The Carbon Emissions Effect of China’s OFDI on Countries along the “Belt and Road”

Abstract

With the continuous practice of the “Belt and Road” initiative, the countries along the “Belt and Road” have achieved rapid social and economic development. However, environmental problems have become increasingly prominent. Around the world, there are comments that China’s “Belt and Road” initiative is a result of resource plundering, transfer of backward production capacity, and environmental degradation of countries along the line. This study quantitatively evaluated the static, dynamic, linear, and non-linear effects of China’s foreign direct investment on the carbon emissions of countries along the line. The results showed that: (1) The direct effect of China’s foreign direct investment on the carbon emissions of countries along the route was significantly negative. (2) The economic scale and industrial structure effects of China’s foreign direct investment increased the carbon emissions of countries along the route. The production technology effect suppressed the carbon emissions of countries along the route and played a leading role. (3) The estimation results of the system generalized method of moments showed that the carbon emissions of countries along the route were significantly affected by the lag period, but the impact was small. (4) The results of the threshold regressive model showed that the GDP and proportion of industrial added value had significant threshold effects on the carbon emissions effect of China’s outward foreign direct investment. When the GDP of countries along the route exceeded 7.2696, China’s outward foreign direct investment carbon emissions reduction effect could not be realized; when the proportion of the industrial added value of countries along the route was lower than 4.0106, China’s outward foreign direct investment carbon emission reduction effect could not be realized. Based on the research conclusion, we concluded that China and countries along the “Belt and Road” should strengthen cooperation on carbon emissions reduction, jointly promote low-carbon construction of industrial parks, accelerate cooperation on green energy projects, and establish a green development fund to achieve sustainable development of the countries along the “Belt and Road”.

1. Introduction

In the context of economic globalization, the participation of Chinese enterprises in the international economy is deepening. In 2013, China launched the “Belt and Road” initiative to promote world economic exchanges. It aims to actively develop economic partnerships with countries along the Silk Road by borrowing the historical symbols of the ancient Silk Road. Moreover, it aims to jointly build a community of interests featuring political mutual trust, economic integration, and cultural inclusion. As of May 2022, China has signed more than 200 cooperation documents with 150 countries and 32 international organizations to jointly build the “Belt and Road”. With the continuous practice of the “Belt and Road” initiative, economic cooperation between China and countries along the “Belt and Road” has become more frequent and in-depth, and social and economic development has been rapid [1]. In 2021, China’s non-financial direct investment in countries along the “Belt and Road” was USD 20.3 billion, up 14.1% year on year. It has made positive contributions to the economic development of the host country. Global warming has become one of the greatest challenges facing mankind in recent decades. Existing studies claimed that carbon dioxide emissions have an important impact on global warming. With the continuous improvement in the economic level, how to reduce carbon dioxide emissions in countries along the line has become an urgent problem to be solved. The “Belt and Road” countries have obtained great support from China in terms of investment, optimized infrastructure conditions, and expanded production scale. However, the resource and environmental problems of the countries along the route are becoming increasingly prominent at the same time. Half of the top ten countries in the world in carbon emissions are countries along the “Belt and Road”, which poses a great challenge to curbing global warming [2]. Therefore, there have been comments made internationally that China’s “Belt and Road” initiative plunders resources, transfers backward production capacity, and deteriorates the environment in countries along the belt and road [3]. These accusations, which are not in line with the facts, undoubtedly hinder the continued progress of the “Belt and Road” initiative. China does have overcapacity, but “overcapacity” may not be low-carbon and high-quality. Outward foreign direct investment (OFDI) is an important way for China to achieve common development with countries along the route. In response to the above comments, we studied the impact of China’s OFDI on the carbon emissions of countries along the route from the perspective of China’s foreign direct investment [4]. Specifically, we analyzed whether China’s “Belt and Road” initiative led to the growth of carbon emissions of countries along the route. Furthermore, how does China’s OFDI affect the carbon emissions of countries along the “Belt and Road”? Moreover, what are the differences in the carbon emission effects of China’s OFDI between different countries?

Considering the indirect impact of China’s OFDI on the carbon emissions of countries along the route, we added explanatory variables such as economic scale, industrial structure, and technological level to evaluate the carbon emissions effect of China’s OFDI on countries along the route. Second, considering that the carbon emissions of countries along the route may have strong time-series-related problems, we took the first-order lag term of the carbon emissions of countries along the route as the explanatory variable and used the system generalized method of moments (SYS-GMM) estimation method to build a dynamic panel model to further analyze China’s OFDI impact on carbon emissions of countries along the route. Finally, considering the heterogeneity of countries along the route. We adopted the threshold regressive model with GDP, proportion of industrial added value (PIAV), and energy intensity (EI) as threshold variables to further analyze the non-linear impact of China’s OFDI on the carbon emissions of countries along the route. Through quantitative assessment of the carbon emissions effect of China’s OFDI on countries along the “Belt and Road”, more reasonable sustainable development policy recommendations were proposed in combination with the background of national conditions of countries along the Belt and Road. This is of great significance to continuously promote the “Belt and Road” initiative and promote the common sustainable development of China and countries along the Belt and Road.

The innovative contributions of this study mainly include the following three points: First, we comprehensively considered the direct and indirect effects of China’s OFDI on carbon emissions of countries along the “Belt and Road”. The scale, structure, and technology effects were simultaneously included in the model to systematically assess the impact of China’s OFDI on the carbon emissions of countries along the line. Second, the previous studies that assessed the impact of OFDI were usually static. This study applied the SYS-GMM estimation method to assess the dynamic impact of OFDI on carbon emissions in China and effectively overcame the endogenous variables problem. Third, with GDP, PIAV, and EI as threshold variables, the threshold regression model was used to examine the carbon emissions utility of OFDI in China under different characteristics of different countries. Thus, we analyzed the possible non-linear characteristics of China’s OFDI carbon emissions utility.

2. Literature Review

The impact of foreign direct investment (FDI) on the carbon emissions of host countries has always been a hot issue of academic concern. However, there are two opposing views on how the inflow of FDI affects the carbon emissions of the host country [5,6].

One view is the pollution haven hypothesis [7]. This hypothesis holds that developing countries tend to lower environmental standards to attract industrial transfer from developed countries to seek economic growth. Therefore, multinational companies have continuously transferred industries with high pollution and energy consumption to developing countries, resulting in the growth of carbon emissions in developing countries [8]. Many studies verified the validity of the pollution haven hypothesis. Singhania constructed a dynamic panel data estimation with 21 developed and developing countries with high carbon emissions as samples. The results show that FDI has a promotion effect on regional carbon emissions [9]. Asghari et al. found that FDI inflows into China will increase local carbon emissions [10]. Zugravu-Soilita found that the pollution haven hypothesis was established only under certain capital intensity and environmental stringency conditions [11].

The other view is the pollution halo hypothesis. This hypothesis holds that FDI positively impacts the host country’s environment. The host country can improve the local environment by absorbing more advanced low-carbon production methods and learning from environmentally friendly management systems [12]. Polloni-Silva et al. took 592 cities in Brazil as research samples and found a negative correlation between foreign direct investment and carbon dioxide [13]. Ahmad found that FDI inflow promoted environmental sustainability and reduced regional carbon emissions [14]. Al-Mulali et al. used the panel data of GCC countries from 1980 to 2009. Their research found that FDI inflow has no significant impact on the environment in the short term, but in the long term, FDI has a significant negative effect on carbon emissions [15].

Other scholars focused on the indirect impact of FDI on the carbon emissions of host countries. They are the economic scale, industrial structure, and production technology effects [16]. First, FDI can indirectly affect carbon emissions through economic growth, forming the economic scale effect of FDI. FDI can expand the economic scale of countries along the route. The expansion of the economic scale means an increase in energy use and the corresponding increase in carbon emissions [17]. Sayari and others built a random effects model based on the GDP and FDI data of 30 European countries. It was found that there is a significant positive correlation between GDP and FDI [18]. Abbes and Mostéfa found that FDI plays a very important role in developing countries. FDI is regarded as the engine of economic growth and development [19].

Second, FDI can indirectly affect carbon emissions through the industrial structure, forming the structural effect of FDI. FDI inflows will adjust the local industrial structure, have an industrial correlation effect on local production enterprises, and ultimately affect carbon emissions [20]. Taking China’s manufacturing industry as the research sample, Sung used the GMM method to separate the structural effect of FDI from the total effect of FDI. Research showed that the inflow of FDI shifts the industrial structure in the direction of high energy consumption, increasing energy consumption [21]. Feng found that FDI inflow can promote the upgrading of inefficient enterprises in countries along the route. Additionally, it promotes the industrial upgrading of countries along the route and improves the resource utilization efficiency of countries along the route [22]. Finally, FDI can indirectly affect carbon emissions through the technology level, forming the technological effect of FDI. FDI can improve countries’ production and technology levels along the route through technology spillover effects. Advanced technology can enable countries along the route to use energy more effectively, thus reducing carbon emissions [23]. The research results of Jane et al. confirmed the technology spillover effect of FDI, and this technology effect is more prevalent in low-carbon-emissions industries [24]. Based on Indian panel data, Fujimori found that the technology effect is the root cause of FDI’s “pollution halo” effect [25].

The above literature systematically studied the carbon emissions effect of FDI, but there are still some limitations. First, most existing studies focused on the impact of FDI inflows on a country’s environment in a single country or group of countries. There are few studies on the impact of OFDI outflow from a single country on the carbon emissions of multiple countries. Second, existing studies mainly assessed the static impact of FDI on regional carbon emissions. Few people analyzed the impact of FDI on carbon emissions from a dynamic perspective. Third, existing studies mainly analyzed the linear effect of FDI on regional carbon emissions. There are few kinds of research on FDI carbon emission effects from a non-linear perspective.

3. Influence Mechanism Analysis

Based on the research framework of Grossman and Krueger, this study analyzed the impact mechanism of China’s OFDI on the carbon emissions of countries along the “Belt and Road” [26]. China’s OFDI can not only directly impact the carbon emissions of countries along the route. It can also indirectly affect carbon emissions through economic scale growth, industrial structure adjustment, and technology progress [27,28]. A diagram representing these effects is shown in Figure 1.

3.1. Economic Scale Effect

The inflow of China’s OFDI has brought capital flow to countries along the route, which is conducive to the expansion of production and economic scale [29]. The expansion of production and economic scale will inevitably lead to an increase in the input of production factors. The energy consumption of countries along the Belt and Road has increased, and the scale of their carbon emissions has also increased accordingly. On the other hand, when economic growth reaches a certain level, economic growth forces the environment to improve. According to the environmental Kuznets curve (EKC), the relationship between per capita income and environmental pollution presents an inverted “U” shape. The living standard of residents also improved yearly. Residents began to pay attention to environmental problems, use clean energy in their lives, and buy environmentally friendly products. Residents hope the government will adopt stricter environmental protection standards to improve environmental problems.

3.2. Industrial Structure Effect

With the entry of China’s OFDI, the industrial structure of countries along the route was greatly affected [30]. When the proportion of the secondary industry is higher than that of the tertiary industry, the carbon emissions of countries along the route will rise. When the proportion of the tertiary industry is higher than that of the secondary industry, the carbon emissions of countries along the route will be reduced. Compared with the secondary industry, the tertiary industry is an obvious low-carbon-emissions industry. The increase in its proportion has a positive effect on reducing the growth rate of overall carbon emissions. In particular, the new generation of information technology, scientific research, technical services, and other tertiary industries have developed rapidly. The combination of various scientific and technological innovations based on Internet technology has profoundly changed people’s lifestyles and production efficiency, effectively reducing the intensity of carbon dioxide emissions. Therefore, in the process of industrial structure adjustment, the impact of changes in industrial structure on carbon emissions is uncertain. Based on the current development stage of China and countries along the route, most countries use China’s OFDI to accelerate industrial expansion. Therefore, the industrial structure effect of China’s OFDI promotes the growth of carbon emissions in countries along the route.

3.3. Production Technology Effect

According to endogenous growth theory, improvement in the technological level is conducive to improving the utilization rate of resources in the production process to reduce resource consumption and carbon emissions [31]. Compared with most countries along the “Belt and Road”, China’s production technology is more advanced. China’s OFDI can introduce green and low-carbon production technologies to countries along the route, which then improves the technological level of countries along the route through the spillover effect of green technology. On the other hand, China’s OFDI strengthens the country’s endogenous development power through the reverse force effect. The country has accelerated technological innovation and improved production methods to reduce carbon dioxide emissions from countries along the route.

4. Methodology and Data

4.1. Index System Construction

Based on the sorting out of the existing academic research on the relationship between FDI and carbon emissions and the analysis of the impact mechanism of China’s OFDI carbon emissions, we selected eight variables. We divided them into explained variables, core explanatory variables, other explanatory variables, and control variables [32,33]. The specific description of each variable is shown in

Table 1. Variable descriptions.

Variable		Symbol
Explained variable	Carbon emissions	C
Core explanatory variable	China’s foreign direct investment	COFDI
Other explanatory variables	Gross domestic product	GDP
	Proportion of industrial added value	PIAV
	Energy intensity	EI
Control variables	Proportion of foreign direct investment	PFDI
	Proportion of urban population	PUP
	Proportion of foreign trade volume	PFTV

.

(1) Explained variable: we used the carbon emissions of countries along the “Belt and Road” over the years examined to reflect the national carbon emissions.

(2) Core explanatory variable: China’s OFDI stock of countries along the route was selected to represent China’s foreign direct investment. It was converted into the actual value of the constant dollar price in 2010, according to the dollar deflator.

(3) Other explanatory variables: We considered the economic scale, industrial structure, and production technology effects of OFDI in China. We chose the GDP of countries along the route to reflect each country’s economic scale level and select the proportion of industrial added value in GDP to reflect the industrial structure of countries along the route. Energy intensity was used to reflect the technical levels of countries along the route.

(4) Control variables: According to the research results of S. Wu [34] and M. Abdouli [35], we selected the proportion of foreign direct investment, the proportion of the urban population, and the proportion of foreign trade as control variables. The proportion of foreign direct investment was expressed using the proportion of FDI stock of countries along the route in their GDP. We added this variable to control the impact of direct investment from countries other than China on the carbon emissions of countries along the route. The proportion of the urban population was expressed using the proportion of the urban population in the total population of countries along the route. The proportion of foreign trade volume was expressed by the proportion of total imports and exports of goods and services in GDP.

4.2. Methodology

We evaluated the carbon emissions effect of China’s OFDI on countries along the “Belt and Road” through static, dynamic, linear, and non-linear econometric models to respond to the international disputes over the “Belt and Road” initiative and promote the sustainable development of the “Belt and Road” initiative.

First of all, according to the research results of Zhao [36] and Copeland [37], we linked the carbon emissions of countries along the route with COFDI. According to the indirect impact of COFDI on carbon emissions, a static panel regression model was established by adding variables such as economic scale, industrial structure, and technical level [38]. We analyzed the carbon emissions effect of China’s OFDI on countries along the route, as shown in Formula (1):

$${\mathrm{lnCO}}_{2\mathrm{it}}={\mathsf{\alpha }}_{\mathrm{i}}+{\mathsf{\beta }}_1{\mathrm{lnCOFDI}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{lnGDP}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{lnPIAV}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{lnEI}}_{\mathrm{it}}+\mathsf{\beta }{\displaystyle \sum }\mathrm{X}+{\mathsf{\mu }}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}+{\mathsf{\delta }}_{\mathrm{it}}$$

(1)

where $\mathrm{i}$ and t represent the country and year, respectively; ${\mathsf{\alpha }}_{\mathrm{i}}$ is the intercept, representing an individual fixed effect; ${\mathsf{\mu }}_{\mathrm{it}}$ is the error term; ${\mathsf{\epsilon }}_{\mathrm{it}}$ is a regional effect; ${\mathsf{\delta }}_{\mathrm{it}}$ is a time effect; ${\mathrm{CO}}_{2\mathrm{it}}$ is the carbon emissions of country i along the route in year t; ${\mathrm{COFDI}}_{\mathrm{it}}$ represents China’s direct investment in countries along the Belt and Road in year t; ${\mathrm{GDP}}_{\mathrm{it}}$, ${\mathrm{IND}}_{\mathrm{it}}$, and ${\mathrm{ENE}}_{\mathrm{it}}$ represent the economic growth, industrial structure, and technological level, respectively. $\mathrm{X}$ is the control variable that affects the scale of carbon emissions, including PFDI, PUP, and PFTV. All variables were logarithmic to eliminate the impact of absolute differences in values.

Second, given the strong time-series-related problems of carbon emissions in the countries along the route, we added a first-order lag term of the explained variable as the explanatory variable based on Formula (1) to build a dynamic panel regression model [39]. We focused on testing whether China’s OFDI increased or reduced carbon emissions of countries along the route under the dynamic panel model, as shown in Formula (2). Building a dynamic panel model can reveal the dynamic change characteristics of carbon emissions and overcome the errors caused by endogeneity. In this dynamic model, ${\mathrm{CO}}_{2\mathrm{it}-1}$ is the first lag term of carbon emissions; the meanings of the rest of the symbols are consistent with Formula (1). Since the independent variable of Formula (2) includes carbon emissions lagging by one period, there may be a two-way causal relationship between China’s OFDI and carbon emissions. The growth of carbon emissions will, in turn, inhibit the growth of China’s ODFI. Therefore, the model inevitably had endogeneity problems. If the traditional ordinary least squares (OLS) and fixed effects (FE) methods are used, the model estimation will be biased. The generalized method of moments (GMM) can overcome the endogeneity problem of dynamic panel model estimation [40]. Differenced GMM (DIF-GMM) estimation and system GMM (SYS-GMM) estimation are two important methods of GMM estimation. Compared with DFI-GMM, SYS-GMM can solve the problem of weak tool variables and improve the estimation efficiency. At the same time, it can also estimate the coefficients of variables that do not change at any time. Considering this, the two-step GMM estimation may lead to bias in the standard deviation of the estimated parameters and affect the estimation results of the parameters. Therefore, we used a one-step SYS-GMM estimation method to estimate the model.

$${\mathrm{lnCO}}_{2\mathrm{it}}={\mathsf{\alpha }}_{\mathrm{i}}+{\mathsf{\beta }}_5{}_{.}{\mathrm{lnCO}}_{2\mathrm{it}-1}+{\mathsf{\beta }}_1{\mathrm{lnCOFDI}}_{\mathrm{it}}+{\mathsf{\beta }}_2{\mathrm{lnGDP}}_{\mathrm{it}}+{\mathsf{\beta }}_3{\mathrm{lnPIAV}}_{\mathrm{it}}+{\mathsf{\beta }}_4{\mathrm{lnEI}}_{\mathrm{it}}+\mathsf{\beta }{\displaystyle \sum }\mathrm{X}+{\mathsf{\mu }}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}+{\mathsf{\delta }}_{\mathrm{it}}$$

(2)

Finally, based on Formula (1), we took into account the heterogeneity of countries along the route. To study the carbon emissions effectiveness of China’s OFDI in different countries, we took GDP, PIAV, and EI as threshold variables to further analyze the non-linear relationship between COFDI and carbon emissions of countries along the route. Using Hansen’s research method as a reference, a threshold regression model of China’s OFDI carbon emissions effect was established, as shown in Formula (3) [41]. In this model, thres represents the threshold variables, including GDP, PIAV, and EI; I indicates the indicating function, with a true value of 1, and a false value of 0; and $\mathsf{\eta }$ is the threshold value.

$$\begin{array}{cc}\hfill {\mathrm{lnCO}}_{2\mathrm{it}}={\mathsf{\alpha }}_{\mathrm{i}}+& {\mathsf{\beta }}_1{\mathrm{lnCOFDI}}_{\mathrm{it}}\times \mathrm{I}\left(\mathrm{thres}\le {\mathsf{\eta }}_1\right)+{\mathsf{\beta }}_2{\mathrm{lnCOFDI}}_{\mathrm{it}}\times \mathrm{I}\left({\mathsf{\eta }}_1<\mathrm{thres}\le {\mathsf{\eta }}_2\right)\hfill \\ & +{\mathsf{\beta }}_{\mathrm{n}}{\mathrm{lnCOFDI}}_{\mathrm{it}}\times \mathrm{I}\left({\mathsf{\eta }}_{\mathrm{n}-1}<\mathrm{thres}\le {\mathsf{\eta }}_{\mathrm{n}}\right)\hfill \\ & +{\mathsf{\beta }}_{\mathrm{n}+1}{\mathrm{lnCOFDI}}_{\mathrm{it}}\times \mathrm{I}\left(\mathrm{thres}\ge {\mathsf{\eta }}_{\mathrm{n}}\right)+\mathsf{\beta }{\displaystyle \sum }\mathrm{X}+{\mathsf{\mu }}_{\mathrm{it}}+{\mathsf{\epsilon }}_{\mathrm{it}}+{\mathsf{\delta }}_{\mathrm{it}}\hfill \end{array}$$

(3)

4.3. Data Processing

Considering the availability of data, we selected the panel data of 42 countries along the “Belt and Road” from 2003 to 2020 to study the carbon emission effects of China’s OFDI. The data were derived from the Statistical Bulletin of China’s Foreign Direct Investment over the years, BP Statistical Review of World Energy, the US Energy Information Administration database, the World Development Indicators database of the World Bank, and the United Nations Conference on Trade and Development database. The carbon emissions data of all countries were from the World Development Indicators Database of the World Bank. For missing data, we used interpolation and the moving average method to complete the data set. The variables involving price factors were calculated and treated with the unchanged price in 2010.

5. Results and Discussion

5.1. Unit Root Test

In order to avoid false regression, this study conducted a unit root test for all variables. If the unit root exists, the panel data is unstable. It is necessary to perform differential processing on the data. In order to improve the accuracy of the test results, this study used LLC, Fisher AD, and Fisher PP test methods on COFDI, GDP, PIAV, EI, PFDI, PUP, and PFTV. The calculation results showed that GDP, PIAV, PFDI, and PFTV had poor stability. After the first-order difference processing, the unit root was checked again. The results showed that the sequence was stable, and the unit root test results are shown in

Table 2. Unit root test results.

	LLC Test	Fisher ADF Test	Fisher PP Test
lnC	0.0000	0.0000	0.0000
DlnC	0.0000	0.0000	0.0000
lnCOFDI	0.0000	0.0000	0.0000
DlnCOFDI	0.0000	0.0000	0.0000
lnGDP	0.0000	0.0039	0.0000
DlnGDP	0.0000	0.0000	0.0000
lnPIAV	0.0007	0.0022	0.0000
DlnPIAV	0.0000	0.0000	0.0000
lnEI	0.0000	0.0000	0.0000
DlnEI	0.0000	0.0000	0.0000
lnPFDI	0.0320	0.0000	0.0112
DlnPFDI	0.0000	0.0000	0.0000
lnPUP	0.0000	0.0000	0.0000
DlnPUP	0.0000	0.0000	0.0000
lnPFTV	0.0552	0.0681	0.0000
DlnPFTV	0.0000	0.0000	0.0000

. Therefore, all variables in this study were first-order single integer sequences.

5.2. Results and Discussion of the Static Panel Regression Model

In order to better reflect the specific relationship between variables, we first used the Hausman test to compare the fixed effect model with the random effects model. The result showed that the chi-square value was 66.06, and the corresponding probability value was 0.0000 (

Table 3. Hausman test results.

Test Summary	Chi-Sq. Statistic	Prob > chi2
Cross-section random	66.06	0.0000

). The probability value was less than 0.05, and thus, the original hypothesis is rejected. Therefore, we chose the fixed effects model, and the regression results are shown in

Table 4. Fixed effects regression results.

Variable	Coefficient
lnCOFDI	−0.0230 ***
lnGDP	0.5975 ***
lnPIAV	0.1426 ***
lnEI	−0.9476 ***
lnPFDI	0.0295 ***
lnPUP	1.2805 ***
ln PFTV	0.0015
Cons	−1.6928 ***
R2	0.905
Prob(F-statistic)	0.0000

*** p < 0.01.

.

The regression results showed that the model fit the whole sample data well, as shown by the R² value of 0.905. The corresponding p-values of PFTV were all greater than 0.1, which had no significant impact on the carbon emissions of countries along the route at the 10% significance level. The p-values corresponding to COFDI, GDP, PIAV, EI, PFDI, and PUP were less than 0.05. Therefore, at the 5% significance level, COFDI, GDP, PIAV, EI, PFDI, and PUP significantly impacted China’s OFDI carbon emissions.

(1) The COFDI influence coefficient was significantly negative. For every 1 unit increase in COFDI, the carbon emissions of countries along the route decreased by 0.2242 units. This showed that COFDI had a carbon emissions reduction effect on countries along the “Belt and Road”. The main reason OFDI in China had a carbon emission reduction effect was that state-owned enterprises were the main force. Compared with local enterprises in the “Belt and Road” countries, these enterprises had advantages such as a larger scale, more advanced environmentally friendly technology, and stricter environmental standards. Since the reform and opening up, China’s state-owned enterprises have been constantly reforming and innovating. The production and operation management system has been constantly improving, production technology has gradually taken the lead, and production emission standards have become increasingly strict. A perfect production and operation management system can effectively improve production efficiency and reduce the redundancy of production factors. Advanced production technology can increase energy intensity and reduce energy consumption. Strict production emission standards can directly reduce enterprise carbon emissions and promote enterprise transformation and upgrading. Therefore, COFDI had a carbon emissions reduction effect on countries along the “Belt and Road”. (2) The coefficient of the GDP explanatory variable representing the scale effect of China’s OFDI economy was significantly positive. It showed that each unit of increase in the economic scale of countries along the route led to an increase of 0.5614 units in the country’s carbon emissions. The main reason for this was that the economies of many of the Belt and Road countries are energy intensive. Fossil fuel is the main consumption of energy produced by countries along the route. With the increase in these countries’ economic scale, fossil fuel consumption also increased rapidly. Take Vietnam, the third largest oil consumer in Southeast Asia, for example. When China’s OFDI drove its economic development, the total carbon emissions of energy-intensive industries increased with the increase in economic scale. (3) The PIAV explanatory variable coefficient representing China’s OFDI industrial structure effect was significantly positive. The greater the proportion of industrial added value to GDP, the greater the regional industrial energy consumption. As a result, carbon emissions also grew rapidly. (4) The coefficient of the EI explanatory variable representing China’s OFDI production technology effect was significantly negative. The increase in energy intensity significantly reduced carbon dioxide emissions. This meant that technological progress could reduce the carbon emissions of countries along the route. (5) Among the control variables, the effect of PFDI on the carbon emissions of countries along the route was significantly positive. However, COFDI had a significant inhibitory effect on the carbon emissions of countries along the route. It can be seen that the quality of China’s investment in countries along the Belt and Road was relatively high. The coefficient of PUP was significantly positive. Population urbanization is transforming human identity from a rural population to urban residents. Moreover, the production and lifestyle of residents have also changed, and energy consumption has increased. Second, population gathering will promote urban construction. A large number of building demands will lead to the growth of various forms of energy consumption, which will increase carbon emissions. PFTV had no significant impact on the carbon emissions of countries along the route.

5.3. Results and Discussion of Dynamic Panel Regression Model

Given the possible time-series-related problems of carbon emissions, we included the one-period lag phase of carbon emissions into the model to build a dynamic panel model. The one-step SYS-GMM estimation method was adopted for the calculation, and the results are shown in

Table 5. GMM regression results.

Variables	Coefficient
L. lnCO2	0.031 ***
lnCOFDI	−0.117 **
lnGDP	0.111 **
lnPIAV	0.321 *
lnEI	−0.123 **
lnPFDI	0.391
lnPUP	0.138 *
ln PFTV	−0.027
Cons	1.023 **
R-squared	0.9013
Wald χ2	0.0002
AR (1)	0.0013
AR (2)	0.0901
Hansen test	0.1267

*** p < 0.01, ** p < 0.05, * p < 0.1.

.

It can be seen from Table 5 that the decision coefficient R² is 0.9013, and the p-value of Wald χ² was less than 0.05. The regression results are reliable. AR test results showed that AR (1) was less than 0.05, and the random disturbance term had first-order autocorrelation. AR (2) was greater than 0.05, and there was no second-order autocorrelation in the random disturbance term. In addition, Hansen’s test results were greater than 0.05. This showed that the model did not have excessive regression, and the effect was good.

The carbon emissions coefficient lagging by one period was significantly positive at the 5% level. This showed that the carbon emissions of countries along the route were significantly affected by their lag of one period, but the impact was small. The coefficient was 0.031. COFDI passed the 5% significance level test, with a coefficient of −0.117. The results showed that China’s OFDI had a carbon emissions reduction effect on countries along the “Belt and Road”. This was consistent with the fixed effect model. Among other explanatory variables, GDP and EI passed the 5% significance level test. PIAV passed the 10% significance level test, which showed that PIAV significantly impacted the carbon emissions of countries along the route. Among the aforementioned variables, the impact coefficients of GDP and PIAV on carbon emissions were significantly positive, which was consistent with the regression result of the static panel. The impact coefficient of EI on carbon emissions of countries along the route was significantly negative, contrary to the regression result of the static panel. The carbon emissions reduction effect gradually emerged because of the continuous accumulation and improvement of production technology. Among the control variables, PUP passed the 10% significance level test. It had a significant positive impact on the carbon emissions of countries along the route, which was consistent with the regression results of the static panel. PFDI and PFTV did not pass the 10% significance level test, which means that they had no significant impact on the carbon emissions of countries along the route.

5.4. Results and Discussion of the Threshold Regression Model

In this study, regression analysis was carried out for three cases of the threshold variables GDP, PIAV, and EI: without any threshold, with one threshold, and with two thresholds. The F-statistic of the threshold test for each case and the p-value obtained using the bootstrap method are shown in

Table 6. Threshold effect test (bootstrap = 0 300 300).

Threshold Variable	Threshold	F-Statistic	Prob.
lnGDP	Single	111.36	0.012
	Double	55.54	0.147
	Triple	47.32	0.167
lnPIAV	Single	98.35	0.000
	Double	40.37	0.247
	Triple	31.33	0.286
lnEI	Single	39.69	0.250
	Double	/	/
	Triple	/	/

. The single threshold test of the threshold variables GDP and PIAV gave a significant result at the 5% level. The corresponding sampling p-values were 0.012 and 0.000, respectively. The double- and triple-threshold effects were not significant at the 10% level. Therefore, GDP and PIAV had important impacts on the COFDI carbon emissions effect, and there was a significant threshold effect. The single-threshold test of the threshold variable EI did not show significance at the 10% significance level, and thus, EI had no threshold effect on COFDI.

According to Hansen’s suggestion, this study only searched the non-repeated values in the threshold variables GDP and PIAV. These non-repeated values were arranged in ascending order, ignoring about 1% of the observed values before and after. Only the middle 98% of the samples were taken as the candidate range of threshold values. In addition, to improve the accuracy of the threshold estimation, we used the “grid search method” proposed by Hansen in the threshold regression to find the threshold values ${\mathsf{\eta }}_1$ and ${\mathsf{\eta }}_2$ of GDP and PIAV. As shown in

Table 7. Threshold estimator (level = 95).

Threshold Variable	lnGDP		lnPIAV
	${\mathsf{\eta }}_1$	95% Confidence Interval	${\mathsf{\eta }}_2$	95% Confidence Interval
$\mathrm{Threshold}(\mathsf{\eta })$	7.2696	7.1384, 7.2800	4.0106	3.9807, 4.0174

, the threshold values of GDP and PIAV were 7.2696 and 0.621, respectively. The threshold values of these two threshold variables were within the 95% confidence interval, and thus, the original hypothesis was accepted. That is, the threshold values of the two threshold variables were equal to the actual threshold values, and the threshold estimates were true and effective.

$\mathrm{Threshold}(\mathsf{\eta })$ As shown in

Table 8. Threshold regression model results.

Variables	lnGDP	lnPIAV
	Coefficient	Coefficient
lnGDP	0.5633 ***	0.5903 ***
lnPIAV	0.1868 ***	0.1199 **
lnEI	0.9795 ***	1.0189 ***
lnPFDI	0.0269 ***	0.0219 ***
lnPUP	1.2436 ***	1.2047 ***
lnPFTV	0.0052	0.0021
$\mathrm{lnCOFDI}({\mathsf{\eta }}_1$ < 7.2696)	0.0286 ***
$\mathrm{lnCOFDI}({\mathsf{\eta }}_1$ > 7.2696)	−0.0019 ***
$\mathrm{lnCOFDI}({\mathsf{\eta }}_2$ < 4.0106)		0.0241 ***
$\mathrm{lnCOFDI}({\mathsf{\eta }}_2$ > 4.0106)		−0.0586 ***
Cons	−1.5237 ***	−1.3217 ***
R-squared	0.9013	0.8821
F test	Prob > F = 0.0000	Prob > F = 0.0000

*** p < 0.01, ** p < 0.05.

, when GDP exceeded 7.2696, the COFDI coefficient was significantly negative at −0.0286. It had a restraining effect on the carbon emissions of countries along the route. When GDP was lower than this level, the carbon emissions reduction effect of COFDI could not be realized. This result was consistent with the previous analysis of the economic scale effect. When the GDP level of the countries along the route was lower than 7.2696, the countries accelerated the expansion of national production and economic scale. China’s OFDI shall be used as much as possible, and carbon emissions products shall not be selected. In turn, the scale of its carbon emissions increased accordingly. When the GDP level of countries along the route was higher than 7.2696, the living standard of residents was relatively good, and residents began to pay attention to environmental problems. Residents gradually used clean energy in their lives and buy environmentally friendly products. Furthermore, residents hope that the government will adopt stricter environmental protection standards to improve environmental problems.

When PIAV was higher than 4.0106, the COFDI coefficient was significantly negative. It had a restraining effect on the carbon emissions of countries along the route. When PIAV was lower than this level, the carbon emissions reduction effect of COFDI could not be achieved. When the PIAV of countries along the route was higher than 4.0106, industrial construction in countries along the route was preliminarily completed. Therefore, when receiving China’s OFDI inflows, these countries were more inclined to use the inflow of investment in the tertiary industry. Therefore, COFDI showed a carbon emissions reduction effect. When the PIAV of countries along the route was lower than 4.0106, countries used COFDI to accelerate industrial expansion, and there was a large inflow of investment in the secondary industry. In turn, this promoted the growth of carbon emissions of countries along the route.

6. Conclusions and Suggestions

Based on the panel data of 42 countries along the Belt and Road from 2003 to 2020, we applied a fixed effects model, SYS-GMM, and a threshold regression model to evaluate the static, dynamic, linear, and non-linear effects of China’s OFDI on the carbon emissions of countries along the Belt and Road. The research found the following: (1) The direct effect of COFDI on carbon emissions of countries along the route was significantly negative, as shown in Table 4. (2) The economic scale and industrial structure effects of COFDI promoted carbon emissions of countries along the route. Production technology effects restrained the carbon emissions of countries along the route, and technology effects played a leading role, as shown in Table 4. (3) The SYS-GMM estimation results showed that the carbon emissions of countries along the route were significantly affected by their lag of one period. However, the impact is small, as shown in Table 5. (4) The threshold regression model results showed that GDP and PIAV had significant threshold effects on the carbon emissions of OFDI in China. When the GDP of countries along the route was higher than 7.2696, COFDI had a restraining effect on the carbon emissions of countries along the route. When the PIAV of countries along the route exceeded 4.0106, China’s OFDI carbon emissions reduction effect could not be achieved, as shown in Table 8.

Based on the research background and the above research conclusions, the following countermeasures and suggestions are proposed. (1) Strengthen cooperation on carbon emissions reduction and reach a consensus on green development. China and countries along the “Belt and Road” should jointly promote all parties to fully implement the United Nations Framework Convention on Climate Change and its Paris Agreement. Moreover, the “maximum common denominator” of carbon emissions reduction should be actively sought for countries along the “Belt and Road”. Through cooperation forums, development conferences, and other forms, China will strengthen dialogue and exchanges and build international consensus on green development. China and countries along the “Belt and Road” should continue to implement the “Belt and Road” carbon emissions reduction cooperation plan. All countries should jointly promote the construction of low-carbon demonstration areas and carbon-emissions-reduction-related material assistance centers to promote the low-carbon and sustainable development of all countries. (2) Accelerate the process of trade openings and jointly promote the low-carbon construction of industrial parks. China and the countries along the line build differentiated fulcrums with the help of the development model and experience of “ports—industrial parks—urban communities”. For example, in the Silk Road Economic Belt, overseas fulcrum cities are built in the form of industrial parks. The fulcrums of the Maritime Silk Road are presented in the form of ports. The industrial park or port can realize the integrated construction of comprehensive utilization of energy resources, green infrastructure, public services, material recycling, environmental protection, water conservation, cleaner production, etc. For example, Brunei Damo Island Petroleum Refining and Chemical Industrial Park focuses on the “Refining and Chemical Integration” project. Qingshan Park in Indonesia mainly focuses on “nickel iron + stainless steel integration”. Malaysia China Kuantan Industrial Park is focusing on the “steel joint project”. Under the condition of fully understanding the unique resources and technological advantages of countries along the line, green construction of industrial parks can be achieved through complementary advantages and reasonable distribution. Finally, the industrial energy consumption intensity of China’s industrial parks along the “Belt and Road” will be 20–30% lower than that of the host country. (3) Accelerate cooperation on green energy projects and optimize the energy structure. The empirical results showed that the economic development and urban construction promoted by the “Belt and Road” countries will lead to the growth of carbon emissions. Countries along the Belt and Road should define their energy structure optimization objectives, establish an energy resource carrying capacity evaluation system consistent with their reality, and effectively reduce the use intensity of resources with greater pollution. The “Belt and Road” countries and China should accelerate the construction of green energy infrastructure projects to provide energy support for low-carbon development. From 2013 to 2021, the cumulative controllable installed capacity of overseas investment of Chinese enterprises in photovoltaic and wind power plants exceeded 10 GW. By 2025, the target of more than 15 GW will be achieved. In addition, investment in coal projects (including coal power and coal mines) has declined since reaching its peak in 2015. It dropped to zero in the first half of 2021. All countries should give full play to the substitutability of clean and renewable energy in the use of energy and increase the proportion of clean energy. (4) Establish a green development fund. China and the “Belt and Road” countries should jointly fund the establishment of a green development fund, as well as constantly increase efforts to promote the development of green finance and strengthen financial support for the “Belt and Road” green projects. China attaches great importance to the role of finance in green and low-carbon development and was one of the first countries to develop green finance. In order to promote the low-carbon and sustainable development of China and the “Belt and Road” countries, all countries should continue to explore, improve, and optimize the green financial policy framework and support the innovative development of green finance. Policy banks and commercial banks in various countries have taken green projects as a priority, and the scale of green credit has continued to expand. It has also continuously launched innovative green financial products, such as green bonds, clean energy investment funds, and green PPPs, and provided effective financial support for the low-carbon development of the “Belt and Road”.