Research Article |
Corresponding author: Ololade Mistura Aromasodun ( meetmistura@gmail.com ) Academic editor: Marina Sheresheva
© 2022 Ololade Mistura Aromasodun.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits to copy and distribute the article for non-commercial purposes, provided that the article is not altered or modified and the original author and source are credited.
Citation:
Aromasodun OM (2022) Determinants of FDI inflows to West Africa: Prospects for regional development and globalization. BRICS Journal of Economics 3(1): 27-51. https://doi.org/10.3897/brics-econ.3.e83129
|
Abstract
This paper examines the determinants of foreign direct investment (FDI) inflow into West Africa. FDI is regarded as the central engine for growth. Such inflows are not often satisfactory, both in terms of their volume and in terms of their sectoral distribution, particularly in developing countries. The study carried out a unit root test using the Im-Pesaran-shin (IPS) method, which revealed that four out of many variables were stationary at first difference, while other variables were stationary at level. Consequently, the Kao co-integration test methodology was used to analyze the long-run relationship. Thus, the regression analysis was carried out using the Panel ARDL method in an equation with a 50-year observation period. Concerning the remaining seven equations with shorter time series observations, the Pooled OLS estimation method was used to analyze the factors determining the inflow of FDI. The results indicate that financial development has a negative effect on FDI flows (and hence on globalization processes) in West Africa, while trade openness, institutional composite index and control of corruption have positive effects on FDI and hence increase globalization tendency. Based on these findings, the study recommends, among other things, that the authorities in West African countries vigorously pursue trade liberalization policy as an effort to globalize the region through FDI inflows. The study examined the macroeconomic determinants on FDI alongside institutional and socio-political determinants that are difficult to study in the case of West Africa as a region. The use of a composite institutional quality index, which combines multiple indicators of institutional quality, is another novelty of this research. Another unique contribution of the study is the use of the Africa Infrastructure Development Index (AIDI), which serves as a composite infrastructure index, as an explanatory variable.
Foreign direct investment, institutional FDI fitness, panel ARDL, pooled OLS, West Africa
Foreign investment is regarded as the central engine for growth. Attracting investment has become the main factor of industrial policy in many countries. Even countries that were formerly inaccessible to foreign investors, such as China, have acknowledged the economic benefits of foreign investment and opened their borders to it.
Regarding West Africa as an FDI recipient region, in 2018, FDI to the region declined by 15 percent to $9.6 billion, the lowest level since 2006. According to UNCTAD (2019), this was mostly owing to a significant decrease in the flow to Nigeria for the second year in a row. Nigeria’s inward FDI dropped by 43% to $2 billion, and the country is no longer the largest recipient of FDI in West Africa. UNCTAD (2019) further reports that Ghana has become the largest FDI receiver in West Africa, despite FDI inflows falling by 8% to $3 billion (see Figure
Regardless of collective initiatives at the regional and continental levels to improve the flow of FDI to West Africa, the task of attracting FDI that is consistent with individual countries’ development goals remains in the hands of the governments, making it critical to identify the major determinants of FDI.
An attempt at assisting policymakers in this regard has been made through various theoretical and, especially, empirical studies on determinants of FDI as reviewed in the next section. However, as also discussed at the end of the next section, such studies at the empirical level are bedevilled with several methodological gaps and pitfalls. One of the limitations of these studies is that they all test some predictions of their models in an ad hoc econometric model controlling for other possible determinants of FDI as GDP per capita, openness, size, etc.
Given this ad hoc formulation and the fact that they use different institutional variables, it is difficult to determine the source of the qualitative and quantitative differences in their results. It would be enlightening for policymakers to know to what extent macroeconomic factors determine FDI in West Africa. The extent to which socio-political factors determine FDI in West Africa has not been empirically tested. Lastly, there is also the need to shed light on the extent to which institutional factors determine the inflows of FDI in West Africa, for which we adopt the composite institutional quality index in this study since most papers in the literature consider only one aspect of a set of institutional factors.
The present study is an attempt directed at addressing all these issues, which the existing studies have failed to address. It examines the impact of macroeconomic determinants on FDI alongside institutional and socio-political determinants which is difficult to study in the case of West Africa as a region. The use of a composite institutional quality index, which combines multiple indicators of institutional quality, is another novelty of this research. Another unique contribution of this study is using the Africa Infrastructure Development Index (AIDI) as an explanatory variable that serves as a composite infrastructure index.
The origins of FDI are not entirely clear. Although various schools of thought have been employed to explain this phenomenon, no superior or general explanation of FDI has emerged.
We can broadly divide theories into two categories: macroeconomic theories and microeconomic theories of FDI. However, for the sake of this study, the theories under review are limited only to macroeconomic ones.
Lipsey (2004) describes the macroeconomic view as a specific type of capital movement across national boundaries, from home nations to host countries, as reflected in balance-of-payments statistics. These flows generate a specific type of capital stock in host countries: the volume of the home country investment in organizations, generally businesses, controlled by a home-country owner or in which a home-country owner has a specified proportion of voting rights. Various macroeconomic theories are reviewed below.
Capital market theory, commonly known as the Currency Area Theory, is one of the first ideas to explain FDI. It is based on the work of
Location-based approach to FDI theory. Although a firm’s behavior (a microeconomic element) influences FDI location in terms of the motives for its location, whether it be the search for resources, markets, efficiency or strategic assets, the overall economic and geographical decision takes into consideration the macroeconomic decision because of its country-level features (Popovici & Calin, 2014). According to them, the theory explains the effectiveness of FDI among nations based on a country’s natural resource endowment, labor availability, local market size, infrastructure, and government policy towards these national resources.
Institutional FDI fitness theory. The term “FDI fitness” was developed by
This section covers studies on the determinants of foreign direct investment outside Africa and then proceeds to review the evidence from Africa. The section concludes with a discussion of the gaps in the empirical studies that this paper aims to fill.
Empirical literature on countries outside Africa. In this category, there are a lot of studies but we limit the review to only recent ones, starting in the early 2000s, to focus on modern methodologies, including the latest datasets.
One of the earliest studies is a paper written by
A further test on determinants of inward FDI was carried out by
Unlike the previous study, which used VAR as an estimating approach, Marcelo and Mario (2004) used an econometric model based on panel data analysis. In order to shed light on FDI in developing nations, they analyzed 38 developing countries (including transition economies) from 1975 to 2000. One of the key results was that FDI is correlates with the level of education, the degree of openness of the economy, political risk and variables related to macroeconomic performance, such as inflation, and the average rate of economic growth. The findings also show that FDI is closely related to stock market performance, which leads to the conclusion that a large portion of direct investment in developing countries is directed to relatively knowledge-intensive activities and that policies aimed at increasing the level of education may induce these investments. In this study, only macroeconomic determinants of FDI were used as variables.
In another study using the Granger causality test on data for the period 1969–2000 for three countries (Chile, Malaysia, and Thailand),
Empirical literature on Africa. Several studies have been conducted on the determinants of FDI inflow in Africa. Again, this review also covers recent studies for the same reasons as stated above when reviewing the non-African countries.
The results obtained by
Using a panel of 69 countries between 1981 and 2005,
Using the same estimation technique as in the previous study by
Gholami et al. (2006) analyze the influence of such factors as GDP, ICT, and openness on FDI inflows in a sample of 23 industrialized and developing countries observed from 1976 to 1999 using the Least Squares Dummy Variables (LSDV) regression analysis technique. According to the study, the existing ICT infrastructure attracts FDI, and higher levels of ICT investment lead to larger levels of FDI inflows in developed nations, while in developing countries, the direction of causation is shifting from FDI to ICT. The study does not consider some other determinants of FDI, such as institutional variables.
Using a different estimation technique compared to that of Gholami et al. (2006),
Another study on FDI conducted by
A study carried out by
In addition,
Using a panel dataset for the period from 1970 to 2010, Anyanwu & Nadege (2015) attempted to establish the determinants of FDI inflows to West Africa. The estimations were made using the OLS and GMM methods. The main findings show that: (i) the quadratic element of real per capita GDP, domestic investment, trade openness, first-year lag of FDI, natural resource endowment and exports, and monetary integration all have positive effects on FDI inflows to West Africa; and (ii) there is a negative relationship between FDI inflows to the sub-region and the loan component of ODA, economic growth, and monetary integration. The use of dummy variables to represent oil-exporting countries does not capture the expected effect of natural resources on FDI.
A large body of empirical literature has been generated to study the determinants of FDI. However, there are few studies on FDI determinants in the context of West Africa as a sub-region. The majority of the previous FDI studies have focused on either Sub-Saharan Africa, Africa as a whole, or a single nation. In addition, there are varying conclusions from existing research on the topic owing to the fact that each studied region has different prevailing economic conditions.
There is a limited amount of research concerning institutional and socio-political determinants of FDI in West African countries. In addition, there is a need for a study based on more recent data to update the existing findings that were based on outdated data sets. The present study meets this need.
The use of a composite institutional quality index, which combines multiple indicators of institutional quality, is another novelty of our research. The majority of the articles in the literature focus on just one or a few institutional variables. In the literature, however, it is suggested that institutional variables are significantly linked to one another (Globerman & Shapiro, 2002). As a result, we use Principal Component Analysis to create a composite index by integrating multiple characteristics of institutions into one component (PCA).
Another unique contribution of the study is the use of the Africa Infrastructure Development Index (AIDI) as an explanatory variable, which serves as a composite infrastructure index. The AIDI data set comprises transport composite index, electricity composite index, ICT composite index, and water supply and sanitation (WSS) composite index.
The Institutional FDI Fitness Theory developed by Wilhems and Witter (1998) is adopted for this study. The words “FDI fitness” refer to a country’s ability to attract, absorb, and retain FDI by responding quickly to threats and opportunities, as well as by being creative and flexible in carving out a niche in which it can compete. According to this theory, nations with high institutional fitness get more FDI than countries with low institutional fitness.
A panel data-based regression model to test for the actual effects of the postulated determinants of FDI is presented in Equation 1 below. Equation 1 shows that FDI is a function of control variables without the inclusion of institutional indicators. It will be estimated using a data set dating back to 1970, as it does not include governance indicators whose data set commences from 1996.
In subsequent equations, each of the afore-mentioned seven governance indicators is added, one at a time, to the benchmark Equation 1. They are included one at a time, instead of two or more featuring simultaneously in an equation, to avoid multicollinearity problems in view of the fact that they are highly inter-correlated. By including these governance indicators, the resulting equations can only be estimated with post-1995 (instead of post-1969) data, as a series of governance indicators start from 1996, with each of the seven governance indicators appearing in an equation.
where: FDI — foreign direct investment, FD — financial development, GRGDP — growth rate of gross domestic product, RGDPPC — real income per capita, URBANPOP — urban population, OPN — trade openness, INF — inflation, INFRA — infrastructure, GOV — institutional variables, POL — political rights, NAT — natural resource, NOVIO — absence of violence, REGQ — regulatory quality, GOVTEFF — government effectiveness, VAC — voice and accountability, CORR — control of corruption, ROL — rule of law.
The basic features of the variables are highlighted based on the results of the descriptive and correlation analyses of policy makers. The main inferential analyses is carried out in the form of a unit root and co-integration test to properly address the time-series features of the data and provide a guide on the methods of estimating the regression equation to be adopted. The study conducts autocorrelation, heteroskedasticity, multicollinearity, normality of distribution of the residuals and stability tests and adopts remedial measures when a test shows there is a problem to ensure that the results obtained lead to reliable conclusions.
The study covers 16 West African countries (Benin, Burkina Faso, Cape Verde, Gambia, Ghana, Guinea, Guinea-Bissau, Ivory Coast, Liberia, Mali, Mauritania, Niger, Nigeria, Senegal, Sierra Leone, and Togo) from 1970 to 2019. The choice of West Africa is due to the fact that limited research was carried out on the region, while the period is chosen based on the availability of data from 1970 onward and also because 2019 is the most recent year of data available at the time of this study.
Foreign direct investment is computed as the % of GDP, the growth rate of real GDP is calculated as the first difference of annual GDP expressed as a percentage of real GDP in the preceding year. The urban population is computed as a percentage of the total population. Gross domestic product per capita is expressed as purchasing power parity, constant for 2010, calculated in US dollars. Trade openness index is computed as total trade, % of GDP, while financial development is expressed as domestic credit to the private sector, % of GDP. The inflation rate is measured in annual percent. The political right is measured in index.
INFRA is an infrastructure composite index that is proxied by Africa Infrastructure Development Index (AIDI). The AIDI data sets comprise of transport composite index, electricity composite index, ICT composite index, and water supply and sanitation (WSS) composite index.
GOV is a composite institutional quality index that combines (through the use of the Principal Component Analysis method) 6 indicators of institutional variables: absence of violence/terrorism, regulatory quality, government effectiveness, voice and accountability, control of corruption, and the rule of law.
The data is obtained from the World Bank database (online), except POL that was obtained from Freedom House.
This section presents and discusses the results of the various analyses conducted in the study. These include descriptive analysis results, unit root results, multicollinearity test, heteroscedasticity test, autocorrelation test, normality test, and the Panel ARDL regression results.
Starting with the descriptive analysis, Table
The mean and median of the variables both measure the central tendency. The result from Table
Variable | Unit of Measurement | Observations | Mean | Median | Standard Deviation | Coefficient of Variation | Min | Max |
FDI | % of GDP | 470 | 3.67 | 1.71 | 8.83 | 5.16 | –11.64 | 103 |
FD | Domestic credit to private sector % of GDP | 449 | 14.56 | 12.31 | 11.47 | 0.93 | 0.4 | 65.74 |
GRGDP | Annual % | 468 | 4.07 | 4.38 | 4.81 | 1.10 | –30.15 | 26.42 |
RGDPPC | Constant 2010 US Dollars | 470 | 2561 | 2144 | 1369 | 0.64 | 931 | 7171 |
URBANPOP | % of the total population | 480 | 39.11 | 39.72 | 11.16 | 0.28 | 13.81 | 66.19 |
OPN | Total trade % of GDP | 468 | 64.91 | 58.76 | 31.07 | 0.53 | 20.72 | 311.35 |
INF | Annual | 427 | 7.41 | 4.36 | 10.96 | 2.51 | –7.8 | 72.84 |
INFRA | AIDI Index | 176 | 16.58 | 14.46 | 9.30 | 0.64 | 3.65 | 50.43 |
GOV | Institution Composite Index | 335 | –0.00 | –0.15 | 1.00 | –6.67 | –2.13 | 3.09 |
POL | An index ranging between 1 and 7 | 480 | 4.13 | 4.00 | 1.80 | 0.45 | 1 | 7 |
NAT | Total natural resources | 455 | 228 | 228 | 131 | 0.58 | 1 | 455 |
REGQ | An index ranging between –2.5 and +2.5 | 336 | –0.62 | –0.57 | 0.40 | –0.70 | –2.02 | 0.34 |
NOVIO | An index ranging between –2.5 and +2.5 | 336 | –0.51 | –0.35 | 0.82 | –2.34 | –2.44 | 1.22 |
GOVTEFF | An index ranging between –2.5 and +2.5 | 335 | –0.78 | –0.80 | 0.47 | –0.59 | –1.88 | 0.37 |
CORR | An index ranging between –2.5 and +2.5 | 336 | –0.61 | –0.69 | 0.52 | –0.75 | –1.7 | 1.14 |
ROL | An index ranging between –2.5 and +2.5 | 336 | –0.67 | –0.67 | 0.55 | –0.82 | –2.01 | 1.04 |
VAC | An index ranging between –2.5 and +2.5 | 336 | –0.39 | –0.37 | 0.60 | –1.62 | –1.55 | 1.00 |
A higher standard deviation value indicates a greater spread in the data. The standard deviation for RGDPPC is about 1369, which is the sole highest of all the variables in the study. The coefficient of variation is the standard deviation divided by the mean. The lower the value of the coefficient of variation, the less spread and less volatile are the data points. The coefficient of variation for FDI is 5.16, which is the highest of all coefficients for the variables covered in the study.
The minimum is the smallest data value, while the maximum is the largest data value. Comparing both minimum and maximum values for all variables in Table
As seen from Table
Stationary | T-Statistic | P-values | Order of Integration | Conclusion | |
FDI | At Level | –4.065 | 0.000 | I(0) | Stationary or I(0) |
FD | At Level | 1.408 | 0.920 | I(1) | Unit root I(1) |
At First Difference | –8.815 | 0.000 | I(0) | ||
GRGDP | At Level | –7.052 | 0.000 | I(0) | Stationary or I(0) |
RGDPPC | At Level | 5.295 | 1.000 | I(1) | Unit root I(1) |
At First Difference | –7.234 | 0.000 | I(0) | ||
URBANPOP | At Level | 6.576 | 1.000 | I(1) | Unit root I(1) |
At First Difference | –4.379 | 0.000 | I(0) | ||
OPN | At Level | –1.206 | 0.113 | I(1) | Unit root I(1) |
At First Difference | –11.027 | 0.000 | I(0) | ||
INF | At Level | –7.410 | 0.000 | I(0) | Stationary or I(0) |
INFRA | At Level | –2.520 | 0.005 | I(0) | Stationary or I(0) |
POL | At Level | –1.598 | 0.054 | I(0) | Stationary or I(0) |
NAT | At Level | –6.627 | 0.000 | I(0) | Stationary or I(0) |
The Kao co-integration test methodology is used to check for the long-run relationship of the dependent variables with their independent variables. The result of the test shows that the t-statistic value is -3.465 with a probability value of 0.0003, which is less than 0.05 significance level in Equation 1. Hence, the null hypothesis is rejected and it is concluded that there is a long-run relationship between the dependent and independent variables. This implies that the Panel ARDL method will be used to estimate both short-run and long-run relationships in Equation 1.
To present and analyze the estimates of Equations 1 to 8 concerning the determinants of FDI in West Africa, two tables of the estimates are first presented. This is followed by an evaluation of the diagnostic statistics and a discussion of the performance of each explanatory variable.
a) Statistic for the model explanatory power and R2 values in the test: The R2 is 0.297 in Equation 1. The R2 is 0.316, 0.298, 0.297, 0.316, 0.304, 0.299 and 0.298 in the Equations 2 to 8, respectively, and their respective F-statistic’s p-values are 0.000 in each case. Thus, these F-statistic values are statistically significant at the chosen 5% critical level. This also means that the models have fairly good fits.
b) Statistics for choosing the best estimator: From Table
Equation 1 (Long Run) | Equation 1 (Long Run) | |||||||||||||||||||
PMG | MG | DFE | PMG | MG | DFE | |||||||||||||||
Variables | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Variables | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | |
GRGDP | 0.045 | 2.03 | 0.042 | 0.341 | 1.46 | 0.145 | 0.271 | 2.45 | 0.014 | URBANPOP | –0.001 | –0.02 | 0.981 | –6.811 | –1.13 | 0.259 | –0.080 | –0.56 | 0.572 | |
OPN | 0.011 | 1.49 | 0.137 | –0.001 | –0.04 | 0.972 | –0.072 | –3.17 | 0.002 | FD | 0.001 | 0.05 | 0.958 | 2.841 | 1.03 | 0.305 | 0.068 | 1.03 | 0.301 | |
RGDPPC | 0.001 | 2.76 | 0.006 | –0.014 | –1.12 | 0.264 | 0.000 | 0.15 | 0.882 | POL | –0.498 | –2.42 | 0.015 | 18.086 | 0.92 | 0.358 | –1.320 | –1.01 | 0.310 | |
INF | –0.010 | –0.82 | 0.412 | –1.791 | –1.03 | 0.304 | 0.000 | 0.01 | 0.996 | NAT | –0.002 | –1.56 | 0.118 | 0.028 | 0.94 | 0.347 | –0.014 | –1.26 | 0.208 | |
INFRA | 0.019 | 1.91 | 0.056 | 0.116 | 4.25 | 0.000 | 0.077 | 1.10 | 0.270 | INFRA | 0.023 | 0.95 | 0.343 | 2.546 | 1.12 | 0.264 | 0.022 | 0.16 | 0.870 | |
Hausman (P-value) | 0.179 | 0.898 | Hausman (P-value) | 0.962 | 0.997 | |||||||||||||||
R2 | 0.296 | R2 | 0.296 | |||||||||||||||||
F(P-value) | 0.000 | F(P-value) | 0.000 | |||||||||||||||||
No of Countries | 16 | 16 | 16 | No of Countries | 16 | 16 | 16 | |||||||||||||
No of Observation | 693 | 693 | 693 | No of Observation | 693 | 693 | 693 |
Table |
|||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Equation 2 | Equation 3 | Equation 4 | Equation 5 | Equation 6 | Equation 7 | Equation 8 | |||||||||||||||
Variables | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value |
Equation 2 | Equation 3 | Equation 4 | Equation 5 | Equation 6 | Equation 7 | Equation 8 | |||||||||||||||
Variables | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value | Coefficient | Z-Statisticst | P-value |
FD | –0.315 | –2.26 | 0.024 | –0.268 | –1.92 | 0.055 | –0.258 | –1.86 | 0.062 | –0.315 | –2.26 | 0.024 | –0.256 | –1.86 | 0.063 | –0.311 | –2.24 | 0.027 | –0.271 | –1.93 | 0.053 |
GRGDP | 0.276 | 1.06 | 0.296 | 0.267 | 1.00 | 0.319 | 0.269 | 1.00 | 0.316 | 0.276 | 1.04 | 0.296 | 0.309 | 1.15 | 0.250 | 0.194 | 0.71 | 0.480 | 0.274 | 1.02 | 0.306 |
RGDPPC | –0.001 | –0.76 | 0.447 | –0.001 | –0.77 | 0.444 | –0.001 | –0.78 | 0.434 | –0.001 | –0.76 | 0.447 | –0.001 | –0.68 | 0.499 | –0.001 | –1.61 | 0.109 | –0.001 | –0.55 | 0.579 |
URBANPOP | 0.271 | 1.67 | 0.096 | 0.189 | 1.19 | 0.236 | 0.180 | 1.10 | 0.273 | 0.271 | 1.67 | 0.096 | 0.145 | 0.93 | 0.354 | 0.215 | 1.32 | 0.188 | 0.133 | 0.79 | 0.432 |
OPN | 0.231 | 3.78 | 0.000 | 0.256 | 4.21 | 0.000 | 0.252 | 4.11 | 0.000 | 0.231 | 3.78 | 0.000 | 0.240 | 3.90 | 0.000 | 0.276 | 4.44 | 0.000 | 0.255 | 4.19 | 0.000 |
INF | –0.096 | –0.38 | 0.702 | –0.147 | –0.58 | 0.565 | –1.166 | –0.65 | 0.513 | –0.096 | –0.38 | 0.702 | –0.214 | –0.85 | 0.397 | –0.348 | –1.31 | 0.194 | –0.209 | –0.81 | 0.419 |
INFRA | –0.482 | 1.74 | 0.082 | –0.260 | –0.98 | 0.325 | –0.210 | –0.84 | 0.400 | –0.481 | –1.74 | 0.082 | –0.108 | –0.43 | 0.665 | 0.093 | 0.35 | 0.730 | –0.115 | –0.41 | 0.679 |
POL | –0.748 | 0.88 | 0.377 | –1.429 | –1.63 | 0.103 | –1.495 | –1.01 | 0.311 | –0.748 | –0.88 | 0.377 | –2.100 | –2.78 | 0.006 | –2.708 | –3.17 | 0.002 | –1.978 | –2.41 | 0.016 |
NAT | 0.054 | 0.31 | 0.757 | 0.209 | 0.18 | 0.861 | 0.035 | 0.20 | 0.841 | 0.053 | 0.31 | 0.757 | 0.025 | 0.14 | 0.887 | 0.053 | 0.29 | 0.776 | 0.015 | 0.08 | 0.933 |
GOV | 4.344 | 1.99 | 0.047 | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – |
ROL | – | – | – | –2.459 | –0.56 | 0.576 | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – |
VAC | – | – | – | – | – | – | 0.821 | 0.18 | 0.859 | – | – | – | – | – | – | – | – | – | – | – | – |
CORR | – | – | – | – | – | – | – | – | – | 8.542 | 1.99 | 0.047 | – | – | – | – | – | – | – | – | – |
REGQ | – | – | – | – | – | – | – | – | – | – | – | – | –5.100 | –1.25 | 0.212 | – | – | – | – | – | – |
NOVIO | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | –4.163 | –1.92 | 0.057 | – | – | – |
GOVTEFF | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | – | –2.673 | –0.58 | 0.560 |
Hausman (P-Value) | – | – | 0.854 | – | – | 0.772 | – | – | 0.645 | – | – | 0.854 | – | – | 0.581 | – | – | 1.000 | – | – | 0.507 |
LM (P-Value) | – | – | 1.000 | – | – | 1.000 | – | – | 1.000 | – | – | 1.000 | – | – | 1.000 | – | – | 1.000 | – | – | 1.000 |
F-Wald (P-Value) | – | – | 0.000 | – | – | 0.000 | – | – | 0.000 | – | – | 0.000 | – | – | 0.000 | – | – | 0.000 | – | – | 0.000 |
Overall R-squared | 0.316 | – | – | 0.298 | – | – | 0.297 | – | – | 0.316 | – | – | 0.304 | – | – | 0.299 | – | – | 0.298 | – | – |
No of Countries | 16 | – | – | 16 | – | – | 16 | – | – | 16 | – | – | 16 | – | – | 16 | – | – | 16 | – | – |
No of Observation | 153 | – | – | 153 | – | – | 153 | – | – | 153 | – | – | 153 | – | – | 153 | – | – | 153 | – | – |
Concerning the test statistics for choosing between the pooled OLS, fixed and random effects methods of panel data estimation, the Hausman test results show that we do not reject the null hypothesis that RE is preferred to FE in Equations 2 to 8 because the p-values are greater than 0.05 level of significance in all cases, being 0.936, 0.851, 0.780, 0.897, 0.214, 0.317 and 0.539. Further testing using the Breusch-Pagan LM method confirms that Pooled OLS is more appropriate than either of Fixed Effects and Random Effects estimation methods in Equations 2 to 8 as the test reports a probability value of 1.000, which, in essence, leads to the rejection of the LM test and confirms pooled OLS as the most suitable method. Accordingly, the evaluation of the results carried out below is based only on the Pooled OLS result for Equation 2 to 8.
c) Multicollinearity test: A multicollinearity test is conducted using the Variance Inflation Factor (VIF) test, and based on the result, there is no high multicollinearity in all 8 Equations as the VIF of all variables is less than 10. Thus, the hypothesis of the absence of multicollinearity in each of the equations is accepted.
d) Heteroscedasticity test: For each of the equations, viz: Equations 1 to 8, White’s Test statistic for heteroscedasticity produces a p-value which is less than the chosen significance level at 0.05, except in Equation 8 where the p-value is higher than 0.05. This shows that the null hypothesis of constant variance is rejected in Equations 1 to 7. The results, therefore, indicate that there is heteroscedasticity in the residuals of Equations 1 to 7. To correct this in the affected seven equations, the standard errors are adjusted using White’s Heteroscedasticity-Corrected Variances and Standard Errors.
e) Test for non-normality of the distribution of the residuals: The Jacque Bera test statistic’s p-value is 0 in each of the models, viz: Equations 1 to 8, which means that the test statistics are significant at a 5% significance level. So, the study fails to reject the null hypothesis of normally distributed error terms, which leads to the conclusion that the residuals are normally distributed.
f) Autocorrelation test: A model is devoid of autocorrelation if the F-statistic of the Wooldridge autocorrelation test is higher than the one corresponding to a 5% level of significance. The reported P-value of the F-statistic is greater than the 0.05 critical significance level for each of Equations 1 to 3 and Equations 5 to 8, while it is less than 0.05 in Equation 4. Thus, the study rejects the null hypothesis of the absence of autocorrelation only in Equation 4 and concludes that there is autocorrelation there, since the probability value is less than 0.05. To correct for this observed autocorrelation, the robust fixed effect regression estimation method was used.
After evaluating the overall diagnostic statistics of the equation, we now proceed to examine the performance of each of the explanatory variables based on three ‘S’ — size, sign, and statistical significance.
a) Financial development (FD): In Equation 1, the coefficient of FD is 0.001 with a p-value of 0.958, while in Equations 2 to 8, the coefficients are -0.315, -0.268, -0.258, -0.315, -0.256, -0.311 and -0.271, respectively, with respective p-values of 0.024, 0.055, 0.062, 0.024, 0.063, 0.027 and 0.053, implying that the positive coefficient is statistically insignificant in the first equation and the negative coefficients are either statistically significant or very close to being statistically significant at the chosen 5% level in the last seven equations. Thus, on the whole, and since most of the coefficients are negative, it can be concluded that financial development in West Africa has a negative effect on FDI inflows. It is also contrary to the findings that are commonly reported in the empirical literature, including the study conducted by
b) Growth rate of GDP (GRGDP): In Equation 1, the coefficient of GRGDP is 0.045 with a p-value of 0.042, while in the Equations 2 to 8, the coefficients are 0.276, 0.267, 0.269, 0.276, 0.309, 0.194 and 0.274, respectively, with respective p-values of 0.296, 0.319, 0.316, 0.296, 0.250, 0.480 and 0.390, implying that the coefficients are positive and statistically significant in the first equation and statistically insignificant at the chosen 5% level in the last seven equations. It is therefore concluded that the GDP growth rate does not affect FDI inflows to the region. It is also contrary to the findings that are commonly reported in the empirical literature, including the study conducted by
c) Real GDP per capita (RGDPPC): In Equation 1, the coefficient of RGDPPC is 0.001 with a p-value of 0.006, while in Equations 2 to 8, the coefficients are –0.001 in each case, with respective p-values of 0.447, 0.444, 0.434, 0.447, 0.499, 0.100 and 0.579, implying that the positive coefficient is statistically significant in the first equation and the negative coefficients are statistically insignificant at the chosen 5% level in the last seven equations. It is therefore concluded that real GDP per capita does not affect FDI. This is also contrary to the findings reported by several previous studies, such as
d) Urban population (URBANPOP): In Equation 1, the coefficient of URBANPOP is -0.001 with a p-value of 0.981, while in the Equations 2 to 8, the coefficients are 0.271, 0.189, 0.180, 0.271, 0.145, 0.215 and 0.133 with respective p-values of 0.096, 0.236, 0.273, 0.096, 0.354, 0.188 and 0.432, implying that the negative coefficient is statistically insignificant in the first equation and the positive coefficients are statistically insignificant at the chosen 5% level in the last seven equations. It is therefore concluded that urban population does not affect FDI. This does not correspond to the findings of
e) Trade openness (OPN): In Equation 1, the coefficient of the OPN is 0.011 with a p-value of 0.137, while in Equations 2 to 8, the coefficients are 0.231, 0.256, 0.252, 0.231, 0.240, 0.276, and 0.255 with respective p-values of 0.000 in each case, implying that the positive coefficients is statistically insignificant in the first equation and statistically significant at the chosen 5% level in the last seven equations. It is therefore concluded that trade openness has a positive effect on FDI. It is also in line with the findings commonly reported in the empirical literature, including the study conducted by Neumayor and Spess (2005) and
f) Inflation (INF): In the Equation 1, the coefficient of the INF is -0.019 with a p-value of 0.056, while in the Equations 2 to 8, the coefficients are -0.096, -0.147, -1.166, -0.096, -0.214, -0.348 and -0.209, with respective p-values of 0.702, 0.565, 0.513, 0.702, 0.397, 0.194 and 0.419, implying that the negative coefficients are statistically insignificant at the chosen 5% level in all eight equations. It is therefore concluded that inflation does not affect FDI. It is also contrary to the findings that are commonly reported in the empirical literature, including the study conducted by De Mello (1997), among others. This can be explained by the same reason adduced in the previous Paragraph (a) that the effects of the combination of political and economic environments of these countries overwhelm and diminish other considerations, including financial development, in the eyes of the portfolio investors.
g) Infrastructure (INFRA): In Equation 1, the coefficients of the INFRA are 0.019 and 0.023 with p-values of 0.056 and 0.343, respectively, while in Equations 2 to 8, the coefficients are -0.482, -0.260, -0.210, -0.481, -0.108, 0.093 and -0.115 with respective p-values of 0.082, 0.325, 0.400, 0.082, 0.665, 0.730 and 0.679, implying that the positive coefficients are statistically insignificant in the first equation and the negative coefficients are statistically insignificant at the chosen 5% level in the last seven equations. Since all the coefficients of infrastructure are insignificant, it is therefore concluded that infrastructure does not affect FDI. This result also contradicts the evidence reported in several previous empirical studies, including
h) Political rights (POL): In the Equation 1, the coefficient of the POL is -0.498 with a p-value of 0.015, while in the Equations 2 to 8, the coefficients are -0.748, -1.429, -1.495, -0.748, -2.100, -2.708 and -1.978 with respective p-values of 0.377, 0.103, 0.311, 0.372, 0.006, 0.002, 0.016, implying that the coefficients are negative and either statistically significant (Equations 1 and 6 to 8) or statistically insignificant (Equations 2 to 5) at the chosen 5% level. It is therefore concluded that there is no robust evidence concerning the effect of this factor, since the evidence based on Equation 1 contradicts that based on the estimates of Equations 6 to 8. It is also contrary to the findings that are commonly reported in the empirical literature, including the study conducted by Dutta and Osei-Yeboah (2013) and
i) Natural resources (NAT): In Equation 1, the coefficient of the NAT is -0.002 with a p-value of 0.118, while in Equations 2 to 8, the coefficients are 0.058, 0.209, 0.035, 0.053, 0025, 0.053, and 0.015 with respective p-values of 0.757, 0.861, 0.841, 0.757, 0.887, 0.776 and 0.933, implying that the negative coefficient is statistically insignificant in the first equation and the positive coefficients are statistically insignificant at the chosen 5% level in the last seven equations. It is therefore concluded that natural resource does not affect FDI. This result also contradicts the evidence reported in several previous empirical studies, including
j) Governance indicators (GOV): In Equation 2, the coefficient of the GOV is 4.344 with a p-value of 0.047, implying that the coefficient is positive and statistically significant at the chosen 5% level. Since the coefficient of governance indicators is significant, it is therefore concluded that governance indicators have a positive effect on FDI.
k) Rule of law (ROL): In Equation 3, the coefficient of the ROL is -2.459 with a p-value of 0.576, implying that the coefficient is negative and statistically insignificant at the chosen 5% level. It is therefore concluded that the rule of law does not affect FDI.
l) Voice and accountability (VAC): In Equation 4, the coefficient of the VAC is 0.821 with a p-value of 0.859, implying that the coefficient is positive and statistically insignificant at the chosen 5% level. It is therefore concluded that voice and accountability do not affect FDI.
m) Control of corruption (CORR): In Equation 5, the coefficient of the CORR is 8.542 with a p-value of 0.047, implying that the coefficient is positive and statistically significant at the chosen 5% level. It is therefore concluded that control of corruption affects FDI.
n) Regulatory quality (REGQ): In Equation 6, the coefficient of the REGQ is -5.100 with a p-value of 0.212, implying that the coefficient is negative and statistically insignificant at the chosen 5% level. It is therefore concluded that regulatory quality does not affect FDI.
o) Absence of violence (NOVIO): In Equation 7, the coefficient of the NOVIO is -4.163 with a p-value of 0.057, implying that the coefficient is negative and statistically insignificant at the chosen 5% level. It is therefore concluded that the absence of violence does not affect FDI.
p) Government effectiveness (GOVTEFF): In Equation 8, the coefficient of the GOVTEFF is -2.673 with a p-value of 0.560, implying that the coefficient is negative and statistically insignificant at the chosen 5% level. It is therefore concluded that government effectiveness does not affect FDI.
Based on the above methodology, the main findings and conclusions relevant to each finding are as follows:
From the foregoing it can be concluded that the evaluation of the factors that determine foreign direct investment and influence globalization processes in West Africa did not yield all the expected results. It is revealed that financial development has a negative effect on FDI flows, while trade openness, governance indicators, as well as control of corruption, have a positive effect on FDI flows to the region.
Based on the findings of this study, as highlighted above, the following policy recommendations are made.
Based on the conclusion that FDI correlates negatively with financial development, West African nations should enhance the quality (including integration into global financial markets) of domestic financing systems to make their economies more attractive for MNCs to invest in them.
The positive effect of trade openness on FDI shows that West African countries should vigorously pursue trade liberalization policy as a potent and deliberate effort to attract FDI inflows, albeit in a way that does not interfere with the development of the domestic economy.
Authorities should also boost high-quality anti-corruption mechanisms to accelerate the globalization process through inbound FDI due to the positive effect of control of corruption on FDI, as well as the composite governance institution index.