Telechargé par Camilla Fiorina

Dterminants of FDI in developing countries

publicité
Determinants of Foreign Direct
Investment in Developing
Countries
A Focus on Sub Saharan Africa Countries and the Role of Public Governance
International Master in Public Finance – 2019/2020
Alessia DE SANTO, Camilla FIORINA
Table of contents
1. INTRODUCTION AND LITERATURE REVIEW
3
2. PUBLIC GOVERNANCE AND FDI
5
3. DATA AND METHODOLOGY
6
4. DESCRIPTIVE STATISTICS
9
5. MODEL SPECIFICATION
14
6. RESULTS
16
6.1 Sub Saharan countries
18
7. CONCLUSION
21
APPENDIX A
23
REFERENCES
25
2
1. Introduction and literature review
Foreign Direct Investments are a global phenomenon which has displayed remarkably constant
growth over the last decades. As reported by Buchanan et al. (2012) the worldwide flow of FDI has
grown by 19% between 1996 and 2006, which is more than double than the 8% growth in the
international trade of goods and services.
Developing countries have been increasing recipient of FDI. While multinational companies benefit
from lower production costs, developing countries are eager to attract FDI inflows, as they play an
important role in nourishing their growth. First, they represent a vital source of funds which creates
job opportunities and increases income. Moreover, productive activity from multinational
enterprises often brings technical and managerial expertise, boost R&D activity and facilitate the
spread of new technologies, contributing in the reduction of the technological gap between
developed and developing economies.
This has led most of developing country to undertake steps to attract high FDI inflows. However,
developing countries differ widely in their ability to attract FDI.
The favorite destination for FDI in the latest decades has been BRIC countries (Brazil, Russia, China
and India) which are very dynamic economies and have been displaying very high growth rates since
the 1990s. In particular, China has been the major FDI recipient in the last 30 years, followed by
India, which according to UNCTAD (2007) is the second most preferred FDI destination.
Countries from Sub Saharan Africa, which are traditionally at the bottom of investors’ preferences
according to the UNCTAD, have registered a more than six-fold increase in FDI inflow since 1996
(Buchanan et al., 2012). The region’s FDI to GDP ratio has improved significantly in the current
century: Azemar and Desbordes (2009) report that, between 1984 and 2004, the median FDI to GDP
ratio was less than 1% in the SSA region, compared to almost 2% in South East Asia and Latin
America. In contrast, we find that over the period 1999-2014, median FDI values represented
around 3,8% of GDP, overcoming the median value for Asian economies – of about 3,2%.
Despite this substantial growth, the SSA region still lags behind in terms of quantity of FDI received
compared to the rest of the world (Figure 1).
While many scholars have been investigating on the basic determinants of FDI, no consensus seems
to have been reached, meaning that there is not a universally accepted set of variables which can
be considered as the true determinants of FDI, especially for developing countries.
As reported by Kumari and Sharma (2017), the existing literature has considered as the
determinants of FDI volumes mainly the following set of variables: market size, trade openness,
infrastructure, return on investment, real labor cost, human capital, exchange rate, inflation,
political risk and government incentives. Most of the studies have found that market size, trade
openness and infrastructure play a positive role in attracting flows of FDI (Na and Lightfoot (2006);
Quazi (2007); Hoang and Goujon (2014)). Others, as Buchanan et al. (2012), find evidences in
support of the centrality of quality of institutions in attracting FDI, concluding that the traditional
policy recommendation of “offering the correct macroeconomic environment” would be ineffective
without adequate institutional reforms.
3
Figure 1 - net FDI inflow (1998-2014)
Source: World Bank Development Indicators. See table 1 in Appendix A for the complete list of countries included in the analysis.
Only a limited number of studies has investigated the sources of the high disparity between SSA
countries and the rest of the world.
Asiedu (2002, 2006) analyzed a panel of 34 SSA countries over the period 1980-200 and confirmed
the importance of market size, infrastructure, trade openness and institutional framework.
Azemar and Desbordes (2009) grouped all the previously mentioned variables under four indicators
of public governance; they perform a panel analysis over a sample of 70 developing countries and
conclude that reasons for the gap in FDI inflows between SSA countries and other regions can be
mostly find in market size and low human capital accumulation. To be able to draw this conclusion,
they first estimate a fixed effect panel model over the entire sample and include a second stage of
the analysis where they decompose the sources of the FDI gap between SSA region and the rest of
developing countries.
Following a similar methodology to that of Azemar and Desbordes (2009), this paper provides an
analysis of the main determinants of FDI based on a more recent dataset, covering the period 19982014 for 66 developing countries. While the study is performed over a sample including developing
countries from all regions, the focus will be, as in Azemar and Desbordes, on investigating the main
factors that drive the disparity of SSA countries in inbound FDI volumes. Because of the important
improvement in FDI volumes directed toward SSA countries, we believe that it would be meaningful
to perform an analysis of FDI determinants over a more recent period of time, to investigate for
possible new dynamics compared to those highlighted in Azemar and Desbordes and the previous
literature.
The remainder of this paper will be composed of the following sections. Section 2 details the four
dimensions of public governance and how they can impact the level of FDI inflow into developing
countries. Section 3 presents the database on which the analysis is performed. Section 4 includes a
graphical representation of the data and the related descriptive statistics. Section 5 define the
econometric model and section 6 report the estimation’s results. Section 7 concludes.
4
2. Public governance and FDI
The OECD defines public governance as “the formal and informal arrangements that determine how
public decisions are made and how public actions are carried out, from the perspective of
maintaining a country’s constitutional values when facing changing problems and environments”.
Public governance plays and important role in influencing the investment decisions since it is
fundamental in designing the necessary rules that will make a country institution trustworthy.
Azemar and Desbordes (2009) distinguish four dimensions of public governance at a
macroeconomic level: the degree of democracy, provision of public good, macroeconomic policies
and security of property right. The following section is dedicated to detail these different dimensions
and discuss their expected effect on FDI level.
The first dimension which is taken into account is the degree of democracy, a form of government
where, thanks to free and competitive elections and the existence of check and balances
mechanisms, governors are elected by citizen and directly accountable to them. Level of democracy
is expected to have a positive effect in attracting foreign capital.
First, democratic states offer more both political and economic stability – in that, for example, they
allow citizens to express dissatisfaction by denying votes to politicians instead than through the use
of violence. Moreover, from an ethical point of view, multinational corporations prefer to invest in
democratic countries, in order not to compromise their image.
These hypotheses are empirically supported. A number of studies (e.g. Busse (2004) and Jakobsen
& De Soysa (2006)) demonstrate that, since the beginning of 1980s, countries which ensure higher
levels of political and civil freedom to their citizens attract higher FDI flow.
Provision of public good is a central dimension as it includes public supply of infrastructure, which
previous literature proved to be an important determinant of FDI: better infrastructures provide
more investment opportunities for companies and facilitate their operations, making them more
eager to invest in a country which offers a dense network of infrastructures and good telephone or
internet coverage.
Moreover, governments are the main provider of health and education, thus determining the level
of human capital of a country. As previously mentioned, higher average human capital often
corresponds to larger FDI inflows. Indeed, as higher human capital improves productivity, it lowers
unit labor costs, therefore decreasing productions costs for multinationals.
Section 1 mentioned some econometric studies which report evidences of a positive and significant
effect of infrastructure and human capital in determining FDI volume into developing countries.
The ability of governments to maintain a favorable and stable macroeconomic environment - mainly
by the use of efficient and sound macroeconomic policies - can play a major role in attracting the
attention of foreign investors. Indeed, volatile inflation and exchange rates or high debt to GDP
ratios increase uncertainty over the feature and consequently investment cost, making companies
desist from investing. Degree of market openness can as well have an impact on inflow of foreign
funds, as it facilitates exports and imports for companies that delocalize their production process in
different countries – as it is often the case for multinationals.
A number of studies have proven empirically that multinational companies favor stable countries to
direct their FDI. Among others, Cevis and Camurdan, (2007) find a negative and significant impact
of inflation on inbound FDI in developing countries. Asiedu (2002) shows that a positive and
significant relation exists between FDI and degree of market openness.
5
The last dimension identified by Azemar and Desbordes (2009) is the capacity of ensuring security
of property rights and contracts, included in the long term, which is fundamental in that it
guarantees to companies the appropriation of gains on their investments. The main tool through
which government can ensure property rights and enforce contracts is the rule of law and its
“monopoly on legitimate violence”. However, some external events as wars can compromise this
ability, thus being negatively associated with the level of FDI.
Some empirical studies, as Globerman and Shapiro (2002), provide evidences that FDI inflows are
higher in countries with a strong rule of law.
3. Data and methodology
The analysis is performed over 66 developing countries pertaining to the period 1998-2014. The
database includes 21 Sub Saharan Africa (SSA) countries, 12 Asian and Pacific countries, 18 Latin
American and Caribbean countries (LAC) and 15 other countries1. The complete list of countries can
be found in appendix A (table A1).
The level of Foreign Direct Investment in each country is proxied by the net FDI inflow obtained from
the World Bank Development Indicators2. It is composed by the sum of equity capital, reinvestment
of earnings and other capital injected into the domestic economy by foreign investors.
Data are converted from current USD to constant USD (base=2010) by multiplying the former by the
ratio of constant 2010 USD GDP to current USD GDP. In this way, the evolution of net FDI inflow
only reflects changes in the level of FDI and not in the price level.
In light of the previously discussed literature, different variables have been considered as possible
determinants of net FDI inflow in developing countries.
A negative FDI dummy, which takes the value of 1 if the FDI net flow is negative in a certain year, is
included to take into account the potential negative signal this may send to investors.
Per capita and total GDP (in constant 2010 USD dollars) are included to account for the market size
of the economy. The sign of their effects on FDI flow is expected to be positive, since a greater
market size offers more investment and operating opportunities. Data are taken from the World
Bank Development Indicators.
As mentioned in the first section of this work, this work considers four dimensions of Public
Governance as determinants of FDI inflow: degree of democracy, public good provision,
macroeconomic stability and security of property rights.
The first dimension, degree of democracy, is proxied by 3 variables:
- The degree of Civil Liberties (extracted from the Freedom House indicator) which consists on
scores given to countries on the basis of Freedom of Expression and Belief, Associational and
Organizational Rights, Rule of Law and Personal Autonomy and Individual Rights. Countries
Countries are grouped by geographical region according to the World Bank classification. Asian countries include country from
South Asia and East Asia and Pacific regions. Group “other” includes countries from Europe and Central Asia, Middle East and North
Africa.
2 https://databank.worldbank.org/source/world-development-indicators
1
6
-
-
obtaining a score of 1 or 2 are considered free; 3,4 or 5 partly free; from 6 to 7 not free. Data
are available on the Freedom House website3.
The POLITY2 index, which defines on a scale from -10 (hereditary monarchy) to 10 (complete
democracy) the concomitant democracy and autocracy qualities of national governing
institutions. Data are obtained from the Polity™ IV project database4, developed by the
Center for Systematic Peace
The POLCONIII index measures the degree of political constraint by estimating the feasibility
of policy change. The variable is constructed by Henisz (2002) by taking into account the
number of independent branches of government with veto power over policy change, and
the extent of the political alignment of the members of the executive and legislative
branches. An updated version of the index is available on the Wharton University of
Pennsylvania website5.
In order to avoid possible multicollinearity between these 3 variables, and to account for possible
nonlinear effects of democracy on level of FDI, 6 dummies variables are created - following Azemar
and Desbordes (2009).
A medium civil liberty dummy takes value 1 when Civil Liberties score is included between 3 and 5,
while when it is equal to 1 or 2, a high civil liberty dummy takes value 1.
The same method applies for POLITY2 and POLCONIII indexes. They are rescaled from 1 to 7 as the
degree of Civil Liberties, so that when their values are greater than 3 and lower than 66, a medium
electoral democracy dummy and a medium political constraint dummy – respectively – take value
of 1. When their values are included between 6 and 7, a high electoral democracy dummy and a
high political constraint dummy, respectively, take value of 1.
To account for the role of public good provision in attracting FDI, proxies for health, education and
level of infrastructure are included in the database.
Data on life expectancy at birth – as a proxy for health – are obtained from the World Bank
Indicators, where it is measured as the number of years a newborn infant would live if prevailing
patterns of mortality at the time of its birth were to stay the same throughout its life.
Level of education is proxied by average years of schooling for people aged 25 or older, taken from
the Human Development Reports of the United Nations Development Programme, which is able to
provide a wide range of observations, gathering values from Barro and Lee (2018), the UNESCO
Institute for Statistics (2019) and ICF Macro Demographic and Health Surveys, UNICEF Multiple
Indicator Cluster Surveys and OECD (2018).
Finally, because of unavailability of data on most measures of infrastructure for SSA countries – as
per capita electrical consumption or per capita stock of fixed telephone mainlines – the number of
fixed telephone subscriptions is used as a proxy for the level of infrastructure. Data are available on
the World Bank Development Indicators.
Nevertheless, this is an imperfect measure of the level of infrastructure for the last years of the
sample, since the spread of mobile phones as the main communication tool has significantly
reduced their use, including in developing countries. However, as displayed in the World Bank, the
average number of fixed telephone users worldwide has been growing until 2010, meaning that an
hypothetical problem exists only for the last 4 years of study; moreover, it could be suggested that
https://freedomhouse.org/report/freedom-world/2020/leaderless-struggle-democracy (Country and Territory Ratings and
Statuses, 1973-2020)
4 https://www.systemicpeace.org/inscrdata.html
5 https://mgmt.wharton.upenn.edu/faculty/heniszpolcon/polcondataset/ (2017 Data Release)
6 the POLITY2 and the POLCONIII indexes are continuous, while the Civil Liberty score can only take integer values.
3
7
in developing countries the spread of mobile phones to the detriment of landline phones has
occurred somewhat later than the world average, further reducing the margin of error.
Macroeconomic stability is measured by the level of Inflation, Foreign Debt Position and Openness
of the economy.
As Azemar (2009) reports, stable and low inflation can be considered as a measure of the capacity
of governments to manage the economy, thus creating a stable and safe economic environment,
which attracts investments. Data on inflation, measured as the annual percentage change in average
Consumer Prices Index, come from the International Monetary Fund7.
The same holds for Foreign Debt position, thus a low debt/GNI ratio is expected to have a positive
impact on FDI inflow. Data on external debt stocks as a % of GNI are available on the World Bank
Development Indicators.
Trade openness, intended as the importance that nondomestic transaction cover in a country’s
economy, is expected to provide more investment opportunities to MNEs, thus to have a positive
impact on FDI inflow. As in previous studies (Kumary and Sharma, 2017) the standard trade
"#$
openness ratio ! %&' ( is used as a proxy for the degree of economy’s openness.
In conclusion, security of property rights is assumed to depend on the rule of law of countries and
the probability of conflict.
The strengthen of the rule of law is measured by the ICRG law and order index, which measures on
a scale from 1 (low) to 6 (high) the strength and impartiality of the legal system - the “Law”
component – and the popular observance of the law – the “Order” component (The PRS group).
Data on ICRG law and order index are available on Wharton University of Pennsylvania website8.
Moreover, the probability of occurrence of a conflict can jeopardize security of property rights. To
account for this eventuality, two dummies are created, according to the UCDP/PRIO9 definition of
conflict intensity. The dummy minor conflict takes value 1 if a conflict involving between 25 and 999
battle-related deaths has occurred in a given year, while the dummy war takes value of 1 if at least
1,000 battle-related deaths are recorded in a given year. Data are available on the UCDP Dataset
Download Center10.
When possible, data are taken in the natural logarithmic form. This allows for an easier
interpretation of the coefficients and for a linearization of the observations, thus reducing the
impact of extreme values. In addition, the explanatory variables are lagged by one year with respect
to the dependent variable, net FDI inflow11. This solution, previously used by Azemar and Desbordes
(2009), allows to reduce possible endogeneity bias and makes the model more realistic in that it
takes into account the fact that companies decide where to direct their investments based on past
information.
The database includes some missing values (see Figure 2). However, since they represent a very
small percentage of total observation and are randomly distributed in the sample, they are labelled
as “NA” and kept in the database, which will therefore result in an unbalanced one.
https://www.imf.org/external/datamapper/PCPIPCH@WEO/OEMDC/ADVEC/WEOWORLD
https://mgmt.wharton.upenn.edu/faculty/heniszpolcon/polcondataset/ (2017 Data Release)
9 Uppsala Conflict Data Program (Uppsala University) and International Peace Research Institute (Oslo)
10 https://ucdp.uu.se/downloads/index.html#armedconflict
11 This is performed directly in the database used for the econometric estimation (see Appendix B2). Observations of the dependent
variable are taken from 1999 to 2014, while observations of the explanatory variable cover the period 1998-2013.
7
8
8
Missing values in the net FDI inflows observations are due to the logarithm transformation12: the
negative values are excluded in order to obtain the logarithm of net FDI inflows. The presence of
negative FDI inflow is taken into account anyway by the negative FDI dummy.
Figure 2 – Distribution of the missing values in the database
4. Descriptive statistics
Before proceeding with the econometric analysis of the relationship between the determinants
described above and the level of net FDI inflow, a descriptive overview of the data will be given in
this section to provide a more complete illustration of the database.
Since the focus of the analysis is to investigate on the FDI inflows weaknesses in Sub Saharan
Countries with respect to the rest of developing countries, the descriptive statistics and the graphic
analysis will be mainly performed over the groups of countries that have been specified at the
beginning of section 3.
We begin with a graphic visualization of variables which allow us to understand both their evolution
overtime and their spread.
The use of a boxplot helps to visualize the distribution of the dependent variable and to check for
the presence of outliers. Figure 3 displays boxplots for the dispersion of net FDI inflows in each of
the four regions. Observations seem to be symmetrically distributed around the mean, although
LAC countries present a slight left skewed distribution while SSA countries are slightly skewed to the
right. The region with the highest number of outliers is Sub Saharan Africa, where 4 countries display
much lower FDI volumes than the others – which is coherent with the skewedness of the
12
Values of net FDI inflow for Myanmar (1998,1999) are NA because of missing values in the deflator for converting
current prices in constant prices.
9
distribution. Dispersion of data is quite homogeneous in the four groups, with countries from the
AP and LAC region displaying a slightly higher variability.
Figure 3 - Boxplots, net FDI inflow by region
Panel A of figure 4 plots the evolution overtime of the FDI inflows in all countries included in the
sample. The logarithm form allows to smooth extreme observations and highlights an average
growth in FDI level in all regions.
Moreover, the regional grouping allows to notice that FDI inflows into SSA countries are much lower
with respect to the other regions of the world, which, on average, present more similar level of FDI.
Panel B takes a closer look to average trends for each region. The growth trend was more stable in
the first half of the period considered, while it slowed down starting from 2007, in conjunction with
the global financial crises.
Even if disparity between SSA countries and the other regions remains significant, it is possible to
notice a reduction of this gap during the whole study period, in particular with respect to LAC
countries. Sub Saharan countries experienced a quite important growth in the inflow of FDI in the
sixteen years considered; moreover, while other regions reported sharp declines in FDI levels
between 2007 and 2010, the amount of FDI injected into this region continued to grow, which
further contributed to cover the distance.
Table 1 reports the regional averages of all variables over the period 1998-2014. It allows to get a
general idea of the existing equilibriums concerning the determinants of FDI under consideration.
The first thing which can be noticed is that SSA countries rank last in the majority of fields. In
particular, an important disparity with respect to the other developing countries, can be found in
the size of markets, as proxied by GDP and GDP pc. Moreover, SSA countries record the worst
performance also in the dimensions related with the provision of public goods, namely life
expectancy, average years of schooling and fixed telephone subscribers.
The latter, in particular, which proxied the level of infrastructure, is much lower than the average of
the other regions.
10
Figure 4 - Net FDI inflows
Panel A –net FDI inflows trend in each country, 1998-2014
Panel B – Average net FDI inflow trend by region, 1998-2014
11
Table 1 - Mean of all variables, by region, 1998-2014
FDI (net inflow)
Presence of negative FDI
GDP
GDP pc
Political Constraints
Civil Rights
Political Rights
Life Expectancy
Av. Years of Schooling
Fixed Telephone Subscribers
Inflation
Foreign Debt
Market Openness
Law and order index
Presence of Minor Conflict
Presence of War
SSA
AP
LAC
Other
19.72
0.03
23.49
6.95
0.25
3.97
2.33
4.01
1.39
11.8
8.12
3.68
-0.55
1.04
0.17
0.05
21.67
0.04
25.58
7.37
0.29
4.13
2.67
4.22
1.79
14.93
7.75
3.58
-0.45
1.13
0.39
0.07
21.03
0.03
24.62
8.28
0.36
3.1
7.17
4.27
1.91
14.04
8.71
3.75
-0.54
0.84
0.06
0.02
21.5
0
24.79
8.27
0.26
4.12
1.61
4.26
2.17
14.73
11.86
3.83
-0.23
1.34
0.13
0.02
*All variables are in log, with the exception for Inflation, Political Constraints, Civil Rights and Political rights.
Concerning the degree of democracy, SSA region performs slightly better. Although the political
constraints on governors and the political rights of citizens are low in SSA countries, for what
concerns the civil liberties of their citizens they outperform Asian countries, as well as countries
which are included in the “Other” category (which covers eastern Europe, north Africa and middle
East). The region which performs better in terms of degree of democracy is Latin America and the
Caribbean, which ranks first in all three indexes.
Asian economies offer, on average, the most favorable economic environment. SSA economies
display the lowest degree of market openness, while the highest average inflation and foreign debt
is recorded in developing countries from eastern Europe, north Africa and middle east.
Finally, the highest incidence of wars and minor conflicts can be found in Asia and the Pacific,
followed by SSA countries
Additional useful information can be obtained by analyzing the variability of data. Total variance in
panel data can be decomposed in two different kind of variability: between variability – that is the
sample variance between countries – and within variability – which instead is intended as the
sample variance within countries, along the years of study.
Since the dataset for this study is unbalanced, the computation of the share of within and between
variability of variables must performed over a balanced subsample. For this purpose, 11 countries
have been excluded13. Since the majority of missing values come from the fact that negative FDI
values have been excluded in order to transform the net inflow of FDI in logarithm, it is preferable
to measure the variability of the net FDI inflow in level. This translates in a slightly higher percentage
of within variability, as the logarithmic form smooth fluctuations of variables over time.
13
Argentina, Armenia, Azerbaijan, Belarus, Gabon, Guinea Bissau, Guyana, Madagascar, Myanmar, Nicaragua and Vietnam.
12
Therefore, we believe that it is more methodologically coherent to compute the variability
composition using the data in level for all variables. To present a more complete analysis, table 2A
in the appendix reports the composition of variance of the logarithmic form of explanatory
variables.
As depicted in Table 2, the dependent variable and the majority of the other variables displays a
higher degree of between variability. These variables include GDP, GDP per capita, life expectancy,
average years of schooling, fixed telephone subscribers and the degree of market openness.
Only inflation and the level of foreign debt present a higher degree of within variability. The risk is
that those variables will not have any significant impact in explaining the variation of FDI inflows.
However, net inflow of FDI still presents a degree of within variability (equal to 15% of total
variability), so it is possible that this fraction of variability is explained by the evolution in price levels
and the amount of foreign debt.
Table 2 – Total, Between and Within Variability
Net FDI Inflow
GDP
GDP pc
Life Expectancy
Av. Years of Schooling
Fixed Telephone Subscribers
Inflation
Foreign Debt
Trade Openness
Law and Order index
Total Variability
% Between
% Within
5.60987899379975e+23
5.18954593097873e+26
8817281380.51
69249.51
6380.26
1292587940688366080
205171.9
1348247.15
74.69
966.49
0.85
0.85
0.94
0.91
0.95
0.9
0.29
0.41
0.84
0.84
0.15
0.15
0.06
0.09
0.05
0.1
0.71
0.59
0.16
0.16
Before proceeding with the choice of the econometric model, it is useful to check for possible
multicollinearity issues in the correlation matrix (Figure 5). In general, it does not report very high
correlation across the explanatory variables, with the exception of telephone subscriptions, which
is very highly correlated with GDP. This could rise some problems when performing the estimation
for their coefficients. However, due to the impossibility to substitute the proxy for infrastructure,
and to the fundamental role that GDP has in explaining FDI patterns, none of the variables could be
dropped.
13
Figure 5 - Correlation Matrix
5. Model Specification
In order to build a suitable model for the purpose of our analysis, we started from the following
unobserved effects model:
𝑦*+ = 𝜇* + 𝛽𝑥*+ + 𝜖*+
Even if this kind of model assumes that 𝛽 coefficients are the same across all units, it differs from
the pooling model in that it allows for a certain degree of heterogeneity. This is made possible by
assuming that the standard error term 𝜀*+ has two components: 𝜇* which is specific to each unit and
𝜖*+ , called “idiosyncratic” error term, which is assumed to be uncorrelated from the set of
explanatory variables, as the standard error term.
Fixed effect models assume that the unit specific error terms is correlated with the regressors.
To be able to perform an OLS estimation on this regression, the error is treated as standard set of n
parameters to be estimated, which translates in a model where the 𝛼 coefficient is specific to each
country and is represented by 𝜇* .
On the other hand, random effects model assume that the unit specific error term is uncorrelated
with the set of regressor as well as the standard error terms 𝜀*+
Fixed effects model seems suitable for the purpose of our analysis, since it is performed over a
sample including exclusively developing country. Since units are not randomly selected, it is very
likely that the individual component of the error term is not independent from the regressors.
Some works from the previous literature (e.g. Azemar and Desbordes (2009)) support this
hypothesis, by making use of fixed effect model for a similar kind of analysis.
14
Once the set entire of regressor is included, the following model is obtained:
𝐿𝑛(𝐹𝐷𝐼<=+ )+* = 𝜇* + 𝛽? 𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝐹𝐷𝐼*+G? + 𝛽H ln (𝐺𝐷𝑃)+G?
+ 𝛽M ln (𝐷𝑒𝑔𝑟𝑒𝑒 𝑜𝑓 𝑑𝑒𝑚𝑜𝑐𝑟𝑎𝑐𝑦)+G?
*
*
+ 𝛽T ln(𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛)+G?
+ 𝛽W ln(𝐿𝑖𝑓𝑒 𝐸𝑥𝑝𝑒𝑐𝑡𝑎𝑛𝑐𝑦)+G?
*
*
+ 𝛽Y ln(𝑇𝑒𝑙𝑒𝑝ℎ𝑜𝑛𝑒 𝑠𝑢𝑏𝑠𝑐𝑟𝑖𝑏𝑒𝑟𝑠)+G?
+ 𝛽_ (𝐼𝑛𝑓𝑙𝑎𝑡𝑖𝑜𝑛)+G?
*
*
+ 𝛽` ln(𝐹𝑜𝑟𝑒𝑖𝑔𝑛 𝐷𝑒𝑏𝑡)+G?
+ 𝛽a ln(𝑇𝑟𝑎𝑑𝑒 𝑂𝑝𝑒𝑛𝑛𝑒𝑠)+G?
+ 𝛽?c ln(𝐼𝑐𝑟𝑔)+G?
+ 𝜀*+
*
*
*
As mentioned in section 3, the degree of democracy is captured through six dummies: two of them
account for the degree of civil liberties (medium and high civil liberties), other two represent the
presence of a more or less democratic regime (medium and high political rights) and the last two
indicate the feasibility of policy change (medium or high political constraints).
As already explained, variables have been lagged by one year and when possible, variables have
been taken in the natural logarithm form.
An individual (country) specific fixed effect is chosen to account for heterogeneity between
developing countries; this enables to account for factors specific to each country that can be
considered constant over time and disentangle their effects from the effect of time varying factors.
For developing countries, there might be some time constant factors which can be correlated with
other explanatory variables and could affect the estimation of the FDI determinants. For example,
many developing countries are rich of natural resources, which might have an important role in
attracting FDI and are very likely to be constant over 16 years covered by our study.
To confirm the choice of the individual fixed effect model we performed both the Fisher and the
Hausman test.
The Fisher test confirmed that the fixed model was to be preferred to the simple pooled regression
due to the presence of significant individual effects. The Hausman test allowed us to compare the
Fixed effect model and the Random effect. Again, the result was in favor of the former regression.
Table 3 reports the results from the two tests
Table 3 – Results of Fisher and Hausman test
F test for individual effects
data:
lnfdinet ~ negfdi + lngdp + lnlifexp + lneduc + lnteleph + inflation +
...
F = 10.704, df1 = 64, df2 = 933, p-value < 2.2e-16
alternative hypothesis: significant effects
Hausman Test
data:
lnfdinet ~ negfdi + lngdp + lngdppc + lnlifexp + lneduc + lnteleph +
chisq = 45.25, df = 17, p-value = 0.0002235
alternative hypothesis: one model is inconsistent
...
15
6. Results
Results from implementing the individual fixed effect model are showed in Table 4.
Column (1) and (2) show the differences in terms of significance of the coefficients when using
normal standard errors or robust standard errors.
As it is possible to see in the first column, our results are partly in line with the traditional literature:
trade openness and market size are the main determinants of the flows of FDI, whereas the proxy
for infrastructure appear to be only mildly relevant.
In accordance with the more recent researches, other factors result relevant. The level of public
good provision is highly significant and have a positive sign in both life expectancy and education.
The political structure result to have an impact only when the political rights are classified as “high”.
Finally, civil freedom of citizens seems to be a relevant element: both medium and civil rights
resulted significant.
When analyzing the results with coefficients robust to heteroskedasticity and serial correlation,
many variables lose their significance. The most traditional determinants however, namely GDP and
trade openness, remain significant, at 1 and 0.1 per cent respectively.
The fact that a larger market size should attract FDI flows is quite straightforward. First of all, the
GDP is an indicator of the degree of development, therefore, since investors will invest in a country
where the perceived profitability of their projects is secured, they will take into account the signal
transmitted by GDP in this regard. Moreover, a larger market size implies, usually, lower distribution
cost (if of course the production and distribution facilities are located where the majority of the
customers will be located) and a greater chance of obtaining better and more specialized inputs.
Trade openness covers a crucial role for developing countries since most of them do not have access
to international capital markets. Due to this issue they often implement several trade policies to
ease the inflows of capital through FDI and the effect of these policies might be captured by the
significance of this indicator. However, as suggested by various previous works (Rodriguez and
Rodrik, 2000; Asiedu,2006) the interpretation of this coefficient should be done with care since
policymakers do not directly control the volume of trade; therefore - when the objective of the
analysis is policy prescription - different measure of openness, that can be considered as directly
influenced by politicians, might result more useful.
The dummy that indicate the presence of negative FDI stays significant also in the heteroskedastic
robust specification. Past studies investigated the phenomenon of agglomeration in foreign
investments. Wheeler and Mody (1992), found that US investments into a country are strongly
conditioned by existing stock of FDI in that country. This could be explained by the presence
agglomeration benefits: some industries rely heavily on the proximity of the suppliers for
intermediate inputs, therefore the presence of an existing network could create a certain
persistence in the investors preferences. Moreover, the possibility of a learning experience is also a
plausible explanation. Kogut and Chang (1996) argue that obtaining a greater familiarity with
operating in a country and a better knowledge of the specific opportunities of a region might create
commitment in the investors. Both these explanation lead to the conclusion that the presence of
negative inflows of FDI can have a negative impact on future investment, which is confirmed by our
results.
16
Also, the strength of the law system is proven to play a positive role in influencing FDI flows in
developing countries. As mentioned before, this is possibly due to the fact that foreign investors
demand the protection of their property rights in order to secure the returns from their investment.
Low security in this field will reduce the feasibility of economic activity and the scope for market
transactions.
In addition, Table 4 shows a positive and significant impact on the dependent variable of higher level
of human rights and citizens freedom. This is in line with the part of the literature that proved that
countries which ensure higher levels of political and civil freedom to their citizens attract higher FDI
flows.
In contrast, and as expected, the presence of both minor conflicts and wars (as defined in section 3)
has a negative effect on FDI, even if it is not strongly significant. A higher incidence of war and
conflicts sends a strong signal to investors regarding the instability of the country, in fact it is often
linked with poor governance and a weak protection of property right.
Surprisingly, when using heteroskedastic robust errors, the three dimensions which proxy provision
of public good – life expectancy, average years of schooling and fixed telephone subscribers - lose
their significance.
At odd with the literature, the infrastructure system does not appear to matter. However this could
be explained by three different reasoning. First, the limited time period that has been taken into
account could affect the estimation since infrastructures seems to be a stable variable over time.
Second, the measure of infrastructure that was chosen (Fixed Telephone Subscribers) resulted
highly correlated with the GDP since usually larger and richer countries are characterized by more
physical infrastructure. In order to understand this issue, we implemented our model in the absence
of GDP and our infrastructure proxy resulted statistically significant and positive14. Lastly, as already
mentioned in Section 3, this measure has some flaws since the spread of mobile phones as the main
communication tool has significantly reduced the use of fixed ones including in developing
countries; according to the World Bank, the average number of fixed telephone users worldwide
has been growing until 2010.
The use of alternative variables would have been useful, for example the number of telephone lines
instead of the number of subscribers would probably be a better source of information but several
problems were encountered when trying to gather these data.
The coefficients in Table 4 are the expression of the direct effects on FDI of the chosen variables.
However, it is meaningful to underline that they leave out the interaction of these various factors,
which is an important element to take into account, especially for countries which want to take
steps to attract more FDI. An example of this interplay could be the link between the level of
democracy of a country and the provision of public goods, or the one between conflicts and the
quality of infrastructures. Indeed, it could be possible that territorial conflicts destroy physical
infrastructure and also that war contribute to the deterioration of public governance.
In this regards, the main risk is that some of the effects of the FDI determinants might be
underestimated.
14
Results of this specification are reported in Appendix A, table A3.
17
Table 4 - Determinants of FDI in developing countries
Dependent variable:
Net FDI inflow
Negative FDI
GDP
Life expectancy
Education
Fixed telephone subscriptions
Inflation
Foreign Debt (% GNI)
Trade openness
ICRG law and order
Medcivrights
Highcivrights
Medpolconst
Highpolconst
Medpolrights
Highpolrights
Minor conflict
War
Observations
R2
Adjusted R2
F Statistic (df = 17; 933)
Signif. codes
default
(1)
robust
(2)
-0.856*** (0.179)
0.547** (0.170)
2.837** (1.016)
0.581 (0.421)
0.155 . (0.082)
-0.002 (0.002)
-0.141* (0.062)
0.635*** (0.142)
0.365* (0.149)
0.370* (0.175)
0.435* (0.216)
-0.026 (0.070)
0.248 (0.156)
0.074 (0.123)
0.327* (0.163)
-0.229* (0.110)
-0.304 . (0.168)
-0.856* (0.380)
0.547* (0.258)
2.837 (2.127)
0.581 (0.595)
0.155 (0.130)
-0.002 (0.003)
-0.141 (0.108)
0.635** (0.200)
0.365* (0.159)
0.370* (0.148)
0.435* (0.207)
-0.026 (0.119)
0.248 (0.189)
0.074 (0.179)
0.327 (0.215)
-0.229 . (0.130)
-0.30 . (0.171)
1,016
0.302
0.241
23.800**
1,016
0.302
0.241
23.800**
0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ‘1
6.1 Sub Saharan countries
Previous section presented an analysis of the general determinants of FDI in developing country.
However, the previous econometric model does not give any information about the sources of
disparity in FDI levels in SSA countries. This section enriches our analysis by filling this gap.
As suggested by other studies (Azemar and Desbordes (2009); Fernandez-Arias and Montiel, 2001),
the specificity of the determinants of FDI flows in sub-Saharan countries can be examined with an
inter-regional study.
18
Differences among various regions considered can be in fact analyzed using the following
decomposition15:
ddddd
ddddd
𝑙 ij
𝑛𝐹𝐷𝐼
−j
𝑙𝑛𝐹𝐷𝐼
eef
jjj
jkj
jj
jlh =
fm+nop q*rr=s=<m=
wv x
𝛽t𝑙𝑛𝑋
−j𝑙𝑛𝑋
ij
jjdddddd
jeef
jkj
jjjl
y*z=G{os|*<} q*rr=s=<m=
+
dddddd
t𝐶
− 𝐶wv x
eef
ij
jkjjl
e+snm+nsop q*rr=s=<m=
+
t𝜀
ddddd
−jl
𝜀wx
eef
v
ij
jkj
•<=€•po*<=q q*rr=s=<m=
The “actual difference” term is the simple average difference in volume of net FDI in region j (which
corresponds alternately to all non-SSA developing countries) and in SSA, in the period considered
(1999-2014). On the right-hand side of the equation, and (i = country i) are respectively the
coefficients and the country specific fixed effects estimated in the previous model (taking into
account all the four groups) and are the time-varying FDI determinants. The last term corresponds
to differences in aggregate FDI residuals and can be interpreted as a measure of the performance
of the benchmark region, that is, whether it has been higher or lower than expected.
dddddeef − 𝑙𝑛𝐹𝐷𝐼
dddddh .
To obtain the share of each components we divided each term by 𝑙𝑛𝐹𝐷𝐼
Results of the decomposition are displayed in Table 5.
A significant part of the inter-regional gap seems to be explained by the market size, which points
out to the big differences, in terms of economic growth, that still exists between SSA and other
developing countries. It is noteworthy, indeed, that SSA countries are reporting major delays in
terms of industrial revolution, which instead, has interested many other developing countries, and
provides many investment opportunities besides the mere exploitation of natural resources.
In line with the literature the good provision of public goods (Education, Health and Infrastructure)
seems to explain a consistent part – more than half – of the unattractiveness of SSA countries for
foreign investors with respect to all the other groups considered.
Indeed, SSA countries still lag well behind the other developing countries in both these 3 indicators,
as confirmed by the descriptive statistics in section 4.
In particular, an higher incidence of infectious disease as HIV and Malaria, has a considerable
negative impact on citizens of SSA countries, which, together with modest educational attainment,
lowers their productivity importantly.
The differences in the degree of democracy and in macroeconomic stability have instead a little role
according to our computations: on average they account respectively for less than 5% and less than
10% of the difference in FDI levels. This is coherent with findings from section 4: SSA countries’
performance in those field is more or less in line with that of the other developing countries.
In addition, also the security of property rights explains only a very small part of the divergencies.
Finally, the contribution of structural factors (due to the country specifics characteristics) seems to
be crucial, particularly with respect to LAC countries. On average, country specific effects seem to
explain at least half of the attractiveness of SSA countries for foreign investors.
As already mentioned, this could be attributed to the fact that SSA region is one of the richest areas
of the world in term of natural resources, especially diamonds, oil, natural gas and metals.
Obviously, the presence of these important resources represents a major investment opportunity,
especially for developed country that lack instead from this kind of richness.
15
This decomposition has been performed using both R and Excel. Computation are reported in Appendix D.
19
However, when studying these results, it clearly appears that the unexplained portion is quite
important and therefore further researches might be necessary, for example by including more
precise proxies for some indicators, as the level of infrastructure or the degree of market openness.
An interesting extension could be to include tax rates on the activity of foreign companies operating
on the national territory. Indeed, as reported by the OECD (2008)16 volumes of FDI are sensitive to
taxation and many countries, especially developing ones, are trying to offer competitive tax
environments in the attempt to increase the FDI inflow toward their economies. Unfortunately, due
to unavailability of data, we weren’t able to include level of taxation among the regressors.
Table 5 - Inter-regional differences in FDI volume, relative contribution of single determinants (%)
16
OECD Observer , 2008. “Tax Effects on Foreign Direct Investment”
20
7. Conclusion
This work aimed at understanding the determinants of the location of FDI in developing countries
and, in particular, to investigate the source of disparity in the ability of attracting FDI that concerns
countries located in Sub Saharan regions. For this purpose, it uses an unbalanced panel dataset
including 66 developing countries and covering the period 1998-2014
The results of our estimation confirm that the size of the markets plays a major role in attracting
FDI inflow. This confirm the idea that a greater market size offers more investment and operating
opportunities. Moreover, the presence of negative FDI in the previous year seems to send a
significant negative signal to investor, which prefer to direct their capital elsewhere.
Following the work of Azemar and Desbordes (2009) and Asiedu (2006), we investigated the role of
public governance in FDI dynamics. This has been done analyzing several dimensions of the
governance of a country: the degree of democracy, the provision of public goods, the
macroeconomic stability and the possibility of ensuring protection of property rights.
Among these dimensions, the degree of democracy and in particular of civil liberties (the guarantee
of freedom of expression and human and civil rights) appears to promote FDI in all developing
counties. Also, the strengthen of the rule of law, which represent the impartiality of the legal system
and its observance, plays an important role in our estimations, confirming previous literature results
(among others Globerman and Shapiro, 2002). In addition, low frequency of conflict, which together
with rule of law enhance security of property right, contributed in attracting foreign capital.
Surprisingly, once robust errors are taken into account, the three dimensions of public good
provision, namely life expectancy, average years of schooling and fixed telephone subscribers, lose
their significance. However, we believe that this issue can be related to the specification of our
model rather than to an actual independence between those variable and level of FDI. In particular,
as already mentioned, the use of fixed telephone subscribers as a proxy for infrastructure is quite
imprecise and leads to some problems at the level of econometric estimation.
Importance of public goods provision is indeed confirmed in the second step of the analysis. To fully
understand the relatively poor FDI performance of SSA countries we implemented an inter-regional
decomposition that allowed us to examine the relative contribution of each determinant to the FDI
gap between SSA e non-SSA regions. What emerged is that the differences can largely be explained
by insufficient provision of public goods in terms of health and education, as well as market size.
However, these deficiencies can be also indirectly related to country specific conditions, as natural
resources provision, that can promote but also undermine the choice of FDI investors towards SSA
countries. On the one hand, richness in terms of resources can be an important element in favor of
investment from the perspective of MNEs. On the other, the greater is the role for these exogenous
factors in attracting FDI, the smaller results the share coming from policy efforts undertaken by the
country.
The empirical work for understanding the determinants of FDI in developing countries and in
particular in Africa should be continued to better disentangle country specific effects from other
influences. This could be crucial for the implementation of adequate policies in developing countries
that aim at a stronger integration in the international market.
Moreover, this kind of analysis might not give a complete overview of the most important
determinants of FDI since the interaction between the various factors is left out and could bring to
an underestimation of the impact of various determinants.
21
Despite its limitation, some useful policy recommendation can be draw from this study. Developing
countries should focus their policy attention in developing their market size and making regulations
more international trade friendly. Moreover, reinforcing security of property rights might be very
useful especially for LAC countries which lag well behind the average of other regions in this field.
Concerning SSA countries, although they are importantly improving their ability in attracting foreign
capital, they should focus their attention on expanding their markets to open new opportunities to
foreign investors, and improve the provision of public good, especially for what concerns health and
infrastructures.
22
Appendix A
Table A1 - Countries in the sample, by region
SSA
AP
LAC
Other
Botswana
Bangladesh
Argentina
Albania
Cameroon
China
Bolivia
Armenia
Congo, Rep.
India
Brazil
Azerbaijan
Ethiopia
Indonesia
Colombia
Belarus
Gabon
Mongolia
Costa Rica
Bulgaria
Ghana
Myanmar
Dominican Republic
Egypt, Arab Rep.
Guinea
Pakistan
Ecuador
Jordan
Kenya
Papua New Guinea
El Salvador
Kazakhstan
Philippines
Guatemala
Moldova
Malawi
Sri Lanka
Guyana
Morocco
Mali
Thailand
Haiti
Romania
Niger
Vietnam
Honduras
Russian Federation
Nigeria
Jamaica
Tunisia
Senegal
Mexico
Turkey
Sierra Leone
Nicaragua
Ukraine
South Africa
Paraguay
Madagascar
Sudan
Peru
Tanzania
Venezuela, RB
Togo
Uganda
Zambia
Table A2 - Variability of variables in their logarithmic form
Ln GDP
Ln GDP pc
Ln Life Expectancy
Ln Av. Years of Schooling
Ln Fixed Telephone Subscribers
Ln Foreign Debt
Ln Trade Openness
Ln Law and Order index
Total Variability
% Between
% Within
2894.19
949.96
18.42
226.55
4278.76
415.19
164.93
119.11
0.98
0.97
0.89
0.96
0.97
0.48
0.82
0.79
0.02
0.03
0.11
0.04
0.03
0.52
0.18
0.21
23
Table A3 - Determinants of FDI in developing country excluding GDP
Dependent variable:
Net FDI Inflow
Negative FDI
Life expectancy
Education
Fixed telephone subscriptions
Inflation
Foreign Debt (% GNI)
Trade openness
ICRG law and order
Medcivrights
Highcivrights
Medpolconst
Highpolconst
Medpolrights
Highpolrights
Minor conflict
War
Observations
R2
Adjusted R2
F Statistic (df = 16; 934)
default
(1)
robust
(2)
-0.862*** (0.179)
4.109*** (0.941)
1.105** (0.390)
0.211** (0.080)
-0.003 (0.002)
-0.149* (0.062)
0.614*** (0.142)
0.364* (0.149)
0.385* (0.176)
0.497* (0.217)
-0.035 (0.071)
0.273. (0.157)
0.045 (0.124)
0.313. (0.163)
-0.207. (0.110)
-0.324. (0.169)
-0.862* (0.376)
4.109* (1.937)
1.105* (0.560)
0.211. (0.119)
-0.003 (0.003)
-0.149 (0.112)
0.614** (0.205)
0.364* (0.157)
0.385** (0.149)
0.497* (0.216)
-0.035 (0.124)
0.273 (0.191)
0.045 (0.189)
0.313 (0.226)
-0.207 (0.129)
-0.324. (0.167)
1,016
0.295
0.234
24.398***
1,016
0.295
0.234
24.398***
Appendix B1 – Database reporting non-lagged data
Appendix B2 – Database reporting lagged data
Appendix C1 – R work file, descriptive statistic
Appendix C2 – R work file, regressions
Appendix D – Excel file and R work file, decomposition
24
References
Asiedu, E. 2006. “Foreign direct investment in Africa: The role of natural resources, market size,
government policy, institutions and political instability”. The World Economy, Vol. 29(1), pp. 6377.
Azémar, C. & Desbordes, R. 2009. "Public Governance, Health and Foreign Direct Investment in
Sub-Saharan Africa". Journal of African Economies, Centre for the Study of African Economies
(CSAE), vol. 18(4), pages 667-709, August.
Buchanan, B. G. & Le, Q. V. & Rishi, M. 2012. "Foreign direct investment and institutional quality:
Some empirical evidence". International Review of Financial Analysis, Elsevier, vol. 21(C), pages 8189.
Croissant, Yves, e Giovanni Millo. 2008. «Panel Data Econometrics in R: The plm Package.» Journal
of Statistical Software, Articles 27 (2): 1-43.
Henisz, W. J. 2002. “The institutional environment for infrastructure investment, Industrial and
Corporate Change”. Volume 11, Issue 2, April 2002, Pages 355–389.
Kogut, Bruce, and Sea Jin Chang (1996) 'Platform investments and volatile exchange rates: direct
investment in the U.S. by Japanese electronic companies,' Review of Eco-nomics and Statistics 78,
221Kumari, R. & Sharma, A. K. 2017. "Determinants of foreign direct investment in developing
countries: A panel data study". International Journal of Emerging Markets.
Marshall, M. G & Gurr, T. R & Jaggers, K. 2019. “POLITY IV PROJECT: Dataset Users’ Manual”. Center
for Systemic Peace.
Pettersson, T. 2019. “UCDP/PRIO Armed Conflict Dataset Codebook v 19.1”.
Wheeler, David, and Ashoka Mody (1992) 'International investment location decisions: the case of
US firms,' Journal of International Economics 33, 57-76
25
Téléchargement