Do FDI Inflows into African Countries Impact Their CO₂ Emission Levels?

Abstract

The emitted levels of CO₂ continue to be a striking topic. These emissions have been growing over the years, thus, making them a predicament to be reckoned with. Eradicating such a predicament has not been easy because finding an optimal determinant has not been achieved by scholars; however, foreign direct investment inflows are known to play a role in such varying instances. Therefore, to analyze the impact that such inflows have on CO₂ emissions, this study employs data from 41 African countries from 2005 to 2019 and aims to assess how foreign direct investment and other variables influence CO₂ emitted levels. Moreover, this study tests the validity of the pollution haven and halo hypotheses on the employed African countries as its two main objectives. After applying the pooled least squares, fixed and random effects models, and the generalized method of moments, the findings revealed that per the adopted African countries, the pollution haven and halo hypotheses do not hold; however, foreign direct investment inflows contribute to the rising and falling levels of CO₂ emissions. In addition, the financial structure and per capita GDP increase the African countries’ CO₂ emitted levels, while trade openness causes a reduction. Based on the aforementioned findings, this study recommends that the government, policy-makers, industries, and interested personnel of this study’s employed countries should: apply and execute policies, laws, and regulations that will deter or punish polluting foreign investment and encourage clean ones; since green finance is making waves but is not well established in most African countries, green financing systems should be initiated and implemented; establish preferential trading policies that will highlight an addition of value via clean technology; and practice carbon capture, usage, and storage.

1. Introduction

As the world evolves into a technologically globalized setting, new predicaments have materialized, requiring the use of finance and its factors to analyze and contemplate solutions [1]. Recent events have showcased climate change, specifically global warming, as a predicament to be reckoned with [2]. This is because certain human activities increase the atmospheric concentrations of greenhouse gases (the most frequent greenhouse gas associated with global warming is carbon dioxide (CO₂)), gradually increasing Earth’s temperature and bringing about global warming.

It is undoubtedly true that CO₂ emissions have been growing these past few years, which is attributable to various causes and reasons [3] to enhance people’s quality of life. Countries, organizations, etc., have prioritized the need for reducing these emissions, but to fully understand and eradicate the predicament (i.e., CO₂ emissions), the causes must be known. Since a sole sector or factor has not been singled out over the years as the contributor to such a phenomenon, researchers have taken it upon themselves to research and discover its various determinants. Due to its linkage to the environment, most existing works in the literature explore its correlation with environmental factors. However, it has been realized that financial factors and activities correlate with this phenomenon.

Over the years, some scholars employed foreign direct investment (FDI), financial development, etc., to explore their correlation with carbon emissions. For instance, Seker et al. [4] looked into how Turkey’s FDI and other factors affected CO₂ emissions from 1974 to 2010. Analyzing data from 17 Asian countries, Khan and Ozturk [5] explored the causal relationship between CO₂ emissions and FDI and a few other variables. Jiang and Ma [6] explored the correlation between financial development and CO₂ emissions for 155 countries, while Wen et al. [7] analyzed financial structures’ impact on emission reduction using data from China’s provinces. Due to the dissimilar factors, data, and the methods employed, the conclusions drawn have been mixed.

Nevertheless, FDI has proven to have a significant impact on CO₂ emissions. Due to their lax environmental regulations, abundant natural resources, and cheap labor, amongst other factors, African countries are known to be attractive for FDIs. It is commonly known that industrialized businesses from developed countries strive to construct facilities in these (African) countries. As a result, more significant FDI can result in higher carbon emission levels in such countries, thus, introducing the infamous Pollution Haven Hypothesis [8]. However, some scholars refute the notion behind the Pollution Haven Hypothesis and conclude that more significant FDI can result in lower carbon emission levels, thus, introducing the opposing infamous Pollution Halo Hypothesis [9].

Do FDI inflows in African countries possess good attributes that contribute to reducing CO₂ emissions? Contrarily, do these inflows pose a threat and cause a rise in their CO₂ emissions? These questions have puzzled some financial, research, and policy-making people.

When industrialized facilities are established in African countries through FDI, they provide benefits, such as clean technologies, but sometimes, the cost outweighs the benefits. These facilities release chemicals into the air, thus, polluting and degrading the environment of the host countries. The emissions cause airborne diseases and increase the atmospheric temperature, which contributes to the rise in the world’s temperature, leads to global warming, and affects the health of inhabitants of such countries. Consequently, this affects the government and policy-makers because the time and resources geared towards the countries’ development must be rerouted to combat the rise in CO₂ emission levels. This slows down the countries’ development, affects the health of the inhabitants, depletes the environment, and distracts the countries from achieving their set aims within their apportioned times.

Governments, policy-makers, researchers, etc., attach much importance to CO₂ emissions due to their intricate nature and effects; many steps have been taken to reduce such a predicament over the years. For instance, the Kyoto Protocol was approved and became efficacious on 11 December 1997 and 16 February 2005, respectively [10]; the Paris Agreement was approved and became efficacious on 12 December 2015 and 4 November 2016, respectively [11]. Other examples include the EU Emissions Trading Scheme [12]; green investments [13]; etc. Although these prevention measures have been implemented previously, with some still in effect, curbing the continuous rise in CO₂ emissions is still not achieved, thus, raising the need for further studies.

This research assesses Foreign Direct Investment and other variables influencing carbon dioxide emissions as its main aim. By adopting such variables, this study analyzes the correlation between the selected variables. Additionally, the objectives operationalized from the main aim of this study are (1) to check for the existence of the pollution haven hypothesis, (2) to check for the existence of the pollution halo hypothesis, and (3) to provide researchers, governments, policy-makers, and interested personnel with an empirical conclusion. Moreover, this study’s results offer insight into how finance (represented by FDI, financial structure, and development in this study) plays a role in the context of carbon dioxide emissions. It adds to the corpus of the existing literature by assessing data from African countries and fusing previously used methodologies and variables into a unified paradigm. The findings of this study reveal the determinants or factors that contribute to the growth and reduction in CO₂ emissions in African countries. In terms of significance, this study and its outcome will theoretically serve as a precedent for future studies and, practicality wise, provide the adopted African countries’ governments, policy-makers, and individuals with empirical evidence and suggestions to aid in improvements to the environment.

With the introduction and background of this study already tackled in Section 1, the rest of this study is structured as follows: Section 2—Literature Review; Section 3—Study Settings; Section 4—Results and Discussion; Section 5—Conclusion and Suggestions.

2. Literature Review

This section elucidates the earlier literature relevant to the current study. Most developed countries, such as Switzerland and the US, are reputable for being industrialized. As time proceeds in the early stages, their economic and financial progress begins to rise together with the environmental deterioration caused by pollution; they rely on traditional forms of energy at those stages. However, they reach a particular turning point, in which their progress corresponds to environmental improvement; they turn to clean ways of operating and renewable energy to reduce their ecological footprint [14,15], thus, generating the underlying idea behind the Environmental Kuznets Curve [16]. After reaching the turning point, some of these industrialized countries export and establish their industrial facilities in the form of FDIs in developing countries, since they (developing countries) have a vast array of natural resources, relaxed environmental laws, plus lower labor costs. As such, these developed countries reduce their pollution levels and improve their environment while the developing countries bear the cost. However, is that always the case (see Section 2.1)?

2.1. Pollution Haven (PHVHP) and Halo (PHLHP) Hypotheses

The pollution haven hypothesis propounds that, as countries advance and develop, they adopt stricter environmental regulations than countries in the developing phase. Consequently, this distorts the existing comparative advantage patterns. As a result, the polluting industries relocate their facilities and activities from developed to developing countries, turning the latter into pollution havens [17,18]. On the other hand, the pollution halo hypothesis asserts that international corporations convey their greener technologies to developing countries through FDI. Green innovations, such as those that reduce pollution or use renewable energy sources and highly energy-efficient ones, lessen the need for traditional energy sources [18,19].

Since the introduction of these hypotheses, numerous studies have been undertaken to determine their viability, some involving two or more countries, while others were country specific. Despite trending areas of research, the conclusions drawn concerning these hypotheses have been dissimilar. Due to brevity,

Table 1. Succinct works of literature on the PHVPH and PHLHP.

Pollution Haven Hypothesis (PHVPH)
Author(s) Names (Year(s))	Chosen Period(s)	Country of Focus	Methodology	Does It Hold?
Al-mulali & Foon Tang (2013) [20]	1980–2009	Gulf Cooperation Council countries	FMOLS	No
Ren et al. (2014) [21]	2000–2010	China	GMM	Yes
Shahbaz et al. (2015) [22]	1975–2012	High, middle, and low-income countries.	FMOLS	Yes
Riti et al. (2016) [23]	1980–2013	Nigeria	ARDL	Yes
Rasit & Aralas (2017) [24]	2000–2010	ASEAN & OECD	Pooled Ordinary Least Square	No
Kathuria (2018) [25]	2002–2010	India	Industry-adjusted abatement expenditure index	No
Shao et al. (2019) [26]	1982–2014	BRICS & MINT	VECM	No
Guzel & Okumus (2020) [27]	1981–2014	ASEAN-5	CCEMG and AMG	Yes
Singhania & Saini (2021) [28]	1990–2016	21 developed and developing countries	GMM and Sym-GMM	Yes
Musah et al. (2022) [29]	1992–2018	G20 countries	DCCEMG	Yes
Pollution Halo Hypothesis (PHLPH)
Elliott & Zhou (2012) [30]	N/A	N/A	Game-theoretic model	Yes
Yildirim (2014) [31]	1980–2009	76 countries	Bootstrap-corrected panel causality test, Cross-correlation analysis	Mixed results
Zugravu-Soilit (2015) [32]	1995–2008	France, Germany, Sweden, United Kingdom	Empirical model	No
Zhang & Zhou (2016) [33]	1995–2010	China	STIRPAT	Yes
Aydemir & Zeren (2017) [34]	1970–2010	G20 countries	Panel data analysis	Mixed results
Zhou et al. (2018) [35]	2003–2015	China	STIRPAT, Diff-GMM, Sys-GMM	Mixed results
Nasir et al. (2019) [36]	1982–2014	ASEAN-5	DOLS	No
Xie et al. (2020) [37]	2005–2014	Argentina, Brazil, China, India, Russia, South Korea, Mexico, Turkey, Indonesia, South Africa, Saudi Arabia	PSTR	Yes
Opoku et al. (2021) [38]	1995–2014	22 countries	GMM	Yes
Caetano et al. (2022) [39]	2005–2018	OECD	ARDL	Yes

“Does it hold?”—either the PHVHP or the PHLHP holds in such conducted studies; FMOLS—Fully Modified Ordinary Least Square; ARDL—Auto regression distributed lag model; GMM—Generalized Method of Moments; VECM—Vector Error Correction Model; CCEMG—Common Correlated Effects Mean Group; AMG—Augmented Mean Group; Sys-GMM—System-Generalized Methods of Moments; DCCEMG—Dynamic Common Correlated Effects Mean Group; ASEAN—Association of South East Asian Nations; OECD—Organisation for Economic Co-operation and Development; BRICS—Brazil, Russia, India, China, and South Africa; MINT—Mexico, Indonesia, Nigeria, and Turkey; G20—The Group of Twenty; STIRPAT —The Stochastic Impacts by Regression on Population, Affluence, and Technology; PSTR—Panel Smooth Transition Regression; DOLS—Dynamic Ordinary Least Squares; Diff-GMM—Different-Generalized Methods of Moments.

summarizes some of the works of literature conducted on these two hypotheses.

2.2. CO₂ Emissions

Carbon dioxide (CO₂) is assuredly one of the most emitted greenhouse gases worldwide. In 2021, it recorded the highest figure ever, with 36.3 billion tonnes, a 6% increase from the previous year [40]. Such an increase affects not only the economic setup but also the financial and, obviously, the environmental setup of countries. The top emitting countries have unquestionably been implementing policies and actions to reduce emissions for the betterment of the environment, country, and the world. An instance is China, a developing country and the largest emitter in the world, setting objectives, such as the dual carbon national target, which by 2030, calls for a peak in carbon emissions and to become carbon neutral by 2060 [41]. With this technique, CO₂ emissions are significantly reduced through market forces.

For over a decade, researchers have been fixated on CO₂ emissions due to the necessity to develop strategies and policies to combat global warming. During this time, a substantial number of studies have developed, many of which have been allotted to examine the determinants. Using the STIRPAT model on data from 2004 to 2013 from China, Zhang and Xu tested and identified a significant effect of land urbanization and finance on carbon emissions [42]. To compare the alikeness and variations in the factors that affect CO₂ emissions, Ouyang and Lin conducted comparison research between China and Japan, at various phases of urbanization. Their findings demonstrated that although CO₂ emissions in China and Japan exhibited stiff development during urbanization processes, there is still substantial variation in variables, such as CO₂ emissions per capita [43]. Hafeez et al., employing data from 1980 to 2016, explored the implications of finance on CO₂ emissions of 52 participants in the OBORI. They applied the FMOLS and DOLS models and uncovered that finance and CO₂ emissions share bidirectional causality in the short run, and finance significantly accelerates environmental deterioration based on their long-run estimates [44].

The steady increase in CO₂ emissions continues to make it a popular research topic due to its impact on the world. As a result, this study purports to establish how foreign direct investment inflows and other factors affect their rise or decline with the study’s chosen dataset.

2.3. FDI vs. CO₂ Emissions

The CO₂ emissions–FDI nexus has been empirically tested by several studies. Still, like many studies, the conclusions drawn have been dissimilar, attributable to different aims, data spans, variables, etc. For instance, by analyzing a panel of data comprising twenty-five provinces in China with a spatial econometric model, Tang et al. studied the variation in the influence of FDI on carbon productivity under various entry routes from 2007 to 2019. Their research concluded that a positive spatial spillover effect is amenable to boosting China’s carbon productivity when FDI enters China through joint ventures. Still, there is a negative spatial spillover effect when FDI moves into China through wholly foreign-owned enterprises, which will impede the improvement in China’s carbon productivity [45]. Kim used VECM to analyze the causal linkages between CO₂ emissions, FDI, and other variables using data from fifty-seven developing nations between 1980 and 2013 and discovered that FDI does not directly cause CO₂ emissions in the short run [46]. Abban et al. evaluated data for BRI nations spanning 20 years and discovered a bidirectional causal link between FDI and CO₂ [47]. Islam et al. studied FDI’s impact on CO₂ emissions in Bangladesh over 44 years while institutional quality was present, revealing that FDI has a negative impact on CO₂ emissions [48]. Although dissimilar, these studies provided the theoretical framework for continuous research, including the current study.

2.4. Other Financial Variables vs. CO₂ Emissions

Finance plays a role in the emissions of CO₂, but with the varieties of finance, extensive study is required to find the other forms or acts that impact CO₂ emissions. In earlier times, most research about environmental degradation (referring to CO₂ emissions) had FDI as the sole financial determinant but began to incorporate others over time, such as financial development and structure, posited to impact carbon emissions significantly [6,7]. To answer the question, “Does financial structure affect CO₂ emissions?”, Yao and Tang employed the G20′s data from 1971 to 2014, applied the STIRPAT method, and uncovered that, based on their analysis, there is significant heterogeneity in the effects of financial structure on carbon dioxide emissions. Moreover, there exists a negative correlation between financial structure and CO₂ emissions in developed countries [49].

Funding via an optimal capital structure (thus, finance activities utilizing either more debt or equity) has always been an issue for countries’ governments and management in companies. In retrospect, for countries, [50] found that there would be an 11.5% reduction in global per capita carbon emissions if all nations’ financial structures were converted to at least 50% equity financing, implying that a good financial structure in the form of equity financing will propel a country’s financial and environmental development. Empirically testing the causal nexus between CO₂ emissions, financial development, and other indicators, Jamel and Maktouf employed the Cobb–Douglas production function and discovered that the neutrality hypothesis connects carbon emissions and financial sector development inflows in their study [51]. One way or another, [52,53] employed financial development or financial structure as their explanatory variables to explore the correlation effect they have on CO₂ emissions. Therefore, this study adopts them as explanatory variables to analyze how other financial variables aside from FDI might play a role in the rise or fall of CO₂ emissions.

Some works of literature have been published examining the relationship between CO₂ emissions and FDI inflows, financial structure, or financial development. However, very few researchers have attempted to link the three in a unified framework. Moreover, there are not enough works that utilized, if not all, almost all African countries as a sample in such a context; most are country specific. Further, some scholars have tested the validity of the pollution halo and haven hypotheses, but the outcomes have been quite dissimilar. Therefore, this study attempts to fill these literary gaps by combining previously adopted models to analyze the relationship between CO₂ emissions and FDI inflows, financial structure, and financial development. Furthermore, this study tests for the validity of both the pollution halo and haven hypotheses by adopting African countries as the study sample. Overall, a unified framework is established to address diverse literary gaps and provide an empirical conclusion to governments and policy-makers to aid in the protection of the environment.

3. Study Settings

The current research adopts data from 41 African countries from 2005 to 2019 as its sample. This research assesses Foreign Direct Investment (FDI) and other variables influencing CO₂ emissions as its primary aim. CO₂ emission per capita, denoted as CEM, is employed as the response variable. The data are sourced from the website of Our World in Data [54]. FDI annual inflows, denoted as FDI, are chosen as the core explanatory variable. In contrast, financial development and financial structure, denoted as FND and FNS, respectively, are chosen as the other explanatory variables.

Contrary to the findings from [50] of an equity-based financial structure being beneficial, this research analyzes the effect a debt-based financial structure will have on CO₂ emissions. Thus, the financial structure in this research is estimated as the private sector’s credit extended to them from banks to analyze how the ease of borrowing provided by the financial systems (i.e., banks) can contribute to CO₂ emission levels. The data for the annual FDI inflows and financial structure of the chosen countries were sourced from the World Bank’s World Development Indicators (WDI) database [55].

Financial development entails various factors due to its multifaceted process, which is why it has been expressed in diverse forms in different literary works. Due to its complex yet expansive coverage nature, this research applies the financial development index proposed by Katsiaryna Svirydzenka [56]. This index concludes by summarizing the “depth, access, and efficiency” of financial institutions and markets (see Figure 1) to address the limitations of single measures as substitutes for financial development. The data of this index are sourced from the IMF Financial Development Index Database [57].

Moreover, to avoid white noise caused by unaccounted factors, provide reasonable model estimates, and follow the works of [37,49,58,59], control variables, namely trade openness, population growth, urbanization, industry value added, and gross domestic product per capita, denoted as TRDO, POPGT, URBNZ, INDVA, and GDPPC, respectively, were employed. The data for the control variables were sourced from the World Bank’s World Development Indicators (WDI) database [55]. A visual representation of the employed variables is depicted in Figure 2.

3.1. Data Normalization

Data used for research sometimes appear sporadic, and this research’s data are no exception. Due to the wide range of figures and different data characteristics, they were normalized to transform into a standard format. As a result, normalizing the data will eradicate the vast differences in the numerical relationships amongst the variables, allowing different types of variables to be evaluated and the data to fall within a narrower specific interval. n number of samples are firstly selected for m number of chosen variables; then, the data are normalized using Equations (1) and (2). Equation (1) is used if the selected variable negatively impacts the CEM; otherwise, Equation (2) is used.

$$X_{it}=\left(max\left\{x_t\right\}-x_{it}\right)÷ \left(max\left\{x_t\right\}-min\left\{x_t\right\}\right), \mathrm{i} = 1,2,3,\dots \mathrm{n}; \mathrm{t} = 1,2,3,\dots \mathrm{m}$$

(1)

$$X_{it}=\left(x_{it}-min\left\{x_t\right\}\right)÷\left(max\left\{x_t\right\}-min\left\{x_t\right\}\right), \mathrm{i} = 1,2,3,\dots \mathrm{n}; \mathrm{t} = 1,2,3,\dots \mathrm{m}$$

(2)

Here, $X_{it}$ and $x_{it}$ connotes the normalized and actual figures for a variable with n in all the years; $max\left\{x_t\right\}$ and $min\left\{x_t\right\}$ connote the maximum and minimum figures of each n in all the years for the variable.

3.2. Study Hypotheses

Howbeit, FDI inflows might harm or protect the host country’s environment by contributing to a rise or fall in CO₂ emissions. Therefore, the analyses carried out in this study are in two folds: one for the haven analysis to test and validate the existence of the PHVHP and the other for the halo analysis to test and validate the existence of the PHLHP. Furthermore, this study analyzes the effect finance (represented by a debt-based financial structure and financial development) has on CO₂ emissions. Therefore, this study puts forth the hypotheses shown in Figure 3.

4. Results and Discussion

The discoveries made during the analysis of the data acquired for this research are highlighted and discussed in this section. Due to their erratic character, the initial data are first normalized for the estimations before being used.

4.1. Descriptive Statistics and Correlation Analysis

A descriptive analysis was performed before the diagnostic analytics to assess changes in the variables’ patterns. The total data set from 2005 to 2019 includes representation from 41 countries (refer to

Table 2. Descriptive analysis of employed variables.

	Mean	Std. Dev.	Skewness	Kurtosis
CEM	0.490	0.320	0.078	1.806
FDI	0.415	0.309	0.429	2.062
FDI1	0.585	0.309	−0.429	2.062
FND	0.540	0.318	−0.227	1.830
FNS	0.512	0.327	−0.053	1.683
POPGT	0.551	0.344	−0.217	1.633
TRDO	0.524	0.312	−0.050	1.843
URBNZ	0.473	0.321	0.089	1.716
INDVA	0.502	0.320	−0.041	1.783
GDPPC	0.540	0.315	−0.218	1.843
Observations	615

). The results of the descriptive analyses’ chosen response, explanatory, and control variables are concentrated on their means, standard deviations, skewness, and kurtosis, as well as 615 observations and 41 cross-sections. Equations 1 and 2 from Section 3.1 were used to normalize the core explanatory variable to determine the validity of the pollution haven and halo hypotheses, accounting for two FDI values being used: FDI and FDI1.

The standard deviation values show that the employed data are fairly clustered around their means. Furthermore, CEM, FDI, and URBNZ have positive skewness outcomes, insinuating a significant number of their distributions lies on the left side of the normal curve, whilst the remaining variables have a significant number of their distributions on the right side of the normal curve. Further, all the variables’ kurtosis values are lower than 3, indicating they have a platykurtic distribution. Moreover, a correlation analysis was conducted, and per the results in

Table 3. Correlation analysis of employed variables.

	CEM	FDI	FDI1	FND	FNS	POPGT	TRDO	URBNZ	INDVA	GDPPC
CEM	1.000
FDI	−0.016	1.000
FDI1	0.016	−1.000	1.000
FND	0.389	0.058	−0.058	1.000
FNS	0.428	0.108	−0.108	0.574	1.000
POPGT	0.020	−0.012	0.012	−0.116	−0.034	1.000
TRDO	−0.105	−0.232	0.232	−0.090	−0.059	−0.116	1.000
URBNZ	0.484	−0.117	0.117	0.564	0.563	−0.048	0.030	1.000
INDVA	−0.009	0.089	−0.089	−0.171	−0.158	0.159	−0.126	−0.172	1.000
GDPPC	0.474	−0.012	0.012	0.361	0.378	0.072	−0.027	0.433	0.081	1.000

, all the variables strongly correlated.

4.2. Diagnostic Statistics

4.2.1. Unit Root Test

Due to the regression analysis carried out in this research to assess the relationship between CEM and the remaining variables, unit root tests were carried out to avoid spurious regression estimates. The LLC and Fisher-type (ADF and PP) tests were applied (see

Table 4. Unit root tests of employed variables.

	Level			1st Difference
	LLC	ADF	PP	LLC	ADF	PP
CEM	0.008	0.714	0.325	0.000	0.000	0.000
FDI	0.000	0.000	0.000	0.000	0.000	0.000
FDI1	0.000	0.000	0.000	0.000	0.000	0.000
FND	0.000	0.001	0.000	0.000	0.000	0.000
FNS	0.000	0.000	0.000	0.000	0.000	0.000
POPGT	0.000	0.000	0.000	0.000	0.000	0.000
TRDO	0.000	0.073	0.000	0.000	0.000	0.000
URBNZ	0.000	0.000	0.000	0.398	0.000	0.000
INDVA	0.000	0.184	0.334	0.000	0.000	0.000
GDPPC	0.000	0.000	0.007	0.000	0.000	0.000

). At that level, most of the variables were stationary under the three tests, but not all. The variables were integrated to order one and retested; all but the urbanization variable under the LLC were stationary.

4.2.2. Regression Estimates

The pooled least squares (referred to as Pooled in

Table 5. Regression analyses.

	Haven			Halo
	Pooled	Fixed	Random	Pooled	Fixed	Random
	CEM
C	0.252 ***	0.391 ***	0.403 ***	0.238 ***	0.326 ***	0.365 ***
	(0.054)	(0.097)	(0.094)	(0.056)	(0.097)	(0.094)
FDI	−0.014	−0.065 *	−0.038	0.014	0.065 *	0.038
	(0.035)	(0.037)	(0.035)	(0.035)	(0.037)	(0.035)
FND	0.005	0.012	0.008	0.005	0.012	0.008
	(0.044)	(0.047)	(0.044)	(0.044)	(0.047)	(0.044)
FNS	0.133 ***	0.187 ***	0.156 ***	0.133 ***	0.187 ***	0.156 ***
	(0.043)	(0.045)	(0.043)	(0.043)	(0.045)	(0.043)
POPGT	−0.005	0.009	0.002	−0.005	0.009	0.002
	(0.031)	(0.032)	(0.031)	(0.031)	(0.032)	(0.031)
TRDO	−0.190 ***	−0.207 ***	−0.198 ***	−0.190 ***	−0.207 ***	−0.198 ***
	(0.036)	(0.036)	(0.035)	(0.036)	(0.036)	(0.035)
URBNZ	0.024	−0.026	0.003	0.024	−0.026	0.003
	(0.061)	(0.067)	(0.063)	(0.061)	(0.067)	(0.063)
INDVA	0.067 **	0.067 *	0.067 **	0.067 **	0.067 *	0.067 **
	(0.034)	(0.034)	(0.034)	(0.034)	(0.034)	(0.034)
GDPPC	0.364 ***	0.433 ***	0.395 ***	0.364 ***	0.433 ***	0.395 ***
	(0.049)	(0.052)	(0.050)	(0.049)	(0.052)	(0.050)
R2	0.421	0.416	0.420	0.421	0.416	0.420
Prob > F	0.000	0.000		0.000	0.000
Prob > chi2			0.000			0.000
Year dummies	Yes	Yes	Yes	Yes	Yes	Yes
Samples	2005–2019
Cross-sections	41
Observations	615

Standard errors in parenthesis; *, **, *** means significant at 10%, 5%, and 1%, respectively.

and PLS hereinafter), fixed effects (referred to as Fixed in Table 5 and FE hereinafter), and random effects (referred to as Random in Table 5 and RE hereinafter) models were conducted utilizing the STATA 17MP software(China)’s in-built functions. The two main objectives of this research are to validate the pollution haven and halo hypotheses. As such, the analyses conducted are in two folds, one for the haven hypothesis and the other for the halo hypothesis. Year dummy variables were introduced in the analyses to control for the year effects.

The outcome for the haven and halo analyses (refer to Table 5) depicts that the constant coefficient for the PLS, FE, and RE are all significant at the 1% significance level. Under the haven analysis, FDI negatively correlates with CEM, while it correlates positively with CEM under the halo analysis under the three models. However, only its coefficient in the FE model is significant (at the 10% level). Under the FE estimations, on an average, ceteris paribus, a 1% increase in FDI inflows will result in a 0.00065 decrease or increase in CO₂ emission levels. Pertaining to the other explanatory variables for the three models’ estimations under the two analyses, FND is insignificant for all, whereas FNS positively correlates with CEM under the 1% significance level.

Moreover, the control variables with significant coefficients are TRDO, INDVA, and GDPPC. TRDO negatively correlates with CEM, and INDVA and GDPPC positively correlate with CEM. The three models under both analyses are deemed a good fit based on the Prob > F and the Prob > chi2; overall, the variables matter jointly in explaining the models. The $R^2$ shows that the data for the employed explanatory and control variables explain approximately 42% of the sample variation in the CO₂ emissions data.

4.3. GMM Estimates

The Hausman test was conducted, and its outcome implied that the random effects model fit well overall; as such, this research presents Equation 3 as a random walk model with a persistent $CEM$ and applies the one-step system GMM (SYM-GMM) estimator to test further how dynamic the panels adopted are.

$$\begin{gathered} CE{M}_{it}={\beta }_i+\vartheta CE{M}_{it-1}+{\theta }_{1}FD{I}_{it}+{\theta }_{2}FN{D}_{it}+{\theta }_{3}FN{S}_{it}+{\theta }_{4}POPG{T}_{it}+{\theta }_{5}TRD{O}_{it}+{\theta }_{6}URBN{Z}_{it} \\ +{\theta }_{7}INDV{A}_{it}+{\theta }_{8}GDPP{C}_{it}+{\mu }_i+{\epsilon }_{it}+{\delta }_{it} \end{gathered}$$

(3)

where $CE{M}_{it}$ is CO₂ emissions; $\vartheta CE{M}_{it-1}$ is lagged order of one of the CO₂ emissions; ${\theta }_{1}FD{I}_{it}$, ${\theta }_{2}FN{D}_{it}$, and ${\theta }_{3}FN{S}_{it}$ are FDI inflows, financial development, and financial structure; ${\theta }_{4}POPG{T}_{it}$, ${\theta }_{5}TRD{O}_{it}$, ${\theta }_{6}URBN{Z}_{it}$, ${\theta }_{7}INDV{A}_{it}$, and ${\theta }_{8}GDPP{C}_{it}$ are population growth, trade openness, urbanisation, industry value added, and GDP per capita; ${\mu }_i$, ${\epsilon }_{it}$, and ${\delta }_{it}$ are the error term, country, and year effects.

Year dummy variables, 574 observations, 41 groups, and 36 instruments (per the STATA 17MP outcome produced during the simulations) were utilized (refer to

Table 6. SYM-GMM with all variables.

	Haven		Halo
CEM	Coefficient	Robust Std. Err.	Coefficient	Robust Std. Err.
C	0.129	0.100	0.073	0.099
CEMi,t−1	0.623 ***	0.047	0.623 ***	0.047
FDI	−0.056 **	0.023	0.056 **	0.023
FND	−0.028	0.041	−0.028	0.041
FNS	0.094 *	0.051	0.094 *	0.051
POPGT	0.003	0.034	0.003	0.034
TRDO	−0.149 ***	0.028	−0.149 ***	0.028
URBNZ	0.053	0.063	0.053	0.063
INDVA	0.035	0.033	0.035	0.033
GDPPC	0.207 ***	0.035	0.207 ***	0.035
No. of obs		574
No. of groups		41
No. of Instruments		36
F-stat		1019.40
Prob > F		0.000
Year dummies		Yes
AR(2)		0.396
Hansen test		0.103

*, **, *** means significant at 10%, 5%, and 1%, respectively.

). The model’s Prob > F is 0.000, indicating that both the overall model and the regression estimates are valid. The random disturbance term in the model does not exhibit a second-order autocorrelation (AR(2) > 0.05), and the model’s effect is good (Hansen test > 0.05).

Under the haven and halo analyses: (1) CEM_i,t−1 is significant at the 1% level with a coefficient of 0.623, denoting that the adopted countries in this research are affected by their lagged order of one pertaining to significant CO₂ emissions, (2) FDI negatively and positively affects CEM with a significant coefficient value of 0.056, (3) FND negatively affects CEM, but its coefficient is insignificant, and FNS with a coefficient of 0.094 significantly affects CEM positively at the 10% significance level, (4) POPGT, URBNZ, and INDVA positively affect the response variable, but their coefficients are insignificant, (5) TROD and GDPPC are the two control variables with significant coefficients, but while TRDO negatively affects CEM, GDPPC positively affects CEM. On the one hand, on an average ceteris paribus, a 1% increase in FDI, FNS, and TRDO will result in 0.00056, 0.00094, and 0.00149 decreases in CEM for the employed African countries in this research, respectively. On the other hand, with all things being equal, a 1% increase in FDI and GDPPC will result in a 0.00056 and 0.00207 increase in CEM.

The significant variables from the GMM estimates from Table 6 were extracted to construct a second one-step system GMM estimate. Per the outcome (refer to

Table 7. SYM-GMM with significant variables.

	Haven		Halo
CEM	Coefficient	Robust Std. Err.	Coefficient	Robust Std. Err.
C	0.177 ***	0.033	0.119 ***	0.032
CEMi,t−1	0.624 ***	0.049	0.624 ***	0.049
FDI	−0.058 **	0.023	0.058 **	0.023
FNS	0.089 *	0.049	0.089 *	0.049
TRDO	−0.151 ***	0.027	−0.151 ***	0.027
GDPPC	0.213 ***	0.037	0.213 ***	0.037
No. of obs		574
No. of groups		41
No. of Instruments		34
F-stat		819.49
Prob > F		0.000
Year dummies		Yes
AR(2)		0.391
Hansen test		0.157

*, **, *** means significant at 10%, 5%, and 1%, respectively.

), all the variables are significant, and the overall model fits, with an F-stat value of 819.49 and a Prob > F of 0.000. The model’s effect is good since it does not contain a second-order autocorrelation per the AR(2) value, and the Hansen test is greater than 0.05. Year dummy variables were applied, and the number of observations, groups, and instruments exploited was 574, 41, and 34 (per the STATA 17MP outcome produced during the simulations), respectively.

4.4. Discussion

This research discovered that FDI inflows negatively and positively impact CO₂ emissions of the selected African countries; these outcomes buttress the points that FDI inflows have both deleterious and advantageous effects, corroborating the findings of Opoku et al. [38] and Huang et al. [60]. The industrial revolution over the years has substantially risen [61,62], tantamount to FDI inflows to African countries [63], because investors have been establishing industries in the form of FDIs. Some industries have been advancing toward and exporting cleaner technologies and manufacturing processes to their host countries, intending to promote sustainability and protect the environment of their host countries. Examples of such industries are M-KOPA, African Clean Energy (ACE), and Daystar power, established in African countries, such as Tanzania, Uganda, Kenya, Nigeria, etc. These industries employ cleaner technologies, such as small-scale wind and hydro energies, solar panels, energy-efficient appliances, etc., to protect their host countries and help with their environmental amelioration, because a sustained environment not only helps the host countries but also yields profitable returns on the investment.

On the contrary, other industries, such as the Volkswagen Group, Nissan, Toyota Motors, etc., maintain and utilize conventional technologies to produce high-end goods to avoid exorbitant costs and expenses. These older technologies tend to be polluting processes, such as generating power and energy from burning coal, oil, natural gas, and petroleum, thus, resulting in high amounts of greenhouse gases, such as CO₂, and chemicals being released into the atmosphere of their host countries. These emissions cause air pollution, which is harmful to the host countries’ inhabitants, i.e., resulting in adverse health ramifications [64] and depleting the ozone layer, contributing to global warming [65]. Overall, the findings have revealed that FDI inflows in African countries possess good attributes that contribute to reducing CO₂ emissions and, at the same time, pose a threat and cause a rise in their CO₂ emissions.

Secondly, the selected African countries’ financial structures have a positive impact on carbon dioxide emissions per capita, indicating that ready access to credit, such as trade credits, loans, nonequity securities purchases, etc., from the financial systems (i.e., banks) results in a rise in CO₂ emission levels. Disparate from the findings of De Haas and Popov [50] and Xu et al. [66], this research infers that a debt-based financial structure contributes to environmental degradation in the form of a rise in CO₂ emissions.

Thirdly, trade openness of the employed African countries negatively impacts CO₂ emissions per capita, which can be justified by technology or technical and composite effects and substantiates the findings of Essandoh et al. [67] and Karedla et al. [68]. Cleaner and more effective technology practices spread throughout partner countries as trade creates a channel for goods exchange and spillover effects [58]. As a result, countries can produce goods by utilizing their comparative advantages. Thanks to the technology spillover effect, they also have access to cleaner or green and more effective operating technologies. Consequently, this will increase trade while reducing CO₂ emitted levels.

Fourthly, in line with the findings of Solarin et al. [8] and Xie et al. [37], this study’s employed countries’ GDP per capita positively impacts their CO₂ emissions while it refutes the findings of Adewuyi and Awodumi [69] and Zubair et al. [70]; thus, the economy (represented by GDPPC) of the employed countries advances at the expense of the environment. As a continent with emerging economies, Africa is known to host developing countries. Generally, at the developing phase of a country, more resources are allotted to the sectors with a quick progression rate. The oil and gas extraction sector in African countries is known to be one of the sectors that contributes greatly to their economies. However, these extraction activities also release high amounts of greenhouse gases, especially CO₂, into the atmosphere, thus, causing a rise in the emitted levels of CO₂. In addition, due to the lack of technological knowhow, such countries rely on traditional forms and technologies to generate power and energy. That is, to generate power at a lower cost for the country, technologies, such as coal-generated plants, are employed, releasing chemicals such as CO₂ into the air and, thus, causing pollution, which interrupts the ecosystem of such a country and the world.

4.5. Hypothesis Testing

In testing for the hypotheses proposed in Section 3.2, this study’s estimations discovered that:

Under the haven analyses using the PLS, FE, RE, and the one-step SYM-GMM, the FDI inflow coefficient is negative, i.e., FDI inflows negatively affect CO₂ emissions. Therefore, the PHVHP does not hold in this study and H₁ is not accepted.

Under the halo analyses using the PLS, FE, RE, and the one-step SYM-GMM, the FDI inflow coefficient is positive, i.e., FDI inflows positively affect CO₂ emissions. Therefore, the PHLHP does not hold in this study and H₂ is not accepted.

The coefficients for FNS revealed there exists a positive relationship between financial structure and CO₂ emissions. Thus, FNS positively impacts CEM. Therefore, H₃ is accepted.

Although the coefficient of financial development under the PLS, FE, and RE was positive, it was negative under the one-step SYM-GMM. Moreover, none of the coefficients were significant. Therefore, H₄ is not accepted.

5. Conclusions and Suggestions

Carbon dioxide emission has been the focal point in conversations and research over the last few decades [65]; however, one impact has not been established due to its intricate nature and effect. Therefore, this study assesses Foreign Direct Investment and other variables that influence carbon dioxide emissions as its main aim by employing panel data from forty-one (41) African countries from 2005 to 2019 (fifteen years). Pooled least squares, fixed and random effects, and the one-step system GMM were applied to analyze the variables. The results obtained denote that regarding CO₂ emissions, a graphical notation can be inferred from Figure 4.

(1) FDI has both a negative and positive impact, (2) financial structure has a positive impact, (3) trade openness has a negative impact, and (4) GDP has a positive impact. The pollution halo and haven hypotheses were not validated in this study. However, it was established that from the standpoint of finance, the financial systems (i.e., banks) and direct investments from foreign investors in the adopted African countries play a role in CO₂ emission levels rising and falling. Moreover, GDP and trade openness contribute to the varying emitted levels of CO₂. Therefore, this study puts forth the following suggestions:

Application and execution of policies, laws, and regulations that deter or punish polluting foreign investment and encourage clean ones. As a hosting nation, the government and policy-makers of this study’s chosen countries should enact and strengthen policies, laws, and regulations requiring polluting foreign investments to pay penalties, such as carbon taxes, to lessen the detrimental effect of FDI on CO₂ emissions. Apart from South Africa, which has a carbon tax system applied, none of the African countries have been enacting such actions [54]. Therefore, the remaining African countries can follow the precedent set by South Africa to aid in mitigating CO₂ emissions. Additionally, investments that deal with clean technologies, green systems, and green development activities should be supported to further aid in attracting clean investments.

Green finance application. Green finance, according to Xie et al. [71], is a future-focused form of investing that aims to advance the financial sector and enhance the environment. Thus, green finance is a strategy for sustaining the environment that encourages consuming less energy and lowering greenhouse gas (GHG) emissions [72]. However, the practice of green finance has not gained ground all over Africa. Therefore, per this study’s outcome, the financial systems (banks) of the selected countries that extend credit to the private sector should apply green finance in the form of green credit to mitigate CO₂ emissions.

Preferential trading policies. The African countries utilized in this study should strategize and implement policies encouraging preferential trade, highlighting an addition of value via clean technology—a country benefits when trade openness is fostered through mutual trade liberalization and removing trade barriers. Therefore, implementing preferential trade policies will assist the chosen African countries in obtaining clean technology, boosting their trade value, and contributing to environmental sustainability. Moreover, the European Union’s (EU) Emission Trading System (ETS) is still in effect, with some African countries, such as Nigeria, Kenya, etc., implementing such a system. With the aim of protecting the environment against emissions caused by greenhouse gasses (GHGs), the remaining non-participating African countries should apply the EU’s ETS to further aid in mitigating highly emitted levels of CO₂ that occurs as a result of trade.

Practice carbon capture, usage, and storage. It is known that to achieve sustainable development and promote the economy of a country, green technology innovation is important [73]. As such, the countries sampled for this study (especially the industries and parties in such countries that use traditional technologies and emit high amounts of carbon) should practice how to capture, use, and store their emitted CO₂. Applying innovative green technologies, such as carbon capture, usage, and storage, will help protect the environment and progress the country’s economy. Thus, the captured carbon can be stored, reused, and recycled to help promote the country’s economy and protect the environment.

Limitations

This paper has some novel conceptual analysis and research insights; however, it has flaws. Some countries were exempt from the study due to data unavailability, so not all the African countries were sampled. The chosen variables are a representation of the factors that affect CO₂ emissions. However, this study did not apply other factors, such as energy consumption. Therefore, the employed variables in this study cannot be considered as the sole impacts of CO₂ emissions. Furthermore, the proxy for financial structure used was solely the credit extended to the private sectors from banks; nevertheless, a composite index of several types of financial structure can be created to give a comprehensive understanding of how financial structure affects CO₂ emissions. The aforementioned ideas should, therefore, be further explored in future research.