The Impact of Highway Transportation Infrastructure on Carbon Emissions in the Yangtze River Delta Region

Abstract

To address the increasingly severe issue of carbon dioxide emissions, the Chinese government has set dual carbon goals: achieving peak carbon emissions by 2030 and carbon neutrality by 2060. Studying the impact of highway transportation infrastructure on carbon emissions is crucial for achieving these dual carbon goals and promoting sustainable development. Using balanced panel data from 41 cities in the Yangtze River Delta region from 2006 to 2019, this paper empirically analyzed the relationship and mechanisms of highway transportation infrastructure’s impact on carbon emissions using fixed effects models, mediating effects models, and threshold effects models. The empirical results indicate: (1) there is a nonlinear inverted-U-shaped relationship between highway transportation infrastructure and carbon emissions; (2) highway transportation infrastructure indirectly affects carbon emissions through foreign direct investment as a mediating variable; (3) the threshold effect model verifies that the impact of highway transportation infrastructure on carbon emissions exhibits threshold effects based on green technological innovation and industrial structure upgrading. This study enriches the research content in related fields to some extent and provides specific policy recommendations for achieving carbon reduction goals and promoting sustainable development.

1. Introduction

Global climate change is one of the most important environmental issues facing humanity today. Reducing carbon emissions and mitigating climate change have become consensus and urgent tasks for countries worldwide in the 21st century. Addressing the issue of carbon dioxide reduction, China committed in the 2015 Paris Agreement to “reduce carbon intensity by 60–65% from the 2005 level by 2030”. Furthermore, at the 75th United Nations General Assembly in 2020, China pledged to peak carbon emissions before 2030 and achieve carbon neutrality by 2060.

Despite China’s firm determination, according to the International Energy Agency (IEA) report “CO₂ Emissions in 2022”, China, as the world’s largest emitter of carbon dioxide, emitted approximately 12.1 billion tons in 2022, accounting for about 32.8% of global carbon emissions. This indicates that China currently faces significant challenges in achieving its dual carbon goals. The national carbon emission targets need to be concretely implemented at the regional level. The Yangtze River Delta region, which accounts for a large proportion of the country’s carbon emissions, has become a key area for national attention on carbon reduction.

In recent years, the transportation sector has become a major contributor to carbon emissions, becoming the second largest carbon-emitting sector [1,2]. Li et al. found that in 2019, carbon emissions from the transportation sector in China accounted for approximately 11% of the country’s total societal carbon emissions, with road transportation contributing 86.76% of the transportation sector’s carbon emissions [3]. Carbon emissions from road transportation far exceed those from other modes of transportation. Despite transportation contributing to increased carbon emissions and global warming through the use of fossil fuels, it makes a significant and visible contribution to economic development [4]. Therefore, it is necessary to empirically analyze the impact of highway transportation infrastructure on carbon emissions to meet various carbon reduction targets.

The Yangtze River Delta region is a fertile plain before the Yangtze River flows into the sea. It is one of the most economically developed regions in China but also one of the regions with the most severe carbon emissions. Rapid industrialization and modernization in the region have increased energy demand and substantial carbon emissions. Hence, this paper focuses on the Yangtze River Delta region to explore the relationship and mechanisms through which highway transportation infrastructure affects carbon emissions. The aim is to provide decision-making references for achieving low-carbon sustainable development in the Yangtze River Delta region under the context of dual carbon goals.

The marginal contributions of this study are: (1) Currently, research literature on the impact of transportation infrastructure on carbon emissions is limited, and there is a lack of explanations regarding the mechanisms through which transportation infrastructure affects carbon emissions. (2) Using the mediation effect model, this study verifies the mediating relationship between highway transportation infrastructure and carbon emissions in the Yangtze River Delta region, with foreign direct investment (FDI) as the mediating variable. (3) Utilizing the threshold effect model, this study examines the nonlinear impact of highway transportation infrastructure on carbon emissions under different constraint mechanisms.

2. Literature Review

2.1. Research on Transportation Infrastructure

According to relevant literature, transportation infrastructure impacts various aspects of the socio-economic landscape. Research has found that transportation infrastructure significantly promotes economic growth in China [5,6]. Xu found that improvements in transportation infrastructure significantly increase exports [7]. Donaldson, studying data from India, found that railway transport reduces trade costs and promotes international trade growth [8]. Ozcan analyzed data from Turkey and found that provinces with higher air traffic volumes and denser road networks tend to attract more foreign direct investment [9]. Xu studied data from Chinese cities and found that the introduction of high-speed rail significantly promotes industrial structural transformation and upgrading [10]. Zong found that transportation infrastructure positively promotes economic agglomeration [11]. This section introduces the impact of transportation infrastructure on socio-economic factors. The next section will discuss the factors that influence carbon emissions.

2.2. Research on Carbon Emissions

In recent years, the study of factors influencing carbon emissions has garnered widespread attention from scholars. Researchers primarily utilize Index Decomposition Analysis (IDA), Structural Decomposition Analysis (SDA), and the STIRPAT model to quantify the factors affecting carbon emissions. IDA decomposes carbon emissions into additive or multiplicative factors, requiring less data and having wider applications. However, IDA cannot quantify the impact of technological changes and efficiency changes on carbon emissions. IDA extended to energy-related carbon emissions studies in 1991 [12]. SDA is based on the input–output model in econometrics. Compared to the IDA method, SDA can more easily distinguish between technical effects and final demand effects. It can evaluate both direct and indirect effects simultaneously but requires extensive input and output data. Some scholars have utilized SDA to study the driving factors of carbon emissions in Beijing. Their research findings indicate that production structure and population growth have a positive impact on carbon emissions in China, whereas reductions in carbon intensity and demand have a negative impact on carbon emissions [13]. The STIRPAT model may be superior to IDA and SDA methods, allowing the examination of more factors than IDA and being less constrained by data than SDA. Due to its strong interpretability and scalability, the STIRPAT model has become a popular method for studying the factors influencing carbon emissions in recent years. Thio et al. used the STIRPAT model to analyze the drivers of carbon emissions in the top 10 emitting countries from 2000 to 2014. The study found that economic growth, population change, and technology innovation are key determinants influencing carbon emissions [14].

In the field of engineering applications, Budennyy et al. introduced an open-source package called eco2AI, which focuses on accurately tracking the energy consumption and corresponding carbon dioxide emissions of models in AI applications [15]. Martínez et al. proposed a new convolutional neural network (CNN) training method for autonomous driving, which significantly reduces training time and decreases the carbon footprint by approximately 96% [16]. Ma et al. applied machine learning algorithms to predict carbon dioxide emissions and found that the Gaussian process regression (GPR) algorithm provided the most accurate CO₂ emission predictions when comparing actual and predicted results [17]. Li et al. utilized three machine learning algorithms to forecast transportation-based CO₂ emissions. By comparing four commonly used statistical metrics (R², MAE, rRMSE, and MAPE), they found that the GBR model performed the best [18].

This section introduced the factors influencing carbon emissions. The next section will discuss how transportation infrastructure affects carbon emissions.

2.3. Research on the Impact of Transportation Infrastructure on Carbon Emissions

Currently, the relationship between carbon emissions and transportation infrastructure is largely overlooked by researchers, and there are few studies on the impact of transportation infrastructure on carbon emissions. Xie et al. argues that highway transportation infrastructure increases carbon emissions; the population scale effect of highway transportation infrastructure helps reduce carbon emissions, while its economic growth effect has a positive impact on carbon emissions [19]. Xiao et al. used spatial econometric models to find that road transportation infrastructure has a significant spatial spillover effect on carbon emissions, with road infrastructure exacerbating carbon emissions in surrounding cities through the industrial agglomeration effect as a mediator [20]. Churchill et al. studied the relationship between transportation infrastructure and CO₂ emissions in OECD countries and found that transportation infrastructure is positively correlated with CO₂ emissions, with economic growth and population being important mediators in the transmission of transportation infrastructure’s impact on CO₂ emissions [21].

In addition to the above research on the impact of highway infrastructure on carbon emissions, other modes of transportation also affect carbon emissions. Zhang and Li found that the introduction of high-speed rail reduces urban carbon emission intensity, with industrial structure adjustment playing a partial mediating role in this process [22]. Dzator et al., through empirical research on developing countries, found that air and railway transport infrastructure promote increases in carbon emissions [23]. Li et al. discovered a short-term and long-term causal relationship between carbon emissions and the development of various transportation modes such as railways, road, airline, and inland waterways. They found that the development of road, airline, and waterway infrastructure leads to long-term increases in carbon emissions, with the positive impact of waterway infrastructure being the strongest [24]. Acheampong et al. using the EU as a research subject found through empirical research that freight and railway transport infrastructure have an inverted-U-shaped relationship with carbon emissions [25].

Table 1. Findings and limitations of the current research.

Author	Findings	Limitations
Xie et al. [19]	highway transportation infrastructure increases carbon emissions	The data being used are relatively old
Xu et al. [26]	there exists an inverted U relationship between highway infrastructure and carbon emissions.	The data being used are relatively old
Churchill et al. [21]	transportation infrastructure is positively correlated with CO2 emissions	Using national-level data
Zhang and Li [22]	high-speed rail reduces carbon emission intensity	The data being used are relatively old
Dzator et al. [23]	air and railway transport infrastructure promote increases in carbon emissions	Using national-level data
Li et al. [24]	the development of road, airline, and waterway infrastructure leads to increases in carbon emissions	The data being used are relatively old
Acheampong et al. [25]	freight and railway transport infrastructure have an inverted-U-shaped relationship with carbon emissions	Using national-level data

shows the findings and limitations of the current research.

In summary, existing studies have several limitations: (1) limited relevant literature, requiring further research; (2) lack of explanation on the mechanism of transportation infrastructure’s impact on carbon emissions; (3) empirical data selection is limited to national or provincial levels, lacking in-depth studies on city-level carbon emissions.

3. Research Hypotheses

Regarding the impact of highway infrastructure construction on carbon emissions, there are currently two main contrasting views. Some studies indicate that during the construction of highway transportation infrastructure, substantial fossil energy and materials are consumed, directly leading to carbon dioxide emissions. Additionally, Chai et al. argue that transportation infrastructure growth stimulates transportation demand, and increased transportation activities consume more fossil energy [27], thereby increasing carbon emissions.

On the other hand, other scholars suggest that with the continuous development of highway transportation infrastructure, there can be a reduction in carbon dioxide emissions. This view arises from the highway transportation infrastructure’s role in reducing transportation time and distance [28], thereby lowering carbon dioxide emissions during transportation processes. Moreover, Xu et al. found through empirical research that due to significant positive externalities, highway transportation infrastructure development can influence carbon emissions by improving regional economic characteristics, leading to a nonlinear inverted U relationship between highway transportation infrastructure and carbon emissions [26]. Based on the above analysis, this study proposes the following research hypothesis:

Hypothesis 1.

There exists an inverted U nonlinear relationship between highway transportation infrastructure and carbon emissions.

Second, good transportation infrastructure can reduce the cost of transporting goods, including reducing transportation time and costs, which can attract more foreign investment. Zhou and Hao argue that improved transportation conditions represent enhanced market accessibility, reduced transportation and logistics costs, and shortened production cycles, thus becoming crucial factors in attracting foreign direct investment (FDI) [29]. Samir and Mefteh have shown through their studies that the enhancement of transportation and logistics infrastructure contributes to increasing attractiveness for FDI [30]. Therefore, better transportation infrastructure facilitates attracting FDI.

FDI has uncertain effects on the environment of host countries. On one hand, the “pollution haven” effect suggests that multinational corporations relocate high-polluting and energy-intensive industries to developing countries to avoid high domestic environmental governance costs, leading to increased local environmental pollution and carbon emissions. On the other hand, the “pollution halo” effect argues that FDI can bring advanced low-carbon production methods and environmentally friendly management systems, as well as being able to enhance local technology and efficiency through technology spillover effects, thereby reducing local environmental pollution and carbon emissions [31]. Hence, the complex mechanisms through which FDI affects carbon emissions may result in a nonlinear relationship between FDI and carbon emissions. Li et al. found through empirical research that in the early stages of development in developing countries, relaxed environmental regulations lead to FDI exerting more of a “pollution haven” effect on carbon emissions. As economies develop and environmental conditions deteriorate, governments pursue green FDI and stricter environmental regulations lead FDI to exhibit more of a “pollution halo” effect, resulting in an inverted U relationship between FDI and carbon emissions [32]. Based on the above analysis, highway transportation infrastructure may indirectly influence carbon emissions through its impact on FDI. Therefore, this study proposes research hypothesis 2.

Hypothesis 2.

Highway transportation infrastructure affects carbon emissions through its mediation of attracting foreign direct investment (FDI).

Finally, Wang et al. argue that transportation infrastructure shortens commuting times, reduces the cost of face-to-face interactions among professionals, expands the dissemination of knowledge and technology, and promotes knowledge spillover [33]. Li and Min suggest that transportation infrastructure facilitates the diffusion of technology and knowledge, enhances resource allocation efficiency in markets, and improves regional innovation levels [34]. Guo et al. propose that various green technology innovations can be widely applied in social production, transportation, and daily life, promoting the development of regional green and low-carbon industrial systems and accelerating the region’s green and low-carbon transformation, thereby reducing regional carbon emissions [35].

Transportation infrastructure can facilitate the dissemination of knowledge related to green technology innovation. Green technology innovation is crucial for reducing carbon dioxide emissions through two main avenues: (1) developing low-carbon and zero-carbon technologies to enhance regional production and energy efficiency, thereby influencing regional carbon emissions; (2) producing a wider variety of low-carbon products and substituting high energy consumption products, thus impacting regional carbon emissions. Therefore, constrained by the level of green technology innovation, the effect of highway transportation infrastructure on carbon emissions exhibits a nonlinear effect.

Changes in industrial structure influence the structure of energy consumption. Zhang finds that transportation infrastructure promotes industrial structural transformation through facilitating labor mobility, reducing transportation costs, and promoting technological progress [36]. Xu et al. suggest that infrastructure investment promotes industrial structure upgrading, although the effects vary across different regions and developmental stages [37].

The tertiary industry, compared to the primary and secondary industries, offers advantages such as lower resource consumption, higher economic efficiency, and greater technological intensity. Therefore, actively promoting the development of the tertiary industry is a crucial pathway to reduce carbon emissions. Some scholars have found through empirical research that industrial structure upgrading is conducive to reducing carbon emissions [38,39]. Thus, depending on the level of industrial structure upgrading, the effect of highway transportation infrastructure on carbon emissions exhibits a nonlinear effect. Based on the above analysis, this study proposes the following research hypotheses:

Hypothesis 3.

The effect of highway transportation infrastructure on carbon emissions exhibits nonlinear variations due to different levels of green technology innovation.

Hypothesis 4.

The effect of highway transportation infrastructure on carbon emissions exhibits nonlinear variations due to different levels of industrial structure upgrading.

4. Materials and Methods

4.1. Variable Description

The dependent variable is ${\mathrm{C}\mathrm{O}}_2$ emissions intensity (variable code: cei). ${\mathrm{C}\mathrm{O}}_2$ emissions intensity is calculated as the ratio of urban carbon dioxide emissions to regional GDP. To balance economic development and carbon emissions, a more realistic and feasible approach is to pursue a reduction in the relative amount of carbon emissions rather than an absolute reduction. Carbon emission intensity is an excellent relative indicator that reflects the relationship between carbon emissions and the level of economic development.

The independent variable is highway transportation infrastructure (road). Extant research identifies two main types of transportation infrastructure development: capital stock and public investment. This paper references the related research by Xie et al. and chooses the length of highways as the measure for highway transportation infrastructure [19].

The mediating variable is foreign direct investment (FDI), and the threshold variables are green technology innovation and industrial structure upgrading. Green technology innovation (gi): measured by the total number of green patent applications. Foreign direct investment (fdi): measured by the actual amount of foreign direct investment utilized in the region for the year. Industrial structure upgrading (is): represented by the percentage of the increased value of the tertiary industry to GDP.

Control variables include R&D investment (desh), represented by the percentage of science and technology expenditure in local budget expenditures; government intervention (gov), measured by the percentage of government fiscal expenditure to GDP; population density (pop), represented by the number of people per square kilometer; economic development (gdpp), represented by per capita GDP as an indicator of economic development.

This paper processes the variables as follows: first, all variables are logarithmized in this paper; secondly, for variables containing 0 values, such as green technological innovations, they are logarithmized after adding 1.

4.2. Data Sources and Descriptive Statistics

This article selects balanced panel data from the Yangtze River Delta (YRD) region’s 41 cities spanning the years from 2006 to 2019. The data in this research are mostly derived from the China Statistical Yearbook, the China Urban Statistical Yearbook, the provincial and municipal statistical yearbooks, EPS global statistical data, The China Carbon Accounting Database (CEADs) and the China Research Data Services Platform. We used the linear interpolation method to fill in missing data.

Table 2. Variable descriptive statistics.

Variable	N	Mean	Std. Dev.	Min	Max
lncei	574	0.392	0.721	−1.511	2.536
lnfdi	574	12.34	1.629	7.220	16.39
lnis	574	3.713	0.189	3.151	4.287
lngi	574	5.454	1.828	0	9.668
lnpop	574	−2.676	0.580	−4.232	−1.008
lngdpp	574	10.70	0.745	8.408	12.20
lnroad	574	9.140	0.556	6.711	10.11
lndesh	574	0.810	0.921	−3.341	3.040
lngov	574	2.324	0.439	0.764	4.997

presents the descriptive statistics for each variable.

4.3. Model Design and Empirical Strategies

4.3.1. Benchmark Regression Model

This study established a panel fixed effects model as the baseline regression model to verify the relationship between highway transportation infrastructure and carbon emissions. The baseline regression model is expressed as follows:

$$ln{cei}_{it}={\alpha }_0+{\alpha }_{1}ln{road}_{it}+{\alpha }_2{ln{road}_{it}}^2+\gamma {control}_{it}+{\mu }_t+{\mathrm{\lambda }}_i+{\epsilon }_{it}$$

(1)

The dependent variable $ln{cei}_{it}$ represents the logarithm of ${\mathrm{C}\mathrm{O}}_2$ emissions intensity. The independent variable $ln{road}_{it}$ represents the logarithm of highway transportation infrastructure; i represents years, and t represents cities; ${control}_{it}$ represents the control variable; ${\epsilon }_{it}$ is a random disturbance term; ${\mathrm{\lambda }}_i$ and ${\mu }_t$ are individual fixed effect (CITYFE) and time fixed effect (YEARFE), respectively; and ${\alpha }_0$ is a constant item.

4.3.2. Mediating Effects Model

This paper refers to the mediating effect model constructed by Xu et al. in empirical research [26] to examine the mediating mechanism through which highway transportation infrastructure influences carbon emissions. First, we analyzed the impact of highway transportation infrastructure on carbon emissions through the baseline regression model results. Second, we examined the effect of highway transportation infrastructure on the mediating variable by using the mediating variable as the dependent variable and highway transportation infrastructure as the independent variable. Finally, we conducted a comprehensive analysis by including the mediating variable, using carbon emissions as the dependent variable and highway transportation infrastructure as the independent variable.

$$ln{M}_{it}={\beta }_0+{\beta }_{1}ln{road}_{it}+\gamma {control}_{it}+{\mu }_{it}+{\mathrm{\lambda }}_{it}+{\epsilon }_{it}$$

(2)

$$ln{cei}_{it}={\pi }_0+{\pi }_{1}ln{road}_{it}+{\pi }_2{ln{road}_{it}}^2+{\pi }_{3}ln{M}_{it}+\gamma {control}_{it}+{\mu }_{it}+{\mathrm{\lambda }}_{it}+{\epsilon }_{it}$$

(3)

Equations (1)–(3) together form the mediation testing equation. First, we tested the regression coefficients ${\alpha }_1$ and ${\alpha }_2$ of Equation (1). The model will continue to be tested according to the existence of intermediary effects if the coefficients are significant. Secondly, we tested the regression coefficients ${\beta }_1$ of Equation (2) and ${\pi }_3$ of Equation (3) in turn; if they are significant, it indicates that the indirect effect is significant. Finally, we tested whether the coefficients ${\pi }_1$ and ${\pi }_2$ of Equation (3) are significant. If they are not significant, meaning the direct effect is not significant, it indicates only the presence of the mediating effect, known as a complete mediating effect; if the direct effect is significant, it is a partial mediating effect.

4.3.3. Threshold Effect Model

Considering the nonlinear impact of highway transportation infrastructure on carbon emissions, a threshold effect model based on Equation (4) was constructed to detect threshold effects.

$$\begin{array}{c}ln{cei}_{it}={{\theta }_0+\theta }_1\ast ln{road}_{it}I\left(c_{it}\le {\rho }_1\right)+{\theta }_{2}ln{road}_{it}\ast I\left({\rho }_1<c_{it}\le {\rho }_2\right)+\\ {\theta }_{3}ln{road}_{it}\ast I\left(c_{it}>{\rho }_2\right)+\gamma {control}_{it}+{\epsilon }_{it}\end{array}$$

(4)

In Equation (4), I(*) represents the index function; ${\rho }_1$ and ${\rho }_2$ are the threshold values; and ${\theta }_1$, ${\theta }_2$, and ${\theta }_3$ represent the threshold regression coefficient levels in different intervals.

5. Empirical Results and Analysis

5.1. Benchmark Regression Results and Analysis

Table 3. Results of the benchmark model.

	(1)	(2)	(3)	(4)
	lncei	lncei	lncei	lncei
lnroad	−0.0231	1.034 *	0.0556	2.119 ***
	(−0.43)	(1.83)	(0.79)	(3.22)
lnroad2		−0.0625 *		−0.118 ***
		(−1.88)		(−3.15)
lnpop			0.295 **	0.439 ***
			(2.58)	(3.59)
lngdpp			−0.105 *	−0.0478
			(−1.75)	(−0.77)
lngov			0.0898 **	0.0965 **
			(2.23)	(2.41)
lndesh			0.0359 *	0.0424 **
			(1.92)	(2.28)
_cons	1.044 **	−3.388	2.048 **	−7.133 **
	(2.17)	(−1.40)	(2.21)	(−2.34)
control	no	no	yes	yes
City/Year	yes	yes	yes	yes
N	574	574	574	574
$R^2$	0.704	0.706	0.713	0.719

Note: *** p value < 0.01, ** p value < 0.05, * p value < 0.1.

presents the regression results of highway transportation infrastructure on carbon emissions. A lower p-value indicates higher credibility, with significance levels typically set at 5% (p = 0.05) or 1% (p = 0.01). Column (1) shows that without the inclusion of control variables, the coefficient of lnroad was not significant, indicating that there was no significant linear relationship between highway transportation infrastructure and carbon emissions. Column (2) includes lnroad² (the squared term of highway transportation infrastructure) to examine whether there was a nonlinear relationship between highway transportation infrastructure and carbon emissions. Without the inclusion of control variables, the results show that the coefficient of lnroad² was significantly −0.0625 at the 10% significance level, and the coefficient of lnroad was significantly 1.034 at the 10% significance level. This indicates that there is indeed an inverted-U-shaped nonlinear relationship between highway transportation infrastructure and carbon emissions. Column (3) examines the linear relationship between highway transportation infrastructure and carbon emissions by adding control variables, and the results show that the coefficient of lnroad was not significant, indicating that there is no significant linear relationship between highway transportation infrastructure and carbon emissions. Column (4) examines the nonlinear relationship between highway transportation infrastructure and carbon emissions by adding control variables and lnroad². The results show that the coefficient of the quadratic term of highway transportation infrastructure was significantly −0.118 at the 1% significance level, and the coefficient of the linear term of highway transportation infrastructure was significantly 2.119 at the 1% significance level. This indicates that there is indeed an inverted-U-shaped nonlinear relationship between highway transportation infrastructure and carbon emissions.

On the one hand, in the early stages of highway transportation infrastructure development, the construction and maintenance of highways consumed a large amount of energy, thereby increasing carbon emissions. On the other hand, as China places increasing emphasis on green transportation, innovations in construction techniques during the building and maintenance of highways have led to significant emission reduction effects. By optimizing energy types, improving heating technology, and enhancing production processes, China has achieved emission reductions of 17.89%, 17.87%, and 35%, respectively [40], significantly reducing carbon emissions. Therefore, hypothesis 1 is verified.

5.2. The Mediating Effects

Columns (1) to (3) of

Table 4. Results of the mediating effects model.

	(1)	(2)	(3)
	lncei	lnfdi	lncei
lnroad	2.119 ***	0.810 ***	1.869 ***
	(3.22)	(3.89)	(2.84)
lnroad2	−0.118 ***		−0.103 ***
	(−3.15)		(−2.75)
lnpop	0.439 ***	0.0491	0.391 ***
	(3.59)	(0.14)	(3.20)
lngdpp	−0.0478	1.156 ***	−0.0676
	(−0.77)	(6.50)	(−1.05)
lngov	0.0965 **	0.738 ***	0.0713 *
	(2.41)	(6.19)	(1.72)
lndesh	0.0424 **	0.154 ***	0.0437 **
	(2.28)	(2.78)	(2.34)
lnfdi			0.323 ***
			(3.12)
lnfdi2			−0.0142 ***
			(−3.01)
_cons	−7.133 **	−8.130 ***	−7.757 **
	(−2.34)	(−2.97)	(−2.54)
control	yes	yes	yes
City/Year	yes	yes	yes
N	574	574	574
$R^2$	0.719	0.455	0.724

Note: *** p value < 0.01, ** p value < 0.05, * p value < 0.1.

present the estimated results of each stage of the three-step mediation effect test. The first step of the mediation effect model examines the overall impact of highway transportation infrastructure on carbon emissions. The results in Column (1) are consistent with those in Column (4) of Table 3, indicating that there is indeed an inverted-U-shaped nonlinear relationship between highway transportation infrastructure and carbon emission. The second step tests whether highway transportation infrastructure affects foreign direct investment (FDI). The results in Column (2) show that the coefficient of the impact of highway transportation infrastructure on FDI was 0.810, and the regression results passed the 1% significance level test, indicating that highway transportation infrastructure has a significant promoting effect on FDI. The third step examines the impact of highway transportation infrastructure on carbon emission when considering the mediating variable (FDI). Column (3) shows that the coefficient of lnfdi² (the squared term of FDI) was −0.0142 at the 1% significance level, and the coefficient of lnfdi was 0.323 at the 1% significance level. This indicates that FDI in the Yangtze River Delta region has an inverted-U-shaped impact on carbon emissions. At this point, the coefficient of lnroad² was −0.103 at the 1% significance level, and the coefficient of lnroad was 1.869 at the 1% significance level, indicating that FDI plays a significant partial mediating role in the impact of highway transportation infrastructure on carbon emissions. The results in Columns (1) to (3) show that FDI is an important pathway through which highway transportation infrastructure has an inverted-U-shaped impact on carbon emissions. Since highway transportation infrastructure can significantly reduce transportation costs, foreign investors are often willing to invest in regions with well-developed highway infrastructure. Highway transportation infrastructure attracts FDI. However, the impact of FDI on carbon emissions is nonlinear. In the early stages, to develop the economy, environmental regulations were relatively lax, attracting high-pollution and high-energy-consuming FDI, which led to an increase in carbon emissions. As the economy developed and the environment deteriorated, the government began to pursue green FDI. At this stage, FDI brought in advanced and environmentally friendly technologies, which helped reduce carbon emissions. This leads to the conclusion that highway transportation infrastructure exerts a partial inverted-U-shaped impact on carbon emissions by enhancing FDI. Therefore, hypothesis 2 is verified.

5.3. The Threshold Effect

Table 5. Threshold effect test.

Threshold Variable	Threshold Number	Fstat	Prob	10%	5%	1%
lngi	Single	52.21	0.0080	25.7893	31.2447	50.9980
	Double	40.37	0.0040	22.7504	25.8009	37.2322
	Triple	15.73	0.4420	42.7912	53.2198	71.0604
lnis	Single	38.95	0.0260	26.8037	31.4796	47.9842
	Double	34.11	0.1040	35.3657	53.8338	86.6399

reports the threshold test results for green technology innovation and industrial structure upgrading. The p-value for the triple-threshold effect test of green technology innovation was 0.4420, which did not pass the 10% significance level test, and thus, the null hypothesis of at least three thresholds in the threshold model should be rejected. The p-value for the double-threshold effect test was 0.0040, which passed the 1% significance level test, indicating that there is a double-threshold effect based on green technology innovation in the impact of highway transportation infrastructure on carbon emissions. The p-value for the double-threshold effect test of industrial structure upgrading was 0.1040, which did not pass the 10% significance level test, and thus, the null hypothesis of at least two thresholds in the threshold model should be rejected. The p-value for the single-threshold effect test was 0.0260, which passed the 5% significance level test, indicating that there is a single-threshold effect based on industrial structure upgrading in the impact of highway transportation infrastructure on carbon emissions.

Table 6. Estimation of the threshold value.

Threshold Variable	Threshold Number	Threshold
lngi	Single	6.7044
	Double	8.6822
lnis	Single	4.1043

shows the threshold values for industrial structure upgrading and green technology innovation. The industrial structure upgrading had a single threshold value of 4.1043, while the green technology innovation had two threshold values of 6.7044 and 8.6822.

Table 7. Results of the threshold effect model.

	(1)	(2)
lncei	lngi	lnis
lnpop	0.222 **	0.251 **
	(1.98)	(2.15)
lngdpp	−0.480 ***	−0.541 ***
	(−16.24)	(−18.48)
lndesh	0.0223	0.0250 *
	(1.64)	(1.76)
lngov	0.0396	0.0494
	(1.03)	(1.24)
lnroad (lngi ≤ 6.7044)	−0.215 ***
	(−3.57)
lnroad (6.7044 < lngi ≤ 8.6822)	−0.233 ***
	(−3.88)
lnroad (8.6822 < lngi)	−0.268 ***
	(−4.46)
lnroad (lnis ≤ 4.1043)		−0.239 ***
		(−3.80)
lnroad (lnis > 4.1043)		−0.281 ***
		(−4.47)
_cons	8.023 ***	8.906 ***
	(18.25)	(20.49)
control	yes	yes
City/Year	yes	yes
N	574	574
$R^2$	0.712	0.686

Note: *** p value < 0.01, ** p value < 0.05, * p value < 0.1.

presents the regression results of the threshold effect model. Column (1) shows that when the level of green technology innovation did not exceed the first threshold (lngi ≤ 6.7044), the coefficient of highway transportation infrastructure was significantly negative at the 1% level, with a value of −0.215. When the level of green technology innovation was between the first and second thresholds (6.7044 < lngi ≤ 8.6822), the negative impact of highway transportation infrastructure on carbon emissions increased to −0.233, significant at the 1% level. When the level of green technology innovation surpassed the second threshold (8.6822 < lngi), the negative impact further increased to −0.268 and remained significant at the 1% level. Column (2) indicates that when the level of industrial structure upgrading was low (lnis ≤ 4.1043), highway transportation infrastructure had a significantly negative coefficient of −0.239 at the 1% level, meaning that the development of highway transportation infrastructure significantly reduced carbon emissions. However, when the level of industrial structure upgrading was high (lnis > 4.1043), the coefficient of highway transportation infrastructure was −0.281 at the 1% significance level, indicating a stronger inhibiting effect on carbon emissions. The results indicate that due to the improvement in green technology innovation and industrial structure upgrading, highway transportation infrastructure has a nonlinear impact on carbon emission reduction.

Overall, as the levels of green technology innovation and industrial structure upgrading increase, the inhibiting effect of highway transportation infrastructure on carbon emissions becomes stronger, resulting in a more pronounced reduction in carbon emissions. With the development of green technology innovation and industrial structure upgrading, on one hand, the carbon emissions generated during the construction and transportation processes of highway infrastructure are increasingly reduced. On the other hand, through the technology diffusion effect of highway infrastructure, green technology innovation and industrial structural upgrading can better mitigate local carbon emissions. This results in highway transportation infrastructure having a more significant impact on reducing carbon emissions. This validates hypotheses 3 and 4.

5.4. Heterogeneity Analysis

Benchmark regression confirmed the inverted-U-shaped relationship between highway transportation infrastructure and carbon emissions. However, the Yangtze River Delta region, encompassing Shanghai, Jiangsu Province, Zhejiang Province, and Anhui Province, comprises 41 cities with significant differences in economic development and urban scale. To investigate whether the relationship between highway transportation infrastructure and carbon emissions varies across different regions due to heterogeneous impacts of economic development levels and urban population sizes, this study divided the sample into subsets based on two perspectives: economic development and urban scale. Using the average per capita GDP as the threshold, cities were categorized into developed and developing categories. Similarly, using the average total urban population as the threshold, cities were categorized into large and small categories. Separate subsample regressions were conducted on the benchmark regression model to explore whether there is heterogeneity in the relationship between highway transportation infrastructure and carbon emissions across regions with different economic development levels and urban scales.

Based on

Table 8. Heterogeneity test results of economic development.

	(1) Developed	(2) Developing
	lncei	lncei
lnroad	3.595 ***	−1.654
	(4.69)	(−0.88)
lnroad2	−0.207 ***	0.0864
	(−4.64)	(0.85)
lnpop	0.369 **	0.00909
	(2.41)	(0.03)
lngdpp	0.0912	−0.473 ***
	(1.01)	(−4.39)
lngov	0.138	0.0210
	(1.24)	(0.46)
lndesh	0.0684	0.0630 ***
	(1.57)	(3)
_cons	−14.92 ***	13.20
	(−4.14)	(1.43)
control	yes	yes
City/Year	yes	yes
N	266	308
$R^2$	0.788	0.695

Note: *** p value < 0.01, ** p value < 0.05, * p value < 0.1.

, in developed cities, lnroad² was significantly negative at the 1% significance level with a coefficient of −0.207, and lnroad was significantly positive at the 1% significance level with a coefficient of 3.595. This indicates an inverted U relationship between highway transportation infrastructure and carbon emissions in developed cities. However, in developing cities, neither lnroad² nor lnroad showed significant coefficients, suggesting that there is no significant relationship between highway transportation infrastructure and carbon emissions in developing cities.

According to

Table 9. Heterogeneity test results of urban scale.

	(1) Large	(2) Small
	lncei	lncei
lnroad	12.28 **	2.030 **
	(2.26)	(2.44)
lnroad2	−0.663 **	−0.118 **
	(−2.27)	(−2.53)
lnpop	−0.153	0.529 ***
	(−0.60)	(3.53)
lngdpp	−0.0259	−0.139
	(−0.40)	(−1.13)
lngov	−0.123 **	0.108 *
	(−1.99)	(1.88)
lndesh	−0.0757 *	0.0550 **
	(−1.96)	(2.32)
_cons	−56.06 **	−5.043
	(−2.19)	(−1.17)
control	yes	yes
City/Year	yes	yes
N	252	322
$R^2$	0.834	0.658

Note: *** p value < 0.01, ** p value < 0.05, * p value < 0.1.

, regardless of whether in large or small cities, lnroad² was significantly negative at the 5% significance level, while lnroad was significantly positive at the 5% significance level. This indicates an inverted-U-shaped relationship between road infrastructure and carbon emissions in both large and small cities.

5.5. Robustness Test

To further study the impact of highway transportation infrastructure on carbon emissions, this paper employed methods of replacing the dependent variable and the independent variable to address causality issues. Specifically, per capita carbon emissions (lnpco²) replaced carbon emission intensity (lncei) as the dependent variable, while road area (lnroada) was used instead of the length of highways (lnroad) as the independent variable. These alternative approaches were incorporated into the baseline regression model, resulting in the outcomes presented in columns (1) and (2) of

Table 10. Robustness test.

	(1)	(2)
	Lnpco2	Lncei
lnroad	1.771 ***
	(2.60)
lnroad2	−0.0953 **
	(−2.47)
lnpop	0.541 ***	0.313 ***
	(4.27)	(3.37)
lngdpp	0.391 ***	−0.193 ***
	(6.05)	(−3.22)
lngov	−0.0184	0.0340
	(−0.44)	(0.84)
lndesh	0.0538 ***	0.0449 **
	(2.79)	(2.47)
lnroada		1.077 ***
		(5.05)
lnroada2		−0.0672 ***
		(−4.50)
_cons	0.677	−0.601
	(0.21)	(−0.65)
control	yes	yes
City/Year	yes	yes
N	574	574
$R^2$	0.548	0.728

Note: *** p value < 0.01, ** p value < 0.05, * p value < 0.1.

.

The regression results in Column (1) show that lnroad² was significantly negative at the 5% level (−0.0953), while lnroad was significantly positive at the 1% level (1.771). This indicates that even after substituting the dependent variable, there indeed was a U-shaped nonlinear relationship between highway transportation infrastructure and carbon emissions.

The regression results in column (2) show that lnroada² (the squared term of highway transportation infrastructure) was significantly negative at the 1% significance level with a coefficient of −0.0672, while lnroada was significantly positive at the 1% significance level with a coefficient of 1.077.

In conclusion, whether replacing the dependent or independent variables, the robustness analysis aligned with previous research findings, indicating that there indeed exists a nonlinear inverted-U-shaped relationship between highway transportation infrastructure and carbon emissions.

6. Conclusions and Policy Suggestions

6.1. Conclusions

China has set dual carbon goals of “carbon peak” and “carbon neutrality”, making it crucial to study the impact of highway transportation infrastructure on carbon emissions to achieve these targets. Based on panel data from 41 cities in the Yangtze River Delta region from 2006 to 2019, this paper empirically analyzed the relationship between highway transportation infrastructure and carbon emissions and their influencing mechanisms. The main conclusions of this study are as follows:

First, there was an inverted-U-shaped nonlinear relationship between highway transportation infrastructure and carbon emissions.

Second, highway transportation infrastructure had a partial mediating effect on carbon emissions through foreign direct investment.

Third, as green technology innovation and industrial structure upgrades improved, the impact of highway transportation infrastructure on carbon emissions exhibited a nonlinear relationship.

Finally, there was an inverted-U-shaped relationship between highway transportation infrastructure and carbon emissions in developed cities, whereas no such relationship existed in developing cities. Regardless of city size, there was an inverted-U-shaped relationship between highway transportation infrastructure and carbon emissions.

6.2. Policy Suggestions

Based on the above research conclusions, the following policy suggestions are proposed to better achieve sustainable development and carbon reduction goals in the Yangtze River Delta region:

First, strengthen investment and construction in highway transportation infrastructure. In the early stages of its development, highway transportation infrastructure may lead to increased carbon emissions. However, as it reaches a certain scale, its further development can result in reduced carbon emissions. Increase investment in green transportation infrastructure and use low-carbon transportation infrastructure construction technologies to ensure that energy conservation and emission reduction are considered from the start. Implement policies to encourage the purchase and use of electric vehicles and other clean energy vehicles by offering purchase subsidies and tax incentives. Implement policies to encourage the purchase and use of electric vehicles and other clean energy vehicles, such as providing purchase subsidies and tax reductions. By promoting the use of clean energy vehicles, transportation-related carbon emissions can be effectively reduced.

Second, enhance efforts to attract foreign direct investment (FDI). FDI plays a critical mediating role in the impact of highway transportation infrastructure on carbon emissions. Therefore, under appropriate policy guidance, efforts should be made to attract FDI and emphasize the “pollution halo” effect of FDI. Offer tax incentives to foreign companies investing in environmental protection, new energy, and other green industries, encouraging them to invest in these areas. By introducing advanced low-carbon technologies and green industries through FDI, economic and environmental sustainability can be promoted.

Third, strengthen investment in green technology research and development and promote industrial structure upgrading. As the level of green technology innovation and industrial structure upgrading improves, the inhibitory effect of highway transportation infrastructure on carbon emissions will become stronger. First, by encouraging green technology innovation and developing green low-carbon technologies, the region’s energy efficiency can be improved, helping to reduce carbon emissions. The government should encourage companies and research institutions to increase investment in green technology R&D by providing research subsidies and special funds. Second, actively promoting industrial structure upgrading is an important way to reduce carbon emissions. The government should offer tax incentives and financial subsidies to attract the development of high-tech, information technology, and financial sectors, thereby reducing the dependency of traditional industries on natural resources and fossil fuels. These measures will help achieve a low-carbon economic transformation and contribute to sustainable development.

Fourth, develop highway transportation infrastructure according to the actual conditions of each city. The relationship between highway transportation infrastructure and carbon emissions may vary depending on local socio-economic conditions. In developed cities, highway transportation infrastructure should be vigorously developed to quickly realize its carbon reduction effects. In developing cities, the development of highway transportation infrastructure should be approached cautiously based on local conditions.

The main limitations of this study are as follows: Firstly, despite China’s vast land area and 293 prefecture-level cities, this research focuses only on the 41 prefecture-level cities in the Yangtze River Delta region due to limitations in data availability. Secondly, the study solely examines the impact mechanisms of foreign direct investment, green technology innovation, and industrial structure upgrading, leaving other potential impact mechanisms awaiting exploration. Regarding future research directions, as more extensive and updated data become available, future researchers can investigate whether the conclusions of this study hold true on a broader scale and attempt to integrate additional variables into the models to comprehensively explore more impact mechanisms.