The Impact of Coordinated Two-Way FDI Development on Carbon Emissions in Belt and Road Countries: An Empirical Analysis Based on the STIRPAT Model and GMM Estimation

Abstract

The Belt and Road Initiative (BRI) promotes significant cross-border investment, raising critical questions about its environmental consequences, particularly regarding carbon emissions. This paper uses panel data from 47 countries that participated in the “Belt and Road Initiative” earlier from 2000 to 2020 to conduct theoretical analysis and empirical research on the relationship between the coordinated development of two-way FDI and carbon emission intensity, dividing it into scale effect, technology effect and structure effect. The coordinated development of two-way FDI can have an increasing or decreasing impact on carbon emission intensity through these three effects. The main findings of this paper are as follows: (1) The improvement of the degree of coordinated development of two-way FDI significantly reduces carbon emission intensity. (2) The improvement of the degree of coordinated development of two-way FDI can enhance the level of technological innovation, while the improvement of the level of technological innovation will increase carbon emission intensity, thereby reducing the carbon emission reduction effect of the coordinated development of two-way FDI. (3) The improvement of the degree of coordinated development of two-way FDI can reduce carbon emission intensity by promoting the upgrading of industrial structure. Based on the above conclusions, this paper puts forward the following suggestions for the subsequent development of countries along the “Belt and Road”: (1) Further increase two-way FDI and promote the coordinated development of two-way FDI. (2) Promote the upgrading of industrial structure and the green transformation of technology. (3) Increase economic freedom to provide a good environment for economic development.

1. Introduction

The Belt and Road Initiative marks China’s first large-scale, global, and comprehensive international economic cooperation project since the reform and opening-up era. This initiative calls on all nations to collaborate in creating a community with a shared future for mankind. Meanwhile, this initiative embodies the vision of sustainable development by promoting shared prosperity and environmental stewardship among participating nations. China’s goal is to establish the Belt and Road as a “path to prosperity” for the world and a “dream factory” for the collective modernization of all nations. Since the early 21st century, propelled by economic globalization and the signing of various regional economic cooperation agreements, investment activities in Belt and Road countries have steadily expanded, with the scale of two-way FDI rising significantly. The Belt and Road Initiative not only drives economic growth but also underscores the importance of sustainable development by integrating green investment practices and low-carbon technologies. Under this framework, signatory nations must prioritize both outbound investments and capital attraction to establish a multifaceted network of connectivity. Such balanced investment strategies are critical for achieving sustainable development goals, as they ensure equitable resource distribution and minimize environmental degradation. Over more than a decade of development, China and its partner countries have built a well-rounded multilateral and bilateral cooperation platform, significantly enhancing the quality and level of trade and investment among the countries along the routes. This collaborative approach fosters sustainable development by leveraging mutual strengths to address global challenges like climate change and resource scarcity. According to data from the China Social Sciences Network, the trade and investment liberalization and facilitation spurred by the Belt and Road Initiative have greatly increased the appeal of the participating countries and regions to high-quality global capital.

This paper builds a basic regression model based on theoretical analysis and starting from STIRPAT, an environmental economics theoretical model. Compared with the basic regression equations constructed by Gong and Liu (2020) [1], Kamal et al. (2023) [2], Mahadevan and Sun (2020) [3], which only rely on theoretical analysis (Zhu et al., 2022; Qiu et al., 2024; Wei et al., 2023; Gong et al., 2021) [4,5,6,7], it has more definite variable relationships and causal paths, which can improve the logical coherence of the argument. It enhances the explanatory power. In addition, compared with previous research methods, this paper incorporates some cases to assist in the analysis and employs more regression models to test the robustness of the results, taking into account many factors such as time control methods, variable representation, extreme values, and the impact of the epidemic. In terms of endogeneity tests, this paper uses four GMM models. These methods can help to expand the research perspective and improve the reliability of the study.

Existing scholarship examining IFDI-OFDI-carbon emissions linkages demonstrates substantial maturation, with numerous investigations dissecting underlying mechanisms. This analysis advances scholarly discourse through dual innovations. Primarily, it designates coordinated two-way FDI development as the explanatory construct. Contemporary research predominantly addresses unilateral FDI while neglecting IFDI-OFDI interdependencies, despite their concurrent operation within economies. Two-way FDI exhibits reciprocal influence, with interaction effects substantively impacting economic trajectories. Isolating IFDI or OFDI fails to unveil their intrinsic correlations (Wang and Duan, 2023; Nie et al., 2022) [8,9]. Concurrently, the Belt and Road Initiative prioritizes mutually advantageous outcomes across participating nations. The initiative mandates concurrent strategic emphasis on outward expansion (“going global”) and inward attraction (“bringing in”). Investigating two-way FDI coordination thus more authentically reflects international cooperation paradigms under this framework. Moreover, employing this integrated metric enables governments to formulate balanced investment strategies rather than unilateral approaches, yielding superior policy relevance compared to isolated FDI examinations (Kongkuah, 2023) [10]. Consequently, this investigation adopts coordinated two-way FDI development as its explanatory nucleus.

However, a significant gap remains in the empirical literature. While numerous studies have examined the unilateral effects of IFDI or OFDI on carbon emissions, the synergistic effect of their coordinated development is often overlooked. This study aims to fill this gap by investigating the core research question: How does the coordinated development of two-way FDI impact carbon emission intensity in BRI countries, and through what mechanisms does this effect operate? To answer this, we construct a coordinated development index for two-way FDI and employ an empirical framework based on the STIRPAT model to decipher the scale, technological, and structural effect pathways.

2. Literature Review

2.1. Related Studies on Sustainable Development

Sustainable development is closely associated with the influence of two-way foreign direct investment (FDI)—comprising both inward and outward FDI—on carbon emissions, particularly in the context of global efforts to address climate change and promote green economic transformation. Two-way FDI can affect carbon emissions through various mechanisms, including technology transfer, industrial structure adjustment, and the interaction of environmental policies across countries. At the same time, the Sustainable Development Goals (SDGs) offer a comprehensive policy framework and evaluation criteria for assessing the environmental and developmental impacts of such investment flows (Ga et al., 2025) [11]. This establishes a clear nexus between sustainable development and the environmental outcomes of two-way FDI.

On the one hand, the sustainable development effect of inward FDI (inflow) can be achieved through the transfer of green technologies. For instance, multinational companies introduce clean production technologies such as electric vehicles and smart grids in the host country. Meanwhile, environmental management experience, such as international standards like ISO14001 [12], can drive enterprises in the host country to reduce emissions. In addition, investment in renewable energy, for instance, European enterprises’ investment in wind power and photovoltaic projects in China can also have a positive impact on promoting low-carbon transformation. However, if the environmental regulations of the host country are lenient, it may become a “pollution haven”. At the same time, energy-intensive FDI (such as in the petrochemical industry) may increase carbon emissions, intensify resource consumption, and thus have the negative impact of potential pollution transfer (Abujder Ochoa et al., 2025; Keşanlı et al., 2025) [13,14]. For instance, concerns have been raised about the potential relocation of energy-intensive manufacturing sectors, such as steel or cement production, to BRI partner countries with less stringent environmental standards, which could exemplify the ‘pollution haven’ effect within the Initiative’s framework.

On the other hand, the sustainable development effect of outward FDI (outflow) also has the positive impact of global green technology cooperation and the negative impact of carbon leakage risk. For instance, when Chinese enterprises acquire new energy companies in Europe, it enables domestic enterprises to obtain low-carbon technologies through overseas mergers and acquisitions, thereby achieving reverse technology spillover. For instance, Apple’s commitment to carbon neutrality in its supply chain can prompt multinational companies to require their overseas subsidiaries to adopt green standards, thereby achieving the construction of a low-carbon supply chain. It cannot be ignored that the situation where developed countries transfer high-carbon industries to developing countries, such as the response measures of the EU’s carbon border tax (CBAM), will lead to the relocation of industries and the transfer of emissions. However, global carbon emissions have not decreased (Arias and Varela-Aldás, 2025; Benović, 2025) [15,16].

Therefore, based on the above analysis, how can sustainable development policies optimize the carbon emission impact of two-way FDI? This is a major issue worthy of study. From the perspective of the policy-making direction of the host country, attracting green FDI, strengthening environmental regulations, and adopting green incentive measures can largely optimize the carbon emission impact of two-way FDI. From the perspective of the policy orientation of the home country, carbon footprint regulation, green diplomacy and climate cooperation can effectively guide responsible investment. Among them, the “Belt and Road” green investment initiative led by China can vigorously promote the development of FDI in low-carbon infrastructure, such as China’s overseas photovoltaic power station projects, etc. This study also takes the “Belt and Road” countries as the research objects, focusing on exploring how to solidly optimize the carbon emission impact of two-way FDI under the framework of sustainable development policies.

2.2. Direct Impact of Two-Way FDI on Carbon Emissions

Currently, the research on the relationship between two-way FDI and carbon emissions can primarily be categorized into three main perspectives. One viewpoint asserts that an increase in either IFDI or OFDI results in higher carbon emissions. Within the IFDI research domain, many scholars support the pollution paradise hypothesis (Khan et al., 2024) [17], which suggests that in the context of trade liberalization, certain developing nations, eager for economic growth, tend to have relatively lenient environmental standards. As a result, multinational corporations prefer to set up production in these countries to cut costs, ultimately exacerbating the environmental degradation in those host nations. AL-Barakani et al. (2022) examining China’s OFDI impact on Belt and Road Initiative nations’ carbon outputs, documented OFDI increases elevating emissions within lower-middle income economies while reducing them within low-income states [18]. Tichá et al. (2024) investigating IFDI-carbon linkages across 67 heterogenous income nations, observed IFDI surges augmenting emissions specifically within low-income territories [19]. Chang et al. (2022) utilizing 1997–2018 Belt and Road panel data through sequential regressions, determined IFDI increments substantially elevated carbon outputs [20]. Concerning OFDI impacts, Qu and Luo (2021) discovered OFDI expansion amplified domestic carbon emissions when utilizing China’s interprovincial panels [21].

The second view holds that IFDI or OFDI will promote carbon reduction. At the IFDI research level, the main supporters are in favor of the “pollution halo” hypothesis (Kaushal et al., 2024) [22], arguing that multinational companies bring advanced technologies and experiences when investing abroad to help host countries reduce environmental pollution. At the OFDI level, the main view is that outward direct investment can adjust the industrial structure and there is a reverse technology spillover effect, which can promote technological development in the home country and thereby reduce environmental pollution (Uzair Ali and Wang, 2024) [23]. Wu and Li (2022) [24] carried out a regression analysis using the dynamic spatial Durbin model with inter-provincial panel data from China covering the period from 2009 to 2018. Their findings revealed that IFDI significantly reduced carbon emission intensity, while OFDI showed no notable effect on carbon emission intensity (Rahim and Chen, 2024; Khan et al., 2022; Rafique et al., 2022) [25,26,27]. Similarly, Dong et al. (2023) [28] applied a spatial econometric model and found that China’s OFDI could notably lower carbon intensity in the host countries. Concurrently, select researchers have probed bidirectional FDI coordination’s nexus with carbon emissions. Ganda (2024) [29] alongside Nie et al. (2022) [9] leveraged a coupled system model, quantifying Chinese provincial bidirectional FDI’s coordinated development extent; applying this as an explanatory variable. Their analyses revealed heightened two-way FDI coordination diminishes environmental contamination; its mitigation extends expressly to carbon discharge. These competing hypotheses provide the foundational theoretical backdrop for our analysis. Our empirical model is designed to test which of these effects, or what combination thereof, dominates in the context of coordinated two-way FDI within BRI countries.

A third perspective postulates phased host-country carbon emission impacts from two-way FDI. Its frequent methodology examines environmental EKCs, wherein pollution and economic advancement exhibit an inverted U-shaped nexus. Multiple scholars extend this framework to FDI analysis by incorporating FDI alongside quadratic terms into models. Raghutla et al. (2024) [30] analyzed 1980–2020 panel datasets spanning Singapore, China, and India; contemporaneously, Liu et al. (2021) [31] scrutinized 1998–2016 provincial panel data across thirty Chinese municipalities. Both empirically substantiated EKC significance employing FDI as principal explanatory variables. Another common method involves selecting an appropriate threshold variable to build a threshold regression model. Additionally, factors like the level of economic development (Nie et al., 2022) [9] and urbanization have also been found to exhibit threshold effects in the relationship between IFDI or OFDI and carbon emissions.

The existing literature presents three contrasting perspectives on the FDI–carbon emissions nexus. The divergence in these findings underscores the complexity of the relationship and suggests that the impact may be contingent on factors such as the level of economic development, environmental regulations, and the type of FDI. Crucially, most studies analyze IFDI and OFDI in isolation, potentially missing their interactive and synergistic effects. This highlights a significant gap: the need for an integrated analysis that captures the coordination between these two investment flows, which is particularly relevant in the context of the BRI’s emphasis on balanced two-way investment.

2.3. Indirect Impact of Two-Way FDI on Carbon Emissions

Beyond analyzing the immediate effects of two-way FDI on carbon emissions, numerous academics have delved into its secondary influence on carbon emissions. For instance, Qu and Luo (2021) [21] found that OFDI can mitigate the amplification effect on carbon emissions by promoting green technological innovation, and it can also worsen carbon emissions by amplifying the economic scale; Kaushal et al. (2024) [22] discovered that IFDI has the potential to lessen the intensification of carbon emissions through the enhancement of governmental quality. Subsequently, an examination of several primary indirect impact routes will be showcased.

A significant body of research has examined the connection between FDI and carbon emissions, with particular attention to one-way FDI and its impact on emissions. Some scholars have focused on understanding the mechanisms at play, while others have explored potential threshold effects, employing econometric techniques like the GMM model, spatial econometric models, and bidirectional fixed effect models. These studies have greatly expanded knowledge of how one-way FDI relates to carbon emissions. Prevailing scholarship predominantly examines unidirectional FDI domains. Employing coordinated two-way FDI development as an explanatory construct constitutes a nascent methodological approach, its application remaining markedly scarce. Conventional one-directional FDI investigations singularly address either IFDI or OFDI impacts upon carbon outputs, whereas empirical reality demonstrates substantive reciprocal interplay between these capital flows. Exclusive analysis of unilateral FDI potentially obscures their interactive consequences for carbon emissions. Moreover, while both IFDI and OFDI follow similar paths to influence carbon emissions, their effects may differ in direction, with positive and negative influences potentially canceling each other out. Analyzing one-way FDI in isolation and merely adding their effects together may result in an underestimation or overestimation of international direct investment’s actual influence on carbon emissions. The introduction of the coordinated development variable helps avoid this by addressing the independent impacts of IFDI and OFDI, mitigating the problem of double-counting, and allowing for the assessment of their combined net effect. As a result, this study adopts the coordinated development variable for bidirectional FDI.

2.4. Summary of Literature Review

Bidirectional FDI coordinated development refers to achieving a balance in scale development between IFDI and OFDI through their benign interaction, so that international direct investment can have the greatest positive impact on the country. According to Dunning’s investment development path theory, a nation’s capacity to attract foreign capital and engage in outward investments is closely linked to its level of economic development. By drawing in foreign investments, developing countries can strengthen their outward investment capabilities. This is primarily driven by the significant technology spillover effect of IFDI, which helps bridge the technological gap between nations, boosting the host country’s economic growth and encouraging domestic firms to expand internationally [9]. The increase in OFDI enhances the technological capacity of the home country and broadens its market reach through reverse technology spillovers, thereby strengthening the home country’s location advantages. According to the international production compromise theory, when enterprises make foreign investment decisions, they mainly focus on ownership advantage, location advantage and internalization advantage, so OFDI in turn can attract IFDI. When two-way FDI promotes further economic development, host country enterprises themselves will also have stronger ownership and internalization advantages, thereby enhancing their ability to discover and utilize foreign location advantages and further expand outward investment.

3. Theoretical Basis

The coordination between IFDI and OFDI is crucial because it moves beyond examining their isolated impacts and captures their synergistic interaction within an economy. A high degree of coordination signifies that a country is not merely a passive recipient of foreign capital (which could be clean or polluting) nor an unchecked exporter of domestic capital (which could lead to carbon leakage). Instead, it reflects a balanced, strategic approach to engaging with the global economy.

This balance is theorized to reduce carbon emission intensity through several mechanisms. Firstly, it helps mitigate the ‘pollution haven’ effect. A country actively engaged in OFDI is less likely to become a dumping ground for polluting industries from abroad through IFDI, as its own outbound investments often reflect a more advanced industrial structure and environmental standards. Conversely, the technological and managerial spillovers from high-quality IFDI can equip domestic firms with the capabilities to undertake OFDI in more technologically advanced and environmentally sustainable sectors, creating a virtuous cycle.

Secondly, coordination promotes industrial upgrading. IFDI and OFDI can reinforce each other in steering the domestic industrial structure towards less carbon-intensive tertiary and high-tech sectors. For instance, IFDI in green technology can boost a country’s capability, which then enables OFDI in similar sectors abroad, further consolidating its comparative advantage in low-carbon industries.

Therefore, the coordinated development index is not merely a statistical combination but a proxy for this strategic, balanced interaction that is hypothesized to be more conducive to sustainable development than the volume of IFDI or OFDI alone.

The deterioration or improvement of the environment is mainly achieved through changes in economic scale, alterations in production technology, and changes in economic structure. These three causative mechanisms—scale effect, technological effect, and structural effect—respectively constitute the foundational framework. Subsequent scholarly inquiries principally derive from this tripartite conceptualization. Consequently, this investigation theoretically scrutinizes the two-way FDI coordination–carbon emissions nexus using this analytical scaffolding.

The extensive impact of synchronized development in two-way Foreign Direct Investment (FDI) implies that enhancing the extent of such development will alter the economic scale of either the host or the originating country, consequently affecting carbon emissions. IFDI has the option to either directly increase its local output through the construction of factories in the host nation or indirectly enhance production by importing funds and technology. In the context of OFDI, a surge in OFDI results in a substantial return of capital to the originating country, thereby enlarging its economic magnitude (Wang and Duan, 2023) [8]. Concurrently, a two-way FDI connection exists, and the deployment of IFDI is expected to invigorate rivalry between international and local capital, foster technological advancements in local businesses, thereby hastening the transition to a global scale; The increase in OFDI expands overseas markets and improves domestic technology, attracting more IFDI. This two-way FDI interaction creates a dynamic cycle that promotes market expansion and optimizes resource allocation, thereby increasing the size of the domestic economy. As the degree of coordinated development of two-way FDI grows, the economy will experience expansion. Typically, economic growth leads to higher carbon emissions; however, as regional GDP increases, the impact on carbon intensity depends on the relative growth rates of the economy and carbon emissions. If the economic growth rate surpasses that of carbon emissions, it will help reduce carbon intensity. Otherwise, carbon intensity will rise. Conversely, as economic growth escalates, inhabitants’ earnings rise, leading to increased environmental requirements. This surge in production capacity will coincide with more rigorous environmental safeguards, consequently diminishing carbon emissions for promoting sustainable development.

The technological impact of the coordinated development of two-way FDI indicates that as the degree of coordinated development between two-way FDI improves, it leads to changes in the technological capabilities of either the host or home country, consequently influencing carbon emissions [32,33]. Regarding IFDI, domestic firms, in their competition with foreign companies, are compelled to boost their investment in technological research and development to maintain market share. At the same time, domestic enterprises can imitate and learn from the technology brought by foreign enterprises, and improve and innovate on that basis to make it more suitable for production in the host country, thereby enhancing their own technological level. For OFDI, multinational companies can acquire technology by directly acquiring leading host enterprises through cross-border mergers and acquisitions, or by establishing subsidiaries or technology alliances overseas; Furthermore, in the case of an “inverse gradient” OFDI, the incoming nations exhibit greater economic growth and more sophisticated technology. By establishing factories in these countries for production, it can receive more advanced production technology feedback, thereby accelerating technological innovation to make it more adaptable to global production (Lee and Zhao, 2023) [34]. Moreover, there is a two-way FDI interaction. When the influx of IFDI contributes to the enhancement of domestic technological capabilities, local companies gain a stronger ownership advantage, thus hastening their global expansion and leveraging OFDI to further boost domestic technological innovation. This cycle of “attracting-investing-attracting again” accelerates the exchange and development of technology. As a result, as the coordinated development of two-way FDI progresses, the level of technological innovation rises, which subsequently impacts carbon emissions. If the technological advancements driven by IFDI and OFDI focus on green and clean innovations, carbon emission intensity will decrease. Conversely, if the emphasis is placed on expanding production capacity with little focus on advancing clean technologies, carbon intensity will increase.

The structural impact of the coordinated development of two-way FDI refers to how an increase in the degree of such coordination can alter the industrial structure of either the host or the home country, which in turn affects carbon emissions. Regarding IFDI, when a nation enforces weaker environmental regulations, it may channel IFDI into high-carbon secondary industries, spurring industrialization and raising carbon intensity. However, if IFDI predominantly flows into the tertiary sector or other clean industries, promoting low-carbon development, carbon emissions intensity will decrease. As for OFDI, when developing nations acquire high-carbon industries from developed countries, such as steel production, they often utilize resource-seeking OFDI to meet domestic demands, accelerating industrialization and pushing the industrial structure toward high-carbon sectors. Conversely, more developed countries can move their high-pollution, energy-intensive production methods or industries abroad by investing in nations with less advanced economies and looser environmental regulations, offering both the space and financial backing necessary to foster clean industries and shift to low-carbon sectors (Wang and Duan, 2023) [8]. In addition, there is a linkage between IFDI and OFDI. The introduction of IFDI can promote the development of target industries through financial support and technological innovation, narrow the gap with foreign industries, thereby promoting OFDI in corresponding industries and squeezing out the share of backward industries for the development of target industries. The increase in OFDI further develops the target industry and enhances the country’s location advantage to attract more IFDI. Through this dynamic process, two-way FDI can prioritize the growth of targeted industries. As the coordinated development of two-way FDI advances, it has the potential to transform a country’s industrial structure. If this transformation leads to a shift toward low-carbon industries, carbon emission intensity will decrease for promoting sustainable development.

Given the consistent growth in economic development among countries along the Belt and Road Initiative, the dissemination of environmental protection concepts through climate agreements like the Paris Agreement, the overall enhancement of industrial structures in these countries, and the implementation of various environmental policies, it is important to also recognize that the green technological innovation capabilities of many countries along the route remain relatively weak. Based on these factors, the following hypotheses are proposed:

• $H_1:$ Enhanced coordination within bidirectional foreign direct investment diminishes carbon emission intensity.

While technological advancements can theoretically lead to either increased efficiency (reducing emissions) or increased scale of production (increasing emissions), the early stages of development and the current industrial focus in many BRI countries suggest that the scale-expanding effect of new technologies might initially dominate their efficiency-improving effect, leading to a potential ‘rebound effect’.

• $H_2:$ Enhanced technological advancements are associated with greater carbon intensity (a rebound effect), and two-way FDI can influence this intensity through technological advancements.

Structural transformation towards tertiary industries is generally associated with lower carbon intensity. Given the BRI’s increasing focus on connectivity and service sector projects, we hypothesize this effect will be significant.

• $H_3:$ Industrial structure advancement ameliorates carbon intensity; two-way FDI correspondingly influences intensity indirectly by propelling such structural upgrading.

The proposed theoretical framework and causal pathways are summarized in Figure 1.

4. Empirical Research

This study began by gathering relevant data, specifically panel data from countries along the Belt and Road. Given data availability, 47 countries with relatively complete data from 2000 to 2020 were chosen as the sample for the analysis (

Table 1. Countries along the Belt and Road covered in this article (47).

Regions	Countries
East and Southeast Asia (9)	Mongolia, Singapore, Malaysia, Indonesia, Thailand, Laos, Cambodia, Vietnam, Philippines
West Asia and North Africa (11)	Iran, Turkey, Jordan, Lebanon, Israel, Saudi Arabia, Oman, United Arab Emirates, Qatar, Bahrain, Egypt
South Asia (4)	India, Pakistan, Bangladesh, Sri Lanka
Central Asia (3)	Kazakhstan, Uzbekistan, Kyrgyzstan
Cis (6)	Russia, Ukraine, Belarus, Georgia, Azerbaijan, Moldova
Central and Eastern Europe (14)	Poland, Lithuania, Estonia, Latvia, Czech Republic, Slovakia, Hungary, Slovenia, Croatia, Bosnia and Herzegovina, Albania, Romania, Bulgaria, North Macedonia

Data source: https://www.yidaiyilu.gov.cn/ (accessed on 31 December 2024).

). For instances of missing data, an interpolation method was employed to fill in the gaps, following the approach outlined by Xiang and Dai (2022) [32]. The primary data sources for this study include the UNCATD database, the World Bank’s WDI database, and the Heritage Foundation (HF).

The sample selection was based on the official list of BRI partner countries, with the primary criterion being the availability of continuous and reliable data for the key variables (IFDI, OFDI, CO2 emissions, GDP, etc.) throughout the study period (2000–2020). Countries with significant data gaps were excluded to ensure the robustness of the panel data analysis.

Regarding the primary factors, the variable elucidated in this study pertains to the carbon emission intensity chosen by Xiang and Dai (2022) [32] and Haq et al. (2022) [33], indicated by ${\mathrm{C}\mathrm{O}}_2$, also employ per capita carbon emissions ${\mathrm{P}\mathrm{C}\mathrm{O}}_2$ as the variable explained in the robustness test, as denoted by, both sourced from the WDI database. In this paper, the STIRPAT model is used as the basic theoretical model, and thus the inherent independent variables of the STIRPAT model are added as control variables, namely population, economic development level and technology. The primary basis for choosing these variables stems from the work of Nie et al. (2022) [9]. In this context, the measurement of population size is determined by the density of the population, economic growth is gauged by the GDP per person, and technological advancement is gauged by the intensity of energy, symbolized by $\mathrm{P}\mathrm{O}\mathrm{P}$, $\mathrm{A}\mathrm{G}\mathrm{D}\mathrm{P}$, and $\mathrm{E}\mathrm{N}\mathrm{E}$ respectively. The data used for the above variables are all from the WDI database. Citing earlier research, this document establishes various control variables, including external world openness, industrial configuration, and the rate of urbanization, as depicted by $\mathrm{O}\mathrm{P}\mathrm{E}\mathrm{N}$, $\mathrm{I}\mathrm{N}\mathrm{D}$and $\mathrm{U}\mathrm{R}\mathrm{B}\mathrm{A}\mathrm{N}$. These variables are also from the WDI database. For the mechanism study section, this paper sets up three mechanism variables: technological innovation, industrial structure advancement, and economic freedom. Technological advancement $\mathrm{S}\mathrm{C}\mathrm{I}$is gauged by the volume of publications in scientific and technological periodicals this year. Industrial structure upgrading $\mathrm{I}\mathrm{N}\mathrm{D}\mathrm{G}$, gauged per $\mathrm{E}\mathrm{C}\mathrm{O}\mathrm{F}$ through tertiary-secondary sector output value ratios within current years, indicates structural progression. Both mediating variables’ datasets originate from WDI archives. Economic freedom (ECOF), sourced from the Heritage Foundation (HF) of the United States, is included as a moderating variable. This index primarily measures aspects of freedom relevant to business operation and investment. While its specific weighting methodology has been debated, it remains a widely referenced proxy for the regulatory and institutional environment affecting investment flows and economic decisions, as mentioned by Raghutla et al. (2024) [30].

In this research, the key variable explained is the synchronized creation of two-way FDI, determined by merging IFDI and OFDI to produce novel variables. The IFDI and OFDI indicators used in this analysis are based on stock data. Stock data is preferred because it provides a clearer picture of the long-term effects of investment on carbon emissions, minimizing the influence of short-term fluctuations typical of flow data. Furthermore, investment returns generally extend beyond a single year, and using annual flow data might fail to capture the full impact of investment on carbon emissions. In terms of computational methods, this document utilizes techniques developed by academics such as Huang et al. (2018) [35] and Gong and Liu (2018) [36] employed the physics-based capacity coupling system model to assess the synchronized evolution of bidirectional foreign direct investment.

We now more clearly state that the Coupling Coordination Degree (CCD) model is adapted from physics and is widely established in sustainability, economic, and environmental studies to quantify the interaction and coordination level between two or more systems, making it suitable for measuring IFDI–OFDI coordination. Therefore, referring to the coupling system model, a bidirectional FDI coupling degree calculation formula is established, as shown in Formula (1), where $\mathsf{\alpha } \mathrm{a}\mathrm{n}\mathrm{d} \mathsf{\beta }$ is the weight of IFDI and OFDI. The Belt and Road Initiative is an open and inclusive regional cooperation agreement that requires the active participation of all co-building countries. Under this initiative, foreign investment is as important as outbound investment, and currently IFDI and OFDI are developing in a distinct synchronous manner, complementing each other. α and β are set to 0.5 to assign equal importance to IFDI and OFDI, reflecting the BRI’s principle of balancing “bringing in” and “going global.” The value of γ (=2) is identified as following the convention set by the foundational methodological paper ([35]) we built upon. Generally, $2\le \gamma \le 5$.

$${\mathrm{C}\mathrm{I}\mathrm{O}}_{\mathrm{i}\mathrm{t}}={\displaystyle \frac{{\mathrm{I}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}\times {\mathrm{O}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}}{{(\mathsf{\alpha }{\mathrm{I}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}\times \mathsf{\beta }{\mathrm{O}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}})}^{\mathsf{\gamma }}}}$$

(1)

Whereas coupling metrics merely gauge interaction intensity between systems, coordination measurements—as established by (Ganda, 2024) [29]—capture both interaction magnitude and individual system advancement levels, specifically two-way FDI scale. Consequently, this analysis further adopts the coordination paradigm Huang et al. (2018) [35] integrated within coupling framework models to construct the two-way FDI coordinated development index. Herein, T signifies the comprehensive investment metric, with computational methodology detailed in Formula (2). According to this formula, a higher value of ${IOFDI}_{it}$, a stronger interaction between the two-way FDI of a country and a larger scale of two-way FDI, implying a higher level of coordinated development. If the interaction is strong but the scale is small, or if the interaction is weak but the scale is large, the degree of coordinated development of two-way FDI will be relatively low.

$$T={\displaystyle \frac{{\mathrm{I}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathrm{O}\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}}{2}}$$

(2)

$${IOFDI}_{it}={DIO}_{it}=\sqrt{{CIO}_{it}\ast T}={({CIO}_{it}\ast {\displaystyle \frac{{IFDI}_{it}+{OFDI}_{it}}{2}})}^{1/2}={\left[{\displaystyle \frac{{IFDI}_{it}\times {OFDI}_{it}}{{(IFDI}_{it}+{OFDI}_{it})/2}}\right]}^{1/2}$$

(3)

The Coupling Degree (CIO) measures interaction strength. The Comprehensive Development Index (T) measures scale. The Coupling Coordination Degree (IOFDI/DIO), our core variable that integrates both interaction and development level.

The variables used in this study, along with their definitions and data sources, are summarized in

Table 2. Definition of main variables.

Types of Variables	Variable		Units	Definition/Measurement	Data Source
Explained variable	Carbon emission intensity	CO2	kg/$	Ratio of carbon emissions to USD GDP in that year	WDI
	Carbon emissions per capita	PCO2	kt	Carbon emissions per capita that year	WDI
Main explanatory variable	Foreign direct investment	IFDI	$’ m	Stock of foreign direct investment	UNCATD
	Foreign direct investment	OFDI	$’ m	Stock of foreign direct investment	UNCATD
	Two-way FDI Coordinated development	IOFDI	/	Calculated based on stock data	This paper calculates
Control variables	Population size	POP	People/km	Population density	WDI
	Economic level	AGDP	$	Gross regional product per capita	WDI
	Energy intensity	ENE	MJ/$	Energy consumption as a quotient of GDP for the year	WDI
	Outward openness	OPEN	/	The quotient of total imports and exports to GDP for the year	WDI
	Industry Structure	IND	/	Proportion of the added value of the tertiary industry in that year	WDI
	Urbanization rate	URBAM	/	Proportion of urban population to total population in the year	WDI
Moderating variables	Economic freedom	ECOF	/	Data released by the Heritage Foundation	HF
Mediating variable	Technological innovation	SCI	article	Number of scientific journal articles in the year	WDI
	Upgrading the industrial structure	INDG	/	The ratio of the output value of the tertiary industry to that of the secondary industry in that year	WDI

.

Table 3. Descriptive statistics.

Variables	Sample Size	Mean	Standard Deviation	Minimum Value	Maximum
Inward FDI stock (IFDI)	987	67,008.608	149,585.96	182	1,985,991
Outward FDI stock (OFDI)	987	29,697.439	103,910.85	0.993	1,379,349
Coordinated development of two-way FDI	987	115.383	152.244	1.409	1275.928
Carbon emission intensity	987	0.356	0.235	0.062	1.889
Carbon emissions per capita	987	6.475	7.029	0.160	47.657
Population size	987	317.426	1038.375	1.584	7965.878
Economic development level	987	10,142.348	13,271.076	279.62	98,041.362
Energy intensity	987	5.417	2.704	1.650	26.480
Openness to the outside world	987	100.313	54.689	24.702	437.327
Industrial structure	987	51.558	9.145	21.632	87.421
Urbanization rate	987	59.782	20.19	18.196	100
Economic freedom	987	60.972	9.257	33.5	89.4
Technological innovation	987	6721.28	14,623.132	7.01	149,212.62
Upgrading the industrial structure	987	1.888	0.874	0.327	13.170

is a descriptive statistical table of the variables used in this study. In terms of the two-way FDI indicator, IFDI and OFDI have larger standard deviations, indicating significant differences in the ability to attract and invest among countries. In terms of carbon intensity, the maximum value reaches 1.889, meaning 1.889 kg of carbon dioxide are emitted for each dollar of GDP produced, while the minimum value stands at 0.062, with a standard deviation of 0.235, highlighting significant disparities in carbon intensity among countries. There is a significant disparity in the standard deviation of per capita GDP, marked by a substantial difference between the highest and lowest figures, highlighting substantial variances in economic growth among nations participating in the Belt and Road Initiative. The threshold effect model could be employed to examine how two-way FDI impacts carbon emissions at various levels. Additionally, there are marked variations in economic freedom and the number of articles published in scientific and technological journals, with the data suggesting that countries with larger economies tend to have more published scientific articles and higher levels of economic freedom. The interpolation method was primarily applied to OFDI stock data for certain countries in earlier years, and to a lesser extent, to the number of scientific publications (SCI) and the industrial structure upgrading index (INDG) for a few country-years. The final balanced panel contains 987 observations after these interpolations.

In constructing the regression equation, this paper builds upon the STIRPAT model and extends it by incorporating variables like the coordinated development of two-way FDI, the level of openness, industrial structure, and the urbanization rate. Subsequently, the paper conducted an f-test with a p value less than 0.01, which rejected the assumption of using a mixed OLS model for regression.

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\omega }}_{\mathrm{i}}+{\mathsf{\upsilon }}_{\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}\mathrm{t}}$$

(4)

where $i\left(i=\mathrm{1,2}\dots ,47\right)$ represents the country, $t(t=\mathrm{1,2}\dots ,21)$ represents time, ${\omega }_i$ denotes individual fixed effects that remain constant over time, ${\upsilon }_t$ signifies time fixed effects that stay uniform across individuals, and ${\mu }_{it}$ stands for the residual term.

Table 4. Baseline regression results.

	(1)	(2)	(3)	(4)
	lnCO2	lnCO2	lnCO2	lnCO2
lnIOFDI	−0.2684 ***	−0.0786 ***	−0.0775 ***	−0.0298 ***
	(0.0130)	(0.0133)	(0.0128)	(0.0106)
lnPOP		0.1998 ***	0.1264 ***	0.4894 ***
		(0.0442)	(0.0446)	(0.0406)
lnAGDP		0.0358 *	0.1136 ***	0.1588 ***
		(0.0183)	(0.0193)	(0.0245)
lnENE		1.2674 ***	1.1817 ***	0.9715 ***
		(0.0384)	(0.0379)	(0.0327)
lnOPEN			0.1584 ***	0.0180
			(0.0306)	(0.0296)
lnIND			−0.0762	0.3220 ***
			(0.0733)	(0.0634)
lnURBAN			1.0124 ***	1.5631 ***
			(0.1160)	(0.1004)
Constant	0.1225 **	3.5104 ***	5.4461 ***	13.3401 ***
	(0.0532)	(0.2419)	(0.5087)	(0.5478)
Country FE	Yes	Yes	Yes	Yes
Time FE	No	No	No	Yes
Hausman	12.52 ***	27.51 ***	32.61 ***	207.33 ***
N	987	987	987	987
R2	0.313	0.747	0.773	0.856

Note: *** represents p < 0.01, ** represents p < 0.05, * represents p < 0.1, the same below.

delineates the sequential regression outcomes examining the two-way FDI coordination-carbon emissions nexus. column 1 incorporates solely the two-way FDI coordinated development variable. column 2 introduces three STIRPAT-model-specific control variables, while column 3 integrates supplementary controls. Acknowledging potential annual policy/event heterogeneity impacting carbon outputs, temporal fixed effects govern column 4, constituting a bidirectional fixed effect paradigm. The analysis reveals two-way FDI coordinated development consistently and significantly attenuates carbon emission intensity across all model specifications. Quantitatively, each 1% elevation in two-way FDI coordination magnitude induces a 0.0298% carbon intensity reduction. To illustrate the economic significance, a 1% increase in the two-way FDI coordination index is associated with a 0.030% decrease in carbon emission intensity. For a country with the sample’s average carbon intensity (0.356 kg per USD of GDP) and average GDP per capita (~$10,142), this implies a meaningful reduction in absolute CO₂ emissions, underscoring the potential environmental benefit of promoting balanced two-way investment flows. This empirically substantiates Hypothesis 1.

Since the start of the 21st century, propelled by economic globalization and the Belt and Road Initiative, countries along the routes have attracted substantial IFDI. The inflow of foreign capital has supplied ample funds for local industry development, introduced advanced machinery and management philosophies, facilitating the expansion of domestic production capacity and elevating economic development levels. Moreover, the growth of domestic industries has stimulated the advancement of foreign investment. Nations along the route have used OFDI to extend their industrial chains regionally, bringing back profits and technologies acquired abroad to further enlarge production scale and enhance economic growth. Presently, two-way FDI primarily channels into the clean sectors along the Belt and Road, where economic growth outpaces carbon emission increases, thereby lowering carbon emission intensity and promoting sustainable development. Simultaneously, as economic development improves, the national income of countries along the route steadily rises, boosting environmental consciousness among their populations. This, in turn, affects foreign investment inflow and outflow policies shaped by their governments, leading to the implementation of more stringent environmental regulations. For example, many countries have already established green development plans. Key carbon-intensive nations along the route—such as India, Indonesia, Vietnam, Russia, Iran, and Turkey—have funneled maximum resources into eco-friendly sectors, thereby counterbalancing the scale expansion effect on carbon emissions and decreasing carbon emission intensity for promoting sustainable development.

From a technological standpoint, regarding IFDI, nations along the route can leverage multiple regional trade agreements, including the Belt and Road Initiative, to more effectively attract IFDI, thereby acquiring advanced technologies from foreign firms, fostering related domestic industries, and enhancing production efficiency. To promote economic development, countries along the route can also use OFDI to invest in more technologically advanced countries, develop new technologies using local resources, and then transfer these technologies back to their own countries. Moreover, under this initiative, both investment and introduction are given equal weight. Especially for some developing countries, after receiving investment from developed countries, their technology has advanced, production efficiency has improved, and they have promoted their enterprises to go global, further enhancing the promoting effect on technological innovation.

Regarding structural impacts, for IFDI, the nations along the route have progressively strengthened environmental regulations, and these countries have started emphasizing environmental concerns when attracting IFDI, speeding up the introduction of service sectors to facilitate industrial structure transformation and upgrading. This process reduces carbon emission intensity while advancing the growth of relevant domestic OFDI industries. For OFDI, in order to support domestic industrial development, developed countries along the route can more easily transfer high-carbon industries to less developed countries under a series of regional trade agreements and provide development space for the clean industry; The developing countries can take advantage of the “reverse gradient” OFDI to learn new management experience in the developed countries, thereby improving their own production efficiency and allowing more funds to be invested in the previously neglected environmental protection industry. At present, developing nations along the route are actively investing in low-carbon sectors abroad to advance industrial restructuring and thus lower carbon intensity. The shift in industrial structure alongside enhanced production efficiency has further drawn IFDI, and the reciprocal dynamics of two-way FDI have driven domestic industrial transformation. Additionally, multiple clean energy projects have been co-financed by countries along the route, decreasing reliance on traditional high-carbon energy extraction industries, which helps reduce carbon intensity. Consequently, as coordinated development of two-way FDI improves, the industrial structure transitions toward low-carbon sectors, diminishing carbon intensity for promoting sustainable development.

The earlier section empirically examined the connection between the coordinated development of two-way FDI and carbon emissions through a baseline model. This study now employs multiple approaches to address the model’s endogeneity and verify its robustness. To tackle the endogeneity issue, the paper initially applies the GMM method, utilizing the lagged terms of variables as instrumental variables, following the approach established. GMM models can be divided into differential GMM models and system GMM models. Differential GMM models use fewer instrumental variables, are simpler, but are less efficient and prone to weak instrumental variable problems. System GMM models have more instrumental variables, are more efficient, can mitigate weak instrumental variable problems, but increase the likelihood of overfitting problems. In practical research, scholars often use both models in combination to determine the optimal approach. These two categories of models can be further classified into one-step and two-step variants according to differing regression techniques. The two-step approach is preferable in cases with heteroscedasticity and generally offers greater efficiency; however, it might underestimate the standard errors of estimated parameters when sample size is limited. Conversely, the one-step method, though less efficient, can still yield consistent estimates. To guarantee the robustness of the regression findings, this study employs all four methods concurrently. The specific regression models of the GMM are shown below, where model 5 is the differential GMM model and model 6 is the system GMM model.

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\alpha }}_1\mathrm{l}\mathrm{n}{{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i},\mathrm{t}-1}+{\mathsf{\alpha }}_2\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\alpha }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}}+{\mathsf{\theta }}_{\mathrm{t}}+{\mathsf{\kappa }}_{\mathrm{i}\mathrm{t}}$$

(5)

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_1\mathrm{l}\mathrm{n}{{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i},\mathrm{t}-1}+{\mathsf{\gamma }}_2\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\gamma }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\iota }}_{\mathrm{i}}+{\mathsf{o}}_{\mathrm{t}}+{\mathsf{\varsigma }}_{\mathrm{i}\mathrm{t}}$$

(6)

Table 5. Regression results of the GMM model.

	(5)	(6)	(7)	(8)
	One-Step Difference	One-Step System	Two-Step Difference	Two-Step System
	lnCO2	lnCO2	lnCO2	lnCO2
L.lnCO2	0.8004 ***	0.9215 ***	0.6923 ***	0.9348 ***
	(0.0621)	(0.0328)	(0.1598)	(0.0429)
lnIOFDI	0.1084 ***	0.0292 **	0.1058 **	0.0320 **
	(0.0302)	(0.0141)	(0.0533)	(0.0136)
Constant		0.5283 **		0.8289 *
		(0.2556)		(0.4281)
Control Variables	Yes	Yes	Yes	Yes
Country FE	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes
N	893	940	893	940
AR (1)	0.000	0.000	0.011	0.000
AR (2)	0.463	0.102	0.738	0.321
Sargan	0.116	0.133	0.117	0.122

Note: ***, **, and * indicate significant at 1%, 5%, and 10% levels, respectively.

presents the outcomes of the GMM regression models, where columns 5 and 7 utilize the difference GMM approach, and columns 6 and 8 apply the system GMM technique. Below the table, results for the AR and Sargan tests are displayed. Specifically, the AR(1) and AR(2) tests assess serial correlation; if the AR(2) test’s p-value exceeds 0.1, it suggests no significant serial correlation exists. The Sargan test evaluates potential over-identification issues, with the null hypothesis stating the model lacks such problems. p-value above 0.1 means the null cannot be rejected, indicating no over-identification. Consequently, after addressing endogeneity, regardless of the method or model, a rise in coordinated development of two-way FDI markedly reduces carbon emission intensity, with no notable changes in significance or direction for the two-way FDI coordination variable. Furthermore, all models passed both AR (2) and Sargan tests, confirming absence of serial correlation and over-identification problems.

To address potential endogeneity arising from reverse causality or omitted time-varying variables, we employ the Generalized Method of Moments (GMM) estimation. We estimate both difference and system GMM models, as specified in Equations (5) and (6). These models include the one-period lagged dependent variable, L.lnCO₂ (i.e., lnCO₂ at time t–1), as an explanatory variable to account for the persistence of carbon emissions over time. The coefficient of this lagged term captures the dynamic adjustment process.

For the robustness test, this paper first considers that the carbon emission intensity may gradually change over time. After controlling for individual fixed effects and other control variables, there is still a certain increasing or decreasing trend. If the potential time trend is overlooked, changes in carbon emissions might be mistakenly linked to the coordinated development of two-way FDI, causing a spurious regression issue. Hence, this study tests the robustness of controlling for time effects by incorporating a time trend variable into the model. The regression outcomes reveal that, across countries along the Belt and Road Initiative, enhancements in coordinated two-way FDI development continue to significantly decrease carbon emission intensity. This confirms that varying approaches to controlling time do not compromise the robustness of the findings.

This study further validates robustness by altering both the dependent variable and the structure of the explanatory variables. In

Table 6. Regression results of the robustness test.

	(9)	(10)	(11)	(12)	(13)
	Time Trend Term	Change the Explained Variable	Change the Explanatory Variable	1% Tail Contraction	Excluding Pandemic Years
	lnCO2	lnPCO2	lnCO2	lnCO2	lnCO2
lnIOFDI	−0.0339 ***	−0.0389 ***		−0.0238 **	−0.0305 ***
	(0.0108)	(0.0120)		(0.0105)	(0.0103)
lnIFDI			0.0582 ***
			(0.0153)
lnOFDI			0.0176 ***
			(0.0050)
lnIFDI*lnOFDI			0.0048 ***
			(0.0013)
Constant	12.2519 ***	12.1400 ***	12.7552 ***	13.5188 ***	13.3406 ***
	(0.5335)	(0.6230)	(0.5522)	(0.5456)	(0.5451)
Control Variables	Yes	Yes	Yes	Yes	Yes
Time FE	No	Yes	Yes	Yes	Yes
Time Trends	Yes	No	No	No	No
Country FE	Yes	Yes	Yes	Yes	Yes
N	987	987	987	987	940
R2	0.845	0.677	0.860	0.856	0.858

Note: ** and *** indicate significant at 5% and 1% levels.

, Column 2, following Xiang and Dai (2022) [32], per capita carbon emissions replace carbon emission intensity in the regression model. Regarding explanatory variables, prior research on two-way FDI coordination, including works by and Gong and Liu (2018) [36], often employed interaction terms rather than coupled coordination models to capture coordinated development. Consequently, utilizing the STIRPAT model, this study amalgamates factors like IFDI, OFDI, and their interplay terms to evaluate the impact of synchronized two-way FDI on the intensity of carbon emissions. To mitigate multicollinearity, both IFDI and OFDI are mean-centered, with the detailed equations provided below:

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\theta }}_0+{\mathsf{\theta }}_1\mathrm{l}\mathrm{n}\mathrm{I}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\theta }}_2\mathrm{l}\mathrm{n}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\theta }}_3\mathrm{l}\mathrm{n}\mathrm{I}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}\times \mathrm{l}\mathrm{n}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\theta }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\phi }}_{\mathrm{i}}+{\mathsf{\rho }}_{\mathrm{t}}+{\mathsf{\pi }}_{\mathrm{i}\mathrm{t}}$$

(7)

Empirical outcomes demonstrate that the two-way FDI coordinated development variable persists in markedly diminishing carbon emission intensity following alternate substitutions of both explained and explanatory variables, confirming this study’s findings demonstrate considerable resilience.

The results in Column (11) of Table 6, which decomposes the coordinated index into IFDI, OFDI, and their interaction term, provide further nuance. The positive and significant coefficients for both lnIFDI and lnOFDI suggest that, individually, the scale of each type of investment might be associated with higher carbon intensity in this context, possibly reflecting scale effects. However, the significantly positive coefficient for the interaction term (lnIFDI*lnOFDI) indicates that the interaction between them has a distinct, additional effect. This supports our core argument that the synergy between two-way FDI—captured by our coordination index (IOFDI)—is a critical factor. While the individual components might show certain pressures, their coordinated development creates a combined effect that ultimately leads to a reduction in carbon intensity, as consistently shown by the negative coefficient of the composite lnIOFDI variable in all other specifications.

To delve further, this part conducts a theoretical-based empirical study to investigate how the synchronized creation of two-way FDI affects carbon emissions. Initially, to examine Hypothesis 2, a novel equation is formulated following the three-step approach.

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\omega }}_{\mathrm{i}}+{\mathsf{\upsilon }}_{\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}\mathrm{t}}$$

(8)

$$\ln {\mathrm{S}\mathrm{C}\mathrm{I}}_{\mathrm{i}\mathrm{t}}={\mathsf{\delta }}_0+{\mathsf{\delta }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\delta }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\zeta }}_{\mathrm{i}}+{\mathsf{\eta }}_{\mathrm{t}}+{\mathsf{\psi }}_{\mathrm{i}\mathrm{t}}$$

(9)

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}=\mathsf{\rho }+{\mathsf{\rho }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\rho }}_2{\mathrm{l}\mathrm{n}\mathrm{S}\mathrm{C}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\rho }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\lambda }}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{t}}+{\mathsf{\phi }}_{\mathrm{i}\mathrm{t}}$$

(10)

Technological innovation, however, has notably driven up carbon emission intensity, exhibiting a rebound effect (e.g., Sorrell, 2009) [37]. This could stem from the fact that, despite numerous environmental protection projects established jointly by Belt and Road countries, the development and investment in new technologies require time, making immediate carbon control through innovation challenging. Moreover, as most selected nations along the route are developing countries with limited innovation capacity, their independent innovation often prioritizes scaling up production for economic growth rather than focusing on energy efficiency and emission cuts, thereby causing carbon intensity to rise.

Table 7. Regression results of the mechanism study.

	(14)	(15)	(16)	(17)	(18)	(19)
	lnCO2	lnSCI	lnCO2	lnINDG	lnCO2	lnCO2
lnIOFDI	−0.0298 ***	0.1110 ***	−0.0337 ***	0.0453 ***	−0.0220 **	−0.0352 ***
	(0.0106)	(0.0262)	(0.0106)	(0.0061)	(0.0108)	(0.0105)
lnSCI			0.0352 ***
			(0.0133)
lnINDG					0.1737 ***
					(0.0570)
lnECOF						0.1858 **
						(0.0756)
lnECOF*lnIOFDI						0.1628 ***
						(0.0351)
Constant	13.3401 ***	8.9043 ***	13.0269 ***	2.2867 ***	13.7373 ***	12.1455 ***
	(0.5478)	(1.3569)	(0.5587)	(0.3168)	(0.5606)	(0.6075)
Control Variables	Yes	Yes	Yes	Yes	Yes	Yes
Country FE	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes
N	987	987	987	987	987	987
R2	0.856	0.785	0.857	0.796	0.857	0.859

Note: ** and *** indicate significant at 5% and 1% levels.

,

Table 8. Results of Sobel tests.

Variable Name	Test Type	Coef	Std	Z	P > |Z|
Technological innovation	Sobel Test	0.004	0.002	2.242	0.025
	Aroian Test	0.004	0.002	2.198	0.028
	Goodman Test	0.004	0.002	2.288	0.022
Upgrading the industrial structure	Sobel Test	−0.008	0.003	−2.820	0.005
	Aroian Test	−0.008	0.003	−2.798	0.005
	Goodman Test	−0.008	0.003	−2.842	0.004

and

Table 9. Bootstrap Test Results.

Variable Names	Effect Type	Coef	Std	Z	P > |Z|	95% Conf. Interval
						Lower	Upper
Technological innovation	Indirect effect	0.004	0.002	1.98	0.047	0.000	0.008
	Direct effects	−0.034	0.011	−3.11	0.002	−0.055	−0.012
Upgrading the industrial structure	Indirect effects	−0.008	0.004	−2.14	0.032	−0.015	−0.001
	Direct effects	−0.022	0.010	−2.21	0.027	−0.041	−0.002

primarily examine the technological impacts of two-way FDI on carbon emissions. This section focuses specifically on exploring the structural effects that two-way FDI exerts on carbon emissions.

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\omega }}_{\mathrm{i}}+{\mathsf{\upsilon }}_{\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}\mathrm{t}}$$

(11)

$$\ln {\mathrm{I}\mathrm{N}\mathrm{D}\mathrm{G}}_{\mathrm{i}\mathrm{t}}={\mathsf{\sigma }}_0+{\mathsf{\sigma }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\sigma }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\theta }}_{\mathrm{i}}+{\mathsf{\kappa }}_{\mathrm{t}}+{\mathsf{\varpi }}_{\mathrm{i}\mathrm{t}}$$

(12)

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\phi }}_0+{\mathsf{\phi }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\phi }}_2{\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{N}\mathrm{D}\mathrm{G}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\phi }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\delta }}_{\mathrm{i}}+{\mathsf{\tau }}_{\mathrm{t}}+{\mathsf{\gamma }}_{\mathrm{i}\mathrm{t}}$$

(13)

The regression outcomes demonstrate that advancing the coordinated development of two-way FDI markedly accelerates industrial upgrading, specifically shifting the structure from secondary to tertiary sectors. Concurrently, this structural transformation plays a pivotal role in lowering carbon emission intensity and promoting sustainable development. Mediation analysis further confirms that industrial upgrading serves as a significant intermediary factor. These findings imply that in Belt and Road Initiative countries, coordinated two-way FDI development fosters industrial restructuring, which in turn mitigates carbon intensity. Hypothesis 3 is thus validated, showing that the structural influence of two-way FDI on carbon emissions is substantial—where the direct pathway contributes 73.6% and the indirect route through industrial upgrading accounts for 26.4%.

The mediation analysis results indicate that industrial structure upgrading is a significant channel. To quantify the contribution of each pathway, we calculated the proportion of the total effect mediated by the indirect effect. The direct effect of lnIOFDI on lnCO₂ from Equation (13) is −0.022 (See Table 7, Column 18), and the indirect effect (via lnINDG) is −0.008 (derived from the product of the coefficient of lnIOFDI on lnINDG in Equation (12) (0.0453, Table 7, Column 17) and the coefficient of lnINDG on lnCO₂ in Equation (13) (−0.1737, Table 7, Column 18), i.e., 0.0453 × (−0.1737) ≈ −0.0079). The total effect is approximately −0.030 (−0.022 + −0.008). Consequently, the direct effect accounts for |(−0.022)/(−0.030)| ≈ 73.3%, and the indirect effect accounts for |(−0.008)/(−0.030)| ≈ 26.7% of the total effect. The figures of 73.6% and 26.4% mentioned in the text are from a similar calculation based on the bootstrap results in Table 9, which provides a more robust estimation of the effects and their confidence intervals.

In countries engaged in the Belt and Road Initiative, safeguarding the environment has emerged as a more prominent aspect of industrial strategy. Across various nations, there is been an increase in the share of output from tertiary industries compared to the secondary sector, coupled with the initiation of multiple joint ventures in the service sector. For example, Saudi Arabia, Russia, Indonesia, Cambodia, etc., have all made efforts in the cultural and tourism industry. They are working with China and their neighboring countries on projects involving service platform construction, digital tourism, creative design and many other fields. In addition, many enterprises are investing in basic service industries such as logistics, digital technology and education along the route. These investment movements have enhanced the caliber of regional economic advancement while significantly diversifying the industrial frameworks within countries along the route. This transformation has supported a transition away from high-emission sectors toward lower-carbon industries, consequently driving down carbon emission intensity.

The theoretical framework suggests that economic freedom could influence the linkage between two-way FDI and carbon emissions for promoting sustainable development. To evaluate Hypothesis 4, this study incorporates an interaction term between economic freedom and the coordinated development of two-way FDI into the baseline regression model, aiming to examine the moderating role of economic freedom. This yields the following equation. In the actual regression process, to reduce the impact of multi-collinearity on the model, the independent and moderating variables were decentralized, and then the interaction term was calculated and added to the model.

$$\ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\xi }}_0+{\mathsf{\xi }}_1\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\xi }}_2\mathrm{l}\mathrm{n}{\mathrm{E}\mathrm{C}\mathrm{O}\mathrm{F}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\xi }}_3\mathrm{l}\mathrm{n}\mathrm{E}{\mathrm{C}\mathrm{O}\mathrm{F}}_{\mathrm{i}\mathrm{t}}\times \mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\xi }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\alpha }}_{\mathrm{i}}+{\mathsf{\sigma }}_{\mathrm{t}}+{\mathsf{\delta }}_{\mathrm{i}\mathrm{t}}$$

(14)

For this moderating effect, under high economic freedom, high government integrity and moderate government low-carbon spending can help guide IFDI into low-carbon industries and boost domestic enterprises to make low-carbon investments abroad or transfer some high-carbon emission industries, thereby improving the local industrial structure. Such an approach will amplify the impact of synchronized two-way FDI development in diminishing carbon intensity. Concurrently, nations with greater economic liberty enjoy increased levels of free trade and investment, easing the challenges in technology exchange and capital movement, thus amplifying the technology overflow and the reverse effect of two-way FDI, along with the introduction of novel technologies. In addition, high economic freedom can attract larger IFDI and expand foreign investment, thus providing sufficient funds for technological innovation both at home and abroad.

Considering policy fluidity during economic advancement, coordinated two-way FDI development’s carbon emission impact may manifest heterogeneity. To holistically capture this relationship, this section adopts per capita GDP—following (Raza et al., 2025; Farabi et al., 2024) [38,39]—as the economic development proxy and threshold variable, examining differential effects across developmental phases. Model construction primarily employed the Bootstrap self-sampling method for threshold quantity determination. Regression outcomes revealed a dual-threshold phenomenon, establishing the subsequent equation:

$$\begin{array}{cc}\hfill \ln {{\mathrm{C}\mathrm{O}}_2}_{\mathrm{i}\mathrm{t}}={\mathsf{\lambda }}_0+{\mathsf{\lambda }}_1& \mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}\left(\mathrm{l}\mathrm{n}{\mathrm{A}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{i}\mathrm{t}}\le {\mathsf{\rho }}_{21}\right)\hfill \\ & +{\mathsf{\lambda }}_2\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}\left({\mathsf{\rho }}_{21}<\mathrm{l}\mathrm{n}{\mathrm{A}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{i}\mathrm{t}}\le {\mathsf{\rho }}_{22}\right)\hfill \\ & +{\mathsf{\lambda }}_3\mathrm{l}\mathrm{n}\mathrm{I}\mathrm{O}{\mathrm{F}\mathrm{D}\mathrm{I}}_{\mathrm{i}\mathrm{t}}\left(\mathrm{l}\mathrm{n}{\mathrm{A}\mathrm{G}\mathrm{D}\mathrm{P}}_{\mathrm{i}\mathrm{t}}>{\mathsf{\rho }}_{22}\right)+{\mathsf{\lambda }}_{\mathrm{n}}\sum \mathrm{l}\mathrm{n}\mathrm{c}{\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\kappa }}_{\mathrm{i}}+{\mathsf{\psi }}_{\mathrm{t}}\hfill \\ & +{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}\hfill \end{array}$$

(15)

Based on the regression outcomes (

Table 10. Results of the double threshold test.

Threshold Number	F-Stat	Prob	Critical Value
			10%	5%	1%
Single threshold	106.85 **	0.0115	58.8887	69.6487	109.5193
Double thresholds	71.42 **	0.0170	46.0106	54.5835	79.6794

Note: ** indicates significant at 5% level.

and

Table 11. Threshold regression and heterogeneity study regression results.

	(20)	(21)	(22)	(23)	(24)
	Threshold Regression	Central and Eastern Europe and CIS	Central and South Asia	East Asia and Southeast Asia	West Asia and North Africa
	lnCO2	lnCO2	lnCO2	lnCO2	lnCO2
lnIOFDI		0.0740 ***	0.0232 **	0.0854 ***	−0.0456
		(0.0140)	(0.0109)	(0.0295)	(0.0369)
lnIOFDI	0.1250 ***
$(lnAGDP\le 5.8979)$	(0.0189)
lnIOFDI	0.0122
$(5.8979<lnAGDP\le 8.4861)$	(0.0104)
lnIOFDI	0.0241 **
$(lnAGDP>8.4861)$	(0.0104)
Constant	12.0385 ***	12.0230 ***	4.7612 ***	35.6433 ***	6.2383 ***
	(0.5244)	(1.2290)	(0.5881)	(3.9632)	(1.4691)
Control Variables	Yes	Yes	Yes	Yes	Yes
Country FE	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes
N	987	420	147	189	231
R2	0.877	0.956	0.959	0.777	0.811

Note: *** and ** indicate significant at 1% and 5% levels.

), the presence of a pronounced double threshold effect tied to economic development levels is evident. In the Belt and Road region, if a nation’s development dips beneath the initial level, improved synchronization of two-way FDI significantly reduces the intensity of carbon emissions. Should the developmental stage fall between the initial and secondary thresholds, enhanced synchronization of two-way FDI results in a slight increase in carbon intensity, albeit without statistical robustness. Once the second threshold is surpassed, advancing coordinated two-way FDI once again results in a significant decline in carbon intensity. This pattern closely mirrors the staged progression of environmental degradation depicted by the Environmental Kuznets Curve (EKC).

Threshold regression outcomes reveal economic development levels exert a significant moderating influence on how coordinated two-way FDI development affects carbon emissions, though this impact may manifest divergences contingent upon geographic dispersal and resource endowment disparities. This study segments Belt and Road nations into four divisions—East Asia and Southeast Asia; Central Asia and South Asia; West Asia and North Africa; plus Central/Eastern Europe and the Commonwealth of Independent States—based on variances in physical location, historical-cultural legacies, and economic configurations, subsequently performing grouped regressions to scrutinize regional heterogeneity concerning coordinated two-way FDI development’s impact upon carbon emission intensity. Results indicate that within East Asia/Southeast Asia and Central Asia/South Asia Belt and Road nations, enhanced coordinated two-way FDI development demonstrably diminishes carbon intensity for promoting sustainable development. Conversely, in West Asia and North Africa, coordinated two-way FDI development reduces carbon intensity yet fails to achieve statistical significance.

The regional heterogeneity, visualized in Figure 2, may be attributed to differences in initial industrial structure, stringency of environmental regulations, and primary types of FDI received (e.g., resource-seeking vs. technology-seeking).

5. Conclusions and Policy Recommendations

Theoretical analyses establish that coordinated development of two-way FDI influences carbon emission intensity via scale, technological, and structural pathways; the directional impact of each pathway remains indeterminate. These pathways collectively contribute to sustainable development by harmonizing economic progress with environmental protection. Consequently, empirical determination becomes imperative regarding the current effect of this two-way FDI coordination on carbon intensity within Belt and Road Initiative nations. Most route-adherent states utilize IFDI introducing low-carbon sectors while employing OFDI potentially relocating high-carbon operations to other regions, a complexity whose net global emission effect requires further study but which contributes to reduced domestic carbon intensity for the investing country. This dual strategy exemplifies how sustainable development can be achieved through strategic investment coordination. Concurrently, supplementary factors, facilitated by two-way FDI, propel the digital economy’s ascent across these nations, rendering industrial structural recalibration markedly more expedient; furthermore, the permeation of green finance concepts assists in channeling capital flows—two-way FDI included—towards low-carbon industries.

Significant mediating roles emerge for technological innovation and industrial structure upgrading. While technological advancements currently amplify carbon intensity, long-term investments in green innovation are essential for sustainable development. Two-way FDI coordination propels technological advancement, yet paradoxically amplifies carbon intensity through this mechanism; conversely, such coordination diminishes carbon intensity by catalyzing industrial structure advancement. Concerning technological innovation’s mediation, a temporal lag exists before two-way FDI-driven novel technologies manifest in production; moreover, innovation within most developing economies frequently prioritizes production efficiency gains, neglecting ecological safeguards. Presently, therefore, technological innovation proves antagonistic towards carbon mitigation. Simultaneously, a rebound phenomenon associated with innovation elevates energy extraction efficiency and productive scale, consequently escalating energy usage and product consumption, thereby elevating carbon intensity.

The preceding conclusions introduce an alternative lens through which to interpret the interplay between capital mobility and environmental stewardship across Belt and Road Initiative regions. Policymakers must prioritize ecological modernization to align industrial growth with sustainable development objectives. Drawing from this, the paper outlines several recommendations aimed at enhancing carbon emission control among participating nations. First, authorities should further expand two-way FDI coordination to reinforce sustainable development goals, ensuring investments align with low-carbon and green growth strategies. Second, it is essential to accelerate industrial structure refinement and ecological technological reform to embed sustainable development principles into economic planning. Third, it is important to enhance economic autonomy and institutional quality to create an enabling environment for sustainable development. Fourth, it is necessary to raise ecological consciousness and implement context-specific environmental policies to mitigate carbon emissions and promote sustainable practices. Fifthly, the study leverages higher economic development stages to deepen two-way FDI coordination, ensuring sustained reductions in carbon intensity and progress toward sustainable development. Ultimately, the Belt and Road Initiative must champion sustainable development by integrating environmental, economic, and social dimensions into its investment framework, setting a global benchmark for inclusive and green growth.

Furthermore, this study focuses on carbon emission intensity within the sample BRI countries. While we note the strategy of relocating high-carbon operations via OFDI, our analysis does not capture the potential transboundary shift in emissions (i.e., the ‘pollution haven’ effect on a global scale). Future research could employ methods like multi-regional input-output analysis to assess the net global carbon impact of two-way FDI coordination under the BRI, accounting for such geographical redistribution of emissions.

This study has several limitations that should be acknowledged. First, while interpolation was used for missing data, it remains an approximation that may introduce measurement error. Second, although we control for major factors, potential omitted variable bias cannot be entirely ruled out. Third, measuring complex concepts like ‘technological innovation’ solely by publication count has inherent limitations, as it may not fully capture green technology adoption or process innovations. Future research could employ more granular data, alternative proxies for key variables (e.g., patent data for innovation), and explore non-linear relationships or additional moderating factors to further unravel the complex FDI-environment nexus.