Sustainable Energy Usage for Africa: The Role of Foreign Direct Investment in Green Growth Practices to Mitigate CO₂ Emissions

Abstract

In line with Africa’s commitment to keeping up with the United Nations Framework Convention on Climate Change, achieving a sustainable future requires balancing economic growth with environmental sustainability. This study investigates the long-term impacts of foreign direct investment, economic growth, agricultural production, and energy consumption on CO₂ emissions across 43 African nations from 1990 to 2021. Despite significant research on the individual effects of these factors, the combined influence on CO₂ emissions remains underexplored. Addressing this gap, this study employs cross-sectional augmented distributed lag estimators (CS-DL and AMG) and updated estimation packages to effectively examine the relationships between variables. Our findings are as follows: firstly, economic growth and energy use was shown to have a significant positive influence on CO₂ in the long term. Also, foreign direct investment significantly promotes CO₂ emissions. Secondly, the causality test shows a unidirectional causal relationship between CO₂ emissions and foreign direct investment. The test also revealed a bidirectional relationship between GDP and CO₂ emissions, as well as between energy consumption and CO₂ emissions. Again, a bidirectional causation was observed between agricultural production and CO₂ emissions. Thirdly, the impulse response analysis shows that GDP will contribute more to emissions over the 10-year forecast period. This study also proposes policy implications to lessen CO₂ across the continent and advocates for the judicious adoption of existing policy frameworks like the 2030 Agenda for environmental Sustainability.

1. Introduction

Foreign direct investment (FDI) plays an essential role in the economic development of several nations, especially those in the developing world [1,2]. It includes bringing technology, experience, and managerial techniques, in addition to financial resources, to domestic businesses or projects through the injection of foreign capital [3,4,5]. Despite the potential of FDI to boost economic growth and development, its role in environmental sustainability, especially carbon emissions, is inevitably complex and varied [6,7,8,9]. This complexity is particularly relevant to Africa, a continent already plagued with economic challenges. Several African nations have shown both good and negative effects of FDI on carbon emissions. For example, Nigeria has experienced substantial environmental challenges, such as oil accidents and gas venting, which have contributed to high carbon emissions, despite experiencing substantial FDI in its oil and gas sector [10]. On the other hand, Kenya has received significant FDI in its renewable energy industry, specifically in geothermal and wind energy projects. These projects have played a significant role in decreasing Kenya’s carbon emissions and encouraging the adoption of sustainable energy practices [11].

The influence of foreign investment on African economies is substantial. This influx of capital, technology, and expertise fuels growth but also brings environmental complexities [12,13]. Nevertheless, according to [14], the complex relationship between FDI and CO₂ presents multiple opportunities and challenges for promoting African sustainable development. Foreign direct investment (FDI) can enhance environmental sustainability by transferring green technologies, improving standards, and investing in green growth initiatives. For instance, Rwanda has utilized FDI profits to fund green development and promote environmental sustainability. In contrast, countries like Angola and the Democratic Republic of the Congo have seen increased carbon emissions due to intensified extraction activities [15] and Luo et al. [16]. Additionally, Nigeria’s oil sector exemplifies the Pollution Haven Effect, where FDI leads to significant environmental degradation and high carbon emissions due to weak enforcement of environmental regulations [17].

As an agrarian continent and a signatory to the Paris Agreement (African Development Bank Group-2016) [18], it is crucial for Africa to thoroughly assess the current and future impact of agricultural production. The sector provides livelihoods for a large portion of the population and contributes significantly to the continent’s GDP [19,20]. However, it is also a significant source of environmental pollution, with fertilizer use, rice cultivation, deforestation, and land-use changes contributing to carbon emissions [21]. While highlighting that agricultural activities can have negative environmental impacts, there are notable positive examples, such as Rwanda’s agroforestry programs that enhance carbon sequestration [22] and Ethiopia’s Sustainable Land Management Program that reduces soil erosion and increases carbon storage [23,24]. As established by the existing literature, emphasizing the need for sustainable, environmentally friendly farming is crucial to offset the sector’s impact on the environment.

Africa, with its rapidly growing economies and population exceeding 1.4 billion [25], is experiencing increasing energy consumption. Economic expansion in Africa fosters social modernization and escalates urbanization, resulting in increased consumption of coal, gas, and oil, which significantly contribute directly or indirectly to environmental pollution [26]. However, as established by Refs. [27,28], energy consumption is closely linked to carbon emissions. Energy-intensive activities like mining and heavy industries in South Africa, for example, have driven economic development [29]. This development is not devoid of its environmental consequences. Also, the on-going rapid urbanization on the continent raises energy needs for infrastructure, transportation, and housing. Countries such as Nigeria and Kenya are undergoing substantial urbanization, which is accompanied by a significant need for more energy [30]. Urban households consume more energy than their rural folks. According to Refs. [31,32], access to electricity, heating, cooling, and mobility increase energy usage. The 43 countries included in this study exhibit a wide range of fossil fuel energy usage patterns. For instance, South Africa, a major coal producer, relies on coal for 77% of its energy needs, leading to significant carbon emissions [33]. Nigeria, one of the largest oil producers in Africa, heavily depends on oil and natural gas for its energy consumption, contributing to substantial CO₂ emissions (SITE). Similarly, countries like Egypt and Algeria also have significant fossil fuel usage due to their abundant oil and gas reserves. On the other hand, countries such as Kenya and Morocco have made considerable strides in incorporating renewable energy into their energy mix. Kenya, for instance, has invested heavily in geothermal and wind energy, reducing its reliance on fossil fuels. Morocco has similarly invested in solar and wind energy projects, aiming to reduce its carbon footprint.

This study is essential as it addresses the critical policy-level problem of balancing economic growth with environmental sustainability in Africa. To attain global sustainability by 2060, for example, it is imperative to reduce intensive agricultural land use by 9%, decrease grazing areas by 30%, and expand forested regions by 11% [25]. Despite the burgeoning academic focus on the environmental impacts of CO₂ emissions, comprehensive evaluations across the continent remain scarce. A thorough continental analysis is essential to address climate change and mitigate CO₂ emissions effectively. Furthermore, economic growth often demands efficient energy use, which is crucial for attracting foreign direct investment (FDI). Countries at different income levels may struggle to balance economic growth with environmental sustainability, particularly in the context of numerous African nations [34]. Understanding the impact of CO₂ emissions on development from a continental perspective is crucial. Additionally, examining the influence of agricultural production on carbon emissions is vital as the link between land-use change and environmental degradation has become a prominent research focus for environmentalists and policymakers [35]. This paper aims to address the following questions:

• Is foreign direct investment detrimental to environmental sustainability in the sampled countries?
• How do energy consumption and economic development affect carbon emissions in the sampled countries?
• Does agricultural production contribute to environmental sustainability in the study area?

In response to the above questions, this research proposes to offer potential solutions by providing empirical evidence on the interactions between FDI, economic growth, energy consumption, agricultural production, and CO₂ emissions across 43 African nations from 1990 to 2021. By utilizing advanced econometric techniques, this study will identify the synergetic effects of these variables and offer insights for policymakers to develop effective strategies for sustainable development. The findings will contribute to the creation of a regulatory framework that promotes sustainable investments and green growth, ensuring that Africa’s economic aspirations do not come at the expense of its environmental health.

This work contributes to the existing body of knowledge in several ways: Firstly, based on the existing literature, minimal studies have researched these variables in Africa. Apart from studies like [36], who concentrated on Eastern Africa, this is the first research to investigate the link between FDI, economic growth, agricultural production, and CO₂ emissions in 43 sampled countries that cut across the length and breadth of Africa. The results of this study have the potential to offer empirical support for a greater awareness of the environmental impact of foreign direct investment and agriculture. Secondly, since Africa is the second fastest-growing economy after Asia, studying the causal impact between these factors might help achieve low carbon emissions, net-zero carbon emissions, and a green growth economy. Verified results may help in understanding and building a financially and politically sustainable framework for foreign direct investment. Thirdly, this study utilizes the most recent econometric estimators to evaluate the influence that agriculture and foreign direct investment have on environmental sustainability in selected African nations. Previous research has exclusively employed methods restricted to monitoring the degree of cross-sectional dependence and has neglected the presence of heterogeneities and collinearities among variables. This study adds to the existing knowledge by incorporating recent estimators, such as the cross-sectional augmented distributed lag (CS-DL) and AMG, which were suggested by [37,38]. Regardless of the level of cross-sectional dependency, collinearities and heteroscedasticity are detected by these approaches. Finally, this research is the first of its kind to investigate impulse response and variance decomposition to establish how much the variation in African countries over ten years delayed carbon emission variability. The use of a holistic approach in the context of Africa makes it easier to understand the complex interplay and factors that contribute to CO₂ emissions, giving stakeholders and policymakers a thorough understanding. To summarize the remaining parts of this research, the following is provided: There is a discussion of the literature review in Section 2. Section 3 discusses the data and methodologies, Section 4 presents the empirical findings and discussion, and Section 5 concludes and discusses policy implications.

2. Literature Review

2.1. Data Sources and Variables

The primary focus of this study is to investigate the relationship between foreign direct investment (FDI), economic growth, energy consumption, agricultural production, and carbon emissions in 43 African countries. The data sources and variables are crucial for understanding the dynamics of these relationships. Data for foreign direct investment (FDI), economic growth (GDP), and agricultural production were obtained from the World Bank Development Indicators (WDI) database [39]. The WDI provides comprehensive data on FDI inflows and outflows, GDP growth rates, and agricultural output, which are essential for analyzing economic and development trends across countries. Carbon emissions and energy consumption data were mined from the US Energy Information Administration (US-EIA) [40]. The US-EIA offers detailed data on energy consumption by type and sector, as well as CO₂ emissions, enabling a thorough examination of the environmental impact of energy use in different countries. With CO₂ being the dependent variable, the independent variables in this study include FDI inflows, GDP, energy consumption, and agricultural production. Further description of the data sources and variables is expanded in Chapter 3 of this paper. The comprehensive and reliable nature of these data sources ensures the robustness and validity of the analysis conducted in this research.

2.2. Theoretical Framework

• Linkages between Foreign Direct Investment and Carbon Emissions

The relationship between foreign direct investment (FDI) and carbon emissions has been extensively explored in the literature. Modernization theory posits that FDI contributes to economic development by bringing in capital, technology, and managerial expertise, which enhance productivity and economic growth [41,42]. This inflow of resources can also lead to increased industrial activity and, consequently, higher energy consumption and carbon emissions. For instance, Refs. [43,44] have demonstrated through panel data analysis that FDI increases industrial activity and CO₂ emissions. Conversely, the Pollution Haven Hypothesis suggests that FDI tends to flow into countries with lax environmental regulations, leading to increased pollution and CO₂ emissions [45,46,47], highlighting that many regions in Africa experience significant environmental degradation due to the relocation of pollution-intensive industries. However, the Porter Hypothesis offers a more optimistic view, suggesting that stringent environmental regulations can stimulate innovation and lead to environmentally friendly technologies and practices, thereby reducing emissions [48]. Recent studies further support these theories. Ref. [48] emphasized the potential of green FDI to reduce CO₂ emissions through sustainable technologies, while Refs. [46,47] pointed out the environmental degradation caused by pollution-intensive industries relocating to Africa. These mixed findings underscore the importance of smart policymaking to maximize FDI benefits while minimizing its environmental impact.

• Economic Growth, Energy Consumption, and Carbon Emissions Nexus

Energy consumption is a key driver of economic growth, fueling industrial and technological development, as posited by the Growth Hypothesis [49,50]. However, this increase in energy consumption often comes at the cost of higher CO₂ emissions. Bing Li et al. [51] and Baloch and Danish [52] demonstrated that the use of fossil fuels significantly contributes to CO₂ emissions, particularly in developing regions. The environmental Kuznets curve (EKC) suggests that in the early stages of economic growth, CO₂ emissions increase, but after reaching a certain level of income per capita, emissions begin to decrease as cleaner technologies are adopted [51,52]. This theory is supported by studies like those of Haldar and Sethi [53] and Khan et al. [54], which emphasize the benefits of renewable energy in reducing emissions. Espoir, Sunge, et al. [50] also highlight that energy consumption increases economic growth and productivity, but the environmental cost of this consumption must be managed through sustainable practices. The recent literature reinforces these findings. Haldar and Sethi [53] and Khan et al. [54] stressed the importance of transitioning to renewable energy to mitigate CO₂ emissions. Musah, Kong, Mensah, Li, et al. [55] and Rezk et al. [56] noted the significant environmental cost of fossil fuel consumption in Africa, underscoring the need for greener energy options. This transition is crucial for Africa’s sustainable future as population growth and increased energy needs continue to drive up CO₂ emissions [57,58].

• The Nexus between Agricultural Production and Carbon Emissions

Agriculture plays a vital role in Africa’s economic expansion, providing revenue and employment opportunities. Espoir, Bannor, et al. [59] highlight that agricultural development stimulates economic growth. However, traditional agricultural practices often lead to deforestation, soil degradation, and increased CO₂ emissions [60,61].

Sustainable agricultural practices, on the other hand, can mitigate these negative impacts. Steinhübel and Minten [62] and Kumara et al. [63] demonstrate that eco-friendly farming methods can lower CO₂ emissions by promoting carbon sequestration. Ashiq et al. [64] reported that modern agricultural technologies can boost production and growth while reducing environmental harm.

Recent studies support these perspectives. Steinhübel and Minten [62] and Kumara et al. [63] highlighted the benefits of sustainable agricultural practices in reducing CO₂ emissions. Refs. [60,61] emphasized the environmental degradation caused by traditional farming techniques, further advocating for the adoption of sustainable practices to balance agricultural productivity with environmental preservation.

2.3. Empirical Framework and Model

To analyze the relationship between FDI, economic growth, energy consumption, agricultural production, and carbon emissions, this study employs the cross-sectional augmented distributed lags (CS-DL) and augmented mean group (AMG) models. The cross-sectional augmented distributed lags (CS-DL) model is widely used to analyze long-run relationships in panel data settings, especially when dealing with cross-sectional dependence. This model augments traditional panel data methods by incorporating cross-sectional averages of the dependent and independent variables, thus accounting for common factors that may influence the relationships under study. Recent studies highlight the efficacy of the CS-DL model. Ref. [65] demonstrated that CS-DL effectively captures the long-term impacts of macroeconomic variables on environmental outcomes, including carbon emissions. Their study found that accounting for cross-sectional dependence significantly improves the accuracy of long-term predictions. Similarly, Ref. [66] used the CS-DL approach to investigate the impact of economic policies on environmental sustainability across Asian economies, underscoring the model’s ability to handle heterogeneity and cross-sectional dependence. The primary advantage of the CS-DL model lies in its ability to address cross-sectional dependence, which is common in panel data involving multiple countries [67]. This makes it particularly suitable for this study, which involves a diverse sample of 43 African countries with varying economic and environmental contexts.

The augmented mean group (AMG) estimator, developed by Eberhardt and Teal [38], is another advanced econometric technique employed in this research. The AMG estimator is designed to handle panel data characterized by cross-sectional dependence and non-stationarity. It extends the mean group (MG) estimator by incorporating common dynamic processes across cross-sections. Recent applications of the AMG estimator underscore its robustness in empirical research. For instance, Herzer [68] used AMG to explore the long-term relationship between trade openness and environmental quality in Latin American countries, demonstrating that AMG provides consistent and reliable estimates even in the presence of cross-sectional dependence. Similarly, [69] applied the AMG estimator to study the effects of renewable energy consumption on economic growth and environmental quality in OECD countries, finding that AMG effectively captures long-term interactions among the variables. The AMG estimator’s strength lies in its ability to accommodate heterogeneous dynamics and cross-sectional dependence, making it an ideal choice for this study’s empirical framework. By using AMG, this research can obtain reliable estimates of the long-run effects of FDI, economic growth, energy consumption, and agricultural production on carbon emissions across different African countries.

By employing the CS-DL and AMG models, this study aims to provide a comprehensive analysis of the interplay between FDI, economic growth, energy consumption, agricultural production, and carbon emissions in 43 African countries. These models’ ability to handle complex data structures ensures robust and reliable estimates, offering valuable insights into policy measures that can foster sustainable development while minimizing environmental impacts.

3. Methodology and Data

This chapter outlines the data employed in the analysis and details the econometric techniques that will be utilized. Specifically, our first test is to determine the presence of cross-sectional dependency in our dataset, after which we will conduct cross-sectional panel unit root tests and employ the Westerlund cointegration test to assess the existence of a long-run relationship between the sampled variables. Subsequently, the estimator procedure will introduce the causality tests and data estimators that will be implemented in the analysis. Figure 1 presents the methodology flowchart.

3.1. Data Description

This part provides a comprehensive overview of the data employed in the analysis, spanning the period from 1990 to 2021 and encompassing 43 African countries. The data for this research were obtained from the World Bank and US-EIA Database. The core variables investigated include energy consumption (quadrillion Btu), which quantifies energy consumption across the selected African countries and is measured in quadrillion British thermal units (Qd. Btu); GDP per capita (constant 2015 USD): this variable is a proxy for economic growth, expressed in constant 2015 US dollars; CO₂ emissions (million metric tons): this variable measures the total amount of CO₂ released by each country, quantified in million metric tons; FDI, which is defined as the net inflow of foreign direct investment as the percentage of GDP; and agriculture production, which is quantified as the total agricultural, forestry, and fishing value added, which is also measured as the percentage of GDP. To mitigate potential issues of heteroscedasticity, which can lead to unreliable results, all variables were transformed using the logarithm. This transformation fosters robust regression analysis by ensuring a more constant variance across the error terms [67].

Table 1. Data description.

Variable	Denoted as	Units	Source
Foreign direct investment	FDI	Net inflow of foreign direct investment (% GDP)	WDI
Carbon emissions	CO2	Million metric tons (MMT)	US-EIA
Energy consumption	E.C.	Quadrillion, British thermal unit (Qd. Btu)	US-EIA
Gross domestic products	GDP	Constant of 2015 US$	WDI
Agriculture production	AG	(% of GDP)	WDI

reports the data sources employed for this study. Also, the sampled countries can be found in Appendix A.

Table 2. Descriptive statistic.

	CO	GDP	FDI	AG	EC
Mean	0.503	3.100	1.293	1.322	−1.301
Median	0.448	3.063	1.280	1.360	−1.316
Maximum	2.682	4.224	1.867	3.461	0.764
Minimum	−1.175	2.280	−2.110	0.240	−3.057
Std. Dev.	0.785	0.401	0.121	0.538	0.746
Skewness	0.621	0.519	−15.909	1.423	0.447
Kurtosis	3.395	2.584	460.983	7.439	3.317
Observations	1376	1376	1376	1376	1376

offers a thorough summary of the descriptive data for each variable that was chosen. It provides a comprehensive overview of the descriptive statistics for all selected variables. In essence, the data employed in this study offer a robust foundation for analyzing the relationships between foreign direct investment, agriculture production, energy usage, economic progress, and CO₂ emissions across a diverse set of African countries over the studied time period.

3.2. Mathematical Methodology

The factors contributing to ecological contamination were identified using the methods described by [70]. We designed the empirical econometric model as follows to achieve this goal:

$$CO=f(fDI,GDP,AG,Ec)$$

(1)

$$lC{O}_{it}={\alpha }_{0i}+{\vartheta }_{1i}lFD{I}_{it}+lGD{P}_{it}+lA{G}_{it}+lE{C}_{it}+{\mu }_{it}$$

(2)

CO₂ denotes carbon dioxide emissions, E.C. refers to energy consumption, GDP represents gross domestic product, FDI means foreign direct investment. AG illustrates agriculture. In addition, α represents the intercept term. The variables i and t represent the 43 countries that were sampled and the time period from 1990 to 2021. The variables we selected for our sample were converted using a logarithmic function, which is represented by the letter ‘l’ in the equation.

3.3. Econometric Methodology

3.3.1. Cross-Sectional Dependence Test

While panel data offer numerous advantages, recent research highlights cross-sectional dependence (CSD) in residuals as a prominent challenge [71]. Ignoring CSD can lead to biased estimates and unreliable conclusions [72]. Understanding the presence of CSD is an important first step in panel data analysis to determine which unit root test to use—whether it is a first-generation or second-generation test.

To address this, we employ various CSD tests, including the Pesaran CD test proposed by Pesaran et al. [73] and the standardized Lagrange Multiplier (L.M.) tests by [74]. Notably, the Pesaran tests are particularly suited for handling large panel datasets with a high number of cross-sectional units (N) and time periods (T). The computation of these tests can be expressed as follows:

$$LM=\sqrt{{\displaystyle \frac{1}{N\left(N-1\right)}}}{\displaystyle \sum _{i=1}^{N-1}{\displaystyle \sum _{j-i+1}^N\left(T_{ij}{\mu }_{ij}^2-1\right)}}\to N(0,1)$$

(3)

$$CD=\sqrt{{\displaystyle \frac{2}{N\left(N-1\right)}}}{\displaystyle \sum _{i=1}^{N-1}{\displaystyle \sum _{j-i+1}^N\left(T_{ij}{\mu }_{ij}^2-1\right)}}\to N(0,1)$$

(4)

Large datasets with long time spans T are suited for Equation (3). In contrast, Equation (4) is specially tailored for situations when many observations occur during a specific period T. The Breusch test Breusch and Godfrey [75] may be more suitable for smaller datasets with defined time spans. Equation (5), which is shown below, is often used to illustrate this test:

$$LM={\displaystyle {\sum }_{i=1}^{N-1}{\displaystyle {\sum }_{j-i+1}^N{T}_{ij}{\mu }_{ij}^2}}\to x^2\left({\displaystyle \frac{N(N-1)}{2}}\right)$$

(5)

The symbol μ² represents the correlation coefficient obtained from the residuals in Equation (5). Thus, the following explanation of the equation might be provided:

$${\mu }_{ij}={\mu }_{ji}={\displaystyle \frac{{\displaystyle {\sum }_{t-1}^T{\epsilon }_{ij}{\epsilon }_{ji}}}{{\left({\displaystyle {\sum }_{t-1}^T{\epsilon }_{ij}^2}\right)}^{\frac{1}{2}}{\left({\displaystyle {\sum }_{t-1}^T{\epsilon }_{ji}^2}\right)}^{\frac{1}{2}}}}$$

(6)

Equation (6) ${\epsilon }_{ij}$ and ${\epsilon }_{ji}$ denote the standard errors.

3.3.2. Pesaran CIPS and CADF Panel Unit Root Test

The CIPS (cross-sectional augmented Im, Pesaran, and Shin) and CADF (cross-sectional augmented Dickey–Fuller) tests proposed by Pesaran [76] are used in this work to examine cross-sectional dependency in panel unit root testing. By using the average lagged level and first-difference terms from every unit in the model, these tests address the potential issue of cross-sectional dependency. The estimation of these tests can be achieved through the equation below:

$$\Delta y_{it}={\vartheta }_i+{\theta }_i{y}_{i,t-1}+{\delta }_i{\overline{y}}_{t-1}+{\displaystyle \sum _{j=0}^p{\sigma }_{ij}\Delta {\overline{y}}_{t-j}+{\displaystyle \sum _{j=1}^p{\beta }_{ij}\Delta y_{i,t-j}+{\mu }_{it}}}$$

(7)

The dependent variable, carbon emissions, is represented by the variable “y_it” in the equation above. The average values of prior levels and the difference between successive values are shown by the terms y_t−1 and Δy_t−j, respectively. The relevant factors are reflected in the coefficients δ and σ. ϑ and $\theta$ stand for the intercept and trends, respectively, while β is the lead coefficient. The CIPS statistics are computed using the cross-sectional augmented Dickey–Fuller (CADF) statistics. Equation (7) may be used to express the CIPS statistics, with ε_ij and ε_ji standing for the standard errors.

$$CIPS={\displaystyle \frac{1}{N}}{\displaystyle \sum _{i=1}^{N}CAD{F}_i}$$

(8)

3.3.3. Westerlund Cointegration Test

We used the Westerlund and Edgerton [77] error correction panel cointegration test to assess the existence of a long-run relationship among the variables. This method, inspired by Bai and Ng [78], may detect possible structural breaks within the panel data series. Panel statistics (P_t, P_a) and group statistics (G_t, G_a) are the four aspects of statistics that are the primary focus of the test. The panel statistics are derived from the following equation:

$$\Delta z_{it}={\beta }_i^{\prime }{d}_i+{\alpha }_i(z_{i(t-1)}+{\delta }_i^{\prime }{y}_{i(t-1)})+{\displaystyle \sum _{j=1}^k{\varphi }_{ij}\Delta z_{i(t-1)}+{\displaystyle \sum _{j=0}^k{\phi }_{ij}\Delta y_{i(t-1)}{\mu }_{it}}}$$

(9)

The equation above represents the adjustment denoted by α_i. The word $d_i$ represents the deterministic component vector, which consists of linear temporal trends and a constant. When z is the K + 1 dimensional vector, it represents an integrated variable. The other factors create an inconvenience in the variable of interest. The Westerlund ECT-based panel cointegration tests may be developed based on the α_i estimations.

$$G_t={\displaystyle \frac{1}{N}}{\displaystyle \sum _{i-1}^N\frac{{\alpha }_i}{SE({\alpha }_i^{\prime })}}$$

(10)

$$G_a={\displaystyle \frac{1}{N}}{\displaystyle \sum _{i=1}^N\frac{T{\alpha }_i}{{\alpha }_i^{\prime }(1)}}$$

(11)

The G_t and G_a statistics represent the average values across all categories in the panel data. These statistics help evaluate the null hypothesis, indicating no cointegration among the variables within any cross-sectional unit. If the null hypothesis is rejected, it suggests that there is cointegration present in at least one group within the panel. The equations below provide the specific formulas for calculating G_t and G_a, which evaluate cointegration across the entire panel:

$$P_t={\displaystyle \frac{\widehat{{\alpha }_i}}{SE(\widehat{{\alpha }_i})}}$$

(12)

$$P_a=T\widehat{{\alpha }_i}$$

(13)

The rejection of the null hypothesis indicates that the entire panel that was sampled did not show cointegration.

$$y_{it}={\phi }_i+{\varphi }_i{y}_{it-1}+{\sigma }_{0i}{x}_{it}+{\sigma }_{1i}{x}_{it-1}+{\displaystyle \sum _{l=0}^{PT}{{\delta }^{\prime }}_{il}{\overline{z}}_{it-l}+{\mu }_{it}}$$

(14)

where $i$ = 1,2,3,4, ……., N, ${\overline{z}}_t=N^{-1}{\displaystyle \sum _{i=1}^N{z}_{it}=({\overline{y}}_t,{\overline{x}}_t,{\overline{f}}_t{)}^{\prime }}$, ${\varphi }_0$ and ${\sigma }_0$ φ and μ represent the intercept and error term, respectively. The unobserved common element with a heterogenous component is represented by f_t. The long-run coefficient estimates may be expressed as follows:

$${\widehat{\theta }}_{CS-ARDL}={\displaystyle \frac{{\displaystyle {\sum }_{l=0}^q{\widehat{\sigma }}_{il}}}{1-{\displaystyle {\sum }_{l=1}^p{\widehat{\varphi }}_{il}}}}$$

(15)

3.4. Estimators

• Augmented Mean Group (AMG)

Developed by Eberhardt and Teal [38], the AMG estimator is a resilient model designed to handle heterogeneous panel data, particularly when cross-sectional dependence is present. It enhances the model by incorporating the time averages of both the dependent variable and the regressors, along with individual dummy variables to account for fixed effects [79,80]. This technique permits varying slopes and intercepts across different panel units, facilitating a robust estimation of the average long-term relationship. It is particularly useful for investigating long-term relationships between variables in panels that exhibit cross-sectional dependence and individual heterogeneity. Initially, a pooled regression model was employed using year dummies calculated by the first-difference OLS method. The AMG equation can be represented as follows:

$$\Delta y_{it}={\varpi }_i+{\varphi }_i+\Delta z_{it}{\sigma }_i{f}_t+{\displaystyle \sum _{t=2}^T{\theta }_i\Delta {Þ}_t+{\mu }_{it}}$$

(16)

AMG Stage 2

$${\widehat{\mathbb{Q}}}_{AMG}=N^{-1}{\displaystyle \sum _{i=1}^N{\widehat{\varphi }}_i}$$

(17)

The first-difference operator and the time variable (used as a dummy parameter) are symbolized by $\Delta$ and $Þ$, respectively. The coefficient of the time variable is denoted by ${\theta }_i$, while ${\widehat{\mathbb{Q}}}_{AMG}$ signifies the coefficient of the AMG estimator.

• Cross-sectional augmented distributed lags (CSDLs)

To evaluate long-term correlations in the presence of semi-weak and weak dependence among parameters, [37] proposed the CSDL estimator. During the estimation process, we incorporated average lags into the model to mitigate strong cross-sectional dependence. In this study, to effectively establish a correlation between our parameters using the CSDL estimator, was used to represent carbon emissions (CO₂) and was used to denote all other regressors. The equation can be expressed as follows:

$$c{l}_{it}=b_i+{\mu }_{il}{z}_{it}+{\displaystyle \sum _{l=0}^{px-1}{\beta }_{il}\Delta {х}_{it}}+{\displaystyle \sum _{l=0}^{p\tilde{r}}{\upsilon }_{y,il}y{\tilde{v}}_{t-1}+{\displaystyle \sum _{l=0}^{p\tilde{k}}{\upsilon }_{х,il}{\tilde{n}}_{t-1}+{\mu }_{it}}}$$

(18)

${\tilde{v}}_{t-1}$ and ${\tilde{n}}_{t-1}$ are the cross-sectional averages. $\mathrm{p}\tilde{\mathrm{k}}=({\mathrm{T}}^{{\displaystyle \frac{1}{3}}})$ represents the regressor’s maximum lag and $p\tilde{r}$ the random lag of the regressed parameters. The unobserved and fixed country effect is marked as $b_i$.

These current estimators (CS-DL, AMG) adopted for this study operate on some key assumptions. We briefly highlight some of these assumptions and how they were accounted for in our work: Both estimators assume some level of parameter homogeneity and account for cross-sectional dependence among the variables adopted for this study. We addressed cross-sectional dependence by employing cross-sectional dependence tests (Pesaran CD and Breusch–Pagan LM tests) to ensure the presence of cross-sectional dependence was adequately captured. The CS-DL method explicitly incorporates cross-sectional averages to account for this dependence, while the AMG accounts for cross-sectional dependence through common factors. Also, both methods require a sufficiently large sample size and high-quality data to produce reliable estimates. To account for this, our study utilized a comprehensive dataset covering 43 African countries over a 31-year period (1990–2021), ensuring a robust sample size. Data were sourced from reliable databases such as the World Bank and US-EIA, and variables were log-transformed to mitigate potential heteroscedasticity. Again, both techniques assume that the variables are either stationary or cointegrated over the long term. We therefore conducted CIPS unit root tests to check for stationarity and Westerlund cointegration tests to verify the existence of long-term relationships among the variables, confirming that our data met these assumptions. Our work assures the robustness of the CS-DL and AMG econometric results by carefully evaluating and resolving these crucial assumptions. It is expedient to say that, though powerful for handling cross-sectional dependence and heterogeneity in panel data, both models can be complex to specify, computationally intensive, and sensitive to model assumptions and sample size.

3.5. Causality Test

For this test, we utilized the Dumitrescu–Hurlin test suggested by [81]. This test helps to determine the causal link existing among variables. This test is based on three main hypotheses: neutral causality, one-way directional (unidirectional) causality, and two-way directional (bidirectional) causality. The equation for causality can be expressed as

$${\mathrm{z}}_{i,t}={\varphi }_i+{\displaystyle \sum _{k=1}^k{\phi }_i^k{z}_{i,t-k}+}{\displaystyle \sum _{k=1}^k{\alpha }_i^k{x}_{i,t-k}+{\mu }_{it}}$$

(19)

Here, $z_{\chi }$ represents the tested variables, $\mathsf{ɤ}$ marks the singular fixed effect, $\varphi$ and $\sigma$ represent the autoregressive parameters and regression coefficients, which vary across sampled groups. Similarly, $k$ offers adequate features about the optimal lag for the whole cross-sectional units. The H_O for this assessment is based on the regression coefficient slope. It links with the singular Wald statistics of Granger non-causality averaged across the cross-sectional units. This can be expressed in the following form:

$$w_{i,t}={\hat{\theta }}_i^{\prime }{R}^{\prime }{\left({\hat{\theta }}_I^{2}R{\left[Y_i^{\prime }{Y}_i\right]}^{-1}{R}^{\prime }\right)}^{-1}R{\widehat{\theta }}_i$$

(20)

see Ref. [81] for further explanation about these parameters.

3.6. Impulse Response and Variance Decomposition

To determine both the immediate and extended effects of a shock to the explanatory variables on the relevant dependent variables, this study utilizes variance decomposition and impulse response analysis. The dynamic characteristics of the employed model facilitate the transmission of these effects between the target variable and other variables within the system. An accurate estimation of the impulse response can be obtained using a reliable companion matrix of the Vector Autoregressive (VAR) model, as proposed by Lanne and Nyberg [82]. The model can be expressed in equation form as follows:

$$z_t={\displaystyle \sum _{j=0}^p{\vartheta }_i{y}_{t-i}+{\epsilon }_t}$$

(21)

Equation (21) ${\vartheta }_i$ represents the function of the simple impulse response, and it can be calculated by transforming Equation (19) to an infinitive vector moving average. This can be expressed as follows in Equation (22)

$${\vartheta }_i=\left\{\begin{array}{l}{I}_K \mathrm{i}=0\\ {\displaystyle \sum _{j=1}^i{\vartheta }_i{z}_{t-j}{G}_{J,} \mathrm{i}=1,2,3,.....}\end{array}\right\}$$

(22)

where $I_K$ denotes the identity element for the companion matrix, $G_{J,}$ symbolizes the converted VAR coefficient matrix infinity vector average, ${\epsilon }_t$ denotes the error term, and $p$ marks the optimal lag. The v-step ahead of the estimating error is derived as

$$z_{it+v}-w\left(z_{it-v}\right)={\displaystyle \sum _{i=0}^{v-1}{\epsilon }_{i\left(t+v-1\right)}{\vartheta }_i}$$

(23)

In this context, $z_{it+v}$ represents the vector variables at a given time $t+v$ and $w\left(z_{it-v}\right)$ denotes the v-step ahead of the projected vector at time t. To orthogonalize the changes in the variables, a matrix p (cross product K × K) is employed, facilitating the identification of the variables’ effects on the forecast-error variance. Consequently, the impact of a variable $b$ on the v-step ahead of the forecast-error variance of the sampled variables c can be derived as follows:

$${\displaystyle \sum _{i-0}^{vl}{\sigma }_{bc}^2={\displaystyle \sum _{i-0}^{vl}{\left(i’R{\beta }_{ib}\right)}^2}}$$

(24)

4. Results and Discussion

4.1. Cross-Sectional Dependence and Panel Unit Root Test

We performed two separate tests developed by Pesaran [83] and Breusch and Pagan [84] to verify the presence of cross-sectional dependence within our data. The findings, shown in

Table 3. Cross-sectional dependence.

	Breusch–Pagan LM	Prob	Pesaran CD	Prob
CO	10,230.020	0.000	82.223	0.000
FDI	1839.034	0.000	6.425	0.000
GDP	10,936.370	0.000	71.957	0.000
AG	4893.850	0.000	18.044	0.000
EC	13,295.250	0.000	106.283	0.000

, reject the null hypothesis of no cross-sectional dependence at all levels of statistical significance. This statistically significant outcome confirms the existence of cross-sectional dependence between the variables in our collected data.

We conducted an analysis on the stationarity of the data series by employing the cross-sectional panel unit root tests proposed by Pesaran [76], specifically CADF and CIPS. According to the results shown in

Table 4. Panel unit root test.

	CIPS	CADF
CO	−6.087 ***	−4.097 ***
FDI	−6.125 ***	−3.263 ***
GDP	−5.724 ***	−4.263 ***
AG	−5.677 ***	−3.344 ***
EC	−6.19 ***	−3.638 ***

Note: *** show the significance level at 1%, respectively.

, the series shows cross-sectional dependency at level I (0).

4.2. Panel Cointegration Test

Using the Westerlund cointegration test proposed by Westerlund [85], we analyzed the presence of a long-run equilibrium relationship among the variables in our sample.

Table 5. Panel cointegration test.

Statistic	Value	Prob
Gt	−2.623	0.146
Ga	−14.028	0.214
Pt	−13.883	0.055
Pa	−13.808	0.000

presents a concise overview of the test results. Given these findings, it is evident that we can easily reject the null hypothesis (H₀), which suggests no cointegration. The research concludes that the variables under study show a cointegrated relationship, suggesting a long-run connection between them.

4.3. Estimate Results

Chudik et al. [86] proposed the cross-sectional augmented distributed lag (CS-DL) model and Eberhardt and Teal [38] the AMG model, both of which we implemented after the Westerlund cointegration test verified a long-term cointegration relationship.

Table 6. CS-DL result.

	CS-DL	Prob	AMG	Prob
GDP	0.357 **	0.015	0.050 *	0.000
FDI	0.890 **	0.039	0.033	0.435
AG	0.052	0.834	0.040 *	0.000
EC	0.834 **	0.000	0.845 *	0.000
RMSE	0.25		0.26

Notes: RMSE for AMG and CS-DL are 0.26 and 0.25, respectively. * and ** denote a 1% and 5% level of significance, respectively.

contains the findings regarding the long-term relationships between energy consumption, FDI, agriculture and economic growth, and CO₂ from our estimators (CS-DL and AMG). The output from the two recent models shows that FDI contributes to increasing CO₂ and the positive effect is statistically significant (0.890 at 5% significant level from the CS-DL). The positive effect on FDI on CO₂ emission is, however, not statistically significant according to the AMG estimator. The results also indicate that, in the long run, GDP promotes the increment in CO₂, with statistical significance (0.357 and 0.050 at 5% and 1% from CS-DL and AMG, respectively). The energy consumption findings mirror those of the GDP CO₂ nexus in the sampled countries. It was revealed that energy consumption in the sampled countries contributes to increasing CO₂ emissions. The positive effect of energy consumption on CO₂ is statistically significant (0.834 and 0.845 at 1% for CS-DL and AMG, respectively). Statistically, agricultural production positively promotes CO₂ emissions (0.040 at 1% significant level). Boosting agriculture production, on the other hand, statistically and insignificantly promotes the increment in CO₂ in the sampled countries. Based on the root mean square error (RMSE) of the estimators, the results from the CS-DL are stronger than the results from the AMG due to the RMSE of the CS-DL being smaller than that of the AMG. Hence, going forward, the conclusive discussion of the results of this research will be based on the CS-DL estimator. The robustness of the CS-DL results coincides with the suggestion of J. Namahoro et al. [67], who asserted that predictions from the CS-DL are better than those of other estimators.

More specifically, the findings indicate that boosting foreign direct investment (FDI) has a significant positive effect on emitted CO₂, whereas a 1% surge in FDI leads to a 0.890% increase in emitted CO₂ in the 43 sampled countries. These results are consistent with those estimated by [87]. According to the authors, the potential transfer of less energy-efficient technologies from foreign firms to host countries with lax environmental regulations could explain this positive correlation. As profit motives often take precedence, these less-efficient technologies might not be replaced over time, contributing to increased pollution. Comparably, Agyeman et al. [88] investigated the role of governance in reducing CO₂ emissions in Africa and found that foreign direct investment (FDI) is a significant factor leading to increased CO₂ emissions. Effective governance and control of FDI inflows are crucial for decarbonization. Other researchers like Ngonadi et al. [89] and Saqib and Dincă [90] arrived at similar conclusions. Economic growth was observed to significantly stimulate the CO₂ increment in the sampled countries, whereas a 1% boost in economic growth translates to a 0.357% increment in emitted CO₂. Our findings coincide with those of Raihan [91], which emphasize that the rise in resource consumption and production activities due to economic growth will lead to environmental degradation. Similarly, the findings by Ifelunini et al. [92] arrived at comparable findings. According to them, economic growth significantly increases CO₂ emissions in West Africa, with the effect being more pronounced in countries with higher initial levels of emissions. Improved governance can moderate this effect. In their study confirming the environment Kuznets curve hypothesis in West Africa, Shobande and Asongu [93] arrived at similar conclusions.

The CS-DL estimator results also reveal that a 1% increase in energy consumption leads to a 0.834% surge in CO emissions in the sampled countries. The continent is known to significantly rely on fossil fuel energy. This reliance on fossil fuels for energy has been growing, with a 4.6% annual increase in CO₂ emissions according to [94]. Similarly, in their study exploring the link between energy consumption and CO₂ emissions in North Africa, Musah, Kong, Mensah, Antwi, et al. [95] found that energy consumption is a significantly positive determinant of CO₂ emissions. It also confirmed bidirectional causality between energy consumption and CO₂ emissions. Also, a recent study focusing on investigating the dynamic effects of energy consumption and economic growth on CO₂ emissions in 23 African countries by [96] arrived at comparable findings. The finding from the CS-DL estimator also indicates that a 1% surge in agricultural production leads to a 0.052 increase in emitted CO in the sampled countries. This increase is, however, statically insignificant. However, a study analyzing agricultural production and CO₂ emissions in the ECOWAS region using panel quantile regression techniques by Dimnwobi et al. [97] found that agricultural production increases the total CO₂ emissions, implying a shift from mechanized farming to traditional methods and the use of biomass as an energy source. The results were statistically significant across various quantiles.

Our research indicates that foreign direct investment (FDI), economic growth, energy consumption, and agricultural production are significant drivers of CO₂ emissions in the selected African countries. Specifically, FDI and economic growth have been shown to increase CO₂ emissions due to industrial activities and urbanization. Similarly, energy consumption, primarily from fossil fuels, and traditional agricultural practices contribute significantly to higher CO₂ levels. These findings highlight the urgent need for policies promoting sustainable development, renewable energy adoption, and environmentally friendly agricultural practices to balance economic growth with environmental preservation. Effective governance and regulatory frameworks are crucial in mitigating the adverse environmental impacts of these activities.

4.4. Causality Tests Results

In

Table 7. Causality test results.

Variables	Statistics	Hypothesis	Variables	Statistics	Hypothesis	Result
CO GDP	3.716 ***	Conservative	GDP CO	5.647 ***	Growth	Bidirectional
CO FDI	2.752	Conservative	FDI CO	3.184 ***	Neutral	Unidirectional
CO AG	3.528 ***	Conservative	AG CO	3.944 ***	Growth	Bidirectional
CO EC	2.952 **	Conservative	EC CO	5.905 ***	Growth	Bidirectional

**, *** denote a 5%, 1% level of significance.

. a panel causality analysis of 43 African countries reveals a complex interplay between CO₂, FDI, and economic growth. Employing the Dumitrescu–Hurlin test proposed by Dumitrescu and Hurlin [81], the study finds a unidirectional causality between CO₂ and FDI. Significant pollution and environmental degradation in certain African countries could prompt governments to implement unpredictable and stringent environmental regulations. This regulatory uncertainty may deter risk-averse foreign investors reluctant to operate in environments with unclear or expensive compliance requirements. In a related study by M. E. Bildirici et al. [98] and Hoffmann et al. [99], the authors also arrived at similar findings. The causality test result also revealed a bidirectional causality between GDP and CO₂; this suggests a feedback loop where growth characteristics can influence emissions and vice versa. Studies by Bildirici (2024) echo these findings. Additionally, a two-way causal link between energy consumption and CO₂ emissions was revealed. These findings are consistent with the findings of [100,101,102,103]. The causality test result also revealed a bidirectional causality between ag and CO₂, which is consistent with the findings of Ponce and Khan [104] and Tapacoba et al. [105], highlighting the need for sustainable farming practices that reduce the carbon footprint of agriculture while maintaining productivity. These findings emphasize the need for sustainable development strategies that promote clean FDI, green technologies, and decoupling growth from emissions. The causality test results for the study are illustrated in Table 7.

4.5. Impulse Response and Variance Decomposition

Within the realm of time series analysis, the impulse response function (IRF) proposed by Lanne and Nyberg [83] serves as a quantitative metric for elucidating the temporal evolution of a variable’s response following an exogenous shock. This function is pivotal in unraveling the dynamic interactions between various system variables. Notably, the IRF establishes a linkage between the impact of a one-standard-deviation shock on an innovation and the subsequent behavior of endogenous variables. Figure 2 visually depicts the impulse response; in the variation in CO₂ on itself, the initial two years witnessed a reduction in CO₂ emissions. This was followed by a slight increase in the third year. Interestingly, a period of relative stability ensued from year three to year ten but remained positive, showing CO has a persistent effect on itself over the 10-year forecast, suggesting that reduction policies need to take a long-term approach to be effective in reducing CO. The positive response of FDI to a CO impulse means that FDI increases initially in response to higher CO. After the third year, the relationship between CO emissions and FDI became neutral but remained slightly positive, indicating no substantial change in CO emissions. Since FDI coincided with a decrease in CO emissions, it has a small, sustained impact in increasing CO over the long run. The policy implication is that efforts to reduce carbon emissions should be coupled with policies that incentivize FDI in low-carbon or sustainable economic activities to avoid locking in high-emission development pathways.

Also, an impulse of GDP to a positive response to CO suggests that higher GDP leads to higher CO initially. However, this trend moderated from year two onwards, with a gradual increase in CO emissions observed throughout the remaining period. Strategies are required to decouple GDP from CO, e.g., funding low-carbon sectors, and regarding the correlation between agricultural and CO emissions, the initial two years demonstrated a positive association. Though the impact lessens with time, it remains positive, indicating agriculture sustains CO emissions. This suggests that higher carbon emissions may have positive impacts on the agricultural sector, potentially due to factors like increased atmospheric CO₂ levels or climate change-induced changes in precipitation and temperature. However, the long-term sustainability and resilience of the agricultural sector in the face of climate change should be a policy priority, as the sector is vulnerable to the negative impacts of environmental degradation. Energy consumption (EC) increases briefly after a positive CO impulse but falls flat from year 2 to year 3. The impulse response suggests that a positive shock to carbon emissions leads to a gradual increase in energy consumption over ten years. This indicates that higher carbon emissions are likely accompanied by increased energy use, potentially in carbon-intensive energy sources. From a policy perspective, this highlights the need to promote energy efficiency and the transition to renewable energy sources to decouple economic activities from carbon emissions.

During the initial two years of the study, a reduction in CO emissions has been demonstrated, potentially due to various policy changes or technological advancements. However, the subsequent years revealed a plateauing effect, suggesting the need for more aggressive strategies to achieve sustained reductions in CO emissions. The relationships between CI emissions and the influencing factors provided mixed results, highlighting the complex interplay of economic activity, resource consumption, and industrial practices in determining overall co levels. Further research is necessary to delineate the specific mechanisms through which these factors influence CO₂ emissions and identify optimal strategies for long-term reduction.

Moving on to the variance decomposition, shocks to one variable may cause variation in other variables in the system. The variance decomposition analysis results for a 10-year forecast period are presented in

Table 8. Variance decomposition test results.

	CO	GDP	FDI	AG	EC
CO	97.608	0.780	0.318	0.096	1.198
GDP	2.777	93.160	3.359	0.284	0.420
FDI	1.087	1.020	97.548	0.034	0.312
AG	0.008	1.106	0.576	98.242	0.068
EC	19.841	3.772	0.374	0.043	75.970

. Based on the sampled data, the innovative shocks in carbon emissions can account for 97.608% on itself. Simultaneously, GDP, FDI, AG, and EC contributed 0.780%, 0.318%, 0.096%, and 1.198%, respectively. FDI represented the most significant explanatory variable to GDP, with a value of 3.359%, whereas innovative shocks in carbon emissions, agriculture, and energy consumption accounted for 2.777%, 0.284%, and 0.420% of the variation. Also, for the studied African countries, innovative shocks in carbon emissions, economic growth, agriculture, and energy consumption explain 1.087%, 1.020%, 0.034%, and 0.312% of the variation in FDI; this implies that carbon emissions will contribute more than GDP, agriculture, and energy consumption. For the case of agricultural production, innovative shocks in carbon emissions, GDP, foreign direct investment, and energy consumption explain 0.008%, 1.106%, 0.576%, and 0.068% of the variation. It shows that for the case of energy consumption, innovative shocks in carbon emissions explain 19.841% of the variation, indicating a significant impact from environmental policies, while GDP contributes 3.772%, showing the influence of economic activities. FDI at 0.374% and agricultural production at 0.043 have minimal effects on energy consumption. In addition, GDP will have the most significant impact on agriculture in comparison to FDI, energy consumption, and carbon emissions. Overall, the findings indicate that carbon emissions and economic growth present a higher level of variation as compared to FDI and agriculture when the sampled variables are compared.

5. Conclusions and Policy Implications

The preceding literature has reasonably studied the influence of economic growth and energy consumption on the emitted CO₂ in Africa, which led to the adoption of policy implications, such as green growth and energy use policy. Little attention was paid to examining the role of foreign direct investment to reduce CO₂ and hence sustainable development. Existing studies showed how agriculture development contributes to economic growth; nevertheless, few studies attempted to detect the influence of agriculture development on CO₂. Responding to these deficiencies, the main goal of this study is to detect the influence of foreign direct investment, energy consumption, economic growth, and agriculture development on CO₂ across a panel of 43 African countries. The most recent econometric estimators and updated codes (CS-DL and AMG) have been employed to detect the relationships among variables. We, furthermore, applied the Dumitrescu–Hurlin causality test to detect the causal relationships between variables. The main findings of this article started by evaluating the variables, by employing the use of cross-sectional dependence, CIPS unit root, and Westerlund cointegration tests. Also, variance decomposition and impulse response analysis were adopted to ascertain how much the total natural resource rent, energy consumption, and economic development can exert on carbon emissions in the 10-year timeframe. The sampled panel dataset covered from 1990 to 2021. The findings of the current study are as follows:

• The cointegration findings verified the existence of long-term links among the variables, and the unit root was rejected in the level for CO₂ but rejected in the first difference for the other variables.
• The estimators reveal that economic growth and energy use have a significant positive influence on CO₂ in the long term. The AMG model also reported similar findings for agriculture production.
• The results of the estimators also reveal that boosting foreign direct investment translates to a positive influence on CO₂ emissions in the long term.
• The causality test reveals a unidirectional relationship between CO₂ emissions and FDI in the sampled countries. The test also revealed a bidirectional relationship between GDP and CO₂ emissions, as well as between energy consumption and CO₂ emissions. Again, a bidirectional causation was observed between agricultural production and CO₂ emissions.
• The impulse response analysis shows that GDP will contribute more to emissions over the 10-year forecast period, indicating the need for policies to decouple economic growth from CO₂ emissions.

Despite the improving socio-economic benefits, these variables can be detrimental to the environment if not well managed.

The profound effect of CO₂ emissions on the environment and the crucial role of econometric models as tools for analysis greatly affect decision making and the management of efforts to reduce global warming. Given this reason, it is crucial to provide a precise analytical approach that can comprehensively analyze CO₂ emissions. Therefore, our results aim to improve legislation on the reduction in carbon emissions and demonstrate the need for an integrated approach to planning green energy upgrades while managing carbon emissions. This technical framework will facilitate the process of economic, agricultural, and energy planning in Africa. It will help to enhance our knowledge of the connections between carbon reductions and the variables being studied. Moreover, this study proposes an economically efficient approach to encourage the use of environmentally friendly, low-emission energy sources and foster sustainable economic development.

Policy Implications

Causal relationships and dynamic interactions that have been observed emphasize the importance of a comprehensive approach to sustainable development. Policymakers should concentrate on strategies that promote renewable energy use by decoupling economic growth from CO₂ emissions and reduce economic activity’s environmental effect; promote energy efficiency by implementing incentives and regulations to encourage the use of energy-saving technology and practices, hence enhancing overall energy efficiency; promote responsible FDI by establishing a conducive environment for the investment of sustainable technologies and industries, while guaranteeing compliance with rigorous environmental regulations; and develop and promote sustainable agriculture practices that reduce environmental consequences and provide food security. By implementing these integrated strategies, African nations can preserve the environment for future generations and accomplish sustainable economic growth. Further research into these causal relationships and policy responses might help policymakers navigate Africa’s complex path to a sustainable future. Strong institutions and governance are needed to promote environmental sustainability and prevent increased energy use and climate change impacts. Enhancing the rule of law, regulatory integrity, government efficacy, and accountability, as well as encouraging stakeholder engagement and public involvement in environmental decision making, are all essential measures. While political organizations in Africa prioritize development and economic progress, environmental sustainability must be included in development plans and initiatives like the 2030 Agenda for environmental Sustainability. This agenda, which has 17 Sustainable Development Goals (SDGs), offers a common framework for peace and prosperity while tackling major global issues, including environmental degradation, poverty, inequality, and climate change. For successful environmentally friendly development, national policies must align with regional and international frameworks and agreements, like the African Union’s Agenda 2063 and the Paris Agreement on climate change. Because economic growth significantly impacts CO₂ emissions, states must set aside funds to invest in a green economy and low-carbon development. Inclusive growth, job creation, innovation, and resilience can be fostered by promoting resource efficiency, renewable energy, the circular economy, green infrastructure, sustainable agriculture and forestry, eco-tourism, and green finance. Africa’s energy demand is expanding at a rapid rate as its population surpasses 1.4 billion. Due to the widespread usage of firewood and charcoal in homes, there is an immediate need for more accessible and environmentally friendly energy sources. It is imperative to support environmental sustainability initiatives by enhancing capacity development, knowledge sharing, and technology transfer. This includes enhancing research and innovation, education and consciousness, data and information systems, technical support, and collaboration among African nations and international allies. These initiatives can be incorporated into Agenda 2063’s flagship programs, which will broaden the availability and scope of renewable energy generation in Africa.

The current study was limited by the unavailability of data for certain countries, the absence of a comprehensive dataset that restricted the analysis timeframe, and other factors that impact carbon emissions, such as technological advancements, changes in land use, climate policies and regulations, and afforestation and reforestation. Furthermore, while the CS-DL and AMG econometric methodologies used in this work are robust, it is important to acknowledge their possible limits. The extensive duration of the data might potentially include instances of structural shifts or alterations in linkages that current methodologies may not comprehensively account for. Factors such as the specification of the model, the quality of the data, the selection of the lag structure, and the sensitivity to outliers may also affect the dependability of the findings. Recognizing these constraints offers an impartial viewpoint on the results and emphasizes the need for future studies to tackle these difficulties. The next research will be one that is conducted with more granular data at regional and local levels to better understand the specific impacts of policy changes on CO₂ emissions.