Sustainable Governance on the Belt and Road Initiative Toward a Carbon-Zero, Regional, Eco-Friendly Logistics Hub: A Difference-In-Differences Perspective

Abstract

The Belt and Road Initiative (BRI) proposed by China in in 2013 prioritizes environmental sustainability and regional economic development from a global perspective. Although the BRI has achieved considerable economic progress in many cities and regions, research on its environmental impacts is still insufficient, with limited attention paid to domestic urban areas in particular. Existing studies have focused primarily on carbon emissions, ignoring the broader environmental impacts of industrial emissions, such as those from smart transportation. To address this gap, this study adopts four major pollutant emissions—carbon dioxide (CO₂), industrial particulate matter, industrial sulfur dioxide (SO₂), and industrial wastewater emissions—as indicators to assess pollution levels in urban environments. Adopting panel data from 281 Chinese cities from 2003 to 2021, this study employs the difference-in-differences (DID) method to estimate the effect of the BRI on urban environmental pollution. This study is based on the following hypotheses: Hypothesis 1. BRI implementation has reduced urban pollution emissions. Hypothesis 2. Advancements in science and technology will drive the implementation of the BRI. Hypothesis 3. A proactive government response can significantly reduce urban environmental pollution. The main findings of this study are as follows. First, BRI implementation significantly reduces urban environmental pollution by 1.05%. Second, the policy effects of the BRI are more pronounced in the eastern and western regions and in larger cities, implying that geopolitical- and market-oriented strategies are important for regional performance. Third, scientific and technological progress positively affects pollution reduction in urban environments. Fourth, the BRI contributes to strengthening government intervention, which subsequently improves sustainable governance, reduces urban environmental pollution, and promotes regional economic cooperation. Our findings will serve as a crucial reference for future policymaking endeavors toward eco-friendly logistics cooperation in the region.

1. Introduction

In 2013, China put forward the Belt and Road Initiative (BRI) with the aim to build a collaborative trade network with countries and regions around the world and usher in a new phase of inclusive globalization [1,2,3]. Significant progress has been made in BRI development. China’s trade in goods with countries along the Belt and Road has surged from USD 1.04 trillion in 2013 to USD 2.9 trillion in 2022 [4]; by June 2023, China had signed more than 200 cooperation documents with more than 140 countries to promote the BRI. However, economic development is not the only requirement of the BRI; environmental protection and green development are fundamental requirements for promoting sustainable mutual prosperity. It is also an important way to achieve green development and promote high-quality economic development. Thus, economic development should be closely connected to environmental protection [5]. Since the BRI was proposed, the Chinese government has attached great importance to the construction of the Green Belt and Road. According to the Belt and Road Eco-Environmental Protection Co-operation Regulation issued by the former Ministry of Environmental Protection, it is necessary to focus on the construction of the Belt and Road eco-civilization to actively promote the transformation of cities along the routes toward zero-carbon eco-cities.

As logistics and transport are important components of urban development, they also play key roles in environmentally friendly urban ecosystems. In 2015, China launched the Vision and Action under the BRI, which identifies 18 provinces and 26 nodal cities with comparative advantages along the Belt and Road as the key provinces for the construction of the Belt and Road. These are each endowed with a unique geographic location and well-developed transport infrastructure and bear the burden of promoting the development of the neighboring economy and facilitating the enhanced Green BRI. Every city can be considered a green logistics hub; thus, urban pollution reduction is essential for building a carbon-neutral, environmentally friendly collaboration network system. In alignment with its objectives, BRI projects have prioritized the development of infrastructure such as highways, railways, ports, airports, and gas pipelines to foster trade and investment among the participating countries [6]. According to data from the Ministry of Commerce of China, in 2024, Chinese enterprises signed new contracts worth CNY 495.15 billion in BRI countries, reflecting an annual increase of 24.1% (equivalent to USD 69.71 billion, an increase of 20.2%). During the same period, completed turnover reached CNY 341.02 billion, marking a growth of 12.9% (equivalent to USD 48.01 billion, an increase of 9.5% [7]). As important logistics hubs for future global trade, cities along the BRI are steadily moving toward the goal of promoting China’s green development and are leading countries along the route toward high-quality and sustainable development. The BRI aims to occupy an important position in the global trading system by expanding trade, which cannot be achieved without logistics and transportation. Nevertheless, the expansion of logistics and transportation will inevitably create environmental problems. However, due to the newly emerging global norm of environmental concerns, most BRI projects should be developed in environmentally friendly ways. Is it feasible for all these BRI projects to be more environmentally friendly? If not, how can we improve the governance of the ecosystem of the BRI? In order to answer these two research questions, this paper aims to evaluate the sustainable performance of BRI projects from an environmental efficiency perspective. For this research objective, it is worth exploring whether the implementation of the BRI impacts the environment of cities along the route.

Many countries along the BRI’s routes are primarily developing or emerging economies, often characterized by rapid economic growth and high carbon intensity [8]. Economic activities in these regions consume large amounts of energy and are expected to increase further, putting significant pressure on the achievement of Sustainable Development Goals (SDGs) related to climate change mitigation [9]. Consequently, resource and environmental issues in the BRI regions have garnered significant attention from the international community. However, numerous scholars have focused on examining the potential negative impacts of the BRI on countries along its route. For instance, some have argued that financial development under the BRI has contributed significantly to environmental degradation in the participating regions of Central Asia, the Middle East, and North Africa [10]. One stream of literature has focused on environmental changes in regions along the BRI’s routes [11,12,13]. What is the domestic side of the BRI? Few studies on the BRI have focused on environmental pollution in cities along China’s domestic routes. China aspires to transition from a participating country to a leader in low-carbon development, aiming to establish a positive benchmark for other countries to reduce emissions. The Chinese government is committed to peaking its carbon emissions by 2030 and achieving carbon neutrality by 2060 [14]. Therefore, it is crucial to investigate whether the BRI has had a stronger positive impact on environmental pollution in Chinese cities along domestic routes. Can the BRI help China advance its green economy and fulfill international commitments? This research question warrants thorough examination as it has significant implications for both China’s domestic development and its global leadership in sustainable practices. In regression analysis, the inclusion of control variables is critical for addressing the potential omitted variable bias. By incorporating control variables, this study ensures that the estimated effects of the BRI reflect their true impacts and are not confounded by external factors. This study employs a range of valid control variables to improve the reliability of the double-difference method (DID), allowing for more robust and credible results. In addition, this study considers the level of science and technology as a moderating variable and empirically explores its key role in the BRI’s impact on urban environments. Moreover, scientific and technological development can improve energy efficiency, promote the adoption of green technologies, and reduce pollution. By including this variable, this study examines how the level of science and technology affects the BRI and whether it plays a positive role in reducing urban environmental pollution. The degree of government intervention is introduced as a mediating variable to explore how proactive governance contributes to the implementation of the BRI. The Chinese government has strong policy-driven capacity and plays a pivotal role in promoting green development through environmental regulations, public investment, and policy implementation. However, can strong government intervention capacity play a positive role in reducing urban environmental pollution? By measuring the mediating effect of government intervention, this study elucidates how the level of government intervention affects the BRI and whether it positively affects urban environmental pollution reduction.

As mentioned above, the main goal of this study is to answer the following core research questions: Does the Belt and Road Initiative (BRI) contribute to environmental improvements in Chinese cities along its route? If not, how can the governance be improved in its ecological systems? As Chinese government has emphasized, ecological systems are very complicated and complex, and thus the governance system and its mechanism should be evaluated with much more reliable methods. Therefore, the main goal of this research is to evaluate the environmental performance of the BRI by focusing on cities that function as key logistics hubs, with an emphasis on their transition toward low-carbon, eco-friendly development. To find out this transition toward the optimal path control by utilizing the difference-in-differences (DID) methodology and incorporating comprehensive pollution indicators, this study aims to empirically verify the BRI’s role in fostering green urban development. Furthermore, it identifies the mechanisms driving these changes, such as proactive government governance and technological innovations, which play critical roles in reducing urban pollution and improving policy outcomes. Ultimately, this study seeks to provide actionable insights for policymakers to enhance the BRI’s implementation, promote sustainable logistics hubs, and share effective practices with other countries to support global green development efforts.

The remainder of this paper is organized as follows. Section 2 presents the literature review and hypotheses; Section 3 describes the methodology, sample selection, and measurement of variables; Section 4 presents the econometric model construction, analysis of empirical results, analysis of impact mechanisms, and heterogeneity; and Section 5 presents the conclusions and policy implications.

2. Literature Review and Hypotheses

2.1. Impact of the BRI on Urban Environmental Pollution

Since its introduction, the BRI has promoted the economic development of countries along its routes, and its environmental impact has attracted considerable attention. Many scholars have selected carbon emissions as a key indicator to assess the impact of the BRI on urban environments. Fan et al. [15] used the production theory of decomposition analysis to decompose the total carbon dioxide emission changes in the BRI countries from 2000 to 2014 into the contributions of seven drivers; the empirical results revealed that China has a clear advantage in terms of carbon emission reduction technology effect and that carbon dioxide emission reduction technology efficiency and carbon emissions will decrease as carbon reduction technologies improve and energy efficiency is optimized. Differences in energy consumption inequality across BRI regions have resulted in varying environmental impacts. Nevertheless, these disparities have had a positive effect on mitigating environmental degradation [10]. Zhen et al. [16] built a model based on the panel data of 290 cities during 2003–2016 and analyzed heterogeneity by dividing the cities into two groups based on their level of economic development. The results showed that a low level of economic development exacerbates urban environmental pollution, whereas a high level of economic development improves urban environmental quality. Saud [12] examined the impact of financial development, foreign direct investment (FDI), economic growth, electricity consumption, and trade openness on environmental quality in 59 BRI countries during 1980–2016. This study found that increased financial development, FDI, and trade openness improve environmental quality. Liu [17] used quantitative analysis with breakpoint regression and found that the construction of the BRI improved the total green factor productivity in key provinces along China’s routes.

Some scholars have used other indicators to measure the BRI’s impact on urban environments. Sueyoshi [18] was the first to empirically analyze PM2.5 and PM10 as measures; the empirical results revealed that regional variability is an important influencing factor and that different regions are affected to different degrees. Cui [19] selected carbon dioxide, sulfur dioxide, industrial dust, and industrial wastewater emissions as environmental indicators, using the panel data of 32 BRI countries during 2006–2018, based on a systematic GMM model, to analyze empirical evidence. The results showed that countries with different levels of economic development have different correlations between economic growth and environmental pollution; differences in environmental pollution indicators were also identified. Some scholars have argued that the BRI may have a negative impact on the urban environment; when China engages in business and trade with countries or regions along the BRI, it may attract enterprises from developed countries to invest in or establish factories in Chinese cities along the route. These enterprises may transfer polluting industries to these cities, resulting in increased energy and resource consumption, higher pollution emissions, and, ultimately, a deterioration of environmental quality in China. This aligns with the “Pollution Haven Hypothesis” [20], also known as the “pollution shelter hypothesis” or the “industrial location relocation hypothesis”. It mainly refers to the tendency of enterprises in pollution-intensive industries to establish themselves in countries or regions with relatively low environmental standards. However, some scholars have opined that the BRI is a way for China to expand its foreign investment and transfer some of its highly polluting and energy-consuming domestic industries abroad, thus reducing pollution emissions in China’s domestic cities [21]. Therefore, is the “Pollution Paradise Hypothesis” applicable to China? This study attempts to refute this hypothesis through a series of environmental pollution indicators and rigorous empirical analyses. The Environmental Kuznets Curve (EKC) hypothesis is one of the most widely cited theoretical frameworks for studying the relationship between economic growth and environmental pollution. This hypothesis suggests that environmental pollution is relatively low when a country is in the early stages of economic development. As the economy grows, pollution increases to a certain level, after which it begins to decline as the economy declines further. Has the economic progress brought about by the BRI had a positive impact on the environment of the cities along the route? Combining these two theories, we propose the following hypothesis:

Hypothesis 1.

BRI implementation has reduced urban pollution emissions.

2.2. Scientific and Technological Development and the Urban Environment

Since the launch of the BRI, government departments in Chinese cities along the route have actively responded to national calls by integrating scientific and technological developments into the global value chain. They exchange markets for technology and successfully achieve exogenous technological absorption. Yang [22] empirically examined the global and heterogeneous effects of urban innovation on environmental pollution using panel data from 271 prefecture-level cities in China from 2005 to 2016. By constructing a geographic neighborhood matrix and an economic geography matrix and applying a dynamic spatial Durbin model, the empirical results demonstrated that an increase in the level of green technological innovation reduces environmental pollution. The BRI has reshaped the existing patterns of regional development inequality, realized the development of “logistics hubs” for inland cities, and significantly enhanced trade promotion effects in western cities [23,24]. It has created a global value chain “conjugate circulation” centered around China, which integrates the eastern and western regions. Through industrial transfer, structural adjustment, and the transformation of economic development models, the BRI has fostered a new equilibrium in development across the eastern, central, and western regions [25]. Conversely, the advancement of science and technology promotes the development of green logistics; cities along the Belt and Road, as logistics hubs, can be effectively transformed into zero-carbon regional eco-logistics centers with the promotion of science and technology. Jayaraman et al. [26] demonstrate that implementing green logistics operations can enhance market share, foster buyer loyalty, and contribute to firms’ financial growth. Similarly, Fang [27] argued that green logistics facilitate the rapid agglomeration of economic factors in the Belt and Road region, enabling economies of scale and promoting both the quantity and quality of regional economic growth, driving them onto a dual upward trajectory. Furthermore, Mahasin [28] emphasized the critical role of green logistics in supporting the environmental and economic sustainability of the BRI countries.

Notably, countries and regions along the BRI are diverse and are often characterized by conflicts and contradictions [29]. The degree of economic development of cities along the route within China also varies significantly; therefore, to determine whether the level of scientific and technological development can promote BRI implementation, this study tests the following hypothesis using empirical evidence:

Hypothesis 2.

Advancements in science and technology will drive the implementation of the BRI.

2.3. Government Governance and the Urban Environment

The BRI, as a new engine for China’s external development, has received close attention from the State and cities along the route have been given more “policy care”. The Chinese government has unparalleled power that favors the promotion of the BRI, especially within China. Gulinar [30] used the data of China’s direct investment in the BRI countries during 2005–2018 to study the relationship between the governance level of the host government and various sub-indicators and China’s location choice for outward foreign direct investment (OFDI); the results revealed that the host country’s government efficiency, regime stability, and other indicators significantly attracted China’s direct investment. Since Xi Jinping’s call in 2012 to build a “green, healthy, smart, and peaceful” Silk Road, China’s OFDI has been dedicated to low-carbon industry development and the promotion of green technologies, as well as increasing China’s participation in global environmental governance. Nearly two-third of the OFDI is invested in infrastructure and energy projects [31], suggesting that China’s strong regime structure can help local governments better administer governance and effectively improve urban environmental issues. Some researchers have found that institutional governance is positively associated with economic growth while improving environmental quality [32,33]. Environmental quality improves when governmental institutions are strengthened to enforce environmental standards and regulations [34]. However, governance is a double-edged sword; corruption, favoritism, and excessive bureaucracy increase the cost of conducting business and discourage investors [35]. Therefore, as the initiator of the BRI, does China’s governance level play a positive role in cities along domestic routes? Based on this, we formulate Hypothesis 3:

Hypothesis 3.

A proactive government response can significantly reduce urban environmental pollution.

3. Methodology and Data

3.1. Model Setting

The DID model is a widely used tool for evaluating policy effects, allowing for an effective assessment of the net impact of a policy on two groups before and after its implementation [36]. First, the DID model accounts for unobservable factors that may or may not be influential, thus facilitating policy evaluation [37]. Second, the DID model enables quasi-natural experiments by adjusting for preexisting conditions, which is particularly valuable because it is often difficult or impossible to ensure the random implementation of a policy across different locations [38]. Therefore, this study employs the DID model to examine the impact of the BRI on urban environmental pollution.

To validate the policy assessment of the BRI on urban environmental pollution in China, this study adopts an approach that treats the BRI as a quasi-natural experiment, designating 2013 as the initiative implementation point. Prefecture-level and nodal cities within the 18 provinces directly affected by the BRI are identified as the experimental group, comprising 143 cities. Conversely, the 138 cities that were unaffected by the initiative constitute the control group (refer to Figure 1).

This study proceeds by constructing a double-difference model as follows:

$${pollution}_{it}={\beta }_0+{\beta }_1{policy}_{it}+{\beta }_2{X}_{it}+{\mu }_i+{\gamma }_t+{\epsilon }_{it}$$

(1)

As seen in Equation (1),where ${pollution}_{it}$ is an explanatory variable denoting the level of environmental pollution emissions in city i in year t, ${policy}_{it}$ is an initiative dummy variable denoting the state of initiative implementation in city i in year t, and $X_{it}$ is the control variable, ${\beta }_0$,${\beta }_1$,${\beta }_2$ is the coefficient set to be estimated, reflecting the treatment effect of the initiative, ${\mu }_i$ denotes a city fixed effect, ${\gamma }_t$ denotes a time fixed effect, i denotes different cities, t denotes a year, and ${\epsilon }_{it}$ is an error term.

For initiative implementation (${policy}_{it}$), this study considers the BRI as a quasi-natural experiment. The initiative treatment effect is captured by the interaction term between the city type dummy variable and the time dummy variable for initiative implementation (${treat}_i$,${period}_t$). Specifically, this study assigns a value of 1 to cities belonging to the experimental group and occurring after the initiation of the initiative, while assigning a value of 0 to all other cities. The time dummy variable ${period}_t$ is set to 0 for periods preceding the initiative implementation and 1 for periods thereafter.

The main factors affecting the level of urban pollution emissions can be categorized as internal and external [39,40]. When a country attains a certain level of economic growth, public awareness of environmental protection tends to increase [41]. Following Liu [3], we incorporate the level of economic development as one of the control variables. Population growth is also a manifestation of urban development, and the demand for energy, transport, and public services from population growth affects the urban environment [42]. Therefore, population is chosen as one of the control variables and FDI is widely recognized as an important variable affecting CO₂ emissions. According to the Pollution Paradise Hypothesis, FDI transfers highly polluting industries, leading to environmental degradation in the host country [20]. Zhao et al. [43] recently argued that industrial structure affects CO₂ emissions both directly and indirectly through improved energy efficiency. Therefore, we select industrial structure as one of the control variables. Finally, as an innovation in this study, government inputs are used as a control variable to examine the impact of government governance on urban environments.

In summary, we select the following five control variables: (1) the economic development level (PGDP), represented by the per capita gross domestic product (GDP); (2) the labor force level (Labor), indicated by the total population of each city at the end of the year; (3) the degree of openness (Open), signifying the level of receptivity to foreign capital, technology introduction, and production development, which is quantified by the amount of utilized foreign capital during the year [44]; (4) the industrial structure (Struc), defined as the ratio of value added from the tertiary industry to that from the secondary industry; and (5) government inputs (Govs), measured by each city’s expenditure on science services as a share of the local GDP for the year.

3.2. Variable Description

When using independent variables to predict the dependent variable, the complexity of the actual problem often involves not only the direct effect of the independent variable on the dependent variable but also the possibility of an indirect effect through the mediating variable, known as mediator variables. By examining the significance of the mediating effect, the mechanism through which the independent variable affects the dependent variable can be explored, thereby increasing its theoretical value and practical application. In this study, while investigating the impact of the BRI on the urban environment, the selection of government intervention as a mediator variable is significant. First, given the dominant role of the government in economic development and environmental governance in China, the level of government intervention directly influences changes in the urban environment. The Chinese government’s actions and policy formulation in the BRI have impacted the urban environment at a scale that goes beyond a single policy effect and involves government intervention and guidance in the market, industry, and society. Therefore, incorporating government intervention as a mediator variable facilitates a deeper understanding of the government’s role in implementing the BRI and the mechanisms underlying its impact on the urban environment.

Consequently, this study selects the degree of government intervention (Govi) as a mediating variable, quantified by the proportion of the government’s annual fiscal expenditure to GDP.

As we transition from discussing mediator variables to exploring the role of moderator variables, it is essential to recognize the intricate interplay between initiative implementation, governmental intervention, and the dynamic influence of science and technology.

The selection of science and technology level as a moderator variable plays a pivotal role in studying the impact of the BRI on the urban environment. First, the advancement of science and technology significantly promotes scientific and technological innovation, enhances industrial efficiency, and bolsters the government’s capacity for environmental governance within the BRI. With advancements in science and technology, governments can more accurately assess the trade-offs among resource utilization efficiency, environmental quality, and economic growth when formulating and implementing environmental protection policies. Consequently, this enables a more effective reduction in urban environmental pollution levels.

In summary, this study incorporates the level of science and technology as a moderating variable, represented by the expenditure on scientific endeavors (Sci).

3.3. Sample Data

Urban pollutants mainly include various aspects such as air, water, solid waste, and soil. The main sources of air pollution are coal combustion, industrial, and vehicle exhaust emissions. To effectively assess the level of air pollution, we select CO₂ and SO₂ emissions as key measurement indicators. In addition, water pollution is an important problem in urban environments, especially industrial wastewater discharge, which contributes significantly to water pollution. Therefore, industrial wastewater discharge is selected as a measure of water pollution. Solid waste pollution includes industrial, construction, and domestic wastes. This study uses industrial soot emissions as an indicator of solid waste pollution. In summary, this study systematically evaluates urban environmental pollution emissions by selecting four indicators: CO₂, industrial SO₂, industrial wastewater, and industrial soot emissions. These indicators consider the main sources of air, water, and solid waste pollution, which can reflect the overall situation of urban pollution more comprehensively and ensure the scientific and rigorous nature of the research results.

We employ the entropy weighting method to determine the weights of four pollutants—CO₂, industrial soot, industrial wastewater, and industrial SO₂—in 281 Chinese cities from 2003 to 2021, with the cities’ environmental pollution emission levels serving as the explanatory variable. The entropy method, recognized for its effectiveness in assigning indicator weights, operates on the principle of assessing each indicator’s contribution to the overarching goal. It determines indicator weights by considering the magnitude of influence that changes in the indicators have on the system as well as the informational content of individual indicators. Using this approach, we comprehensively evaluate the relative importance of each indicator, thereby mitigating potential biases stemming from subjective human assignments and enhancing the objectivity and precision of the evaluation outcomes.

This study builds on Shi’s [45] research methodology, which employed the entropy value method to ascertain the pollutant emissions of each city and subsequently deduce the pollution emission levels of 281 cities. This method effectively addresses concerns related to overlapping indicator information and subjectivity in weight determination, thereby providing a reliable foundation for the analysis and evaluation of urban environments. The specific steps of the entropy method are as follows.

1.

Collection and organization of raw data.

Taking the evaluation of pollutant emissions in city i over m years as an example, the evaluation indicator system includes n indicators. This constitutes the problem of a comprehensive evaluation using n indicators for m samples, forming the initial data matrix:

$$\begin{gathered} X=\left[\begin{array}{ccc}{x}_{11}& x_{12}\dots & x_{1n}\\ \begin{array}{c}{x}_{21}\\ \cdots \end{array}& \begin{array}{c}{x}_{22}\\ \dots \end{array}& \begin{array}{c}{x}_{2n}\\ \dots \end{array}\\ x_{m1}& x_{m2}\dots & x_{mn}\end{array}\right], \\ X = {\left\{x_{ij}\right\}}_{m\ast n}(0 \le i \le n, 0 \le j \le m) \end{gathered}$$

(2)

Here, $x_{ij}$ represents the value of the jth indicator in the ith year for city x. Given the varying units of measurement across different indicators, data standardization is applied to eliminate the influence of units. All indicators in this study are negative, meaning that lower values correspond to better evaluations.

$$X{\prime }_{ij}={\displaystyle \frac{max{X}_{ij}-X_{ij}}{max{X}_{ij}-min{X}_{ij}}}$$

2.

Calculate the proportion $P_{ij}$ of the ith sample value for the jth indicator relative to the sum of all sample values for that indicator.

$$p_{ij}={\displaystyle \frac{{x\prime }_{ij}}{\sum _{i=1}^n{x\prime }_{ij}}}$$

(3)

$$0 \le p_{ij}\le 1, i=1, \dots , n, j=1, \dots , m$$

3.

Calculate the entropy value of the jth indicator.

$$e_j=-{\displaystyle \frac{1}{\ln k}}{\sum }_{i=1}^m{p}_{ij}\ln p_{ij}$$

(4)

$$0 \le p_{ij}\le 1, i=1, \dots , n, j=1, \dots , m$$

4.

Calculate the difference coefficient.

$$d_j=1-e_j$$

(5)

$$0 \le d_j\le 1, i=1, \dots , m$$

5.

Calculate the weights.

$$w_j={\displaystyle \frac{1-e_j}{\sum _{j=1}^n{d}_j}}$$

(6)

$$0 \le d_j\le 1, i=1, \dots , m$$

6.

Calculate the evaluation value of pollutant emission levels for the ith year.

$$s_i={\sum }_{j=1}^n{X\prime }_{ij}{w}_j$$

(7)

$$0 \le S_i\le 1, i=1, \dots , n, j=1, \dots , m$$

Specific results are detailed in

Table 1. Weighting results of entropy method.

Indicator	Weight
Industrial wastewater emissions	0.1873338
SO2 emissions	0.1159741
Industrial soot emissions	0.2093118
CO2 emissions	0.4873802

Source: Authors’ calculations based on statistical yearbooks at all administrative levels and the China Carbon Accounting Database.

.

This study focuses on 281 prefecture-level cities in China during 2003–2021; provinces with missing data are excluded to ensure authenticity. The experimental group comprises prefecture-level cities and key node cities under the jurisdiction of the 18 provinces and regions classified as impacted by the BRI, as outlined in the Vision and Action document, totaling 143 cities. The reference group comprises the remaining prefecture-level cities in China, amounting to 138 cities (excluding those with missing data). Data sources include the China Urban Statistical Yearbook 2003–2021 and official statistical yearbooks released by each province and city. Carbon emission data are sourced from statistical yearbooks at all administrative levels and from the China Carbon Accounting Database. The missing data are supplemented and enhanced using linear interpolation.

Table 2. Descriptive statistics of variables.

VarName	Obs	Mean	SD	Min	Median	Max
pollution	5339	0.471	0.066	0.305	0.476	0.770
did	5339	0.187	0.390	0.000	0.000	1.000
treat	5339	0.509	0.500	0.000	1.000	1.000
post	5339	0.368	0.482	0.000	0.000	1.000
PGDP	5339	10.317	0.835	8.309	10.396	11.970
Labor	5339	5.875	0.671	3.850	5.924	7.205
Open	5339	2.578	1.612	0.000	2.462	6.501
Struc	5339	86.685	8.714	60.780	88.060	99.570
Govs	5339	0.003	0.005	0.000	0.001	0.027
Govi	5339	0.175	0.190	0.009	0.137	6.041
Sci	5339	9.538	1.960	0.000	9.633	15.529

Source: Authors’ elaboration based on the China Urban Statistical Yearbook 2003–2021 and the official statistical yearbooks released by each province and city.

presents the descriptive statistics of the model variables.

4. Empirical Results

4.1. Correlation Test

A correlation test is conducted to ascertain whether there exists a correlation between the independent and dependent variables, serving as a preliminary step to the regression analysis. The primary objective is to assess whether the selected independent variable exhibits a statistically significant correlation with the dependent variable, thereby determining its suitability for regression analysis. If there is a significant correlation between the independent and dependent variables, the chosen independent variable can explain the variation in the dependent variable, which shows that the modelling is justified.

First, a Pearson’s correlation coefficient matrix test is conducted prior to the regression analyses; the results are presented in

Table 3. Correlation coefficient matrix.

	Pollution	did	PGDP	Labor	Open	Struc	Govs
pollution	1.000
did	−0.068 ***	1.000
PGDP	0.274 ***	0.347 ***	1.000
Labor	0.453 ***	−0.120 ***	−0.031 **	1.000
Open	0.503 ***	−0.088 ***	0.543 ***	0.454 ***	1.000
Struc	0.290 ***	0.087 ***	0.707 ***	−0.020	0.516 ***	1.000
Govs	0.014	0.021	0.002	0.037 ***	0.103 ***	0.044 ***	1.000

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01, ** p < 0.05.

. The primary explanatory variable (did) exhibits a negative correlation with pollution, aligning with the expected hypothesis and suggesting a significant effect of initiative implementation on urban pollution emissions. However, given that the correlation coefficient matrix only assesses the relationship between bivariate variables and does not account for the interference of the control and potential variables, these results are indicative and require further regression analysis for a precise determination.

4.2. Collinearity Test

Multicollinearity tests are conducted to address concerns related to data collinearity, typically by using the variance inflation factor (VIF). The VIF represents the ratio of variance in the presence of multicollinearity among the explanatory variables to the variance in its absence. A higher VIF indicates more severe covariance. Generally, when 0 < VIF < 10, multicollinearity is absent; when 10 ≤ VIF < 100, strong multicollinearity is present; and when VIF ≥ 100, severe multicollinearity is observed.

Table 4. Collinearity test.

	VIF	1/VIF
PGDP	3.047	0.328
Open	2.495	0.401
Struc	2.218	0.451
Labor	1.508	0.663
did	1.355	0.738
Govs	1.024	0.977
Mean VIF	1.941	.

Source: Authors’ calculations.

presents the results of the multicollinearity tests of the model; the VIF value of each variable is less than 10, indicating that the indicators selected in this study do not generally exhibit collinearity issues.

Variance inflation factor.

4.3. Model Fitting Test: Hausman and F-Test

In a panel data analysis, the selection of an appropriate model is crucial. Panel data are typically analyzed using mixed, fixed, or random effects models. This study relies heavily on Hausman and F-tests to determine the most suitable model for data analysis. The results of the Hausman test guide the choice between the random effects model and the fixed effects model, whereas the F-test results aid in determining whether to opt for the fixed effects model or the mixed effects model. Following these tests, we select the model best suited to the data characteristics for subsequent analyses to ensure the accuracy and reliability of the results. The results of these tests are presented in

Table 5. Hausman test and F-test.

Hausman Test			F-Test
chi2 Statistic	p Value	Result	chi2 Statistic	p Value	Result
103.58	0.000	reject	73.32	0.000	reject

Source: Authors’ calculations.

.

The Hausmann and the F-tests’ result statistics are significant in rejecting the original hypothesis; therefore, this study should choose the fixed effect model for the next step of the analysis.

4.4. Benchmark Regression

We employ stepwise fixed regression to test our hypotheses and enhance the reliability of our research findings. We initially use mixed ordinary least squares (OLS) regression and random effects models for analysis. Subsequently, we integrate the results of the Hausman test and F-test from the preceding section to identify the most appropriate model. After addressing endogeneity concerns stemming from individual and time effects, we ultimately select a fixed effects model for the regression analysis. This decision is informed by a combination of multiple test results, ensuring a high degree of accuracy and credibility for our analyses.

The regression results in the third column of

Table 6. Benchmark regression.

	(1) OLS	(2) RE	(3) FE
Variables	Pollution	Pollution	Pollution
Did	−0.0078	−0.0103 ***	−0.0105 ***
	(−1.49)	(−8.74)	(−3.79)
PGDP	0.0071	0.0068 ***	0.0077
	(1.60)	(7.44)	(1.06)
Labor	0.0341 ***	0.0371 ***	0.0329 ***
	(9.15)	(13.01)	(10.56)
Open	0.0096 ***	0.0004	−0.0004
	(19.97)	(0.79)	(−0.26)
Struc	0.0009 ***	−0.0002 *	−0.0004
	(4.07)	(−1.89)	(−1.13)
Govs	−0.3784	−0.6596 ***	−0.6672 ***
	(−1.21)	(−8.65)	(−3.15)
Constant	0.0977 *	0.2051 ***	0.2364 ***
	(1.70)	(11.30)	(3.16)
Observations	5339	5339	5339
R-squared	0.337	0.0582	0.0592
Number of groups	281		281
Area	NO	NO	YES
Year	NO	NO	YES
Number of id		281

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01, * p < 0.1.

reveal that the core explanatory variable (did) is negatively correlated with the dependent variable pollution at the 1% significance level, with a negative coefficient and a coefficient value of −0.0105, indicating that the initiative effect reduces 0.0105 units of pollution on average, indicating that the BRI reduces urban pollution emissions.

In the same column, the regression coefficient for the level of economic development (PGDP) is positive but not statistically significant. This indicates that an increase in the level of economic development is positively correlated with an increase in urban pollution emissions; however, it is not possible to accurately judge the relationship between the level of economic development and urban environmental pollution. This result may reflect the notion that industrialization and urbanization processes spurred by economic development may elevate pollution emissions to some extent. However, this does not imply that economic development inevitably leads to increased pollution. Factors such as environmental awareness, technological advancements, and government intervention in pollution control may counterbalance the potential impact of economic development on pollution emissions. In subsequent empirical analyses, we will examine these factors to better understand the relationship between economic development and pollution emissions. Our goal is to provide more accurate and effective guidance for the coordinated development of environmental protection and economic progress.

The regression coefficient for the labor force level (Labor) is 0.0329 and significant at the 1% level, indicating a positive and statistically significant correlation between labor force size and urban pollution levels. This suggests that higher labor force numbers correspond to increased urban pollution levels, a relationship that is supported by statistical significance.

On the other hand, the regression coefficients for both openness to the outside world (Open) and industrial structure (Struc) are negative but not statistically significant. This finding suggests a negative correlation between both the variables and urban pollution emissions, albeit without statistical significance. Although increased openness to the outside world and optimized industrial structures may contribute to reduced urban pollution emissions, they do not emerge as major influencing factors in this analysis.

Conversely, at the 1% significance level, the control variable (Govs) is negatively correlated with the dependent variable (pollution) with a negative coefficient and a coefficient value of −0.6672; a unit of government inputs reduces pollution by an average of 0.6672 units, which demonstrates that the government inputs play a positive role in reducing the city’s pollutant emissions. A comparison of the variables’ regression coefficients elucidates that the coefficient of the control variable “government inputs” (Govs) is −0.6672. This negative value indicates that government inputs are negatively correlated with urban pollution emissions and that the absolute value of this coefficient is the highest among all variables. This implies that increasing government input has the most significant effect on reducing urban pollution emissions. Thus, the government plays a crucial role in the management of urban pollution emissions.

Government inputs can effectively manage urban pollution emissions in several ways. First, increased government input can increase investment in scientific and technological research and development, such as cleaner energy technologies and equipment upgrades, thereby reducing pollutant emissions. Second, increased government investment can also actively respond to the government’s call to promote the implementation of policies and environmental regulations; for example, China’s State Council issued the Action Program to Promote Large-Scale Equipment Renewal and Consumer Goods Trade-in, which calls for accelerating the development of “new productive force” and promoting scientific and technological innovation in China, which can reduce pollution emissions. Third, increased government investment can increase environmental awareness and public participation, promote active public participation in environmental protection, and reduce urban pollution. In summary, the significant impact of government investment in science and technology on the reduction of urban pollutant emissions underscores the pivotal role of government intervention in environmental protection. Measures such as increased investment in science and technology, the promotion of technological innovation, strengthened implementation of environmental policies, and heightened public awareness of environmental issues are essential for achieving the dual objectives of environmental protection and sustainable economic development.

The analysis of the empirical results based on the benchmark regression confirms that the implementation of the BRI significantly reduces urban pollution emissions, thus validating Hypothesis 1.

Heterogeneity Analysis

Before examining the mechanism underlying BRI’s impact on urban pollution emissions, we meticulously analyze and summarize the empirical findings derived from the benchmark regression. Our focus shifts to delving into the specific ramifications of the BRI on urban pollution emissions through two dimensions of heterogeneity analyses: geographic division and city size. Through these analyses, we aim to obtain a comprehensive understanding of the variations in pollution emissions across different regions and cities of varying sizes under the BRI’s implementation. This endeavor enables us to identify the mechanisms driving the BRI’s environmental impact, thereby providing more targeted recommendations and insights for future environmental policy formulation.

First, mainland China can be broadly classified into three major economic zones: eastern, central, and western. This classification, officially announced in 1986 through the “Seventh Five-Year Plan” and adopted at the Fourth Session of the Sixth National People’s Congress, serves as the foundation for our analysis. To explore the potential disparities in the influence of the independent variable on the dependent variable among samples from different regions, we employ a group regression approach. According to the results in

Table 7. Heterogeneity analysis (1).

	Eastern China	Central China	Western China
Variables	Pollution	Pollution	Pollution
did	−0.0127 ***	−0.0099 *	−0.0150 ***
	(−4.96)	(−1.87)	(−5.80)
PGDP	0.0109	−0.0027	0.0153 ***
	(1.45)	(−0.40)	(3.69)
Labor	0.0057 *	0.0487 ***	0.0470 ***
	(1.71)	(4.91)	(9.36)
Open	−0.0003	0.0031	−0.0011 *
	(−0.17)	(1.49)	(−1.86)
Struc	0.0011 *	−0.0004	−0.0013 ***
	(1.90)	(−1.20)	(−6.50)
Govs	−0.4031 ***	−0.8001 ***	−0.5420 ***
	(−3.58)	(−3.52)	(−3.85)
Constant	0.2596 ***	0.2332 ***	0.1320 **
	(3.44)	(3.00)	(2.55)
Observations	1900	1900	1539
R-squared	0.130	0.0799	0.103
Number of groups	100	100	81
Area	YES	YES	YES
Year	YES	YES	YES

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

, in the eastern and western regions, the independent variables exhibit negative correlations with the dependent variables at the 1% significance level. However, in the central region, the core explanatory variables are significant at the 10% level. This divergence may mirror variations in economic development, environmental governance, and policy support across different regions during the BRI’s implementation. As a relatively underdeveloped region of China, the western region has enjoyed a full policy dividend owing to the early support of various policies, such as “Western Development”. These policy benefits not only promote the improvement of urban infrastructure but also attract a large amount of talent to the west, creating favorable conditions for policy implementation. Hence, the empirical results show that the policy effects of the BRI are most significant in the western cities. By contrast, the eastern region is usually at the forefront of economic development, with strong financial strength and a huge potential for scientific and technological development. These advantages make eastern cities uniquely positioned for policy promotion and implementation.

However, the central region may lag behind, with relatively limited government investment and policy implementation in environmental protection, aligning with the “central collapse theory”. In addition, geographic location and resource endowment may influence BRI’s efficacy in each region. Natural conditions and resource utilization disparities may underlie the differential effects of the BRI across regions. While the coefficient for the central region is negative, it remains significant at the 10% level, indicating a potential role for the BRI in environmental improvement, albeit less robustly than that in the other regions. Therefore, it is imperative for the government to bolster support and investment in environmental protection and cleaner production technologies in the central region, foster industrial upgrades and structural adjustments, and mitigate pollution emissions. In addition, crafting targeted environmental protection policies and measures, enhancing environmental governance and pollution prevention, and bolstering economic development support for the central region are recommended. Encouraging more investments and resources, promoting economic development and industrial upgrading, and ensuring the coordinated advancement of national economic growth and environmental protection are essential considerations.

Following the analysis of geographic heterogeneity, we scrutinize city size (population size) to ascertain whether the BRI’s impact on urban pollution emissions varies significantly across different city sizes.

Based on the Circular on Adjusting the Criteria for Classifying the Size of Cities issued by the State Council, cities with a population exceeding 5 million are categorized as megacities, those with a population ranging from 5 million to 1 million are classified as large cities, and those with a population less than 1 million are deemed small- and medium-sized cities. According to the empirical findings presented in

Table 8. Heterogeneity analysis (2).

	Megacities’	Large Cities’	Small- and Medium-Sized Cities’
Variables	Pollution	Pollution	Pollution
	−0.0145 ***	−0.0105 ***	−0.0060 ***
	(−4.27)	(−3.76)	(−3.46)
PGDP	0.0054	0.0081	0.0173 ***
	(0.70)	(1.23)	(3.50)
Labor	0.0188 *	0.0544 ***	0.2184 ***
	(1.96)	(12.77)	(7.09)
Open	−0.0002	−0.0004	0.0072 **
	(−0.14)	(−0.34)	(2.22)
Struc	−0.0006	−0.0003	−0.0008
	(−1.43)	(−1.05)	(−1.53)
Govs	−0.7954 ***	−0.6487 ***	0.2051
	(−2.73)	(−3.07)	(0.34)
Constant	0.3766 ***	0.1033	−0.6256 ***
	(3.12)	(1.30)	(−5.18)
Observations	1839	3276	224
R-squared	0.0464	0.0813	0.330
Number of groups	118	200	17
Area	YES	YES	YES
Year	YES	YES	YES

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

, the coefficients associated with mega-cities, large cities, and small- and medium-sized cities all exhibit negative and significant values at the 1% level. This suggests that the BRI exerts a substantial influence on reducing urban environmental pollution emissions across cities of all sizes. However, the magnitude of the impact varies across different city categories. Specifically, megacities demonstrate the most pronounced effect, followed by large cities, while small and medium-sized cities display the least impact.

The empirical results reveal that the BRI has the greatest policy impact on megacities. This can be attributed to the abundance of human resources and the highly efficient logistics systems in these cities. As key nodes along the domestic routes of the BRI, megacities are well positioned to play a leadership role in driving the transformation of China’s cities into eco-friendly logistics hubs. By spearheading this transition, these cities can play a critical role in advancing China’s efforts to develop a sustainable economy, serving as models for other cities to follow and achieve green development.

4.5. Moderation Effect

Following the heterogeneity analysis, we further explore the moderating effect of the level of science and technology as a moderating variable. This step reveals whether the level of science and technology positively influences the BRI’s impact on reducing urban pollution emissions. The level of science and technology plays a key role in environmental governance, and a high level of science and technology expenditure may imply more research, development, and application of environmental protection technologies as well as more effective pollution control measures. Therefore, we expect that an increase in science and technology will further enhance the effectiveness of the BRI in reducing urban pollution. Next, we test this hypothesis by analyzing its moderating effect and explore its implications for environmental protection policies.

This study examines whether a moderation effect exists by incorporating an interaction term between the moderating and core explanatory variables. The empirical findings presented in

Table 9. Moderation effect.

	(1)
Variables	Pollution
did	−0.0038
	(−0.73)
Inter	−0.0008 *
	(−1.68)
Sci	0.0014 ***
	(3.47)
PGDP	0.0115 ***
	(10.60)
Labor	0.0191 ***
	(5.82)
Open	−0.0010 **
	(−2.15)
Struc	−0.0005 ***
	(−5.44)
Govs	−0.6114 ***
	(−5.97)
Constant	0.3128 ***
	(18.79)
Observations	5339
R-squared	0.912
Area	YES
Year	YES

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

indicate that the coefficient of the interaction term (Inter) between the BRI (did) and science and technology level (Sci) is negative at the 10% significance level, exhibiting the same sign as that of the moderated variable. This suggests the presence of an amplified moderation effect, signifying that as the level of science and technology increases, the negative impact of the BRI on urban environmental pollution emissions becomes more pronounced. In essence, an increase in the level of science and technology fosters a stronger effect of the BRI in mitigating urban environmental pollution.

This finding underscores the pivotal role of science and technology in environmental protection. A heightened level of investment in science and technology signifies an increased allocation of resources toward the research, development, and application of environmental protection technologies. Consequently, this promotes innovation and the adoption of cleaner production technologies and bolsters the enforcement of pollution control measures. Therefore, augmenting the level of science and technology not only stimulates economic growth but also effectively curbs environmental pollution, thereby providing critical support for enhancing urban environmental quality and fostering sustainable development.

This conclusion has significant implications for the governmental formulation of environmental protection policies and the advancement of scientific and technological innovation. It is imperative for the government to ramp up investment in science and technology, incentivize scientific and technological breakthroughs, and encourage the development and application of environmental protection science and technology. Concurrently, the government should foster a conducive developmental milieu for the environmental technology industry through policy backing, regulatory measures, and steering enterprises toward augmenting their investments in environmental protection. Collectively, these efforts can forge a path toward building a model of green and low-carbon urban development.

In summary, the implementation of the BRI will be more effective with improvements in science and technology, which will promote the reduction of urban pollution levels. As the level of science and technology continues to improve, the application of new technologies and innovations will provide more possibilities for urban environmental management, which will enhance the implementation and effectiveness of the policy, which also verifies Hypothesis 2.

4.6. Mediation Effect

After confirming that the level of science and technology has a positive effect on reducing urban environmental pollution, this study further explores the influence of the degree of government intervention (Govi) on urban environmental pollution and reveals its potential mechanism through an analysis of mediating effects. Government financial expenditure is crucial in environmental governance; the ratio of government financial expenditure to GDP reflects the intensity and efficiency of government intervention. Analyzing the mediating effect of the degree of government intervention on urban environmental pollution helps us comprehensively understand the mechanism of government policies in environmental protection and provides theoretical and empirical support for the formulation of more precise environmental policies.

To explore the pathway through which the independent variable influences the dependent variable and address the endogeneity issue inherent in traditional three-step methods, this study adopts the two-step mediation analysis proposed by Jiang [46]. The first step is the regression of did with pollution, which is the previous benchmark regression, and the result is significant. In the second step, the dependent variable is replaced with the mechanism variable (Govi), combined with the empirical results in

Table 10. Mediation effect.

	(1)	(2)
Variables	Govi	Govi
did	0.1581 ***	0.1149 ***
	(8.91)	(12.21)
PGDP		−0.0109
		(−0.40)
Labor		−0.1392 ***
		(−8.57)
Open		0.0022
		(0.84)
Struc		0.0190 **
		(2.48)
Govs		8.0797
		(1.38)
Constant	0.1453 ***	−0.5974 *
	(7.95)	(−1.68)
Observations	5339	5339
R-squared	0.0991	0.304
Number of groups	281	281
Area	YES	YES
Year	YES	YES

Source: Authors’ calculations. t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

; results indicate that the core explanatory variable (did) is still significantly positively correlated with Govi at 1% significance level, thus establishing a mediating effect. Therefore, the implementation of the BRI is conducive to increasing the level of government intervention (Govi).

With the implementation of the BRI, inter-regional economic cooperation has deepened, which may lead to greater government investment and intervention in environmental protection and pollution management. First, inter-regional cooperation may bring more resources and technical support, enabling governments to better implement environmental protection policies and projects and to enhance the capacity and level of environmental governance. Second, the promotion of the BRI may have facilitated cooperation and communication between governments and enterprises, making the design and implementation of environmental policies more coordinated and efficient. Moreover, the implementation of the BRI may lead to the development of related industries and projects, and the government’s supervision and support for these industries and projects will increase accordingly, thus increasing investment and intervention in environmental protection. Therefore, the implementation of the BRI may be conducive to increasing the level of government intervention, leading to the implementation of more effective environmental governance measures, which may subsequently inhibit the development of urban environmental pollution.

In summary, based on the empirical results of the mediating effect, it can be concluded that government governance is conducive to reducing urban environmental pollution, which also highlights the importance of government governance and verifies Hypothesis 3.

4.7. Parallel Trends Test

The pivotal hypothetical assumption in the evaluation of policy effects using the DID method is the parallel-trend hypothesis. This suggests that the trajectories of urban pollution emission levels in both the treatment and control groups would have remained consistent in the absence of the BRI and would not have exhibited systematic divergence over time.

The parallel trend test involves employing the event study method to regress initiative implementation either forward or backward by “t” years. This provides an estimate of the significance level of the coefficients, thereby determining the validity of the parallel trend identification. In Figure 2, the vertical line represents a 95% confidence interval. “Pre1” denotes the initiative implementation year (current), while “Pre_2”, “Pre_3”, “Pre_4”, “Pre_5”, and “Pre_6” represent the years preceding initiative implementation. Similarly, “Post_1”, “Post_2”, “Post_3”, “Post_4”, and “Post_5” denote the years following the initiative’s implementation.

From Figure 2, we observe that the confidence intervals of Pre_2 to Pre_6, representing the pre-initiative implementation years, intersect at the coefficient = 0 level, indicating non-significance. Although Post_1 and Post_2 show similar crossings, Post_3, Post_4, and Post_5, reflecting the years after initiative implementation, exhibit negative coefficients, indicating significance, at least at the 5% level. This suggests a negative correlation between the dependent variable and dummy variable post-initiative implementation, implying a significant initiative effect and lag effect post-implementation.

In summary, the findings indicate that the BRI primarily drives changes in urban pollution emissions following the implementation of the initiative.

4.8. Robustness Test

4.8.1. Lagged Processing

To mitigate the endogeneity issue arising from bidirectional causation and to explore potential time lags in the impact of the independent variables on the dependent variable, this study employs first- and second-order lags for all explanatory variables. First- and second-order lags involve comparing the current period’s data with the data lagged by one and two periods, respectively, to assess the robustness and reliability of the model. The robustness test aims to identify potential outliers, autocorrelation, and heteroskedasticity issues in the model, which can then be adjusted to enhance prediction accuracy. The results of the re-regression are presented in

Table 11. Lagged processing.

	First-Order Lag	Second-Order Lag
Variables	f_Pollution	f2_Pollution
did	−0.0180 ***	−0.0190 ***
	(−5.63)	(−5.22)
PGDP	0.0086	0.0057
	(1.34)	(0.89)
Labor	0.0303 ***	0.0257 ***
	(14.12)	(9.05)
Open	−0.0021 *	−0.0027 ***
	(−1.88)	(−2.98)
Struc	−0.0007 **	−0.0007 ***
	(−2.50)	(−3.39)
Govs	−0.4317 *	−0.4701 **
	(−1.84)	(−2.14)
Constant	0.2764 ***	0.3399 ***
	(4.34)	(4.95)
Observations	5058	4777
R-squared	0.0679	0.0747
Number of groups	281	281
Area	YES	YES
Year	YES	YES

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

. The core explanatory variables remain significantly negatively correlated with the dependent variable at the 1% significance level, further confirming the robustness and reliability of the model.

4.8.2. Excluding Outliers

To enhance the generalizability of our findings and minimize the influence of outlier samples, we exclude municipalities such as Beijing, Shanghai, Chongqing, and Tianjin from the analysis. This step allows us to evaluate whether our conclusions are applicable to an ordinary sample. As illustrated in

Table 12. Excluding outliers.

	(1)
Variables	Pollution
did	−0.0108 ***
	(−3.94)
PGDP	0.0079
	(1.08)
Labor	0.0337 ***
	(10.95)
Open	−0.0004
	(−0.31)
Struc	−0.0004
	(−1.16)
Govs	−0.6740 ***
	(−3.14)
Constant	0.2276 ***
	(3.06)
Observations	5263
Number of groups	277
Area	YES
Year	YES
R-squared	0.0615

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01.

, after excluding municipalities, the regression results still demonstrate a significant negative correlation between the core explanatory variable (did) and pollution at the 1% significance level. This indicates that our conclusions are applicable to a broader sample, further bolstering the reliability and generalizability of our research findings. When conducting the analysis by excluding municipalities, we acknowledge that these areas may possess distinct characteristics in terms of economy, policies, and population. Therefore, by excluding these special samples, we can evaluate the impact of the BRI on urban environmental pollution more accurately, thereby enhancing the generality and reliability of our conclusions.

4.8.3. Robustness Testing Based on Model Setting: Propensity Score Matching Double-Difference (PSM-DID) Modelling

The DID approach is employed to estimate causal effects by comparing the differences between the treatment and control groups. However, if there are preexisting differences between these groups, confounding factors may be introduced, leading to inaccurate estimates of causal effects. PSM-DID can help address such differences and enhance the accuracy of the causal effect estimation. Hence, it is essential to conduct robustness test using a multi-temporal PSM-DID model.

To enhance sample quality and result credibility, this study initially applies Propensity Score Matching (PSM) to match samples, with a caliper value restricted to 0.01 and using nearest-neighbor matching with a 1:1 ratio. Figure 3 illustrates that while there is considerable deviation between the two kernel density curves before matching, the distance between the mean lines shortens post-matching, with the curves nearly overlapping. Consequently, differences between the treatment and control groups become less pronounced. Figure 4 demonstrates that after matching, the distances of the points are closer to 0, indicating a reduction in the covariate differences.

Table 13. Balance sheet of differences.

Variable	Matched	Treated	Control	%bias	|bias|	t	p > |t|	V (C)
PGDP	U	10.38	10.26	14.60		5.320	0	1.030
	M	10.35	10.46	−13.20	9.600	−4.830	0	1.11 *
Labor	U	5.680	6.078	−62.30		−22.70	0	1.51 *
	M	5.739	5.701	6	90.30	2.120	0.0340	0.85 *
Open	U	2.305	2.861	−35.10		−12.78	0	1.62 *
	M	2.376	2.576	−12.60	64	−4.640	0	1.69 *
Struc	U	86.42	86.96	−6.200		−2.280	0.0230	1.60 *
	M	86.37	88.25	−21.60	−245.2	−8.160	0	1.79 *
Govs	U	0.00299	0.00314	−3.200		−1.160	0.244	1.10 *
	M	0.00302	0.00313	−2.200	30.10	−0.830	0.409	1.32 *

Source: Authors’ calculations. t-statistics in parentheses, * p < 0.1.

shows the covariate differences before and after matching, denoted by U for the unmatched samples and M for the matched samples. The changes before and after matching indicated significant reductions in all covariate differences, with the t-test transitioning from originally significant to non-significant. This suggests an improvement in the covariate similarity between samples and the alleviation of sample selection bias.

Subsequently, the matched samples are subjected to a regression analysis. As shown in

Table 14. PSM regression.

	(1)
Variables	Pollution
did	−0.0107 ***
	(−3.58)
PGDP	0.0062
	(0.85)
Labor	0.0294 ***
	(5.58)
Open	−0.0005
	(−0.31)
Struc	−0.0003
	(−0.76)
Govs	−0.6664 ***
	(−2.86)
Constant	0.2655 ***
	(2.88)
Observations	2552
Number of groups	279
Area	YES
Year	YES
R-squared	0.0476

Source: Authors’ calculations. t-statistics in parentheses, *** p < 0.01.

, the regression results reaffirm that the core explanatory variable (did) retains a significantly negative coefficient at the 1% significance level, which is consistent with the previous benchmark regression conclusion. These results demonstrate robustness and further underscore the significant inhibitory effect of the BRI on urban pollution emissions.

4.8.4. Placebo Tests

Despite controlling for numerous city characteristic variables, the possibility remains that unobserved factors influence the assessment results of the BRI. To mitigate the impact of potential variables on the relationship between did and the dependent variable, this study employs a counterfactual assumption for robustness testing. Specifically, the experimental and control groups are randomly disrupted and an equal number of groups are extracted to form a new “experimental group”. Similarly, the policy time points are adjusted and the pseudo-experimental group interacts with the time dummy variable to create a new did variable. After repeating these experiments 500 times, the kernel density of the coefficients is plotted (see Figure 5). The results demonstrate a distribution that generally follows a normal distribution, indicating random sample selection. Moreover, there is no evidence of endogeneity or human interference. The actual regression coefficients fall within a low probability of rejection range, confirming the validity of the placebo test.

5. Conclusions

Based on data from 281 prefecture-level cities in China from 2003 to 2021, this study considers the BRI as a quasi-natural experiment between participating and non-participating cities using 2013 as the comparative time point. Consequently, this study empirically evaluates the mitigation effects of the BRI on cities along domestic routes from a China-wide perspective using the DID method. After conducting parallel trend tests and a series of robustness tests, our empirical findings consistently support political promotion policies for participating cities to be more effective in reducing their emissions. Moreover, the results of the moderating effects, mediating mechanism tests, and heterogeneity analyses indicate that the BRI has a significant impact on reducing urban environmental pollution. There is a negative correlation between the implementation of the BRI and the level of urban pollution, implying that the initiative reduces the level of urban pollution on average. Notably, the regression results with government inputs (Govs) as a control variable indicate that increasing science and technology inputs can significantly reduce urban pollution emissions, implying that market-oriented promotion policies are more effective than regulatory policies. In addition, the heterogeneity analysis indicates that the policy effects of the BRI are more pronounced in the eastern and western regions as well as in larger cities. In the eastern region, the initiative’s significant impact is driven by demand-pull market forces. Larger cities in this region benefit from strong market dynamics and heightened demand for green development, which facilitate more effective implementation of the BRI. Conversely, policy-driven regulations play a pivotal role in mitigating environmental pollution in the western regions. Local governments in these areas have enforced stricter environmental regulations and implemented sustainability-focused initiatives, leading to substantial improvements in environmental outcomes. By contrast, the central region faces challenges due to weaker policy support and limited market-driven demand for sustainable practices, resulting in a relatively smaller impact of the BRI on urban environmental pollution. From the perspective of the new development paradigm, which emphasizes a domestically driven economic cycle complemented by international integration, the central region must leverage its unique geographical advantages. By capitalizing on its strategic position as a domestic market hub, the central region can integrate more effectively with the BRI and better contribute to the construction of the new development framework. Simultaneously, the central government should allocate more policy support to cities in this region. As a crucial agricultural base and comprehensive transportation hub in China, the central region plays a significant role in the national economy; strengthening the development of this region is essential for aligning with China’s broader development strategy and ensuring its pivotal role in national growth and prosperity. Further analyses of the moderating effects emphasize the positive role of S&T progress in reducing urban environmental pollution.

Furthermore, the mediation effect results indicate that the BRI enhances government intervention, thereby improving governance levels and reducing urban environmental pollution. In summary, this study underscores the crucial role of government-led governance in environmental protection and provides valuable insights into the formulation of more effective environmental policies in developing countries to foster logistics hubs in their regional economies. Even if governments suffer at the initial stage of trial and error, government-led (not regulation-based) promotional policies on air pollution can certainly result in more harmonized sustainable development of hub cities and their surrounding regions.

In conclusion, the empirical findings not only confirm the BRI’s effectiveness in reducing urban pollution but also underscore the pivotal role of government intervention and green technology innovation in advancing sustainable development. However, the initial success of the BRI in our research does not guarantee its successful performance unless the market-oriented systematic support by all the neighboring cities and provinces, because governance by government-led policies is only effective at the initial stage. To avoid continuous trial and error, the BRI should harmonize environmentally friendly regulations with stronger and collaborative partnerships with local markets. For this harmonized system of collaboration networking alongside the BRI, factors such as competition and collaboration with neighboring cities and regions, capitalizing on the opportunities presented by national development strategies, and various policy incentives to advance the establishment of sustainable, eco-friendly logistics hubs can result in stronger governance for sustainable performance. The logistics hubs of the BRI can be sustainable only when the market-driven power to create logistics demand from this collaborative network creates value. To better integrate eco-friendly logistics hubs into the future of the BRI and ensure sustainable green development under low-carbon policies, policymakers should increase investment in green transport infrastructure with a focus on supporting electric public transportation systems, low-carbon freight transport networks, and logistics facilities that rely on renewable energy. At the same time, governments should encourage the logistics industry to adopt green technologies through incentives such as tax breaks, subsidies, and technical assistance. These investments will not only significantly reduce the emissions generated during transportation but also promote long-term environmental sustainability. Policymakers should encourage cities along the Belt and Road to strengthen regional cooperation and technology transfer, particularly in areas such as low-carbon transportation and sustainable logistics platforms. By promoting cross-regional cooperation, expertise and best practices can be shared to enhance the overall capacity for implementing green logistics solutions. In addition, governments should play an active role in accelerating the transition to green logistics by launching low-carbon logistics projects and stimulating private sector participation in financing sustainable transport and logistics projects through incentives such as tax breaks and subsidies. Finally, for regions facing developmental challenges, particularly in terms of infrastructure and technology, governments should introduce transitional support policies, including measures such as subsidies for the procurement of green technologies, capacity-building projects for sustainable logistics, and green infrastructure development. By supporting lagging regions in their transition to sustainable logistics, policymakers can reduce regional disparities and ensure more balanced and equitable implementation of the BRI. These insights provide valuable guidance for policymakers in crafting strategies for the optimal transition from traditional urbanization issues toward an inclusive carbon-zero economic system, not only along the Belt and Road routes but also other regional hub cities in developing countries. By fostering low-carbon, resource-efficient cooperation network systems, these cities can reduce emissions and environmental impacts, while supporting economic growth. This is essential for achieving carbon-neutrality targets and advancing the broader mission of the BRI to promote sustainable development in all participating countries. The transformation of major urban centers into environmentally friendly logistics hubs will be key to driving the shift toward greener and more sustainable global value chains.