crisis.
11
Clinton R. Shiells, FDI and the Investment Climate in the CIS Countries 2003

14

FDI inflow to Uzbekistan and Belarus has been extremely limited, except the invested
capital in the construction of the Yamal pipeline in Belarus. Net FDI inflow in Uzbekistan
accounted less than 1 percent of GDP in 2005.

Figure 2. Commonwealth of Independent States (CIS) and Central and Eastern Europe and the
Baltic States (CEB): FDI Inflows, 19922005
Average annual FDI inflow
as a percentage of GDP (1992-2005)

Cumulative, in US$ Per Capita 1992-2005

Central Europe (EU)


Central Europe (EU)

Southeast Europe

Southeast Europe

CIS

CIS

Estonia

Azerbaijan

Czech Republic

Estonia

Hungary

Lithuania

Croatia

Bulgaria

Slovakia

Kazakhstan

Slovenia

Czech Republic

Poland

Georgia

Latvia

Armenia

Azerbaijan

Hungary

Bulgaria

Moldova

Lithuania

Croatia

Kazakhstan

Latvia

Romania

Bos.&Herz.

Serbia and Mont.

Romania

Macedonia

Slovakia

Bos.&Herz.

Poland

Albania

Albania

Georgia

Kyrgyzstan

Russian Federation

Tajikistan

Armenia

Serbia and Mont.

Ukraine

Macedonia

Moldova

Ukraine

Belarus

Turkmenistan

Turkmenistan

Slovenia

Kyrgyzstan

Russian Federation

Tajikistan

Belarus

Uzbekistan

Uzbekistan

0
0.0%

5.0%

10.0%

15.0%

1000

2000

3000

4000

5000

6000

20.0%

Sources: IMF, International Financial Statistics; IMF, World Economic Outlook Database; and authors' calculations.

According to Lankes and Venables (1996, 1997) motivation for foreign firms to invest in
CIS countries has been primarily resource seeking, however FDI inflow to more advanced
countries with transition economy has been more often efficiency seeking, i.e. focused on
product exports based on low labor cost. CIS countries with large domestic market such as
15

Russia and Ukraine may be potential for market seeking FDI which is oriented to produce and
sale within domestic market. Undoubtedly, this will require significant improvement in the
investment climate.
As seen from the Figure 2 & Figure 3 (Appendix A) throughout the transition period the
main receivers of the foreign direct investment were Russia, Kazakhstan, Ukraine and
Azerbaijan. Three of these countries are rich in oil. It suggests that natural resource
endowments can be one of the main factors affecting FDI flows to the CIS countries. When
FDI inflows are normalized with respect to population or GDP Azerbaijan and Kazakhstan
become leaders by a big margin over the rest of the countries. (Figure 6)
In the below Table A. CIS countries are placed in the matrix of Performance and Potential
indices, where Performance index captures the influence on FDI of factors other than market
size, assuming that, other things being equal, size is the "base line" for attracting investment.
These other factors can be diverse, ranging from the business climate, economic and political
stability, the presence of natural resources, infrastructure, skills and technologies, to
opportunities for participating in privatization or the effectiveness of FDI promotion. The
Inward FDI Potential Index captures several factors (apart from market size) expected to
affect an economys attractiveness to foreign investors (Appendix F).12 Matrix is based on
2002-2004 data.13

12
13

UNCTAD
Performance and Potential Indices are shown for three-year periods to offset annual fluctuations in the data.

16

Table A. Inward FDI Performance and Potential

HIGH FDI

HIGH FDI

LOW FDI

PERFORMANCE

PERFORMANCE

(Front-runners)

(Below potential)

Kazakhstan

Belarus, Ukraine,

POTENTIAL

Russian Federation
(Above potential )

(Under-performer)

Armenia, Azerbaijan,

Uzbekistan

LOW FDI

Georgia, Kyrgyzstan,

POTENTIAL

Moldova, Tajikistan

Source: UNCTAD 200614


Kazakhstan is front runner of all CIS countries with high FDI performance and high
potential. Although FDI inflow to Russia is the highest among all CIS countries (Figure 3) due
to the size of the economy, the performance of FDI is low. Belarus, Ukraine and Russia are
economies with high inward FDI potential. Depending on governments policy they have a
big potential to become attractive for foreign investors. Azerbaijan with highest FDI
performance is the best of all 141 economies. Armenia, Georgia, Kyrgyzstan, Moldova and
Tajikistan are in the group of above potential. Performance is the ratio of a countrys share in
global FDI inflows to its share in global GDP and it doesnt imply that those countries are
doing their best in attracting FDI inward. They still have empty rooms for improvement.
However, they are doing relatively well in attracting FDI. The worst of all is Uzbekistan and
under-performing in FDI inward. Performance and Potential indices of inward FDI are not
calculated for Turkmenistan due to lack of data.

14

The indices are calculated for 141 economies all over the world. For methodology see Appendix F.

17

1.4. Literature overview


There has been voluminous research and written works on FDI, but chosen works are
appropriate for this paper. Two different sets of literature will be discussed below. The first of
the groups includes those works that related to the determinants of FDI flows to transition
economies, while the second of them includes those which focus on explaining the
agglomeration or self-reinforcing effect of the FDI inflows.
Most of the existing literature on the determinants of the Foreign Direct Investment flows
to transition economies focuses primarily on the determinants and the effects of foreign direct
investment flows to Central and Eastern Europe. Smarzynska and Wei (2002) analyzed the
effects of corruption on FDI at firm level evidence in European transition economies. Bevan et
al. (2003) examine the relationship between institutional development and FDI inflow to
European transition economies. Holland and Pain (1998) analyzed the factors affecting
aggregate inflows of FDI in the ten accession economies plus Croatia over the five year period
from 1992 to 1996. Resmini (2000) also uses a sample of only ten countries.
There are two works that include CIS countries in their sample. Garibaldi et al. (2001)
make an analysis of the determinants of FDI and portfolio investment flow to 25 transition
countries. In their work the panel data includes the year period from 1992 to 1999, and OLS
model is used for estimation. Campos and Kinoshita (2003) worked in greater detail the role
of natural resources, infrastructure and agglomeration economies. The panel data is estimated
by a fixed effects model and system GMM. Even though the panel data includes the time
period 1990-1998, only 46 observations used in the regression for the CIS countries due to the
data incompleteness.

18

There is much evidence on the value of agglomeration economies, however the existing
studies focus on FDI in the United States or US FDI abroad. A work by Mody and Wheeler
(1992) makes a strong case for agglomeration in US investors location decisions. Ries, Head
and Swenson (1995) found that industry-level agglomeration economies play a significant role
in the location choice of Japanese manufacturing FDI in the United States. Barrell and Pain
(1999) found similar results on U.S. investment in Europe. A recent work by Cheng and Kwan
(2000) confirmed an analogous effect of agglomeration in China. They used in their work 29
regions of China in panel data model. Additionally, as mentioned before Campos and
Kinoshita (2003) tested for self-reinforcing effect.

19

2. DATA and Estimation


The data used comprise a panel of 12 transition CIS countries between 1992 and 2005. The
observations in the complete panel is 168 (=12x14).15 For the summary of data look Table 7
(Appendix E). Year 2006 is not included in this study, because sources used to pool the data
for independent and dependent variables are not sufficient . The definitions of the variables
used in our estimation is given in Table1 (Appendix C). Our dependent variable is natural log
of FDI stock constructed from a series of FDI inflows reported in the World investment
reports, UNCTAD 2006. Independent variables are pooled from different sources reported in
Table 1.
The cross-sectional aspect of our study is important. Transition from planned to market
economy started in the early 1990s after the break-up of Soviet Union. However foreign
investors were cautious in the beginning period. The past studies on FDI in transition were
mostly limited to the more advanced countries in Central Europe and Eastern Europe countries
which have been major receipients of FDI in the region. There has been conducted very few
numbers of studies related to the determinants of FDI in CIS countries due to the difficulty of
obtaining sufficiently long series of FDI data. In this study I hope to introduce more
heterogeneity and incorporate motives of investment in CIS countiries.
The time series aspect is important for this study. This is because during the time period
covered in data, transition economies have had lots of reforms in every sector of economy.
Cross-sectional data helps us to avoid taking into account changes of the reform variables. In
addition to this, the agglomeration or self-reinforcing of FDI can only be observed using time
series data. If there is an agglomeration effect then newly made investment would be an
increasing function of past investment.
15

The estimation data are unbalanced as some variables for certain observations are missing, mostly for 19921993. This is the period when countries only obtained their independence and early transition period.

20

2.1. Description of Variables


Foreign investors choose a location for investment based on expected profitability related
with each possible location. Return on investment is in turn affected by various locationspecific factors and by the type of investment motives. For instance resource-seeking investors
will look for a country with abundant natural resources. For this reason the location of FDI is
closely related to a countrys comparative advantage.
The independent (explanatory) variables are chosen according to empirical findings and
theoretical findings of the authors listed literatures in the reference section. In order to see
whether corruption, which is very general issue in CIS countries, has an impact on attracting
foreign direct investment in transition economies an additional explanatory variable control
of corruption is also included. According to Wei (2000a, 2000b), the corruption in a host
country substantially worsens inflow of FDI. Unfortunately I dont have a chance to compare
the result of my estimate, because he did not include CIS countries in his sample.
Based on the approach and explanation described by Garibaldi et al. (2001) it was not
imposed any normalization on the endogenous variable, that is on the FDI data, in other words
unscaled FDI data was used (i.e. simply log of FDI measured in millions of US dollars), while
controlling for the alternative scale variable, i.e. log of GDP on the right-hand side. If we look
more in details, Garibaldi et al. (2001) claim that the standard approach is to normalize the
data using GDP evaluated at market exchange rates, but this could introduce distortions due to
the large fluctuations of both exchange rates and real output in the sample. Normalizing by the
population also imposes significant impact on the relative levels of foreign direct investment
for the current sample.
21

In the model following explanatory variables are used, in Table 4. (Appendix C) you can
find the detailed definition and the sources for independent variables.
The first group of variables includes macroeconomic performance variables such as
Inflation and Fiscal Balance. Many economists and policymakers view macroeconomic
stability as a favorable condition for attracting FDI. Most investors would prefer
macroeconomic stability over instability in order to secure their revenues and profitability.
Only very few investors can invest in a country with instable economy and politics taking on
risk, this is in order to make an advantage of certain conditions and making unforeseen profit.
In general, inflation should have a negative effect on the FDI inflow. None of the investors
would like to loose their money through inflation in the host country and depreciation of
domestic money. In the model Log of inflation (Ln_Inflation) is used in order to scale the
variables, since inflation was too high in the early periods of transition, i.e. as big as 4000 or
more. The fiscal balance is also one of the indicators of the macroeconomic stability (Fischer,
Sahay and Vgh, 1997), and it should have a positive effect on the FDI, the reason is surpluses
are better than deficits in the eyes of investors. If an investor foresee high fiscal deficit in
certain country, then he expects the government can impose more taxes in the future, so
investor will be discouraged.
The CIS countries such as Azerbaijan, Kazakhstan and Russia are rich in oil and gas. They
receive much FDI in resource-based industries. Thats why natural resource endowments are
also expected to have a positive effect on the FDI. In order to test the effect of natural
resources I used a dummy for the richness of the countries in natural resources. Data is taken
from De Melo and others (1997), where they use 0, 1, 2 dummy variables for poor,
intermediate and rich countries respectively.

22

In order to operate successfully good infrastructure is important condition for foreign


investors regardless of the type of FDI. To test the effect of infrastructure I used main
telephone lines as infrastructure variable. Availability of main telephone lines is important to
facilitate communication between the home and host countries.
The EBRD index of FOREX and trade liberalization is also used as an explanatory
variable. That index covers two aspects of the economy, i.e. exchange rate market restrictions
and trade restrictions. Multiple exchange rates and convertibility restrictions worsen inflows
of FDI, since the investors will face difficulties with repatriation of profits and also when
importing intermediate goods. However, trade liberalization may have an ambiguous effect on
the FDI. Trade barriers may increase inflows from market-seeking type of investors, since
they want to serve local market and to overcome the trade barriers and in the case of trade
being open they would simply export their products. At the same time trade restrictions will
deter vertical type of investors and export-oriented investors, since the former will have
obstacles with buying intermediate goods and the latter will face difficulty with selling the
product in other markets.
One another explanatory variable of the regression is market (investor) perceptions, as
captured by country risk ratings, which is regularly published in investor magazine
Euromoney. This is an overall indicator of investor perceptions that gives equal weight to a
countrys economic and political/institutional performance.16 I decided to include market
perceptions as one of the explanatory variables, whether or not the perceptions matter even in
the presence of variables that directly control for fundamentals.
Some of EBRD selected transition indicators have also been used as independent variable
to account for their impact on FDI in the transition countries. Those indicators are EBRD
16

Garibaldi, P., N. Mora, R. Sahay and J. Zettelmeyer (2001), What Moves Capital to Transition Economies?

23

index of small-scale privatization, EBRD index of large-scale privatization, EBRD index of


banking sector reform and EBRD index of reform of non-banking financial institutions. Bevan
and al.(2003) have also used those indices in their estimation model which comprises a sample
of 11 European transition countries. All those EBRD indices are expected to have positive
impact on attracting FDI.

2.2. Estimation Methodology and Partial adjustment model of the FDI


Agglomeration economies arise when there are benefits from locating close to other
economic units due to positive externalities. This is when foreign investors are attracted to
host countries by more existing foreign investment. In this situation new foreign investors
have less knowledge about host country and its environment, and then they view the
investment decisions by others as a good signal of favorable conditions and follow previous
investors. Foreign investors reduce uncertainty.
In order to test the agglomeration effect I build a model of current FDI stock which is
related to the past FDI stock and also to other time-invariant and time-variant explanatory
variables. For this purpose a partial adjustment model of FDI a la Chow (1967) used by
Campos and Kinoshita (2003) and Cheng and Kwan (2000) will be employed. In the model I
will use capital stock rather than investment flow, because the profitability of investment
depends on the marginal return to capital that is generally a decreasing function of the stock of
capital.17

17

Cheng and Kwan (2000), model has been slightly changed in this paper, see Appendix B for further
clarifications.

24

Let

Yit be the actual stock of FDI in the country i at time t and Yit* is the corresponding

equilibrium or desired stock. The flow of investment will adjust Yit towards Yit* according to
the below process:

d ln Yit / dt = (ln Yit* ln Yit ), 0 < < 1

(1)

Above equation states that percentage change in the FDI stock level is proportional to the
gap between log of the actual FDI stock level ( ln Yit ) and log of the desired FDI stock level
( ln Yit* ).
Since d ln Yit

= dYit / Yit , then equation (1) can be rewritten as below:

dYit / dt = Yit (ln Yit* ln Yit ), 0 < < 1

(1a)

Now it is possible to see the agglomeration effect or self-reinforcing effect of FDI stock.
This effect is also called the positive feedback effect. Since
it suggests that

Yit is positively related to Yit ,

Yit will affect positively the future level of Yit . In other words, todays FDI

stock level is affected positively by the past level of the stock of FDI. Moreover, this effect
diminishes as the gap between actual and desired level of FDI decreases. It suggests gradual
adjustment.
An attractive point of the model is that Yit* is not static and it will shift over the time, that
is Yit* Yi* . Assuming a particular level of the equilibrium stock, Yit* = Yi* for all t, (1) can be
solved as a differential equation to yield the Gompertz growth curve:

Yit = exp(ln Yi* exp( t ))


25

(2)

Equation (2) describes the natural growth of the FDI stock which would have prevailed
had there been no change in factors that shift the equilibrium stock. Therefore Equation (1)
combines two elements that account for the observed accumulation of FDI. First, the selfreinforcing effect and the adjustment effect drive the FDI stock to reach an equilibrium level,
and second, the equilibrium level itself shifts ( Yit* Yi* ) as a result of changes in the
environment.
For the purpose of empirical work, equation (1) can be rewritten as follows ( yit = ln Yit ):

yit yit 1 = ( yit* yit ),


After regrouping above equation we get:

) yit 1 + (
) y*
1+
1 + it

yit = (

For the system of adjustment to be stable (non-explosive) the term

(3)

1
) must be bigger
1+

than 0 (zero) and smaller than 1.


The determinants of yit* must be specified in order to estimate the model given by
equation (3). Equation (3) tells that current FDI stock is determined by past FDI stock and
other explanatory variables.
In order to estimate other time-variant and time-invariant explanatory variables of the FDI
stock, Panel data Random Effects model will be conducted:

yit = X it + it ,

26

where X it = Z it + N i and

it = i + it .

These tell that explanatory variables include both

time-variant and time-invariant variables. Additionally, it includes the country-specific (i )


attributes.
In order to test the agglomeration effect of the FDI stock, System GMM model (dynamic
panel data) will be employed:
1
) y + X it + it ,
1 + it 1

yit = (

where it

= i + it

1
If there exist agglomeration effect or a positive feedback effect, then (
) should be
1+
positive.

3. Panel Data Fixed Effects vs. Panel Data Random Effects


It is time to estimate the variables that affect the FDI. Panel data is used for estimating the
explanatory variables excluding the agglomeration effect. The use of panel data model is
appropriate, because it covers both time and cross sectional dimensions. For this purpose the
below model that described above is used:

yit = X it + it ,
where X it = Zit + Ni and

(4)

it = i + it .

In other words, explanatory variables include

both time-invariant ( N i ) and time-variant ( Z it ) variables. Additionally,


country-specific (i ) attributes, or

it includes

the

i represents the unobserved heterogeneity among

countries in the sample.


27

This is a general form of panel data model. The model can be estimated by several models.
It should be decided which model specification to use out of available options: random effects
model, fixed effects model, pooled OLS and the Hausman-Taylor formulation. As it is very
important to use the right model, each of the above mentioned models and their advantages
and disadvantages will be discussed below.
Theoretically, Fixed-Effects estimator takes

i as fixed or as country-specific constant

term, and it does not change over time. However it does not necessarily imply that i is nonstochastic18. The fixed-effects estimator subtracts from the equation to be estimated the over
time average of the equation for each country19. Because of that transformation the term

disappears from the estimation and the coefficients are estimated according to the time
variation within each cross-sectional unit, in our case within each country. The advantage of
this approach is avoiding any potential correlation between
variables. Because

i (fixed effect) and explanatory

i is not included in the equation of our estimation anymore. For this

reason it is better to use fixed effects method, if this correlation is too high. The disadvantage
of the fixed-effects model is that time-invariant variables will not be estimated, because timeinvariant variables will not be included in the equation when their average over the time (is
equal to itself) is subtracted from the equation. Additionally, the variables with very little
variation over the time will be estimated inefficiently.
If in the model there are time-invariant variables or variables with little variation over the
time, then it is better to use one of the following models: Random Effects model, Pooled-OLS,
the Hausman-Taylor formulation.

18

Greene (2003), Econometric Analysis, Fifth Edition

19

Neumayer E. (2003). Are Socioeconomic Factors Valid Determinants of Suicide? Controlling for
National Cultures of Suicide with Fixed-Effects Estimation, Cross Cultural Research

28

When Pooled OLS model is used in the estimation of the model, unit fixed effect is
ignored and excluded from the equation. However, if
then exclusion of the term

i is correlated with both X it and yit ,

i will bias the estimation. Consequently, it is unlikely that

unbiased coefficients will be obtained as a result of pooled OLS, especially if N is small. 20


The Random Effects model is also sometimes described as a regression with a random
constant term. It is assumed that the intercept i is a random outcome variable. So, i enters
the error term. Composite error term of the random effects model will be

it = i + it .

Where i is random effect (not fixed-effect anymore), i will be incorporated into the overall
error term. The major disadvantage of this model is driven by the assumption that, the
unobserved effect

i is uncorrelated with each independent variable ( X it ). Following the

assumption, if there is a high correlation between the terms

X it andi ,

the estimated

coefficients will be inconsistent, even though they may still yield efficient results.
A strategy that should prove more appropriate than pooled OLS in the presence of timeinvariant variables is adapting the random effects model. 21
In order to overcome the issue of random effects inconsistency, Hausman and Taylor
(1981) suggest to use the instruments for the variables that are likely to be correlated with the
error term which encompasses the random effects. So, Hausman and Taylor models
assumptions are:
1. Some of the explanatory variables

X it

can be correlated with i term.

20

Thomas Plmper , Vera E. Troeger, Philip Manow (2005). Panel data analysis in comparative politics:
Linking method to theory. European Journal of Political Research
21
Thomas Plmper , Vera E. Troeger, Philip Manow (2005). Panel data analysis in comparative politics:
Linking method to theory. European Journal of Political Research

29

2. None of the explanatory variables

X it

are correlated with it term.

Unfortunately, the correlation between some terms of

X it andi

is unobservable and

thats why it requires some logical approach to correctly specify the Hausman-Taylor model
with instrumental variables.
In the presence of time-invariant variables, Pooled OLS model is worse choice than the
Random Effects model. The Hausman-Taylor model is difficult to perform correctly,
consequently Random Effects model would be the best choice. But the results of Random
Effects model can be inconsistent. For this reason, before using the model it is required to
check for the consistency.

3.1. The estimation of the Model

yit = X it + it ,

(4)

Where X it = Zit + Ni and it = i + it . In other words, the independent variables include


both time-variant ( Z it ) and time-invariant ( N i ) variables. Additionally,
country-specific (i ) attributes, or

it includes

the

i represents the unobserved heterogeneity among

countries in the sample.

yit = {Ln_FDI};

Zit ={Ln_GDP, Ln_Inflation, External_Liberalization, Fiscal_Balance, Corruption,


Risk, Small_Privatization, Large_Privatization, BankReform_IntRate,
SecurityMarket_NonBank, Ln_TelLines, Structural_Change }. All these variables are
changing over time and across countries. Note: Structural_Change is a dummy and in
1992-1998 takes the value of 0, for 1999-2005 takes the value of 1 and same for all
countries;
30

Ni ={ Natural_Resource}. This variable does not change over time, but does so

across countries.
The

main

model

regression

will

include

all

explanatory

variables

except

Structural_Change (Table 2). One another model is built excluding Corruption and
Structural_Change (Table 3).
Hausman's specification test, which tests the appropriateness or the consistency of the
random-effects estimator, was used in order to check for the consistency of random effects
estimation. Primarily the regression was run with both fixed-effects model and random-effects
model. After running both models the Hausman test was executed. Under the null hypothesis,
Hausman test indicates that difference in coefficients obtained from two estimations in not
systematic and random effects estimator is efficient. I got following results after running
Hausman test: chi2(11)=97.24 and Prob>chi2 = 0.000 (Table 8). Results tell that the null
hypothesis is accepted. It suggests that the random effects model is consistent and it should be
used. Consequently, the regression was run with Random Effects model.
Using both iterated GLS and LR test heteroskedasticity was detected in the model. After
correcting for the standard errors, further estimations were conducted.
Instrumental variable was also used in order to fix the problem of endogeneity22. Since
GDP can affect the FDI, so we can face an endogeneity problem. That is why instrumental
variable population was used, because it is highly correlated with GDP, and has lower
22

Although the estimators implemented in Hausman Taylor (xthtaylor) and Random-effects IV


regression (xtivreg) use the method of instrumental variables, each command is designed for very
different problems. The estimators implemented in Random-effects IV regression (G2SLS) assume
that a subset of the explanatory variables in the model are correlated with the idiosyncratic error or

it

term. In contrast, the Hausman-Taylor estimator assume that some of the explanatory variables are

correlated with the individual-level random-effects, i , but that none of the explanatory variables are
correlated with the idiosyncratic error

it .
31

correlation with FDI. G2SLS random-effects Instrumental Variable regression was run using
the Ln_Population as an instrument. But estimation results did not change. (Model.1 and
Model2 in the Table 4)
In a following regression, dummy variable Structural_Change was introduced in the
model to detect structural change. Once again population was instrumented to Ln_GDP in
model4. (Model.3 and Model.4 in the Table 4)
Unfortunately, there is no any technique that would test for the endogeneity problem in the
random effects model. Any of the techniques that are used to test the endogeneity issue in the
fixed effects model did not work for random effects model.
Finally, all these models were run once again excluding the variable Corruption in order to
increase the number of observations (Table 5). The data for explanatory variable Corruption is
available for 1996-2005 periods.

3.2. Estimation Results


The explanatory variable Control of Corruption turned out to be insignificant. This result
can be explained by the fact that the investors coming to the region already may know how to
deal in the situation of high corruption environment. That is why corruption may not affect
their investment decision. This result can also be explained by the fact that numbers of
observations for explanatory variable Corruption were not enough to get better estimate.
However, in order to be sure about the corruption effect it was run OLS regression to
detect cross-country effect of corruption on FDI. It was run to check once again the effect of
corruption, because as mentioned above in some studies corruption had effect on inward of
FDI but CIS countries were not included in those studies. I decided to check in different way
and Annual FDI inflow per capita (1996-2005) is used as dependent variable. Accordingly
32

GDP per capita, Fiscal balance and inflation are used as explanatory variables together with
corruption variable. The answers were again insignificant (for this reason the results were not
reported in details in the paper), even when I got significant result on corruption it had a
negative sign which is logically wrong.

23

Another approach was made replacing corruption

independent variable with Rule of Law explanatory variable. However, the outcome was same
with Control of Corruption i.e. with unexpected sign (negative) and insignificant. Thats why
it was not reported in details.
As corruption explanatory variable found to be insignificant, alternative model was
constructed where all explanatory variables kept same as main model except corruption
variable. Corruption is excluded in the alternative model. Below described results for each
explanatory variables were from alternative model (Table 5, Model 1).

Market size GDP is significant at 1% level and positively affects the FDI. 1% increase in
the GDP will increase the FDI by 0.26%. It implies that when GDP increases, the value of
FDI as a share of GDP will decrease, since GDP will grow faster than FDI. But here one must
be careful with the interpretation. It does not imply that GDP grows faster than FDI across
time. Change in FDI and GDP has two dimensions: cross country and across time. So, the
change in the GDP can be seen not only across time, but also cross country. In other words
countries with higher level of GDP may not have proportionally high level of FDI.

Inflation rate (measured in %) is also significant at 1% level and its sign is as expected.
Most of the investors would prefer macroeconomic stability over instability in order to secure
their revenues and profitability. That is why an increase in inflation as an indicator of
instability will deter the investment. One should be careful in interpreting the result, because I

23

For different models in testing the corruption variable, I got correlation coef.s such -0.28, -0.17. Those
coefficients are fairly high. But the sign of correlation is opposite of expected, which explains the fact that
regardless the significance of coefficients the result for corruption is unexpected.

33

used natural logarithm of the inflation rate. So, 1% change in inflation will decrease the FDI
by 0.29%.

The EBRD index of FOREX and trade liberalization also has a positive effect and is in
most cases significant (Table 3 and 5). However its significance is not robust, because in
models where Corruption is included this index estimation becomes mostly insignificant
(Table 4). When structural change is included in the model, this variable became significant at
1%. Unit change in external liberalization index implies to 26.1% change in inflow of FDI.
(Index range: from 1 to 4)

Fiscal balance (measured in % of GDP) is found to be significant at 1% level and sign is


positive as expected. One unit change (from 5% to 6%) must cause 3.3% increase in FDI. In
all models fiscal balance variable stays very significant explanatory variable.

Natural resources endowment indicator is also significant at 1% level and its sign is also
correct. As expected, countries rich in natural resources received more direct investment on
average, 82.4% compared to intermediate countries and 161.3% compared to poor countries.
One unit increase in natural resources indicator must cause 92.2% increase in FDI. This
indicator is significant at 1% for all models.

Market (investor) perceptions variable (noted as Risk) has also positive effect and
significant in most cases. However in the model where structural change is included, the
variable became insignificant. Unit change in index implies to 3.3% change in FDI. (Index
range: from 1 to 100)
All above described results24 are consistent with the findings of other papers such as
Garibaldi et al. (2001) and Campos and Kinoshita (2003).

24

Except variable control of corruption that was not used in previous studies for CIS countries.

34

The EBRD index of large-scale privatization has a negative and significant at 1% level
effect on FDI inflow. One unit decrease in large-scale privatization indicator should cause
49.8% increase of FDI inflow. Almost in all models this variable stays very significant. At
first sight this result seems contradictory to our expectation. However, if we analyze
thoroughly the reason it becomes logical. In CIS countries most of the large-scale
privatizations have been carried out through selling the businesses to management or some
insider local people (For instance, privatization by vouchers program in Russia). Since law
enforcement is weak, foreign investors are discouraged to invest in those businesses. It is
better for foreign investors to invest in state owned companies. Because the state is more
credible than private sector, also it provides guarantee for investors and most of the
investment incentive programs have been directed to this sector. This result is also consistent
with the findings of other papers. Garibaldi et al. (2001) found that insider privatizations
(privatization as a sale or gifts to the management and/or workers of a previously state owned
enterprise) discourage direct investment.

The EBRD index of reform of non-banking financial institutions has a positive and
significant (5% level) effect. In the model where structural change included, the variable
becomes significant at 1%. One unit increase in this EBRD indicator should cause 50.8%
increase of FDI inflow. Bevan and al. (2003) found that this is not significant in their model. It
can be explained in the following way: since in European Transition countries banking reform
was intense and fast, the reform of non-banking financial institutions can not influence the
decision of foreign investors; in contrast to this, the banking reform in CIS countries has been
very slow that foreign investors should take non-banking financial institutions reforms into
consideration in their investment decision making process.

35

The EBRD index of small scaled privatization has a positive as expected and significant at
1% level effect on FDI. One unit increase in this EBRD indicator should cause 48.9% increase
of FDI inflow.

The EBRD index of banking sector reform variable is turned out to be insignificant in the
model. Bevan and al. (2003) found that this indicator has positive impact on FDI in their
model of 11 European transition countries. Such finding following our estimation can be
explained by the fact that this reform in CIS countries is not fast and intense enough to affect
the FDI.
The dummy variable Structural_Change is introduced to detect the structural change. This
explanatory variable is positive and significant at 1% level in both model main and alternative
(no corruption). It tells that there is clear structural change or positive shift in the model
starting from year 1999. Such positive shift can be explained by two factors:
1. In 1999 CIS countries began a steady recovery after 1998 Russian financial crisis
making the region favorable for foreign investment.
2. Rapid oil price increase began to attract huge resource-seeking investment.

4. Dynamic Panel Data


In order to detect the agglomeration effect of the FDI stock, Dynamic Panel Data Model
(System GMM and Difference GMM) was employed. From the derivation of the partial stock
adjustment model it will be obtained:
1
) y + X it + it ,
1 + it 1

yit = (

(5)

36

Where it

= i + it

yit = {Ln_FDI}, yit 1 ={ Lag_Ln_FDI}

X it ={Ln_GDP, Ln_Inflation, Risk, Fiscal_Balance, External_Liberalization,


Small_Privatization, Large_Privatization, SecurityMarket_NonBank, Ln_TelLines,
Natural_Resources}

The explanatory variable Corruption is excluded from the model in detecting


agglomeration effect, because of two reasons. Firstly, it is because of increasing the number of
observations of the model, since the data for this variable is available for 1996-2005 periods.
Secondly, this explanatory variable is insignificant in explaining the dependent variable. The
EBRD index of banking sector reform is also excluded, because of its insignificancy.

1
If there is an agglomeration effect or a positive feedback effect, then (
) should be
1+
positive. There are several issues while dealing with this type of the model because

yit 1 and it will be correlated. In Fixed-Effects model they will be correlated in finite or fixed
amount of T, and in Random-Effects model it is obviously correlated, because random term

i enters the equation for each time period in group i .


Arellano and Bond (1991) refined a two steps procedure based on differencing and
instrumenting to solve that problem. The first step of the process consists of differencing the
equation to remove the fixed effects.

yit

1
)yit 1 + X it + it
1+

=(

(6)

In equation (6), yit 1 and it are contemporary correlated. And in the second step they
suggested that either the two period lagged difference or the two period lagged level of

37

dependent variable could be used as an instrument for yit 1 .Since both are correlated with
the latter term while being uncorrelated with it , both instruments will lead to a consistent
estimator. This estimation method is called Difference GMM.
Arellano and Bover (1995) showed that if the original equation in levels were added to the
system of the first differenced equations, giving rise to a System-GMM estimator, the
efficiency of the estimation would rise. 25
The model to detect agglomeration effect is estimated with both Difference and System
GMM model. Different sets of instruments (Table6) were used for the estimation procedure.
In Model (1) variables Ln_Population, Ln_Inflation, Ln_TelLines and External_Liberalization
are considered as strictly exogenous and used as instruments to difference equation. Ln_FDI

and Ln_GDP are used as GMM-style instruments. That is, their lagged values from 1st
through 4th period are used as instruments. The model (1) is called Difference GMM model.
Model (2) uses identical instruments, but level equation is included in the estimation. That is
why it is called System GMM model. In Difference GMM model Natural Resources variable
and constant is dropped, because they are constant over the time period and the change is zero.
In models (3) and (4) External_Liberalization is excluded from the list of instruments. In
Models (5) and (6) Ln_TelLines and External_Liberalization is excluded from the list of
instruments. Models (7) and (8) includes Ln_Population and Ln_Inflation as instrument but
lagged values from 1st through 3rd period are used as instruments, while in Models (9) and (10)
only 1st and 2nd period lagged values are employed as instruments. In all of the Models Hansen
test indicates the validity of the instruments.
Estimation results showed significant (at 1% level) agglomeration effect and its value lies
between 0 and 1 as expected. However, some of other explanatory variables became
25

See APPENDIX D for details.

38

insignificant in some models. It is necessary that the first-differencing operation not only
eliminates unobserved country-specific effects but also time-invariant explanatory variables
for which only cross-sectional information is available. As mentioned above in all difference
GMM models Natural_Resource (time-invariant) explanatory variable is dropped out of the
equation. Even though this variable appears in System GMM model, it is highly insignificant.
The reason is System GMM includes conditions of moments of both difference and level
equations. Since, conditions of moments of difference equation do not include

Natural_Resource, the System GMM estimation can be biased. Even though System GMM
estimation is more efficient than Difference GMM, the efficiency gain from imposing level
moment conditions does not come free; we do need extra assumptions and the violation of
which may lead to bias.
Furthermore, as shown by Blundell and Bond (1998) and Ahn and Schmidt (1995, 1997),
under a Random-Effects model, the first-differenced GMM estimator can suffer from serious
efficiency loss, for there are potentially informative moment conditions that are ignored in the
first-difference approach.26
Campos and Kinoshita (2003) found a significant agglomeration effect and it is particular
for both non CIS and CIS transition economies. And the evidence of agglomeration effect in
my estimation is consistent with the results of other papers.

26

Cheng and Kwan (2000)

39

5. Conclusion

Foreign direct investment can play a significant role in raising a countrys technological
level, creating new employment, and promoting economic growth. For this reason many
countries are actively trying to attract FDI in order to promote host countries economic
development, particularly at times when the countrys domestic growth prospects appear
weak. Foreign direct investment has been an important, relatively stable source of private
financing in some transition economies. However, most of the CIS countries witnessed lower
and less stable inflows.
The analysis presented in this paper has enabled identification of several key determinants
of FDI inflows to the transition economies of Commonwealth of Independent States countries.
For this purpose an unbalanced panel data covering the years 1992-2005 for CIS countries
was used. The regression results indicate that the pattern of inward direct investment in
transition economies can be well explained in terms of a standard set of economic
fundamentals.
After thorough examination of theoretical and empirical studies, potential determinants
were chosen for further estimation of the model. In order to present the agglomeration effect
of FDI theoretical framework was deployed. Derivation of Partial Adjustment Model of FDI is
followed by Cheng and Kwan (2000) with slight changes. Fixed and Random Effects models
were tested for the model and Random effects model was employed to estimate the panel data
model as it fitted better than fixed effects model.

40

By utilising a well detailed dataset which permits to identify FDI inflows to CIS countries
with transition economies, I have found that FDI inflow is not influenced by Corruption in
host countries. Results showed that Control of Corruption has no significant effect. Although
corruption has a significant negative effect on inward FDI according to previous literatures, it
is not significantly applicable for CIS countries. Main hypothesis is rejected for CIS transition
economies. Then I made another approach by replacing corruption explanatory variable with
Rule of Law27 variable. But the replaced variable also was insignificant with unexpected sign
(negative). For this reason it was not reported in the paper. Alternative way to test the effect of
corruption is by qualitative research. Even if it is more time consuming, it is possible to get
complete and detailed results for tested hypothesis like conducted questionnaire survey by
Aristidis Bitzenis (2006) for Bulgarian transition economy.28

By excluding Control of Corruption from the main model, alternative model was
constructed and all explanatory variables effects were described according to alternative
model. Natural resources endowment indicator explanatory variable is significant and have
expected sign. It is significant at 1% in all tested models. Second hypothesis is accepted and
test results tell that natural resources play very important role in attracting FDI in CIS
countries.
The macroeconomic indicators: Market size (GDP), Inflation rate, Fiscal balance, Main

Telephone lines explanatory variables are all significant and have expected sign. The EBRD
index of FOREX and trade liberalization has also positive effect and in most cases significant.
But its significance is not robust, because in models where Corruption is included this index

27

The variable law and order that assesses the strength and impartiality of the legal system and popular
observance of the law [Source: International Country Risk Guide]
28
Questionnaire survey concluded that foreign MNEs looked upon bureaucratic or administrative issues and the
regulatory environment, together with corruption, political and macroeconomic instability, as the most decisive
barriers in their decision to undertake FDI projects in a Balkan country.

41

estimation becomes mostly insignificant. Market (investor) perceptions variable also has
positive effect and significant in most cases. All these results were consistent with the results
of other papers. The results for above explanatory variable including natural resource
endowment are consistent with the findings of other papers such as Garibaldi et al. (2001) and
Campos and Kinoshita (2003).

EBRD index of banking sector reform turned out to be insignificant in the model. The
EBRD index of small scaled privatization and EBRD index of reform of non-banking financial
institutions have both positive and significant effect. Bevan and al. (2003) found reform of
non-bank financial institutions insignificant for European transition economies. It can be
explained that in European Transition countries banking reform was fast and the reform of
non-banking financial institutions cant effect the decision of foreign investors; unlike in
European transition countries the banking reform in CIS countries has been very slow that
foreign investors should take non-banking financial institutions reforms into consideration in
their investment decision making process. The EBRD index of large-scale privatization has
negative and significant effect on FDI.
The dummy variable Structural_Change was introduced to detect for structural change
and the result is positive and significant. It means that there is a structural change or positive
shift in the model beginning from the year 1999. It can be explained by two factors whereas
first after Russian financial crisis in 1998 CIS countries started to recover and region became
more favorable for foreign investors and second is the increase of oil price influenced
investors decision making about FDI location.
In order to check for agglomeration effect of FDI in CIS countries Derivation of Partial
Adjustment Model was used empirically. For this purpose Difference GMM and System
GMM models were employed. The result is that there is an agglomeration effect. The result
42

was consistent with other studies. The outcome of test tells that new investors mimic past
investment decisions by other investors in choosing where to invest.
This research can be more extended by considering industry specific FDI determinants.
The industry level data analysis for European countries showed that industry specific
determinants do matter however extensive research has yet to been done. Analysis of such
determinants can be very useful in order to attract FDI into some specific industry, which is
relevant for the overall sound FDI attraction strategy. Besides, there is under examined issue
related special economic zones, their efficiency in the scope of FDI attraction, and their
influence on foreign investors decision making about FDI location which may contribute to a
better knowledge of the FDI determinants.

43

References
Andreas Johnson, 2006, FDI inflows to the Transition Economies in Eastern Europe:

Magnitude and Determinants, CESIS Paper No. 59


Arellano, M., and S. Bond, 1991, Some Tests of Specification for Panel Data: Monte

Carlo Evidence and an Application to Employment Equation, Review of Economic Studies,


Vol. 58, pp. 27797.
Aristidis Bitzenis, 2006, Decisive FDI barriers that affect multinationals' business in a

transition country, Global Business and Economics Review, Volume 8, Number 1-2 / 2006,
pp. 87 - 118
Artige, Lionel, Nicolini, Rosella (2005), Evidence on the Determinants of Foreign Direct

Investment: The Case of Three European Regions, Barcelona.


Bevan, Alan A., and Saul Estrin, 2000, The Determinants of Foreign Direct Investment in

Transition Economies, CEPR Discussion Paper No. 2638


Bevan, Alan, Estrin, Saul, Meyer, Klaus, 2004. Foreign investment location and

institutional development in transition economies. International Business Review 1, 4364.


Bevan, Alan A., Estrin, Saul, 2004. The Determinants of Foreign Direct Investment into

European Transition Economies. Journal of Comparative Economics 32, 775-787.


Beyer J., (2002), Please invest in our countryhow successful were the tax incentives

for foreign investment in transition countries?, Communist and Post-Communist Studies 35,
pp. 191-211
Brenton, Paul, DiMauro, Francesca, Lcke, Matthias, 1999. Economic integration and

FDI: An empirical analysis of foreign investment in the EU and in Central and Eastern
Europe. Empirica 26, 95121.
Cheng, L., and Y. Kwan, 2000. What Are the Determinants of the Location of Foreign

Direct Investment? The Chinese Experience, Journal of International Economics, Vol. 51, pp.
379400.
Claessens, S, D. Oks and R. Polastri (1998), Capital flows to Central and Eastern Europe

and Former Soviet Union, World Bank working paper.


44

Clinton R. Shiells, FDI and the Investment Climate in the CIS Countries 2003
Daniel C., J.J. Reid, 1998 Observations on the Relations between the Transition of

Eastern European/Central Asian Economies, Economic Growth, and Direct Foreign


Investment, The Social Science Journal, Volume 35, Number 3, pages 455-459.
De Melo, M., C. Denizer, A. Gelb, and S. Tenev, 1997, Circumstance and Choice: The

Role of Initial Conditions Policies in Transition Economies, World Bank Policy Research
Working Paper 1866 (Washington: The World Bank).
Dunning J. H., 1993, Multinational Enterprises and the Global Economy. New York:
Addison-Wesley.
Esanov, A., M. Raiser, and W. Buiter, 2001, Natures Blessing or Natures Curse: The

Political Economy of Transition in Resource-Based Economies, EBRD WP No. 65, (London:


European Bank for Reconstruction and Development).
European Bank for Reconstruction and Development, 1999-2006, Transition Report 19992006, London.
Fabry Nathalie, Sylvain Zeghni (2002), Foreign direct investment in Russia: how the

investment climate matters, Communist and Post-Communist Studies 35, pp. 89303
Findlay, Ronald. "Relative Backwardness, Direct Foreign Investment, and the Transfer of

Technology: A Simple Dynamic Model." Q.J.E. 92 (February 1978)


Garibaldi, P., N. Mora, R. Sahay and J. Zettelmeyer (2001), What Moves Capital to

Transition Economies? IMF working paper WP/02/64.


Explaining Foreign Direct Investment in Transition

Holland, D., Pain, N. (1998): The Diffusion of Innovations in Central and Eastern

Europe: A Study of the Determinants and Impact of Foreign Direct Investment. National
Institute of Economic Research, Discussion Paper, London, June.
Isayev U., Foreign Investment During the Transition: How to Attract It, and How to Make

the Best Use of It.General Director Foreign Investment Agency, Kyrgyz Republic., Session VII
Ishaq, Mohammed (1999), Foreign direct investment in Ukraine since transition,
Communist and Post-Communist Studies 32, pp. 91109.

45

Kinoshita, Y. And N. Campos (2003), Why does FDI go where it goes? New evidence

from the transition economies, WDI working paper n. 573


Lansbury, M., N. Pain and K. Smidkova (1996), Foreign Direct Investment in Central

Europe since 1990: An Econometric Study, National Institute Economic Review, May, 104114.
Lankes, H.-P., Venables, A.J. Foreign Direct Investment in Economic Transition: The

Changing Pattern of Investments. Economies of Transition, 1996,


Lim Ewe-Ghee (2001): Determinants of, and the Relation between, Foreign Direct

Investment and Growth: A Summary of the Recent Literature. IMF Working Paper
(WP/01/175), Washington, November.
Lipschitz, L., T. Lane and A. Mourmouras (2002), Capital flows to transition economies:

master or servant?, IMF working paper, WP/02/11.


Magnus Blomstrm, Ari Kokko, The Economics of Foreign Direct Investment

Incentives, 2003
Meyer, Klaus E., 1998. Direct Investment in Economies in Transition. Edward Elgar,
Aldershot.
Meyer, K.E. (1995), Direct Foreign Investment in Eastern Europe: The Role of Labor

Costs, Comparative Economic Studies, Winter, 37 (4), 69-88.


Mohsin Habib, Leon Zurawicki Corruption and Foreign Direct Investment Journal of
International Business Studies, Vol. 33, No. 2. (2nd Qtr., 2002), pp. 291-307
Resmini, L., 2000, The Determinants of Foreign Direct Investment in the CEECs
Economics of Transition, Vol. 8 (3), pp. 66589.
Shiells, C., FDI and the Investment Climate in the CIS Countries, IMF Policy
Discussion Paper
UNCTAD, 2002-2006, World Investment Report 2002-2006.
World Bank, 2002. Transition: The First Ten Years. World Bank, Washington.

46

APPENDIX A29
Figure 3.
Average Annual FDI Inflow (1992-2005)
(in millions of USD)

Tajikistan

Kyrgyzstan

Uzbekistan

Moldova

Armenia

Turkmenistan

Belarus

Georgia

Azerbaijan

Ukraine

800
600
400
200
-

Kazakhstan

1 800
1 600
1 400
1 200
1 000

Russia = 4.683 mil. USD

29

Source: Data for Population: The World Bank, World Development Indicators. Other Data:
UNCTAD online database on FDI, authors calculations.
47

Figure4.
FDI inflow stock 2005
(in millions of USD)
30 000
25 000
20 000
15 000
10 000

Russia = 132.491 mil. USD

48

Tajikistan

Kyrgyzstan

Uzbekistan

Moldova

Armenia

Turkmenistan

Georgia

Belarus

Azerbaijan

Kazakhstan

Ukraine

5 000

Figure5.
Average Annual FDI inflow (1992-2005)
(in millions of USD)
140000
120000
100000
80000
60000
40000

Af
ric
a

So
ut
hEa
st
As
ia
C
en
tra
lE
ur
op
e

Am
er
ic
a

So
ut
h

U
ni
te
d

St
at
es

C
So
IS
ut
he
as
tE
ur
op
e

20000

NB: Southeast Europe: Albania, Bosnia and Herzegovina, Bulgaria, Croatia, Macedonia,
TFYR, Romania, Serbia and Montenegro.

49

50
Uzbekistan

Tajikistan

Kyrgyzstan

Belarus

Turkmenistan

Moldova

Ukraine

Armenia

Russia

Georgia

Kazakhstan

Azerbaijan

Figure6.
Average Annual FDI inflow per capita
(1992-2005)

160

140

120

100

80

60

40

20

APPENDIX B
Derivation of a partial stock adjustment model done by Cheng and Kwan (2000)30:

Following Chow (1967), it is assumed that the flow of investment serves to adjust

Yit towards Yit* according to the following process:

d ln Yit / dt = (ln Yit* ln Yit ), 0 < < 1

(1)

In empirical applications, (1) is replaced by its discrete version (where lower case letters
stand for logarithmic values, e.g., yit = ln Yit ),

yit yit 1 = ( yit* yit 1)

(3)

which, after collecting terms, becomes

yit = (1 ) yit 1 + yit*

(4)

For the adjustment process described by Eq. (4) to be stable (i.e. non-explosive) and nonfluctuating,

(1 ) must be a positive fraction

Note: Cheng and Kwan replace the right hand side term

ln Yit of equation (1) by yit 1 in

equation (3). But ln Yit = yit .


Derivation of a partial stock adjustment model done in this paper:

Let Yit be the actual stock of FDI in country i at time t and Yit* the corresponding
equilibrium or desired stock. The flow of investment will adjust Yit towards Yit* according to
the following process:
30

L.K. Cheng, Y.K. Kwan / Journal of International Economics 51 (2000) 379 400, see pages 382
and 383

51

d ln Yit / dt = (ln Yit* ln Yit ), 0 < < 1

(1)

For empirical work equation (1) can be rewritten as follows ( yit = ln Yit ):

yit yit 1 = ( yit* yit ),


After regrouping we get:
1

) yit 1 + (
) y*
1+
1 + it

yit = (

For the system of adjustment to be stable (non-explosive) the term


than zero and smaller than 1

52

(3)

1
) must be bigger
1+

APPENDIX C

Table1. Definition of Variables


Ln_FDI:

Natural log of the Cumulative FDI stock (million US$)


(Source: UNCTAD, World Investment reports)

Lag_Ln_FDI:

lagged value of Ln_FDI

Ln_GDP:

Natural log of GDP (current US$) (Source: The World Bank,


World Development Indicators)

Ln_Inflation:

Natural log of inflation (Source: The World Bank, World


Development Indicators)

Corruption:

Control of corruption. (Percentile rank). (Source: World Bank


Institute Governance & Anti-Corruption.). The higher the
rank, the lesser the corruption level.

Risk:

Market perceptions of country risk rated by the Euromoney


magazine. The Euromoney ratings are computed on the basis
of assessments of country risk experts using the following
nine weighted categories: economic performance (25
percent), political risk (25 percent), external debt indicators
(10 percent), debt in default or rescheduled (10 percent),
credit ratings (10 percent), access to bank finance (5 percent),
access to short-term finance (5 percent), access to capital
markets (5 percent), and discount on forfeiting (5 percent).
The higher the rank, less risky the country. (Source:
Euromoney magazine)

53

External_Liberalization:

EBRD index of forex and trade liberalization which measures


the openness. (Source: EBRD, Transition reports)

Natural_Resource:

Natural resource endowment: =0 if poor, =1 if moderate, and


=2 if rich (Source: De Melo and others (1997))

Fiscal_Balance:

fiscal balance (in percent of GDP) (Source: EBRD,


Transition reports.)

Ln_Population:

Natural log of population (Source: The World Bank, World


Development Indicators)

Small_Privatization:

EBRD index of small-scale privatization

Large_Privatization:

EBRD index of large-scale privatization

BankReform_IntRate:

EBRD index of banking sector reform

SecurityMarket_NonBank: EBRD index of reform of non-banking financial institutions


Ln_TelLines:

Natural log of Number of main Telephone Lines per 1000


people. (Source: The World Bank, World Development
Indicators)

Structural_Change:

time dummy used for detecting the structural change, for


1992-1998 takes the value of 0, for 1999-2005 takes the value
of 1.

54

APPENDIX D.
Difference GMM Model31

The GMM approach starts with the first-differenced version of the model.
yit

1
)yit 1 + X it + it
1+

=(

(6)

in which the country-specific effects are eliminated by the differencing operation. Under
the assumption of serially uncorrelated level residuals, values of y lagged two periods or more
qualify as instruments in the first-differenced system, implying the following moment
conditions:
E ( yit s it )=0

t=3,..., T and s 2

(7)

But GMM estimation based on (7) alone can be highly inefficient. In most cases, it is
necessary to make use of the explanatory variables as additional instruments.
Here the issue of endogeneity due to reverse causality becomes critical. For strictly
exogenous explanatory variables both past and future X it are valid instruments:
E (X it s it )=0

t=3,..., T and all s

(8)

But using (8) for s<2 will lead to inconsistent estimates if reverse causality exists in the
sense that E ( X ir it ) 0, for r t . To allow for this possibility, one may assume X to be
weakly exogenous, i.e. E ( X is it )=0, for s<t , which implies the following subset of (8):
E (X it s it )=0

t=3,..., T and s 2

(9)

Equations (6)(9) imply a set of linear moment conditions to which the standard GMM
methodology applies. The consistency of the GMM estimator hinges on the validity of these
moment conditions, which in turn depends on maintained hypotheses on the level residuals

31

See Cheng and Kwan (2000)

55

being serially uncorrelated and the exogeneity of the explanatory variables. It is therefore
essential to ensure that these assumptions are justified by conducting specification tests.
System GMM Model

Following Blundell and Bond (1998), one can increase the first-differenced moment
conditions (7)(9) by the level moment conditions
E ( it yit 1 )=0

t=3,..., T

(10)

which amounts to using lagged differences of y as instruments in the level Equation (5). In
addition, for strictly exogenous explanatory variables, there are level moment conditions
E ( it X it s )=0

t=2,..., T and all s

(11)

For weakly exogenous explanatory variables, the appropriate level moment conditions would
be
E ( it X it s )=0

t=3,..., T and s 1

(12)

The BlundellBond system GMM estimator is obtained by imposing the enlarged set of
moment conditions (7)(12). By exploiting more moment conditions, the system GMM
estimator is more efficient than the first-differenced GMM estimator that uses only a subset
(7)(9). The validity of the level moment conditions (10)(12) depends on a standard random
effects specification of the level equation in (5), plus additional assumptions on the initial
value generating process and the absence of correlation between country-specific effects and
the explanatory variables in first-differences. (See Blundell and Bond (1998) for details)

56

APPENDIX E
Table2 : Random Effect Panel Data Model: Ln_FDI is dependent
variable
(1)
(2)
(3)
(4)
Ln_FDI
Ln_FDI
Ln_FDI Ln_FDI
0.898
0.757
0.764
0.727
Ln_GDP
(5.67)*** (7.24)***
(7.64)*** (8.21)***
-0.509
-0.464
-0.364
Ln_Inflation
(15.56)*** (9.47)*** (7.02)***
0.149
0.131
External_Liberalization
(1.23)
(1.16)
0.053
Fiscal_Balance
(4.45)***
Natural_Resource

(5)
Ln_FDI
0.565
(5.73)***
-0.352
(6.98)***
0.198
(1.78)*
0.049
(4.23)***
0.679
(3.26)***

(6)
Ln_FDI
0.518
(4.64)***
-0.133
(2.32)**
0.36
(3.16)***
0.168
(9.84)***
0.67
(2.67)***
0.018
(2.48)**

(7)
Ln_FDI
0.466
(4.44)***
-0.141
(2.47)**
0.26
(2.16)**
0.153
(8.25)***
0.624
(2.81)***
0.016
(2.24)**
0.017
(1.87)*

(8)
Ln_FDI
0.497
(8.30)***
-0.165
(2.50)**
0.281
(2.15)**
0.128
(6.01)***
0.529
(5.02)***
0.01
(1.24)
0.032
(3.22)***
-0.119
(0.68)

(9)
Ln_FDI
0.473
(8.23)***
-0.129
(2.02)**
0.265
(2.12)**
0.128
(6.27)***
0.552
(5.48)***
0.015
(1.96)**
0.029
(3.04)***
0.426
(1.85)*
-0.517
(3.44)***

(10)
Ln_FDI
0.471
(8.21)***
-0.113
(1.74)*
0.192
(1.39)
0.123
(5.98)***
0.525
(5.10)***
0.013
(1.75)*
0.028
(2.89)***
0.408
(1.78)*
-0.55
(3.61)***
0.27
(1.22)

(11)
Ln_FDI
0.377
(5.54)***
-0.119
(1.88)*
0.22
(1.63)
0.132
(6.45)***
0.663
(5.75)***
0.011
(1.51)
0.018
(1.72)*
0.468
(2.07)**
-0.586
(3.92)***
0.076
(0.33)
0.503
(2.46)**

7.791
(3.55)***
158
12
0.82

7.897
(3.25)***
117
12
0.85

8.909
(3.92)***
117
12
0.86

8.272
(7.29)***
117
12
0.86

8.252
(7.63)***
117
12
0.87

8.181
(7.57)***
117
12
0.88

9.892
(7.82)***
117
12
0.88

Corruption
Risk
Small_Privatization
Large_Privatization
BankReform_IntRate
SecurityMarket_NonBank
Ln_TelLines
Constant
Observations
Number of country
R2

-0.13
-0.04
161
12
0.4

4.914
(2.04)**
158
12
0.74

4.184
(1.77)*
158
12
0.74

4.97
(2.37)**
158
12
0.77

Note: Absolute value of z-statistics in parentheses


* significant at 10%; ** significant at 5%; *** significant at 1%

(12)
Ln_FDI
0.226
(2.35)**
-0.128
(1.87)*
0.136
(0.90)
0.122
(5.42)***
0.758
(6.01)***
0
(0.05)
0.023
(2.09)**
0.582
(2.21)**
-0.484
(3.02)***
-0.241
(0.89)
0.696
(3.11)***
0.429
(1.93)*
11.092
(8.26)***
107
12
0.89

Table 3: Random Effect Panel Data Model. Ln_FDI is dependent variable and Corruption is excluded from explanatory variables

Ln_GDP

(1)
Ln_FDI
0.898
(5.67)***

(2)
Ln_FDI
0.757
(7.24)***
-0.509
(15.56)***

(3)
Ln_FDI
0.764
(7.64)***
-0.464
(9.47)***
0.149
(1.23)

(4)
Ln_FDI
0.727
(8.21)***
-0.364
(7.02)***
0.131
(1.16)
0.053
(4.45)***

(5)
Ln_FDI
0.565
(5.73)***
-0.352
(6.98)***
0.198
(1.78)*
0.049
(4.23)***
0.679
(3.26)***

(6)
Ln_FDI
0.459
(5.01)***
-0.314
(6.42)***
0.114
(1.08)
0.04
(3.47)***
0.63
(3.43)***
0.043
(4.06)***

(7)
Ln_FDI
0.446
(4.96)***
-0.281
(5.29)***
0.02
(0.17)
0.038
(3.39)***
0.634
(3.54)***
0.039
(3.72)***
0.256
(1.58)

(8)
Ln_FDI
0.499
(7.36)***
-0.277
(5.44)***
0.226
(1.90)*
0.047
(4.18)***
0.604
(4.66)***
0.035
(3.28)***
0.496
(2.75)***
-0.596
(3.42)***

(9)
Ln_FDI
0.496
(6.90)***
-0.281
(5.42)***
0.236
(1.81)*
0.048
(4.21)***
0.611
(4.41)***
0.035
(3.24)***
0.513
(2.82)***
-0.572
(3.17)***
-0.118
(0.52)

(10)
Ln_FDI
0.416
(4.73)***
-0.281
(5.42)***
0.215
(1.64)
0.047
(4.11)***
0.73
(4.42)***
0.027
(2.38)**
0.531
(2.92)***
-0.589
(3.20)***
-0.216
(0.93)
0.434
(1.90)*

-0.13
(0.04)
161
12
0.4

4.914
(2.04)**
158
12
0.74

4.184
(1.77)*
158
12
0.74

4.97
(2.37)**
158
12
0.77

7.791
(3.55)***
158
12
0.82

9.105
(4.60)***
158
12
0.84

8.834
(4.55)***
158
12
0.84

7.904
(5.53)***
158
12
0.85

8.057
(5.28)***
158
12
0.85

9.491
(5.24)***
158
12
0.86

Ln_Inflation
External_Liberalization
Fiscal_Balance
Natural_Resource
Risk
Small_Privatization
Large_Privatization
BankReform_IntRate
SecurityMarket_NonBank
Ln_TelLines
Constant
Observations
Number of country
R2

Note: Absolute value of z-statistics in parentheses


* significant at 10%; ** significant at 5%; *** significant at 1%

58

(11)
Ln_FDI
0.265
(3.31)***
-0.295
(5.52)***
0.261
(2.04)**
0.037
(3.30)***
0.922
(7.24)***
0.033
(2.79)***
0.489
(2.72)***
-0.498
(2.89)***
-0.552
(1.78)*
0.508
(2.26)**
0.58
(3.17)***
10.072
(7.90)***
148
12
0.87

Table 4: Random Effects and G2SLS Panel Data Models


Ln_FDI is dependent variable
(1)
(2)
(3)
Ln_FDI Ln_FDI Ln_FDI
0.226
0.226
0.321
Ln_GDP
(2.35)** (2.38)** (3.21)***
-0.128
-0.128
-0.084
Ln_Inflation
(1.87)*
(1.83)*
(1.24)
0.136
0.136
0.251
External_Liberalization
(0.90)
(0.95)
(1.64)
0.122
0.122
0.097
Fiscal_Balance
(5.42)*** (5.51)*** (4.08)***
0.758
0.758
0.795
Natural_Resource
(6.01)*** (6.13)*** (6.46)***
0.006
0.006
0.004
Corruption
(0.05)
(0.05)
(0.03)
0.023
0.023
0.01
Risk
(2.09)** (2.09)** (0.90)
0.582
0.582
0.283
Small_Privatization
(2.21)** (2.21)** (1.01)
-0.484
-0.484
-0.441
Large_Privatization
(3.02)*** (3.12)*** (2.82)***
-0.241
-0.241
-0.041
BankReform_IntRate
(0.89)
(0.91)
(0.15)
0.696
0.696
0.634
SecurityMarket_NonBank
(3.11)*** (3.12)*** (2.90)***
0.429
0.429
0.292
Ln_TelLines
(1.93)*
(1.91)*
(1.32)
0.478
Structural_Change
(2.61)***
11.092
11.092
9.672
Constant
(8.26)*** (8.26)*** (6.85)***
107
107
107
Observations
12
12
12
Number of country
0.8919
0.8919
0.8993
R2
Note: Absolute value of z-statistics in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%
For both G2SLS models Ln_GDP is instrumented
and Ln_Population is an instrument.
Model (3) & (4) check for structural change in the model

(4)
Ln_FDI
0.321
(3.18)***
-0.084
(1.24)
0.251
(1.64)
0.097
(4.09)***
0.795
(6.52)***
0.004
(0.03)
0.01
(0.90)
0.283
(1.02)
-0.441
(2.93)***
-0.041
(0.15)
0.634
(2.94)***
0.292
(1.32)
0.478
(2.82)***
9.672
(6.85)***
107
12
0.8993

Table 5: Random Effects and G2SLS Panel Data Models


Ln_FDI is dependent variable. Corruption is excluded from explanatory variables

Ln_GDP
Ln_Inflation
External_Liberalization
Fiscal_Balance
Natural_Resource
Risk
Small_Privatization
Large_Privatization
BankReform_IntRate
SecurityMarket_NonBank
Ln_TelLines

(1)
Ln_FDI
0.265
(3.31)***
-0.295
(5.52)***
0.261
(2.04)**
0.037
(3.30)***
0.922
(7.24)***
0.033
(2.79)***
0.489
(2.72)***
-0.498
(2.89)***
-0.552
(0.98)
0.508
(2.26)**
0.58
(3.17)***

(2)
Ln_FDI
0.265
(3.43)***
-0.295
(5.57)***
0.261
(2.14)**
0.037
(3.38)***
0.922
(7.29)***
0.033
(2.88)***
0.489
(2.75)***
-0.498
(2.91)***
-0.552
(0.98)
0.508
(2.26)**
0.58
(3.19)***

10.072
(7.90)***
148
12
0.8712

10.072
(7.90)***
148
12
0.8712

Structural_Change
Constant
Observations
Number of country
R2

(3)
Ln_FDI
0.298
(3.93)***
-0.204
(3.75)***
0.333
(2.75)***
0.031
(2.87)***
0.999
(8.26)***
0.011
(0.90)
0.375
(2.20)**
-0.477
(2.95)***
-0.382
(1.65)*
0.563
(2.66)***
0.484
(2.79)***
0.791
(4.38)***
9.281
(7.67)***
148
12
0.8872

Note: Absolute value of z-statistics in parentheses


* significant at 10%; ** significant at 5%; *** significant at 1%
For both G2SLS models Ln_GDP is instrumented
and Ln_Population is an instrument.
Model (3) & (4) check for structural change in the model

60

(4)
Ln_FDI
0.298
(3.85)***
-0.204
(3.77)***
0.333
(2.75)***
0.031
(2.83)***
0.999
(8.27)***
0.011
(0.90)
0.375
(2.20)**
-0.477
(2.99)***
-0.382
(1.65)*
0.563
(2.68)***
0.484
(2.79)***
0.791
(4.31)***
9.281
(7.67)***
148
12
0.8872

Table 6: Dynamic Panel Data Model (System GMM and Difference GMM) with different sets of instruments
Ln_FDI is dependent variable

Lag_Ln_FDI
Ln_GDP
Ln_Inflation
Risk
Fiscal_Balance
External_Liberalization
Small_Privatization
Large_Privatization
SecurityMarket_NonBank
Ln_TelLines
Natural_Resource
Constant

(1)
Diff.
GMM
Ln_FDI
0.572
(13.71)***
-0.095
(0.56)
-0.149
(2.84)**
0.011
(0.93)
0.019
(2.85)**
-0.122
-1.07
-0.086
(0.49)
-0.146
(1.18)
0.034
(0.13)
1.772
(4.41)***
dropped
dropped

Number of country
Observations
# of instruments
Hansen Test
ar (1)
ar (2)
Note:

12
122
91
1
0.205
0.73

(2)
Sys.
GMM
Ln_FDI
0.594
(9.01)***
0.231
(2.24)**
-0.103
(2.84)**
0.013
(0.83)
0.029
(2.79)**
0.034
-0.46
0.214
(1.65)
-0.221
(2.30)**
-0.328
(1.00)
0.249
(2.34)**
0.082
(0.39)
2.42
(2.18)*
12
136
118
1
0.21
0.73

(3)
Diff.
GMM
Ln_FDI
0.573
(13.68)***
-0.09
(0.51)
-0.15
(2.84)**
0.01
(0.89)
0.019
(2.93)**
-0.168
-1.26
-0.069
(0.41)
-0.122
(0.92)
0.063
(0.26)
1.804
(4.23)***
dropped
dropped
12
122
90
1
0.199
0.796

(4)
Sys.
GMM
Ln_FDI
0.59
(8.56)***
0.179
(2.26)**
-0.123
(2.07)*
0.013
(0.83)
0.029
(3.07)**
-0.05
-0.28
0.191
(1.45)
-0.208
(1.98)*
-0.252
(1.14)
0.324
(2.26)**
0.162
(1.35)
3.471
(2.64)**
12
136
117
1
0.189
0.711

(5)
Diff.
GMM
Ln_FDI
0.569
(16.75)***
-0.094
(0.54)
-0.149
(2.77)**
0.01
(0.87)
0.019
(2.94)**
-0.171
-1.32
-0.064
(0.36)
-0.12
(0.91)
0.072
(0.30)
1.855
(4.97)***
dropped
dropped
12
122
89
1
0.197
0.794

(6)
Sys.
GMM
Ln_FDI
0.571
(8.09)***
0.187
(2.16)*
-0.133
(2.12)*
0.012
(0.81)
0.026
(3.02)**
-0.046
-0.26
0.153
(1.01)
-0.202
(1.73)
-0.184
(0.87)
0.52
(2.24)**
0.16
(1.09)
2.747
(2.06)*
12
136
116
1
0.179
0.66

Robust t-statistics in parentheses


* significant at 10%; ** significant at 5%; *** significant at 1%

(7)
Diff.
GMM
Ln_FDI
0.504
(9.92)***
-0.351
(1.56)
-0.141
(2.59)**
0.015
(0.92)
0.012
(1.90)*
-0.188
-1.43
0.02
(0.10)
-0.059
(0.31)
-0.146
(0.74)
3.256
(5.32)***
dropped
dropped
12
122
71
1
0.195
0.752

(8)
Sys.
GMM
Ln_FDI
0.562
(6.64)***
0.23
(2.36)**
-0.136
(2.00)*
0.015
(0.86)
0.03
(3.41)***
-0.044
-0.24
0.232
(1.67)
-0.22
(1.90)*
-0.419
(1.73)
0.491
(2.28)**
0.062
(0.35)
2.282
(1.46)
12
136
98
1
0.17
0.89

(9)
Diff.
GMM
Ln_FDI
0.437
(6.93)***
-0.6
(2.27)**
-0.128
(2.36)**
0.019
(0.98)
0
(0.05)
-0.105
-0.69
0.26
(0.70)
-0.237
(0.84)
-0.296
(1.55)
4.439
(6.74)***
dropped
dropped
12
122
50
1
0.23
0.74

(10)
Sys.
GMM
Ln_FDI
0.55
(5.93)***
0.248
(2.36)**
-0.151
(2.11)*
0.015
(0.79)
0.031
(3.64)***
-0.069
-0.37
0.27
(1.42)
-0.26
(1.74)
-0.497
(1.91)*
0.542
(2.53)**
0.023
(0.12)
2.152
(1.18)
12
136
77
1
0.16
0.97

Table 7. Summary Statistics

Variable

Obs

Mean

Std.
Dev.

Min

Max

Ln_FDI

161

20.31

2.01

15.76

25.61

Ln_GDP

168

22.78

1.65

20.64

27.36

Ln_Inflation

165

3.81

2.21

-0.51

9.66

External_Liberalization

168

2.70

1.20

1.00

4.30

Fiscal_Balance

167

-5.16

7.37

-54.70

7.52

Natural_Resource

168

0.92

0.86

0.00

2.00

Risk

168

27.82

8.80

9.46

54.16

Corruption

120

17.28

10.17

2.00

54.40

Small_Privatization

168

2.97

0.96

1.00

4.00

Large_Privatization

168

2.28

0.88

1.00

3.67

BankReform_IntRate

168

1.76

0.59

1.00

3.00

SecurityMarket_NonBank

168

1.64

0.51

1.00

3.00

Ln_TelLines

158

4.77

0.53

3.55

5.82

Structural_Change

168

0.50

0.50

0.00

1.00

Table 8. Hausman Tests Result

---- Coefficients ---(b)


(B)
(b-B)
sqrt(diag(V_b-V_B))
FE
RE
Difference S.E.
---------------------------------------------------------------------------------------------------Ln_GDP
-.1606198 .0787609 -.2393808 .195646
Ln_Inflation

-.1231355

-.1216429

-.0014926

Risk

.0082186

.0219403

-.0137217

External_L~n

-.0243757

.0637747

-.0881504

Fiscal_Bal~e

.1114575

.1080227

.0034347 .

Corruption

.0108631

-.0027688

.0136319 .

Small_Priv~n

.4875965

Large_Priv~n

-.1090685

-.4097657

BankReform~e

-.6639906

-.2977887

-.3662018

.1337149

SecurityMa~k

.7095038

.8411936

-.1316898

Tel_Lines

.0194548

.005785

.6786225 -.1910261

.3006972 .1670199

.0136698 .0023304

---------------------------------------------------------------------------------------------------b = consistent under Ho and Ha; obtained from xtreg


B = inconsistent under Ha, efficient under Ho; obtained from xtreg
Test: Ho: difference in coefficients not systematic

chi2(11) = (b-B)'[(V_b-V_B)^(-1)](b-B)
=
Prob>chi2 =

97.24
0.0000

(V_b-V_B is not positive definite)

63

APPENDIX F32
Inward FDI Performance index - Methodology

It is the ratio of a countrys share in global FDI inflows to its share in global GDP. A value
greater than one indicates that the country receives more FDI than its relative economic size, a
value below one that it receives less (a negative value means that foreign investors disinvest in that
period).

IND i

FDI i / F w
GDP i / G w

Where,
INDi

The Inward FDI Performance Index of the ith country

FDIi

The FDI inflows in the ith country

FDIw

World FDI inflows

GDPi

GDP in the ith country

GDPw

World GDP

Inward FDI Potential index - Methodology

It is an average of the values (normalized to yield a score between zero, for the lowest scoring
country, to one, for the highest) of 12 variables (no weights are attached in the absence of a priori
reasons to select particular weights):

GDP per capita, an indicator of the sophistication and breadth of local demand (and

of several other factors), with the expectation that higher income economies attract
relatively more FDI geared to innovative and differentiated products and services.

The rate of GDP growth over the previous 10 years, a proxy for expected economic

growth.

32

The share of exports in GDP, to capture openness and competitiveness.

UNCTAD

64

As an indicator of modern information and communication infrastructure, the

average number of telephone lines per 1,000 inhabitants and mobile telephones per 1,000
inhabitants.

Commercial energy use per capita, for the availability of traditional infrastructure.

The share of R&D spending in GDP, to capture local technological capabilities.

The share of tertiary students in the population, indicating the availability of high-

level skills.

Country risk, a composite indicator capturing some macroeconomic and other

factors that affect the risk perception of investors. The variable is measured in such a way
that high values indicate less risk.

The world market share in exports of natural resources, to proxy for the availability

of resources for extractive FDI.

The world market share of imports of parts and components for automobiles and

electronic products, to capture participation in the leading TNC integrated production


systems (WIR02).

The world market share of exports of services, to seize the importance of FDI in the

services sector that accounts for some two thirds of world FDI.

The share of world FDI inward stock, a broad indicator of the attractiveness and

absorptive capacity for FDI, and the investment climate.

65

You might also like