Carbon Emissions and Influencing Factors in the Areas Along the Belt and Road Initiative in Africa: A Spatial Spillover Perspective

Abstract

The carbon dioxide spillover effects and influencing factors of the “Belt and Road Initiative” (BRI) in African countries must be assessed to evaluate the effectiveness, promote low-carbon transmissions in African countries, and provide recommendations for achieving the 2030 Sustainable Development Goals. This novel study employs carbon dioxide emission intensity (CEI) and per capita carbon dioxide emissions (PCE) as dual indicators to evaluate the spatial spillover effects of 54 BRI African countries on their neighboring countries’ carbon emissions from 2007 to 2023. It identifies the key factors and mechanisms affecting these spillover effects using the spatial differences-in-differences (SDID) model. Results indicate that since the launch of the BRI, the CEI and PCE of BRI African countries have significantly increased, largely due to trade patterns and industrialization structures. Greater trade openness has further boosted local economic development, thereby increasing carbon dioxide’s spatial spillover. Government management and corruption control levels show some heterogeneity in the spillover effects, which may be attributed to long-standing issues of weak institutional enforcement in Africa. Overall, this study reveals the complex relationship between BRI African economic development and environmental outcomes, highlighting the importance of developing sustainable development strategies and establishing strong differentiated regulatory regimes to effectively address environmental challenges.

1. Introduction

Reportedly, China has brought USD 421.15507.08 million in trade to the “Belt and Road Initiative” (BRI) African countries. By August 2022, the BRI had signed cooperation treaties with 52 African countries and organizations. Through this initiative, many prestigious Chinese state-owned enterprises have contributed to building numerous landmarks worldwide, including 6000 km of railways and highways, 20 ports, and 80 large power facilities. Between 2008 and 2021, China contributed USD 24 billion in aid to 22 countries that are “almost entirely” debtors of BRI infrastructure plans, including Angola, Egypt, Nigeria, Sudan, South Sudan, Kenya, and Tanzania [1,2,3]. However, trade and economic prosperity have also led to an increase in carbon dioxide (CO₂) emissions, posing a challenge to the United Nations Sustainable Development Goals (SDGs) for climate change mitigation.

Evidence from multiple perspectives—including carbon emissions [4,5], climate change [6,7], and renewable energy [8]—can be employed to assess the environmental impact of the BRI. However, most studies on the impact of its carbon emissions have focused on high-emission countries along the route, such as China, the United States, and Southeast Asia. Discussions about African countries often only cover a few high-emission nations, such as South Africa, Egypt, and Algeria. In addition, the factors influencing these impacts are not well understood [9], and insufficient attention has been given to how neighboring countries might also affect their own environmental quality. Furthermore, research on spatial spillover effects among African countries is scarce.

To address this gap, our study established a spatial econometric model for 54 African countries to examine the spatial spillover effects on carbon emissions from a global perspective. Based on this, we selected two indicators—CEI and PCE—to assess the various factors influencing carbon emissions and created a carbon spatial distribution map to evaluate the dynamic changes in carbon emissions from 2007 to 2023. From the perspective of African countries along the BRI route, this study offers policy recommendations to help governments deeply understand the impact of carbon emissions, providing an objective basis for China and African governments to implement differentiated policies and carbon reduction measures, and offering a reference for the sustainable development of Africa’s green economy.

The remainder of the article is structured as follows. Section 2 reviews the literature and proposes three hypotheses. Section 3 offers a comprehensive overview of the data and methodology adopted in the study. Section 4 explains the results, Section 5 provides the heterogeneity analysis, Section 6 describes the mediation effect analysis, Section 7 entails a discussion, and Section 8 highlights the conclusion and implications.

2. Literature Review and Hypothesis

2.1. The BRI and the Spillover Effect of Carbon Dioxide Emissions

Table 1. Literature on the influencing factors of carbon emissions in BRI countries.

Countries	Period	Model	Explain Variables	Influencing Factors	Reference
61BRI	2009–2016	LMDI	BRI × Post	CI, EXPORTS, SHAREPOL, POP, URB, DIST	[7]
65BRI	2008–2018	SDID	TCE, PCE, CEI	GDP, PGDP, URB, INS, OPD, FDI	[10]
67BRI	2013–2015	MRIO	PTT	CI, GDP, STR, lRE, INF	[11]
56 BRI	2005–2018	2TSF	lnCI	COFDI, GDPPC, EC, IND, URB, TRADE, PSI	[12]
65BRI	2005–2020	SDID	TFCEP	CI, GDP, STR, RE, INF	[13]
74BRI	2000–2018	SDM	CDE	FDI, PU, LU, EU, PIAV	[14]
22 BRI	2000–2019	Regression	CEI	LABOR, OPENNESS, ENERGY	[15]
32African	2002–2021	STIRPAT	INSTQ	REN, INSTQ, TO, UB	[16]
66BRI	2015–2019	DEA	CEE	CEQ, GDP, HLT, MED	[17]
61 BRI	1990–2020	CS-ARDL	ECG	FNT, GEN, NTR	[18]

Note: (1) Methods: DEA (data envelopment analysis), CS-ARDL (cross-section augmented Dickey–Fuller test with auto regressive distributed lag), SDM (spatial Durbin model), 2TSF (two-tier stochastic frontier models), SDID (spatial differences-in-differences), STIRPAT (a stochastic impacts by regression on population affluence and technology), MRIO (multiregional input-output); (2) Explained variables: CEE (the trade-embodied carbon emission efficiency), ECG(economic growth), CEI (carbon intensity), CDE (carbon dioxide emissions), lnCI (carbon intensity), TFCEP (total-factor carbon emission performance), PTT (pollution terms of trade). Influencing factors: CEQ (carbon inequality), GDP (gross domestic product), IND (industry development proportion of value added of industry in GDP), HIT (healthy level life expectancy at birth), MED (medical level domestic general government medical expenditure), FNT (FinTech), GEN (renewable energy generation), NTR (the optimum usage of natural resources), FDI (foreign direct investment), PU (population urbanization), LU (land urbanization), EU (economic urbanization), PIAV (portion of industrial added value), COFDI (The logarithm of the stock of China’s OFDI in BRI countries), GDPPC (GDP per capita), lnEC (energy consumption), INDI (industrialization), URB (urbanization), RADE (trade openness), PSI (political stability), CI (carbon intensity), EXPORTS (China’s exports to the partner), SHAREPOL (export share of polluting industries),DIST (distance between the partner and China), GDP (lnGDPP), lnSTR (dustrial structure), lnRE (annual renewable electricity production), lnINF (infrastructure development level), GVCs (Global Value Chains), REN (renewable energy), INSTQNSTQ (institutional quality index), TO (trade openness), UB (urbanization), PGDP (per capita GDP), URB (urbanization level), INS (industrial structure), OPD (opening degree), URB (urban population of total population), LABOR (labor force), POP (population), CR (corruption control).

summarizes the relevant literature [7,10,11,12,13,14,15,16,17,18,19]. Theoretical research on environmental issues related to the BRI has expanded, which can be categorized as follows. Some scholars have proposed that stringent environmental regulations, as suggested by the “pollution haven hypothesis,” could lead to the relocation of pollution-intensive industries from China to countries or regions with less stringent regulations. In contrast, the “pollution halo hypothesis” suggests that multinational companies are more likely to spread green technology in host countries and adopt unified environmental standards, which can help reduce emissions and promote energy conservation in these countries, similar to the positive environmental spillover effects of productivity and economic growth in host countries. These two hypotheses can coexist [10,11]. Wang et al. (2023) [12] empirically demonstrated that China’s Outward Foreign Direct Investment (OFDI) has a dual impact on carbon intensity in BRI countries, with an average net reduction efficiency of approximately 3%. The net effect of China’s OFDI varies per the economic development level of the host country, industrialization, energy expenditure, and urbanization.

In addition, studies have shown the spatial correlation of CO₂ emissions among BRI countries. For example, green development is promoted by reducing PCE and CEI, although CO₂ emissions in partner countries have increased [10]. Some scholars have also confirmed that the impact of OFDI on the total factor productivity of carbon emissions (TFCEP) in BRI countries varies across regions and exhibits spatial autocorrelation. Among these, Southeast Asian countries have the most significant impact, followed by Central Asian, West Asian, European, and African countries. Ethiopia, Laos, Slovenia, Singapore, and Latvia have relatively high levels of green development, while Algeria, Syria, Tunisia, India, Qatar, and Ukraine have relatively low levels [13]. These studies have focused on all BRI countries, with fewer African countries involved, making it difficult to provide comprehensive empirical evidence on the spatial spillover effects of carbon emissions in BRI African countries. Therefore, we propose

Hypothesis 1. 

The BRI may have spillover effects on carbon emissions across African nations.

2.2. Key Factors of the BRI Affecting CO₂ Spillover Effects in Africa

Strict environmental regulations in one country may lead to heavy polluting enterprises relocating to “adjacent” areas without such regulations, exacerbating pollution in these regions. The research factors listed in Table 1 most commonly include economic development, urbanization, trade, education, industrial structure, and technology [11,12]. For example, Wei et al. (2023) [14] found that, particularly after the implementation of the BRI, foreign direct investment reduced this direct impact and enhanced the benefits of carbon reduction in neighboring areas. Urban systems are important spaces for CO₂ emissions, and OFDI affects carbon emissions through scale effects, technology spillovers, and industrial division of labor. These findings highlight the importance of integrating emission reduction policies in terms of population, economy, urbanization, and industrial and foreign investment, and provide strong evidence for joint governance of carbon emissions among BRI countries. Similarly, a study used spatial econometrics and the super-EBM model to determine the carbon emissions efficiency of 136 nations between 2000 and 2019. It showed that energy efficiency and carbon reduction efforts should focus on Uzbekistan, Iran, Mongolia, Ukraine, and Angola. On average, developed rather than developing nations are more efficient in reducing carbon emissions. While factors like industrial structure and the proportion of electricity users limit carbon emission efficiency, urbanization level, foreign commerce, and renewable energy effectively increase it and affect the overflow effect of CO₂ in this region [20].

Accordingly, we propose

Hypothesis 2. 

Economics, urbanization, industrialization, and technology are the key factors impacting the spillover effects of CO₂ in Africa under the BRI.

2.3. The Influence of the BRI on CO₂ Spillover Effects in African Regions Depends on Institutional Quality

Many African nations have pledged commitment to the 2030 Sustainable Development Goals, the Agenda for Sustainable Development, and the Agenda 2063, all of which emphasize reducing CO₂ and other harmful emissions. These initiatives reflect a strong political will among African leaders to combat climate change. For instance, Kwakwa et al. (2023) [16], using panel data from 32 African countries between 2002 and 2021, demonstrated that improvements in institutional quality, measured by indicators such as corruption control, rule of law, political stability, regulatory quality, government effectiveness, and voice and accountability are associated with reductions in CO₂ emissions. However, many African countries still face significant implementation challenges due to fiscal constraints, governance inefficiencies, and corruption. For example, Nigeria has historically subsidized gasoline and diesel to support economic growth. While such subsidies may offer short-term gains, they also risk increasing long-term emissions by reinforcing fossil fuel dependency. Ghana’s national environmental policy emphasizes sustainable development by encouraging community involvement in natural resource management and promoting renewable energy. Similarly, although South Africa remains heavily reliant on coal, it has made notable progress in developing solar and wind energy. Government policy has accelerated growth in these sectors, contributing to CO₂ reduction [21]. Accordingly, we propose:

Hypothesis 3. 

The effect of the Belt and Road Initiative (BRI) on CO₂ spillover in African regions is moderated by institutional quality and corruption control indicators.

2.4. Gaps in the Literature

The institutional treatment in BRI and non-BRI countries has not emphasized marginal effects in different regions. Spatial heterogeneity analysis and influencing factors should be performed to explore the effects of the BRI in the future. More importantly, the SDID model should be extended to other areas, such as the impact of CO₂ spillover effects in African regions depending on economics, industrial structure, institutional quality, and corruption control indices. This study confirmed the BRI’s important role in the environmental outcomes–economic development link in Africa. Such analyses help determine the key factors and mechanisms affecting these spillover effects. Although studies are increasing, the evidence from Africa remains insufficient. This study addresses these identified gaps.

3. Methods

3.1. Study Area, Data Source, and Methodology

Egypt was the first among Africa’s 54 nations to join the BRI in 2016. Morocco and Madagascar joined in 2017 after signing cooperation agreements. Malawi, the most recent African nation to join the program, enlisted in 2022. Only Swaziland and Mauritius have not joined the BRI, as shown in

Table A1. Situation of African countries joining the Belt and Road Initiative.

Country	The National Abbreviation	Time Joined
Egypt	EGY	2016
Madagascar	MDG	2017
Morocco	MAR	2017
Algeria	DZA	2018
Angola	AGO	2018
Burundi	BDI	2018
Cabo Verde	CPV	2018
Cameroon	CMR	2018
Chad	TCD	2018
Congo, Rep.	COG	2018
Cote d’Ivoire	CIV	2018
Djibouti	DJI	2018
Ethiopia	ETH	2018
Gabon	GAB	2018
The Republic of The Gambia	GMB	2018
Ghana	GHA	2018
Guinea	GIN	2018
Kenya	KEN	2018
Libya	LBY	2018
Mauritania	MRT	2018
Mozambique	MOZ	2018
Namibia	NAM	2018
Nigeria	NGA	2018
Rwanda	RWA	2018
Senegal	SEN	2018
Seychelles	SYC	2018
Sierra Leone	SLE	2018
Somalia	SOM	2018
South Africa	ZAF	2018
South Sudan	SSD	2018
Sudan	SDN	2018
Togo	TGO	2018
Tunisia	TUN	2018
Uganda	UGA	2018
Tanzania	TZA	2018
Zambia	ZMB	2018
Zimbabwe	ZWE	2018
Benin	BEN	2018
Comoros	COM	2018
Equatorial Guinea	GNQ	2018
Lesotho	LSO	2018
Liberia	LBR	2018
Mali	MLI	2018
Botswana	BWA	2021
Burkina Faso	BFA	2021
Central African Republic	CAF	2021
Eritrea	ERI	2021
Congo, Dem. Rep.	COD	2021
Guinea-Bissau	GNB	2021
Niger	NER	2021
Sao Tome and Principe	STP	2021
Malawi	MWI	2022
Mauritius	MUS	Not joined
Eswatini	SWZ	Not joined

in Appendix A. In this study, 43 African nations that signed agreements before 2020 are included in the experimental group. The control group includes 11 African nations: 9 that signed after 2020 and 2 that did not sign at all. The data used for this study spans the period 2007–2023. Figure 1 shows CO₂ emissions and the geographical distribution of nations in the BRI and those that are not.

3.2. Assessment Indicators and Data Sources

Complete and balanced panel datasets spanning 54 nations from 2007 to 2023 were acquired after screening and deleting countries with missing data. Every variable was adjusted to a 2010 fixed-price base year to guarantee uniform measurement standards.

Table 2. Summary statistics of key variables.

Variable	Description	Unit	Observation	Mean	SD	Min	Median	Max
CEI	CO2 emissions intensity	kg CO2/USD	918	0.369	0.296	0.003	0.298	4.071
PCE	Per capita CO2 emissions	Ton CO2/person	918	1.085	1.735	−0.039	0.384	9.986
DID	Policy × time	-	918	0.279	0.449	0.000	0.000	1.000
PGDP	Per capita GDP	USD	918	2.585	3.159	0.171	1.336	19.850
URB	Urban population	person	918	0.928	1.489	0.004	0.390	12.137
EDU	School enrollment, primary	%	918	0.980	0.352	0.085	0.989	4.302
TI	Scientific and technical journal articles	piece	918	0.939	2.737	0.000	0.084	28.227
NI	Government effectiveness	-	918	−0.636	0.681	−2.350	−0.702	1.050
CC	Control of corruption	-	918	−0.656	0.652	−1.940	−0.705	1.800
PS	Political stability and absence of violence/terrorism	-	918	−0.630	0.903	−3.310	−0.490	1.200
Trade	Sum of exports and imports of goods and services	% of GDP	918	0.752	0.554	0.027	0.646	7.210
INS	Industry construction	% of GDP	918	27.796	24.670	−14.240	23.415	270.040

describes the study variables.

The dependent parameter (CEI, PCE) comprises the CO₂ emissions of regions or countries along the BRI. CEI measures CO₂ emissions relative to gross domestic product (GDP). Locations with lower CEI are often predicted to see improved product competitiveness following the enforcement of an international carbon price. Thus, to comprehensively analyze the regional spillover effects of carbon emissions in BRI countries, this study incorporates both PCE and CEI indicators, acknowledging the limitations of employing a single CO₂ emission metric [10].

In the two-fold difference model, the central explanatory variable is the interaction term BRIi × POSTt. Its value is 1 if area i became part of the BRI at time t, and 0 otherwise. The BRI, which Egypt and China joined in 2016, may thus indirectly impact carbon emissions. This leads to the following policy time points: 2016, 2019, 2020, and 2021.

Control variables comprise economic growth’s significant impact on CO₂ emissions. During the early growth phases, emissions increased significantly with PGDP [15,19,20]. However, policies and funding to support high-tech research and development in BRI countries in Africa are limited, likely due to their weak economic power. In this study, scientific and technical research (IT) is measured by articles in scientific and technical journals [21].

An inverted U-shape is evident between urbanization (URB) and CO₂ emissions [22]. Rapid urban economic growth has increased the demand for resources, fuel, and infrastructure, further increasing carbon emissions. In this study, URB was measured by the total urban population [23]. Education for poverty reduction and various low-carbon technology trainings are also key factors in urban population growth and mitigating carbon emissions. In this study, the lever of education for poverty reduction is measured by primary school enrollment rates (EDU) [24].

Regarding institutional variables, quality improvement in the environment, including CO₂ abatement, necessitates government effectiveness (NI). High-quality institutions, regulations, the rule of law, along with efficient governance, prompt industries and firms to observe eco-friendly practices [25]. Many African countries struggle to effectively implement environmental policies due to fiscal constraints, misplaced priorities, and corruption control (CC) [21]. In this study, the institutional variables were selected as the NI and CC.

Trade implies that many countries participate in international trade and become part of the global economy, whereas industries that rely heavily on technology prioritize the use of cutting-edge tools to boost output and efficiency. Industrial structure (INS) not only directly impacts carbon dioxide emissions but also indirectly affects emissions through improved energy efficiency [26]. This study selected Trade and INS as the mediating variables.

3.3. Data Description

The model, equations, and main variables are presented in Table 2. Due to data availability, the sample period for the study was 2007–2023, for which data of 54 African households were obtained, including 52 from countries along the BRI. The CO₂ emissions data, including CEI and PCE, were collected from the International Energy Agency (https://www.iea.org/data-and-statistics/data-tools/energy-statistics-data-browser?country=WORLD&fuel=CO2%20emissions&indicator=CO2BySource (accessed on 5 December 2024)). Additionally, socio-economic data were totally collected from the World Bank (https://databank.worldbank.org/id/adc593a?Report_Name=WDI (accessed on 5 December 2024)), and all variables were converted into the constant prices of the 2010s. Missing data were interpolated, resulting in balanced panel datasets.

3.4. Spatial Autocorrelation

Based on Li et al. (2024) [13], we construct four distinct spatial weight matrices to provide a comprehensive representation of the spatial spillover effects among various BRI countries and enhance the robustness of the empirical findings. These matrices include a conventional weight matrix based on an inverse geodesic distance (W1), inverse geodesic distance square matrix (W2), economic distance weight matrix (W3), and economic–geographic nested matrix (W4). The equation for calculating the matrix is as follows:

$$W_1=\{\begin{array}{c}{\displaystyle \frac{1}{\hspace{0.17em}{d}_{ij}}},i\ne j\\ 0,i=j\end{array} W_2=\{\begin{array}{c}{\displaystyle \frac{1}{d_{ij}^2}},i\ne j\\ 0,i=j\end{array} W_3=\{\begin{array}{c}{\displaystyle \frac{1}{\sqrt{gd{p}_i-gd{p}_j}}},i\ne j\\ 0,i=j\end{array} W_4=0.5W_1+0.5W_3$$

(1)

where dij denotes the geographical distance from the capital of BRI nation i to that of BRI nation j.

To evaluate these spatial relationships, this study uses the global Moran’s I, which is used for worldwide data. As shown in Equation (2), N stands for the sum of all samples. The spatial weight matrix is represented by Wij, which is 1 when i and j are adjacent, and 0 otherwise. In addition, the z-statistic, which measures the discrepancy between the actual and expected values of Moran’s I and then divides it by the standard deviation, confirms the relevance of the index [10].

$${\displaystyle \frac{n·{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^n{w}_{ij}·{{\displaystyle \sum }}_{j=1}^n{\left(x_i-\overline{x}\right)}^2}}$$

(2)

3.5. The SDID Model

Compared to the benchmark DID model, the SDID model effectively compensates for the insufficient observation of the intensity and direction of policy effects on the control group by examining spatial spillover effects. We established the following conventional double-difference model and analyzed policies related to BRI based on the literature [27,28,29].

$$Y_{it}={\beta }_0+\alpha BRIi\times POSTt+{\beta }_1\mathrm{Xit} +\mu i+\upsilon t+\epsilon it$$

(3)

In Equation (3): i signifies country, t signifies year, and Yit indicates carbon emissions. The effect on carbon emissions is indicated by the estimated coefficient α for the key explanatory parameter, the interaction term αBRIi × POSTt. The set of control variables is represented by Xit, the temporal fixed effect by υt, the individual fixed effect by μi, and the random error term by εit. This high connectivity may mean that changes in elements within a single region can affect both its own economic development and that of neighboring countries. To address this, we constructed an SDID Model (4) [10,13].

Model (4) uses spatial lag terms from Model (3), with W being the spatial weight matrix, $\rho$ denoting spatial auto-correlation of CO₂ emissions, $\beta$ the spillover effects of BRI on carbon intensity in African regions or countries, δ the spillover effects of other control variables, and λ the spatial correlation between random disturbance terms.

$$Y_{it}=\rho \mathrm{W}\times \mathrm{CDit}+\alpha BRIi\times POST\mathrm{t}+\beta \mathrm{WBRIi}\times \mathrm{POSTt}+\mathsf{\gamma }\mathrm{Xit}+\delta \mathrm{W}\times \mathrm{Xit}+\mu +\nu +\left(\mathrm{I}-\lambda \mathrm{W}\right)-1\epsilon$$

(4)

4. Empirical Results

4.1. Impact of BRI on Carbon Dioxide Emissions

This study evaluated the change in carbon emissions for the period 2007–2023 for 54 African countries. Figure 1a–c illustrates that since the policy implementation in 2013, the CEI and PCE in BRI countries in Africa are significantly higher than in non-BRI countries. Note that during this period, although the growth trajectory remained relatively unchanged, that for PCE declined from 2016. Since 2016, the BRI has emphasized green development, and economic and technological progress have had an increasingly critical impact on improving PCE.

4.2. Results of the Spatial Auto-Correlation Test

In this study, we employed the global Moran’s I index to analyze the spatial autocorrelation of carbon dioxide emission levels from 2007 to 2023; the results are shown in Figure 2. The global Moran’s I index ranges from −1 to 1, with larger absolute values indicating stronger spatial relationships. If the index is greater than 0, it suggests a positive spatial clustering pattern; if the index is less than 0, it indicates negative spatial correlation. When z-statistic > 2.58 (corresponding to a 99% confidence level) or Z > 1.96 (95% confidence level), it demonstrates statistically significant spatial clustering. It can be observed that global spatial correlations of CEI (see Figure 2 CEI) steadily decreased for W1 from 0.074 to 0.003, for W2 from 0.252 to 0.037, and for W4 from 0.054 to 0.004, a result that is statistically significant at the 1% level (z-statistic < 2.58); for PCE (see Figure 2 PCE), the values of W1 and W2 were between 0.015 and 0.018 or 0.065 and 0.09—both of the changes are irregular, whereas the values of W3 were between 0.272 and 0.199, both of which are greater than the spatial correlation values obtained in relation to CEI. Moran’s I index values were found to be positive at the significant level of 1% and the z-statistics were greater than 2.58, indicating the presence of positive spatial correlation for PCE in the carbon emissions indicators. This finding means that a country with high CO₂ emissions is always surrounded by other large carbon emitters, and vice versa. The spatial autocorrelation results also reveal the presence of spatial spillover effects among countries with respect to CO₂ emissions, which reinforces the necessity of using an SDID model when investigating the pattern of CO₂ emissions in order to avoid biased results.

Figure 3 and Figure 4 illustrate the local Moran LISA clustering maps for CEI and PCE for each country in 2016, 2017, 2018, 2019, 2021, and 2022. CEI is shown in Figure 3.

The results of the spatial weight matrix analysis are similar—all are within the low-low and high-high clustering in the first and third quadrants. With the implementation of the BRI, the national carbon emission intensity changes from high-high to low-low, showing positive spatial correlations between the BRI and the regions or countries it runs through. For PCE, as indicated in Figure 4, most countries are within the first and third quadrants. In addition, as the BRI is implemented, the PCE in various nations changes from high-high to low-low and high-low. The most significant changes occur in 2022, indicating that with a changing population and economy, the carbon intensity of African countries shows spatially related characteristics, with significant differences in different regions. This strengthens the need to use SDID models when investigating CO₂ emission patterns to prevent biased outcomes.

4.3. Benchmark Regression Results and Analysis

4.3.1. Impact of the BRI on CO₂ Emissions

Table 3. Baseline Regression Results for CEI Variable.

CEI Model	SDID Model			Regular DID Model
	W1	W2	W4
	(1)	(2)	(3)	(4)	(5)
Main
BRI × POST	0.0777 ***	0.0488 **	0.0556 ***	0.0357	0.0306 **
	(3.3023)	(2.1856)	(2.6922)	(0.0255)	(0.0137)
PGDP	−0.0412 ***	−0.0442 ***	−0.0367 ***		−0.0381 **
	(−8.2677)	(−8.4784)	(−7.9871)		(0.0153)
URB	−0.0054	−0.0069	−0.0093		−0.0205
	(−0.3233)	(−0.4137)	(−0.5848)		(0.0125)
EDU	0.0028	0.0045	0.0139		−0.0034
	(0.1208)	(0.1915)	(0.6374)		(0.0123)
TI	−0.0076 *	−0.0158 ***	0.0042		−0.0175 ***
	(−1.8625)	(−3.8575)	(0.9898)		(0.0026)
NI	−0.0504	−0.0694 **	−0.0289		−0.0842 *
	(−1.5978)	(−2.1878)	(−0.9445)		(0.0440)
CC	−0.0048	−0.0074	−0.0181		−0.0439 *
	(−0.1626)	(−0.2470)	(−0.6064)		(0.0227)
W × BRI × POST	1.1749 ***	0.4488 ***	0.7347 ***
	(5.2115)	(5.8774)	(5.5523)
PGDP	−0.0022	0.0011	−0.0279 ***
	(−0.0804)	(0.1256)	(−2.6564)
URB	−0.6225 ***	−0.3005 ***	−0.1370
	(−2.9688)	(−4.0906)	(−0.6537)
EDU	−0.1897	−0.0314	−0.2682
	(−0.7673)	(−0.3660)	(−1.5165)
TI	0.6644 ***	0.2664 ***	0.4732 ***
	(11.6301)	(10.4016)	(11.8806)
NI	0.5567 **	0.1502 *	0.7484 ***
	(2.3737)	(1.7101)	(2.9597)
CC	−0.1869	−0.0844	−0.1068
	(−0.9657)	(−1.0606)	(−0.5093)
Spatial rho	−0.8872 ***	−0.2017 ***	−0.6983 ***
	(−4.5078)	(−2.6623)	(−6.7555)
Variance sigma2_e	0.0153 ***	0.0161 ***	0.0139 ***
	(21.0266)	(21.3416)	(24.0230)

t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

and

Table 4. Baseline Regression Results for PCE Variable.

PCE Model	SDID Model			Regular DID Model
	W1	W2	W3
	(6)	(7)	(8)	(9)	(10)
Main
BRI × POST	0.1605 **	0.1173 *	0.1279 **	−0.2056	0.0755 *
	(2.4466)	(1.9277)	(2.2332)	(0.1312)	(0.0376)
PGDP	0.2329 ***	0.2260 ***	0.2324 ***		0.2434 ***
	(16.8123)	(15.9625)	(18.8503)		(0.0360)
URB	0.1285 ***	0.1018 **	0.1443 ***		0.1414 ***
	(2.7698)	(2.2449)	(3.2269)		(0.0318)
EDU	0.1375 **	0.0770	0.0804		0.0998 *
	(2.1169)	(1.2082)	(1.2978)		(0.0482)
TI	−0.0751 ***	−0.0693 ***	−0.0650 ***		−0.0734 ***
	(−6.7814)	(−6.3614)	(−6.2445)		(0.0114)
NI	0.1521 *	0.1983 **	0.2207 ***		0.1453 **
	(1.7322)	(2.3051)	(2.6736)		(0.0653)
CC	0.1184	0.0818	0.1658 **		0.1504 *
	(1.4329)	(0.9996)	(2.1241)		(0.0849)
W × BRI × POST	1.8863 ***	0.5961 ***	1.0668 ***
	(3.0051)	(2.8720)	(5.5857)
W × PGDP	0.0403	−0.0003	−0.3789 ***
	(0.4875)	(−0.0107)	(−6.3254)
W × URB	1.2712 **	0.7140 ***	0.6045 ***
	(2.1831)	(3.5839)	(2.8490)
W × EDU	1.9049 ***	0.6125 ***	1.1398 ***
	(2.7791)	(2.6283)	(3.9722)
W × TI	−0.4735 ***	−0.2434 ***	−0.1011 **
	(−3.0124)	(−3.5091)	(−2.2077)
W × NI	−1.6239 **	−0.5727 **	0.2589
	(−2.4881)	(−2.3994)	(1.0089)
W × CC	1.6057 ***	0.5468 **	0.4812 *
	(2.9494)	(2.5038)	(1.7668)
Spatial rho	−0.2820 *	−0.0312	0.5305 ***
	(−1.7147)	(−0.4686)	(8.7147)
Variance sigma2_e	0.1184 ***	0.1184 ***	0.1029 ***
	(21.3512)	(21.4219)	(20.5898)

t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

report the regression results; the SDID model and conventional DID model are used, respectively. Columns (1) to (3) use the SDID model that includes W1, W2, W3, or W4 and three matrices. Columns (4) and (5) use the conventional DID model. From columns (1) to (3), we can see clearly that the estimated coefficients for BRI × POST are positive under the three weights of W1, W2, W3, or W4. The coefficients of W × BRI × POST under three different weights are also significantly positive, indicating that the BRI has significantly strengthened the CEI and PCE among BRI countries. Columns (4) and (5) show that the BRI × POST coefficients based on the conventional DID model are also all positive, indicating that the BRI has continuously enhanced the carbon dioxide spillover in BRI countries; even without considering spatial correlation, it is noteworthy that the estimated coefficient of BRI × POST in the SDID model is significantly higher compared to the conventional DID model.

For the W3, the estimation result of PCE rho is positive, and the variables have a significant impact, which is significantly higher than that of other matrices. Therefore, regarding spatial correlation, the BRI greatly influences increasing CO₂ emissions across the African regions along the route. The initial finding that the practice of the BRI elevated carbon emissions in participant nations, consistent with our results when estimating Moran’s I, confirms that African countries geometrically adjacent to each other have significant spillover relationships.

Furthermore, to explore the dynamic influence of the BRI over carbon emissions annually during the research duration, the interaction item DID was divided into six years after 2016 with a multi-period SDID model.

Table 5. Regression results of dynamic SDID model.

Year	CEI	PCE
2016	−0.0097	0.1540 ***
	(0.0126)	(0.0487)
2017	−0.0055	0.1224 **
	(0.0110)	(0.0457)
2018	0.0008	0.1088 **
	(0.0101)	(0.0439)
2019	0.0046	0.1338 **
	(0.0111)	(0.0485)
2020	0.0076	0.1201 *
	(0.0151)	(0.0586)
2021	0.0322 **	−0.0358
	(0.0128)	(0.0529)
W*Y
_cons	0.3659 ***	0.6815 ***
	(0.0402)	(0.0739)
N	918	918
r2	0.7757	0.9553

t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

displays relevant results. The present analysis revealed that its impact on CEI has steadily increased since 2016, with the initiation of African countries joining the BRI. Since many BRI participants are less developed or developing countries, the practice of this initiative promoted economic advancement and exacerbated CO₂ emissions generated by the host country. In contrast, the effect on PCE was significant in the first year, and PCE drastically decreased among participating nations. PCE was found to be significant at a level of 1% or 5%. The impact on PCE and CEI proves that the practice of the BRI promotes the active adjustment of policies to promote green development after discovering the influence of the host state on increasing the CO₂ spatial spillover effect. Finally, these countries participate in reducing national emissions and low-carbon economic development.

4.3.2. Marginal Influence of the Determinants of CO₂ Emissions

This study estimated the indirect and direct influences of independent variables on carbon emissions. Figure 5a,b presents relevant results. Regarding the marginal effects of PGDP on CEI and PCE, the marginal effects of per capita GDP on CEI and PCE exhibit stronger impacts on W1 and W2 across the four matrices. Within the same matrices, W1 and W2, their spillover effects demonstrate certain differences, with the total effect of CEI being negative while the total effect of PCE is positive at the 1% significance level. The direct effect of PCE outweighs its indirect effect, indicating that the rapid economic growth of the domestic BRI is a key factor in increasing its spatial spillover of carbon dioxide to neighboring countries. Strict implementation of differentiated low-carbon policies and promoting green economic development are essential pathways for reducing carbon emissions in African BRI countries.

For URB and EDU, the marginal effects of both on CEI and PCE demonstrate similar overall trends across the four matrices; the total effect on CEI emissions indicators is negative, implying that the practice of the BRI accelerated the urbanization of emission reduction efficiency, improved the energy-saving education level, and reduced carbon emissions in participating African nations. However, the total impact on PCE emission indicators is positive at the 1% or 5% significance level, and the indirect effect is greater than the direct effect. Alongside Africa’s growing population and the increasing size of its cities, the rapid increase in per capita emissions, domestic use of non-renewable energy fuels, and lower education levels among impoverished populations hinder the widespread adoption of energy-saving technologies and have indirectly contributed to an elevated risk of carbon dioxide spatial spillover from African BRI countries to their neighboring nations.

For IT, the marginal effects of both CEI and PCE show similar overall trends across three matrices. It exhibits positive marginal effects on CEI and more pronounced negative effects on PCE. The total effect and indirect effect are positive at the 1% significance level, with the indirect effect being greater than the direct effect.

Against this backdrop, the application of green emission reduction technologies under the BRI has optimized and reduced carbon emission levels in the local and surrounding regions. Although the impact on the carbon intensity index is relatively weak, neighboring areas may still benefit from the demonstration effect of learning green low-carbon emission reduction technologies, thereby reducing the spatial spillover effects of carbon dioxide.

However, as previously mentioned, high-quality institutional systems characterized by national NI and CC compel businesses and industries to adhere to environmental practices. For national institution management, the marginal effects of the two indicators, CEI and PCE, differ across the three matrices, with CEI showing a positive marginal effect for the W4 matrix and PCE for the W3 matrix, where indirect effects are greater than direct effects. The marginal effect of CEI is positive at the 1% or 5% significance level. This enhancement in the enforcement of green economic system policies prompts countries to formulate differentiated sustainable development policies, implement carbon tax reforms, and reduce spatial spillover from high-emission countries to neighboring nations, thereby alleviating the growing problem of carbon dioxide spatial spillover.

Corruption issues in the use of environmental funds by governments are institutional challenges for African BRI countries. This study found that CC has opposite directional responses to the marginal effects of the two indicators, CEI and PCE, across the three matrices, where the positive marginal effect of PCE far exceeds the negative effect on CEI, with indirect effects being greater than direct effects, significant at the 1% or 5% level. This indicates that corruption in the use of energy-saving and emission-reduction funds by governments severely restricts high-emission countries’ investments in new energy technologies and high-tech environmental talent training. This hinders the green and sustainable development of BRI African countries and indirectly exacerbates spatial spillover to neighboring countries, worsening the situation of carbon dioxide spatial spillover in Africa.

4.4. Robustness Check

4.4.1. PSM-DID

The model was further estimated using a combined approach of propensity score matching (PSM) and SDID. Nearest-neighbor matching with replacement was performed at a 1:3 ratio, yielding the Logit regression results. The results indicate that the t-tests for all covariates were insignificant, suggesting that after PSM, the null hypothesis of no significant difference between the treatment and control groups cannot be rejected, and the sample’s self-selection bias was significantly reduced. Further regression analyses were conducted for Model (4) under four matrices. The PSM-SDID regression results are presented in

Table 6. (a) PSM-SDID estimation results for CEI; (b) PSM-SDID estimation results for PCE.

(a)
Variable	CEI
	W1	W2	W3	W4
	(11)	(12)	(13)	(14)
LR_Direct
x	0.0559 **	0.0378 *	0.0177 *	0.0480 **
	(0.0224)	(0.0227)	(0.0244)	(0.0207)
PGDP	−0.0508 ***	−0.0524 ***	−0.0378 ***	−0.0403 ***
	(0.0052)	(0.0054)	(0.0054)	(0.0048)
URB	−0.0023	−0.0104	−0.0160	−0.0244
	(0.0167)	(0.0172)	(0.0183)	(0.0153)
EDU	0.0063	0.0010	0.0094	0.0197
	(0.0226)	(0.0229)	(0.0256)	(0.0215)
TI	−0.0294 ***	−0.0290 ***	−0.0228 ***	−0.0075 *
	(0.0049)	(0.0050)	(0.0047)	(0.0041)
NI	−0.0695 **	−0.0745 **	−0.1070 ***	−0.0361
	(0.0326)	(0.0329)	(0.0358)	(0.0316)
CC	−0.0026	−0.0081	−0.0428	−0.0251
	(0.0283)	(0.0285)	(0.0307)	(0.0278)
LR_Indirect
x	0.6516 ***	0.4085 ***	−0.2160 ***	0.4475 ***
	(0.1249)	(0.0650)	(0.0685)	(0.0737)
PGDP	−0.0134	−0.0011	−0.0117	−0.0127 *
	(0.0168)	(0.0084)	(0.0276)	(0.0069)
URB	−0.5752 ***	−0.3527 ***	−0.1515 *	−0.2612 **
	(0.1289)	(0.0668)	(0.0828)	(0.1210)
EDU	−0.0290	−0.0008	−0.3365 ***	−0.1150
	(0.1306)	(0.0758)	(0.1103)	(0.1069)
TI	0.3656 ***	0.2374 ***	0.0183	0.2879 ***
	(0.0418)	(0.0237)	(0.0171)	(0.0260)
NI	0.2909 **	0.1320 *	0.0895	0.2377
	(0.1172)	(0.0708)	(0.0892)	(0.1461)
CC	−0.1198	−0.0869	0.1207	0.0021
	(0.1019)	(0.0673)	(0.0998)	(0.1186)
LR_Total
x	0.7074 ***	0.4463 ***	−0.1984 ***	0.4956 ***
	(0.1312)	(0.0715)	(0.0763)	(0.0766)
PGDP	−0.0642 ***	−0.0535 ***	−0.0495 *	−0.0529 ***
	(0.0166)	(0.0091)	(0.0276)	(0.0070)
URB	−0.5776 ***	−0.3631 ***	−0.1674 *	−0.2856 **
	(0.1311)	(0.0679)	(0.0878)	(0.1224)
EDU	−0.0226	0.0002	−0.3271 ***	−0.0952
	(0.1275)	(0.0735)	(0.1103)	(0.1036)
TI	0.3362 ***	0.2085 ***	−0.0045	0.2804 ***
	(0.0427)	(0.0242)	(0.0168)	(0.0259)
NI	0.2214 *	0.0574	−0.0176	0.2016
	(0.1170)	(0.0723)	(0.1002)	(0.1456)
CC	−0.1224	−0.0950	0.0779	−0.0230
	(0.0981)	(0.0663)	(0.1050)	(0.1201)
N	918	918	918	918
r2	0.0146	0.0105	0.0624	0.0337
(b)
Variable	PCE
	W1	W2	W3	W4
	(15)	(16)	(17)	(18)
LR_Direct
x	0.1323 **	0.1114 *	0.2179 ***	0.0925 *
	(0.0648)	(0.0621)	(0.0681)	(0.0592)
PGDP	0.2487 ***	0.2379 ***	0.2207 ***	0.2543 ***
	(0.0144)	(0.0146)	(0.0152)	(0.0133)
URB	0.1585 ***	0.1293 ***	0.2189 ***	0.1637 ***
	(0.0462)	(0.0459)	(0.0542)	(0.0443)
EDU	0.1173 *	0.0803	0.1758 ***	0.1058 *
	(0.0620)	(0.0615)	(0.0681)	(0.0591)
TI	−0.0635 ***	−0.0611 ***	−0.0679 ***	−0.0781 ***
	(0.0118)	(0.0118)	(0.0119)	(0.0114)
NI	0.1782 **	0.2023 **	0.2607 ***	0.1279
	(0.0889)	(0.0883)	(0.0967)	(0.0878)
CC	0.1052	0.0807	0.2055 **	0.1358 *
	(0.0771)	(0.0767)	(0.0822)	(0.0785)
LR_Indirect
x	1.3267 ***	0.5378 ***	2.4504 ***	0.3285 *
	(0.5003)	(0.1992)	(0.5456)	(0.3170)
PGDP	0.0896	0.0187	−0.5178 ***	0.0315
	(0.0662)	(0.0243)	(0.1867)	(0.0271)
URB	1.6654 ***	0.8831 ***	1.4288 ***	0.8042
	(0.5324)	(0.2044)	(0.5188)	(0.5241)
EDU	1.2668 **	0.5581 **	2.5553 ***	0.5849
	(0.5671)	(0.2368)	(0.7359)	(0.4541)
TI	−0.3495 ***	−0.2377 ***	−0.2839 ***	−0.2338 **
	(0.1278)	(0.0680)	(0.1038)	(0.0975)
NI	−1.2095 **	−0.5523 **	0.8146	−2.3760 ***
	(0.4892)	(0.2211)	(0.5948)	(0.6442)
CC	1.3042 ***	0.5719 ***	1.1912 *	1.4048 ***
	(0.4315)	(0.2106)	(0.6213)	(0.5027)
LR_Total
x	1.4590 ***	0.6492 ***	2.6684 ***	0.4209 *
	(0.5283)	(0.2221)	(0.5882)	(0.3339)
PGDP	0.3383 ***	0.2567 ***	−0.2971	0.2858 ***
	(0.0678)	(0.0278)	(0.1957)	(0.0293)
URB	1.8239 ***	1.0125 ***	1.6476 ***	0.9679 *
	(0.5464)	(0.2116)	(0.5559)	(0.5386)
EDU	1.3842 **	0.6384 ***	2.7310 ***	0.6907
	(0.5726)	(0.2382)	(0.7719)	(0.4540)
TI	−0.4130 ***	−0.2988 ***	−0.3518 ***	−0.3119 ***
	(0.1282)	(0.0685)	(0.1087)	(0.0994)
NI	−1.0312 **	−0.3500	1.0754	−2.2481 ***
	(0.4996)	(0.2337)	(0.6560)	(0.6565)
CC	1.4094 ***	0.6526 ***	1.3967 **	1.5406 ***
	(0.4327)	(0.2157)	(0.6670)	(0.5196)
N	918	918	918	918
r2	0.3056	0.4055	0.0823	0.4847

t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

a,b, showing that the estimated coefficients for the direct, indirect, and total effects of BRI × POST were all positive and statistically significant at the 1% or 5% level, consistent with the decomposition results obtained earlier. This effectively reduces the treatment effect bias caused by issues such as sample self-selection and policy endogeneity and confirms the strong robustness of the previous conclusions.

4.4.2. Placebo Test

To test the robustness of the benchmark regression results, this paper further conducts a placebo test. Specifically, we randomly generated a false treatment variable 1000 times and repeatedly ran the benchmark regression. As shown in Figure 6, the kernel density distribution of the placebo effect coefficients estimated from the 1000 placebo tests is presented. The distribution of these placebo effect coefficients is centered around 0, with the vast majority concentrated within the range of [−0.02, 0.04], indicating no significant effect under random treatment. In contrast, the estimated result of the core model in this paper is 0.037, which significantly deviates from the distribution of the placebo effects and lies on the right tail (marked by a red dashed line). This suggests that the core findings of this study are unlikely to be caused by random factors or potential omitted variables, further enhancing the robustness and credibility of the main research conclusions.

Furthermore, the key premise of the placebo trial DID model is that the common trend assumption should be fulfilled by the experimental group and controls, implying that there must be no significant inter-group differences in CO₂ emissions in the absence of an implemented BRI. We first performed a placebo trial to examine the validity of the results. Pre-implementation data were used, with 2012 as the start year of the BRI. The rest of the settings were the same as those elaborated in the prior section.

Table A2. Regression results when assuming 2012 as the starting time.

Variable	(1)	(2)	(3)	(4)
	CEI	PCE	CEI	PCE
DID	−0.0819	0.0486	−0.1327	0.0165
	(0.0690)	(0.0479)	(0.0892)	(0.0462)
Control variables	No	No	Yes	Yes
Individual effect	Yes	Yes	Yes	Yes
Time effect	Yes	Yes	Yes	Yes
Observations	918	918	918	918

Note: Standard errors in parentheses. Models (1), (2) present the estimation results when control variables, Models (3), (4) present the estimation results when control variables are added.

lists the hypothetical influences over PCE, CEI, and TCE. The coefficient of the interaction term DID was barely significant regardless of whether the control variable was added, indicating that the sample met the common trend assumption before implementing the BRI. Suggestively, the effects during the 2013–2023 period were indeed caused by the BRI and not by other factors.

4.4.3. Sensitivity Analysis

We further estimated the lag phase I trial and hypothetical effects on carbon emissions, and

Table A3. Regression results for CEI when the lagged dependent variable by one period.

Variables	CEI
	W1	W2	W4
	(5)	(6)	(7)
Main L.y1	1.3072 ***	1.3088 ***	1.2376 ***
	(72.5333)	(74.4583)	(69.6856)
X(DID)	0.0543 ***	0.0480 ***	0.0317 ***
	(5.4530)	(5.1910)	(3.5038)
Wx x	1.6344 ***	0.3332 ***	0.2656 ***
	(16.9188)	(10.3635)	(4.5149)
Spatial rho	3.5115 ***	0.8424 ***	0.1510 ***
	(26.7649)	(16.8130)	(2.9227)
Variance sigma2_e	0.0031 ***	0.0029 ***	0.0027 ***
	(24.1421)	(22.5387)	(21.9589)

Note: *** p < 0.01. CEI is the lagged dependent variable by one period. Standard errors in parentheses. Models (5), (6), and (7) present the estimation results when control variables are added.

and

Table A4. Regression results for PCE when the lagged dependent variable by one period.

Variables	PCE
	W1	W2	W3
	(8)	(9)	(10)
Main Ly2	1.6176 ***	1.5298 ***	0.9285 ***
	(76.6335)	(72.1744)	(41.1933)
x	0.1806 ***	0.0902 ***	0.3294 ***
	(4.9530)	(2.6661)	(9.6365)
Wx x	1.9080 ***	−0.4898 ***	1.8588 ***
	(5.4201)	(−4.1888)	(16.2124)
Spatial rho	2.8098 ***	0.6179 ***	0.3754 ***
	(21.1687)	(11.8473)	(6.8176)
Variance sigma2_e	0.0362 ***	0.0376 ***	0.0376 ***
	(21.3523)	(21.9436)	(21.7690)

Note: *** p < 0.01. PCE is the lagged dependent variable by one period. Standard errors in parentheses. Models (8), (9), and (10) present the estimation results when control variables are added.

detail the results. The coefficients of CEI, regardless of whether the W1, W2, or W4 matrix was adopted, were positive and significant at a level of 1%. Those of PCE, consistent at W1, W2, and W3, also exhibited significant positivity at a level of 1% upon incorporation of control parameters. Thus, the effect of the BRI on carbon emissions was verified to be plausible.

We further estimated the long-term and short-term trials and assumed influence over carbon emissions, as detailed in

Table A5. Regression results for CEI when short-term and long-term effects.

Variables	CEI
	W1	W2	W4
	(11)	(12)	(13)
SR_Direct x	0.3281	0.1633 ***	0.0329 ***
	(0.0701)	(3.8030)	(3.7236)
SR_Indirect x	−1.0029	2.6398	0.3213 ***
	(−0.2145)	(1.6047)	(4.6950)
SR_Total x	−0.6748 ***	2.8031 *	0.3543 ***
	(−11.8724)	(1.6627)	(4.9613)
LR_Direct x	−0.5850	−0.2543	−0.1144 ***
	(−0.0726)	(−0.1432)	(−3.0868)
LR_Indirect x	0.1418	−0.0780	−0.6766 ***
	(0.0176)	(−0.0439)	(−3.3963)
LR_Total x	−0.4432 ***	−0.3323 ***	−0.7909 ***
	(−13.9376)	(−9.5553)	(−3.8178)
r2	0.0946	0.1490	0.3492
N	864	864	864

Note: * p < 0.1, *** p < 0.01. SR refers to short-term effects, LR refers to long-term effects. Standard errors in parentheses. Models (11, (12), and (13) present the estimation results when control variables are added.

and

Table A6. Regression results for PCE when short-term and long-term effects.

Variables	PCE
	W1	W2	W3
	(14)	(15)	(16)
SR_Direct x	0.0816 **	0.1377 ***	0.4215 ***
	(2.0941)	(3.9595)	(11.2562)
SR_Indirect x	0.4683 ***	−0.3839 ***	3.1093 ***
	(4.5119)	(−4.8441)	(9.1195)
SR_Total x	0.5499 ***	−0.2462 ***	3.5308 ***
	(5.8928)	(−3.1290)	(9.7400)
LR_Direct x	−2.1195	−4.3101	−6.0918
	(−0.0223)	(−0.0454)	(−0.0465)
LR_Indirect x	3.0768	0.4449	−1.4205
	(0.0323)	(0.0044)	(−0.0108)
LR_Total x	0.9573 ***	−3.8651	−7.5123 ***
	(5.7393)	(−0.1192)	(−4.6281)
r2	0.8442	0.9368	0.6129
N	864	864	864

Note: ** p < 0.05, *** p < 0.01. SR refers to short-term effects, LR refers to long-term effects. Standard errors in parentheses. Models (14), (15), and (16) present the estimation results when control variables are added.

. The CEI and long-term effect coefficients, regardless of whether the W1, W2, or W4 matrix was adopted, are significantly negative at the 1% level, while their short-term effects are negative in W1 and positive in W2 and W4. The long-term coefficient of PCE, in W1 and W3, differs from the CEI coefficient, and is significantly positive and negative at the 1% level. Its short-term effect is negative in W2 and positive in W1 and W3, and all are significant at the 1% level. These highly agree with prior findings. The minor estimate variations among the four weight matrices verify that the baseline regression outcomes are robust.

Given the probable impact of regression outcomes by the control variable changes, the additional variable “political stability and absence of violence/terrorism” (PS) and new matrix W5 (See the Supplementary Materials for the equation) were used to verify the influence of the BRI over carbon emissions, as seen in

Table A7. Regression results for CEI when increase control variables.

Variables	CEI
	W1	W2	W4
	(17)	(18)	(19)
Main x	0.0872 ***	0.0528 **	0.0635 ***
	(3.7932)	(2.4008)	(3.1093)
PGDP	−0.0512 ***	−0.0522 ***	−0.0405 ***
	(−9.7425)	(−9.4482)	(−8.3130)
URB	−0.0322 *	−0.0257	−0.0353 **
	(−1.8927)	(−1.4971)	(−2.1305)
EDU	0.0048	0.0008	0.0155
	(0.2103)	(0.0354)	(0.7188)
TI	−0.0113 ***	−0.0197 ***	0.0029
	(−2.6517)	(−4.5355)	(0.6563)
NI	−0.0553 *	−0.0694 **	−0.0281
	(−1.7971)	(−2.2220)	(−0.9298)
CC	−0.0076	−0.0106	−0.0240
	(−0.2620)	(−0.3564)	(−0.8149)
PS	0.0345 ***	0.0282 ***	0.0275 ***
	(4.8598)	(3.9398)	(4.1043)
Wx x	1.3022 ***	0.4882 ***	0.8131 ***
	(5.8933)	(6.4536)	(6.2022)
PGDP	−0.0779 ***	−0.0137	−0.0546 ***
	(−2.6296)	(−1.5060)	(−4.8921)
URB	−1.1208 ***	−0.4202 ***	−0.4717 **
	(−5.0464)	(−5.4124)	(−2.2217)
EDU	−0.0647	−0.0059	−0.2002
	(−0.2673)	(−0.0694)	(−1.1446)
TI	0.6790 ***	0.2745 ***	0.4973 ***
	(12.0750)	(10.8192)	(12.5743)
NI	0.5021 **	0.1423	0.3866
	(2.1910)	(1.6440)	(1.4988)
CC	−0.2408	−0.1077	−0.0127
	(−1.2737)	(−1.3708)	(−0.0607)
PS	0.4200 ***	0.1207 ***	0.2858 ***
	(5.3858)	(4.1603)	(6.2679)
Spatial rho	−1.0314 ***	−0.2425 ***	−0.8026 ***
	(−5.1722)	(−3.2014)	(−7.7352)
Variance sigma2_e	0.0145 ***	0.0156 ***	0.0135 ***
	(20.9101)	(21.2994)	(23.7929)
LR_Direct x	0.0561 **	0.0379 *	0.0481 **
	(2.4928)	(1.6672)	(2.3176)
PGDP	−0.0508 ***	−0.0524 ***	−0.0403 ***
	(−9.7980)	(−9.7388)	(−8.4269)
URB	−0.0026	−0.0105	−0.0245
	(−0.1549)	(−0.6150)	(−1.5929)
EDU	0.0063	0.0010	0.0197
	(0.2811)	(0.0429)	(0.9164)
TI	−0.0293 ***	−0.0289 ***	−0.0075 *
	(−6.2699)	(−6.0258)	(−1.9010)
NI	−0.0681 **	−0.0729 **	−0.0349
	(−2.1925)	(−2.3347)	(−1.1641)
CC	−0.0022	−0.0079	−0.0248
	(−0.0719)	(−0.2554)	(−0.8252)
PS	0.0242 ***	0.0242 ***	0.0218 ***
	(3.2729)	(3.2659)	(3.1866)
LR_Indirect x	0.6535 ***	0.4088 ***	0.4486 ***
	(5.1986)	(6.3105)	(6.0124)
PGDP	−0.0135	−0.0011	−0.0127 *
	(−0.8097)	(−0.1266)	(−1.8306)
URB	−0.5757 ***	−0.3523 ***	−0.2615 **
	(−4.4711)	(−5.2928)	(−2.1473)
EDU	−0.0285	−0.0004	−0.1147
	(−0.2173)	(−0.0054)	(−1.0583)
TI	0.3638 ***	0.2356 ***	0.2864 ***
	(8.2972)	(9.7395)	(10.7314)
NI	0.2987 **	0.1365 *	0.2479
	(2.3790)	(1.8279)	(1.6185)
CC	−0.1314	−0.0947	−0.0111
	(−1.2932)	(−1.4411)	(−0.0935)
PS	0.2042 ***	0.0971 ***	0.1536 ***
	(4.6585)	(3.9114)	(5.8304)
LR_Total x	0.7095 ***	0.4467 ***	0.4967 ***
	(5.3621)	(6.2529)	(6.3872)
PGDP	−0.0643 ***	−0.0535 ***	−0.0530 ***
	(−3.9100)	(−5.9008)	(−7.5432)
URB	−0.5783 ***	−0.3629 ***	−0.2860 **
	(−4.4110)	(−5.3506)	(−2.3244)
EDU	−0.0222	0.0006	−0.0950
	(−0.1727)	(0.0078)	(−0.9073)
TI	0.3345 ***	0.2067 ***	0.2789 ***
	(7.4365)	(8.2295)	(10.3496)
NI	0.2306 *	0.0636	0.2129
	(1.8774)	(0.8564)	(1.4024)
CC	−0.1336	−0.1026	−0.0359
	(−1.3679)	(−1.5585)	(−0.2945)
PS	0.2285 ***	0.1213 ***	0.1754 ***
	(5.1959)	(4.8129)	(6.5847)
r2	0.0146	0.0105	0.0337
N	918	918	918

t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

,

Table A8. Regression results for PCE when increase control variables.

Variables	PCE
	W1	W2	W3
	(20)	(21)	(22)
Main x	0.1476 **	0.1132 *	0.1268 **
	(2.2718)	(1.8701)	(2.2164)
PGDP	0.2505 ***	0.2387 ***	0.2401 ***
	(16.9037)	(15.7391)	(18.2517)
URB	0.1758 ***	0.1312 ***	0.1627 ***
	(3.6620)	(2.7925)	(3.5332)
EDU	0.1341 **	0.0842	0.0827
	(2.0861)	(1.3271)	(1.3361)
TI	−0.0682 ***	−0.0629 ***	−0.0576 ***
	(−5.7756)	(−5.3320)	(−5.0906)
NI	0.1608 *	0.1980 **	0.2313 ***
	(1.8518)	(2.3136)	(2.7956)
CC	0.1256	0.0875	0.1642 **
	(1.5372)	(1.0754)	(2.1071)
PS	−0.0599 ***	−0.0446 **	−0.0301 *
	(−2.9964)	(−2.2749)	(−1.6728)
Wx x	1.7081 ***	0.5423 ***	1.0734 ***
	(2.7403)	(2.6125)	(5.6288)
PGDP	0.1966 **	0.0301	−0.3730 ***
	(2.1319)	(1.0316)	(−4.7368)
URB	2.2162 ***	0.9299 ***	0.5906 **
	(3.5405)	(4.3584)	(2.5126)
EDU	1.6504 **	0.5695 **	1.1418 ***
	(2.4252)	(2.4541)	(3.9771)
TI	−0.4874 ***	−0.2550 ***	−0.1069 **
	(−3.1092)	(−3.6726)	(−2.3021)
NI	−1.5154 **	−0.5589 **	0.2782
	(−2.3431)	(−2.3513)	(1.0836)
CC	1.7303 ***	0.5926 ***	0.4699 *
	(3.2042)	(2.7201)	(1.7233)
PS	−0.7971 ***	−0.2128 ***	−0.0023
	(−3.6128)	(−2.6682)	(−0.0263)
Spatial rho	−0.3489 **	−0.0529	0.5295 ***
	(−2.0785)	(−0.7914)	(8.5360)
Variance sigma2_e	0.1157 ***	0.1170 ***	0.1025 ***
	(21.3937)	(21.4187)	(20.6002)
LR_Direct x	0.1323 **	0.1115 *	0.2176 ***
	(2.0434)	(1.7937)	(3.1881)
PGDP	0.2487 ***	0.2379 ***	0.2208 ***
	(17.3634)	(16.2478)	(14.4403)
URB	0.1583 ***	0.1293 ***	0.2186 ***
	(3.4537)	(2.8311)	(4.0654)
EDU	0.1173 *	0.0803	0.1755 **
	(1.8982)	(1.3061)	(2.5625)
TI	−0.0636 ***	−0.0614 ***	−0.0685 ***
	(−5.5906)	(−5.3653)	(−5.8702)
NI	0.1825 **	0.2069 **	0.2675 ***
	(2.1259)	(2.4400)	(2.9236)
CC	0.1054	0.0806	0.2029 **
	(1.2402)	(0.9523)	(2.2092)
PS	−0.0526 ***	−0.0438 **	−0.0321
	(−2.6114)	(−2.2029)	(−1.6393)
LR_Indirect x	1.3248 ***	0.5376 ***	2.4450 ***
	(2.6387)	(2.6918)	(4.4474)
PGDP	0.0900	0.0190	−0.5153 ***
	(1.3708)	(0.7882)	(−2.7854)
URB	1.6672 ***	0.8854 ***	1.4219 ***
	(3.1996)	(4.3864)	(2.7434)
EDU	1.2666 **	0.5592 **	2.5516 ***
	(2.2333)	(2.3601)	(3.4298)
TI	−0.3600 ***	−0.2439 ***	−0.2914 ***
	(−2.9455)	(−3.7298)	(−2.9166)
NI	−1.1768 **	−0.5378 **	0.8815
	(−2.3530)	(−2.3364)	(1.4176)
CC	1.2588 ***	0.5485 ***	1.1327 *
	(3.0323)	(2.7275)	(1.8516)
PS	−0.5958 ***	−0.2011 ***	−0.0348
	(−3.2586)	(−2.5777)	(−0.1928)
LR_Total
x	1.4571 ***	0.6491 ***	2.6626 ***
	(2.7495)	(2.9128)	(4.4933)
PGDP	0.3387 ***	0.2570 ***	−0.2945
	(5.0249)	(9.1869)	(−1.5164)
URB	1.8255 ***	1.0147 ***	1.6404 ***
	(3.4148)	(4.8558)	(2.9591)
EDU	1.3839 **	0.6395 ***	2.7271 ***
	(2.4149)	(2.6755)	(3.4921)
TI	−0.4237 ***	−0.3053 ***	−0.3600 ***
	(−3.4438)	(−4.6067)	(−3.4233)
NI	−0.9944 **	−0.3310	1.1490 *
	(−1.9658)	(−1.3962)	(1.7033)
CC	1.3642 ***	0.6291 ***	1.3356 **
	(3.2507)	(2.9787)	(1.9918)
PS	−0.6484 ***	−0.2449 ***	−0.0669
	(−3.4847)	(−3.0186)	(−0.3551)
r2	0.3056	0.4055	0.0823
N	918	918	918

t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

and

Table A9. Regression results when change the matrix.

Variables	(23)	(24)
	CEI	PCE
Main
x	0.0539 ** (0.0229)	0.0993 * (0.0589)
PGDP	−0.0520 *** (0.0057)	0.2556 *** (0.0147)
URB	−0.0349 * (0.0183)	0.1228 *** (0.0466)
EDU	−0.0066 (0.0241)	0.0881 (0.0619)
TI	−0.0155 *** (0.0049)	−0.0644 *** (0.0121)
NI	−0.0309 (0.0351)	0.0426 (0.0903)
CC	−0.0610 * (0.0332)	0.2455 *** (0.0855)
Wx
x	0.3639 *** (0.0691)	0.4930 *** (0.1775)
PGDP	−0.0123 * (0.0068)	0.0722 *** (0.0214)
URB	−0.1060 (0.1149)	−0.4095 (0.2943)
EDU	0.1324 (0.0891)	−0.7020 *** (0.2308)
TI	0.1213 *** (0.0224)	−0.0443 (0.0585)
NI	−0.2072 * (0.1059)	−0.6785 ** (0.2711)
CC	0.0357 (0.0956)	1.0406 *** (0.2464)
Spatial rho	−0.1599 *** (0.0546)	−0.3224 *** (0.0508)
Variance sigma2_e	0.0172 *** (0.0008)	0.1136 *** (0.0053)
LR_Direct
x	0.0475 ** (0.0234)	0.0818 (0.0608)
PGDP	−0.0522 *** (0.0055)	0.2554 *** (0.0144)
URB	−0.0312 * (0.0174)	0.1462 *** (0.0452)
EDU	−0.0092 (0.0233)	0.1176 * (0.0609)
TI	−0.0180 *** (0.0046)	−0.0635 *** (0.0118)
NI	−0.0264 (0.0364)	0.0723 (0.0938)
CC	−0.0632 ** (0.0314)	0.2025 ** (0.0809)
LR_Indirect
x	0.3216 *** (0.0606)	0.3838 *** (0.1403)
PGDP	−0.0035 (0.0063)	−0.0068 (0.0140)
URB	−0.0921 (0.0994)	−0.3658 (0.2277)
EDU	0.1246 (0.0833)	−0.5699 *** (0.1957)
TI	0.1112 *** (0.0202)	−0.0154 (0.0459)
NI	−0.1796 ** (0.0899)	−0.5541 *** (0.2086)
CC	0.0405 (0.0836)	0.7721 *** (0.1932)
LR_Total
x	0.3691 *** (0.0665)	0.4656 *** (0.1506)
PGDP	−0.0557 *** (0.0079)	0.2485 *** (0.0172)
URB	−0.1234 (0.1008)	−0.2196 (0.2250)
EDU	0.1154 (0.0851)	−0.4523 ** (0.1940)
TI	0.0932 *** (0.0211)	−0.0789 * (0.0465)
NI	−0.2059 ** (0.0966)	−0.4819 ** (0.2181)
CC	−0.0227 (0.0909)	0.9746 *** (0.2055)
N	918	918
r2	0.0051	0.4925

t statistics in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01.

.

The coefficients, boundary spillover effects, and effect tests of CEI were positive and significant at a level of 1% in W1, W2, and W4 or W5, while the coefficient for PCE, consistent at W1, W2 and W3 or W5, was significantly positive at the 1% or 5% levels when a control variable was added. These estimates align with prior findings. The small estimate variation between the two matrices confirms that the benchmark regression outcomes are robust.

5. Heterogeneity Analysis of the BRI Policy Effect

Institutional quality is a crucial factor influencing the economic development and carbon dioxide spatial spillover of a country or region. Most African countries along the BRI are economically underdeveloped, with significant disparities in institutional levels among them. Therefore, it is necessary to further categorize BRI countries based on their institutional development levels. According to Equation (4), we separately evaluated the policy impacts of the BRI on countries with varying levels of NI and CC. The effects of the BRI on different national governance systems and corruption risks are presented in

Table 7. Heterogeneity analysis of the policy effect.

Variable	NI				CC
Group	Above Average		Below Average		Above Average		Below Average
x	CEI	PCE	CEI	PCE	CEI	PCE	CEI	PCE
	(19)	(20)	(21)	(22)	(23)	(24)	(25)	(26)
	0.1295 **	1.6984 ***	0.3478 **	−0.3746	−0.0918 ***	4.3167 ***	0.1915	0.6442
	(0.0655)	(0.4662)	(0.1674)	(0.3381)	(0.0264	(1.0188)	(0.1634)	(0.4453)
N	442		476		425		493
r2	0.0017	0.1724	0.0342	0.6898	0.0298	0.0221	0.0004	0.6297

t statistics in parentheses ** p < 0.05, *** p < 0.01, for the country clusters see the Supplementary Materials.

. Columns (19) to (20) represent regression results with high-average regulatory institutions as the dependent variable for CEI and PCE, respectively, while columns (21) to (22) represent regression results with low-average regulatory institutions as the dependent variable for CEI and PCE, respectively. As shown in Table 7, the BRI significantly promotes carbon dioxide spatial spillover in countries with higher economic institutional management levels. This indicates that economies with superior economic institutional management exhibit greater stability, inclusivity, and a more favorable business environment, enabling the BRI to better drive robust economic growth. However, rapid economic development simultaneously leads to increased carbon dioxide spillover. According to the “Pollution Haven Hypothesis,” the heterogeneity in institutional management levels among African countries may also result in pollution-intensive industries relocating from China or other nations to BRI African countries or regions with looser regulations, indirectly confirming this phenomenon in BRI African countries [30].

Table 7. Columns (23) to (24) represent regression results with high-average corruption levels as the dependent variable for CEI and PCE, respectively, while columns (25) to (26) represent regression results with low–average corruption levels as the dependent variable for CEI and PCE, respectively. Conversely, in BRI African countries with controlled high corruption levels, although PCE shows positive adjustments, CEI levels are significantly reduced. This fully demonstrates that with the implementation of the BRI, corruption in environmental protection funds has been effectively curbed. Increased investments in green BRI initiatives, widespread adoption of new energy sources and technologies, enhanced low-carbon education, and the implementation of low-carbon policies and technologies have collectively suppressed carbon dioxide spillover in BRI countries. According to the “Pollution Halo Hypothesis,” as the green BRI accelerates, multinational corporations are more inclined to promote green technologies in host countries, facilitate training in low-carbon technologies, and encourage the extensive use of new energy technologies, thereby reducing carbon dioxide spillover [31]. The observed differences in green environmental fund investments and management among BRI African countries also indirectly support this phenomenon, aligning with the “Pollution Halo Hypothesis.” The rapid development of the green BRI in Africa will further encourage African nations to establish more refined and differentiated management measures to prevent environmental corruption, reduce spatial carbon dioxide spillover, and promote green and low-carbon development.

6. Mediation Effect Analysis of the BRI in African Countries

To further investigate the channel mechanisms of carbon emission intensity in African countries along the BRI, two mediating variables were set based on research Hypotheses 1: economic and trade growth (Trade) in the regions along the route, measured by the GDP of these regions, and the level of infrastructure development (INS), measured by the industrial structure advancement index (INS) of the regions along the route. The study reports regression results under the W1 spatial weight matrix, as presented in

Table 8. Mediation effect analysis of BRI on CEI and PCE in African countries.

Variable	Trade			INS
	Trade	CEI	PCE	INS	CEI	PCE
Model	(27)	(28)	(29)	(30)	(31)	(32)
BRI × POST	1.0329 **	0.6730 ***	2.6249 ***	47.7420 **	0.7208 ***	1.1725 **
	(0.4538)	(0.1228)	(0.5419)	(19.6376)	(0.1319)	(0.4646)
Trade		0.0591 **	−0.0557
		(0.0269)	(0.0636)
INS					−0.0037 **	0.0135 **
					(0.0016)	(0.0063)
Control var.	Yes	Yes	Yes	Yes	Yes	Yes
Period fixed	Yes	Yes	Yes	Yes	Yes	Yes
Regional fixed	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.0360	0.0136	0.0715	0.0549	0.0197	0.3222
Sample size	918	918	918	918	918	918

t statistics in parentheses ** p < 0.05 and *** p < 0.01.

. Columns (27)–(29) indicate that the coefficients of Trade with CEI and PCE are 1.0329, 0.6730, and 2.6249, respectively, with the partial mediating effect on CEI being more significant. This suggests that the BRI can significantly promote rapid economic and trade growth in the regions along the route, improve income levels, and consequently increase carbon emission intensity in neighboring regions. This conclusion aligns with the Environmental Kuznets Curve (EKC) hypothesis, which posits that when income levels in the regions along the route reach a certain threshold, residents will increase their demand for clean and low-carbon products, thereby pressuring governments to strengthen environmental regulations. The resulting “income effect” will compel enterprises to transition toward the production of low-carbon and clean products [32].

Columns (30)–(32) show that the coefficients of INS with CEI and PCE are 47.7420, 0.7208, and 1.1725, respectively, with significant partial mediating effects on both CEI and PCE. This indicates that the implementation of the BRI helps optimize the industrial structure of countries along the route. In the initial stages, large-scale infrastructure construction, coal projects, and the export of high-carbon agricultural by-products in these regions led to carbon dioxide spillover effects in neighboring African countries. This contradiction between economic development and the environment aligns with the “Pollution Haven Hypothesis” [31]. However, with the release of the China–Africa Cooperation Declaration on Climate Change in 2021, China will further expand investments in low-emission projects in Africa, such as renewable energy, energy-saving technologies, and green low-carbon industries, while ceasing new overseas coal-fired power projects. This will assist African BRI countries in promoting industrial structure optimization and upgrading, achieving green, low-carbon, and high-quality development.

7. Discussion

Previous studies have employed CEI and PCE to quantify the spatial spillover effects among neighboring countries along the BRI, revealing strong spatial correlations in carbon dioxide emissions among these nations. The BRI was found to exacerbate carbon dioxide emissions in partner countries while promoting green economic development by reducing PCE and CEI [10]. However, as the research focused only on select African countries, the spatial spillover effects and influencing factors specific to BRI-affiliated African nations remain unclear. The findings of this study exhibit certain differences and innovations compared to the aforementioned research. African countries demonstrate geographical proximity and significant spillover correlations in carbon dioxide emissions, exhibiting a “shared prosperity, shared decline” characteristic. Economic level, urbanization, industrial structure of countries, technology, and government governance are identified as key drivers of carbon dioxide spillover effects in Africa. A novel discovery is that per capita GDP exerts a marginally positive influence on CEI and PCE, while information technology exhibits a suppressive effect. These conclusions provide important new insights for advancing carbon emission reduction efforts and accelerating green economic development in Africa.

China’s outward foreign direct investment (OFDI) and its impact pathways on carbon emissions in BRI countries have attracted extensive scholarly attention and research [33,34,35]. However, research on the spatial spillover effects of the BRI on Africa’s CO₂ emissions and its underlying mechanisms remains scarce [36,37]. This paper innovatively examines the influence of trade and industrial structure on two CO₂ spatial spillover indicators—CEI and PCE—in African BRI countries, along with their impact pathways. It proposes that due to the relatively underdeveloped economies of African BRI nations, increased trade openness has primarily stimulated local economic growth, thereby exacerbating CO₂ spatial spillover effects to neighboring countries. Consequently, as the BRI policy steadily advances, facilitating trade connectivity and industrial spillovers among participating countries, governments should guide active cross-border cooperation. Promoting mutual trade would foster technology sharing, optimize domestic resource allocation, achieve industrial upgrading, and collectively realize carbon emission reduction goals.

Alternately, high-quality institutions compel businesses and industries to adhere to environmental practices. Globally, many African countries are signatories to the United Nations 2030 Agenda or SDGs. However, carbon dioxide emissions have continued to rise over recent years. Despite research confirming the impact of political systems on carbon emissions, significant disagreement remains. Acheampong (2022) [25] found that high-level democracy measures drive environmental degradation measured by increasing carbon emissions in sub-Saharan Africa, while Jahanger et al. (2022) [38] confirmed that authoritarian regimes exacerbate carbon emissions, whereas democratic regimes reduce them. Kwakwa (2023) [16] demonstrated that renewable energy and NI, including CC, rule of law, regulatory quality, political stability, accountability, and government effectiveness, significantly curb emissions. Existing studies are limited to a sample of 32 countries and have not examined the differences in national institution management levels and corruption control on the spillover of CO₂ among neighboring countries. This study’s innovation lies in discovering that the BRI has significantly promoted the spatial spillover of CO₂ in countries with higher economic institutional management levels. In BRI African countries with high corruption control levels, although PCE shows a positive moderating effect, CEI levels significantly decrease. This fully illustrates that with the implementation of the BRI, issues of corruption in environmental funds have been effectively curbed. Increased green investments, widespread adoption of new energy and technology, improved low-carbon education levels, and the implementation of low-carbon policies and technologies collectively suppress the CO₂ spillover of BRI countries. This validates Recep Ulucak’s (2020) [39] view that sound national institutions are crucial for the deployment of renewable energy and indirectly promote its adoption and development through various differential regulations, effectively addressing climate change.

8. Conclusions, Implications, and Limitations

8.1. Conclusions

Many African countries participating in the BRI are characterized by large populations, lagging economic development, and imbalanced industrial structures. Assessing the carbon dioxide spillover effects and their driving factors is therefore crucial. This study is the first to employ an SDID model using dual indicators—CEI and PCE—to evaluate the spatial spillover effects and underlying mechanisms of BRI African countries and neighboring nations. The results indicate that since the implementation of the BRI, both CEI and PCE in participating African countries have significantly increased.

Notably, the study identifies trade openness and imbalanced industrial structures as key factors exacerbating carbon dioxide spillovers from BRI countries to their neighbors. While increased trade openness stimulates economic growth in BRI African nations, foreign direct investment predominantly flows into large-scale infrastructure projects with high carbon emission intensities, such as oil, gas, and transportation. This pattern exacerbates transboundary carbon emissions from countries pursuing high-carbon development. This study also provides novel empirical evidence that green technologies can mitigate this spillover effect. The BRI has facilitated the transfer of advanced emission-reduction technologies from host countries to their neighbors. The widespread application of new energy technologies further optimizes industrial structures and accelerates progress toward a greener BRI in Africa.

Additionally, this study identifies, for the first time, the significant role of government corruption in hindering the adoption of low-carbon green technologies and sustainable development. As African BRI countries are experiencing rapid economic growth, attention must be paid not only to introducing green technologies but also to the impact of government corruption on low-carbon development. The findings of this study offer valuable guidance for promoting the low-carbon transition of African countries and collectively achieving the 2030 SDGs.

8.2. Implications

Based on these findings, we propose three key policy recommendations. First, a green BRI in Africa should be actively promoted by establishing a mechanism for shared but differentiated environmental responsibilities. Countries along the route should develop regional environmental policies tailored to their overall carbon emission levels and spatial distribution.

Second, political governance and corruption control in African BRI countries should be assessed to inform differentiated environmental financing policies. Leveraging China’s green technology spillover effects and the demonstration effect of advanced technologies can incentivize countries with lower governance standards to improve their environmental systems and regulatory capabilities, thereby achieving mutual benefits and win-win outcomes.

Third, the establishment of domestic carbon emission trading systems should be promoted in African BRI countries. Market-based carbon trading mechanisms can guide high-emission enterprises into the carbon trading financial market, facilitating optimized industrial structures and fostering a transition to green and innovative production models, ultimately advancing green development within the BRI in Africa.

8.3. Limitations

Despite tremendous effort, this study still faced some limitations due to data availability. For instance, the specific contributions of key mediating factors—educational structure and technological advancement—could not be fully evaluated, and differences in economic, cultural, and regional factors were not fully considered. Future research will involve the collection of more extensive data through surveys and case studies, which will better capture the green development challenges faced by African BRI countries. More in-depth and systematic analyses will support the formulation of effective policies for sustainable development and low-carbon growth on the continent.