Study on the Spatial Characteristics and Spillover Effects of Carbon Emissions in the Yangtze River (Main Stream) Basin

Abstract

Carbon emissions reduction is crucial to global climate governance and sustainable development. By 2060, China envisioned being carbon-neutral, and it has adopted a series of policies and measures for environmental management, especially in the main stream of Yangtze River basin, where China’s carbon emissions are centered. The spatial distribution characteristics and agglomeration effects of carbon dioxide (CO₂) emissions in the main stream of Yangtze River basin are analyzed from 2010 to 2019 based on the perspective of local (city and state) administrative regions, and uses the spatial Durbin model to examine the influencing factors and spatial spillover effects of carbon emissions. The findings discovered from the extensive research are as follows: First, carbon emissions in the main stream of Yangtze River basin present a fluctuating upward trend, and CO₂ emissions in the lower reaches are significantly higher than those in the middle and upper reaches, which are closely related to the economic volume. Secondly, carbon emissions have a significant positive spatial correlation among prefecture-level cities, and carbon emissions show a high-high concentration in downstream regions and low-low concentration in upstream regions. Thirdly, regional economic development level, secondary industry development level, and population density have considerable influence on CO₂ emissions, among which the Kuznets hypothesis is evidenced by the interaction between economic progress and carbon emissions. Therefore, strengthening regional cooperation efforts and collaborating to promote low-carbon development are the vital ways to achieve carbon emissions reduction.

1. Introduction

Large-scale utilization of fossil energy sources has greatly promoted economic development; meanwhile, it has caused a dramatic increase in global carbon dioxide emissions, which accounts for most greenhouse gases. CO₂, as the main component of greenhouse gases, increases the Earth’s temperature by absorbing long-wave radiation reflected by the Earth, and then causes the melting of permafrost and the continuous rise in sea level and other environmental deteriorations, with unimaginable consequences. Xie et al. believed that it is a common phenomenon that carbon emissions increase along with economic growth. Reports published by various institutions can confirm this view (Xie et al.) [1]. Data from China’s National Bureau of Statistics (NBS) showed that China’s GDP has grown rapidly from 54 trillion yuan in 2012 to 114 trillion yuan in 2021, and according to China Carbon Accounting Database (ceads) data, China’s total carbon emissions (million tons) grew from 9858.99 in 2012 to 14,093 in 2019. In response to this challenge, President Xi announced that “China will strive to peak CO₂ emmisions and achieve carbon neutrality at 2030 and 2060 by forceful policies” at the UNGA’s 75th session (2020). The “double carbon” target reflects the “common area principle” and the principle based on the development stage in addressing climate change, and also demonstrates the positive attitude of a responsible country in addressing climate change. In order to achieve this goal, China has introduced and implemented a series of measures, such as non-fossil energy development and utilization, low-carbon transformation, dual substitution, and forest cities. The main stream of Yangtze River basin plays a leading role in China’s high-quality development. Its winds form east to west and across 9 provinces, 2 municipalities, and 74 prefectures, many of which are developed areas. In accordance with the China Statistical Yearbook, the main stream of Yangtze River basin will have 606 million people overall in 2020, which is 42.92% of China’s entire population, and a yearly GDP of USD 47.15 trillion, which is 46.42% of China’s entire GDP. Therefore, the main stream of Yangtze River basin is the key for China to fulfill the double carbon target as scheduled.

From the existing literature, CO₂ emissions as a kind of greenhouse gas brought about by development have significant clustering characteristics and spatial spillover effects (Tong X et al.) [2]. Especially, China, as the second-largest economy in the world, is also a large CO₂-emitting country, and a number of studies have found that there is a positive spatial correlation between CO₂ emissions in various provinces of China (Sun L et al.) [3] and meaningful regional spillover effects of economic growth and proportion of tertiary production on CO₂ emissions (Song M et al.) [4]. However, previous scholars mainly studied and analyzed the effects of economic growth, urbanization rate, energy terminal efficiency, and energy investment level, on CO₂ emissions, revealing some of the patterns, while the conclusions also differed due to the differences in research objects and different focus and methods. At present, most studies on CO₂ emissions in China are conducted at the provincial level, while there are few studies on the Yangtze River basin with prefecture-level cities as the samples. However, the GDP of the Yangtze River basin accounts for nearly half of China’s total, so the CO₂ emissions in the main stream of Yangtze River basin is the focus of China’s CO₂ emissions. The entire Yangtze River basin is used as the research issue in this article, which analyzes the spatial correlation and influencing factors of the Yangtze River basin at the local (city and state) level from a global perspective. This analysis can serve as a theoretical foundation and decision-making guide for the government as it develops and enhances its CO₂ reduction policy, further advancing China’s “double carbon” goal.

The article’s marginal contribution largely consists of the following: first, due to the lack of literature on the spatial characteristics of CO₂ emissions in the main stream of Yangtze River basin and its relationship with economic growth, this study examines the spatial heterogeneity and trend of CO₂ emissions in the main stream of Yangtze River basin from 2010 to 2019 and describes the close relationship between CO₂ emissions and the level of economic development, thus filling a research gap. Second, it comes to the conclusion that CO₂ emissions in the main stream of Yangtze River basin exhibit significant spatial autocorrelation and that the effect of economic development on carbon dioxide emissions is shown by the Kuznets curve.

The rest of this paper is organized as follows: Section 2 contains the literature review and research hypotheses; Section 3 contains the research design; Section 4 contains the empirical results and analysis; Section 5 contains further discussion; Section 6 contains the conclusion and policy recommendations.

2. Literature Review and Research Hypothesis

2.1. Characteristics of CO₂ Emissions’ Spatial Distribution

Research on the characteristics of CO₂ emissions’ spatial distribution is conducive to understanding the distribution differences of CO₂ emissions in different spaces, which is crucial for the formulation and implementation of macro-CO₂ emissions reduction policies (Du et al.) [5]. Previous research has revealed that there are significant regional clustering patterns for global CO₂ emissions, the type of clustering has a close relationship with geographical location and income level, and the stages of CO₂ emission that are different in the countries that have disparate income levels and CO₂ emissions of the country can be directly influenced by its own economy and the economies of neighboring countries (Wang K et al.) [6]. China is one of the largest CO₂ emitters, and in recent years, a large number of academics have investigated the spatial spillover effect of China’s CO₂ emissions. The study found that the spatial correlation between CO₂ emissions in Chinese provinces is positive, which is mainly manifested as low-low (L-L) and high-high (H-H) patterns. It proved that the L-L aggregation pattern is mainly in the western, less developed provinces, the H-H aggregation pattern is in the eastern developed provinces, and the central region shifts mainly show low-high or high-low patterns (Sun L et al.) [3]. Inner Mongolia, Shanxi, Sichuan, Shandong, Henan, and Shaanxi CO₂ emissions show H-H clustering characteristics, and Qinghai, Gansu, and Ningxia CO₂ emissions show L-L clustering characteristics (Gong W et al.) [7]. The analysis of CO₂ emission data from 2000 to 2015, Liang S confirmed that the overall CO₂ emission intensity in China demonstrated a fluctuating decreasing trend, provinces with low CO₂ intensity became more prevalent over time, and the clustering characteristics developed toward the L-L model with improving spatial differences (Liang S et al.) [8]. Meanwhile, the distribution pattern of CO₂ emission efficiency has the tight junction with distribution pattern of the state of neighboring provinces, and provinces that are surrounded by higher state provinces are more likely to shift upward, whereas provinces that are boxed in by lower-state provinces are less likely to shift upward. All the above studies proved that CO₂ emissions are spatially dependent, and local correlations exhibit spatial clustering characteristics and significant spillover effects (Tong X et al.) [2]. Accordingly, Hypothesis 1 is put forward:

Hypothesis 1.

CO₂ emissions in the main stream of Yangtze River Basin of China have the characteristics of spatial agglomeration, with L-L and H-H accumulation characteristics in the upstream and downstream regions, respectively.

2.2. Factors Affecting the Scale of CO₂ Emissions

While analyzing the spatial distribution characteristics of CO₂ emissions, previous scholars have also explored the impact of disparate influencing factors on CO₂ emissions, which can be summarized as follows: first, the relation between economic progress and CO₂ emissions. Previous scholars have found that China’s economic growth is positively correlated with CO₂ emissions, and among the many factors affecting CO₂ emissions, indicators such as density of population, growth in the economy, fossil energy proportion, and the proportion of three industries have direct or indirect effects on CO₂ emissions (Lu et al.) [9]; therefore, GDP and the primary energy supply required by industries may be the cause of CO₂ emissions (Ming et al.) [10]. Some scholars studied the role of economic growth on CO₂ emissions based on 181 countries, which shows that 12% of the countries have clear evidence to support the environmental Kuznets curve hypothesis. A total of 27% of the countries have future income growth that will reduce CO₂ emissions (Narayan et al.) [11]. It has also been found that there is an inverted U-shaped relationship between economic growth and environmental quality, with economic growth bringing about an initial phase of environmental deterioration, which later turns into a phase of improvement (Grossman et al.) [12]. There is a significant spatial correlation between CO₂ emissions and economic growth, and the relationship between CO₂ emissions and economic growth in China has an inverted U-shaped curve development trend, which is basically consistent with the Kuznets curve hypothesis (Xu H et al.) [13]. The second is the impact of industrial restructuring on CO₂ emissions. Some scholars based on Chinese data found that improving industrial production factors and the level of industry both have a negative impact on energy-dependent industrial structure and, thus, indirectly reduce CO₂ emissions, so efficient energy use and an accelerated improvement of industrial structures are key measures to reduce CO₂ emissions in energy-rich regions (Wu et al.) [14]; modernization of industrial structures can be an efficient technique to cut local CO₂ emissions and has the potential to make a strong impact on CO₂ emissions. Because of the aspiring modernization process, the impact of industrial modernization on CO₂ reduction will diminish slightly, but the impact on CO₂ reduction will remain significant (Li Z and Zhou Q) [15]. The third is other factors affecting CO₂ emissions. Some scholars have found in their studies that population size and residential consumption all have positive effects on CO₂ emissions, especially in China; residential consumption accounts for the largest share of CO₂ emissions from consumption, and this growth rate is accelerating, and the pulling effect of consumption structure on CO₂ emissions is more obvious. In addition, the income effect contributes most to the growth of household CO₂ indirect emissions (Kong et al.) [16]; (Cao et al.) [17]), and the production structure only increases indirect CO₂ emissions slightly (Wu L) [18]. On these grounds, this paper proposes Hypothesis 2:

Hypothesis 2.

There is an inverse U-shaped relationship between the level of regional economic development and CO₂ emissions, with secondary sector development having a tending effect on CO₂ emissions and a significant regional spillover effect between the two.

2.3. Spatial Spillover Effect of CO₂ Emissions

In today’s economic globalization, are CO₂ emissions an isolated problem as a greenhouse gas? Do CO₂ emissions in one area have an impact on the surrounding area? Research reveals that CO₂ emissions are not only regionally dependent, but also have spill-over effects on neighboring countries ((Ragoubi H and Mighri Z [19]);(Mahmood) [20]). Researchers’ views on the regional spillover effects of CO₂ emissions can be roughly divided into two categories. One category is the positive effect, i.e., economic development in one region significantly increases the CO₂ emissions of neighboring regions. The most significant spillover effect is the spillover effect of industrial development (Song M et al.) [4], and the spillover effect of CO₂ emissions is more obvious because industrial relocation in developed regions deepens the regional link between each province’s economy and CO₂ emissions (Li et al.) [21]. Some scholars also found that, among the factors influencing the spatial spillover effect of CO₂ emissions, economic factors are an important factor of positive influence; each 1% increase is accompanied by a 0.38–0.43% rise in CO₂ emissions in the surrounding areas (Lv et al.) [22], and industrialization also has a remarkable positive impact on the CO₂ intensity of neighboring countries (Ding et al.) [23]. In addition, with changes in the size of final demand, the structure of final costs and the size of exports in most regions have the same positive regional spillover effect on CO₂ emissions growth in other regions (Meng et al.) [24]. The other type of effect is negative, i.e., the decision and development path of one region reduces CO₂ emissions of neighboring regions. As the Chinese economy enters a new normal, the concept, mode, and pace of development are changing; many regions are actively exploring the greening and cleaning of energy use; a large number of low-carbon-development cities have emerged to actively implement low-carbon-development strategies, which effectively reduce industrial emissions, promote technological innovation, and optimize energy use efficiency, and the low-carbon policies of these cities will show a negative impact on neighboring cities (Ren et al.) [25]. In addition, some scholars found that capital-saving technological changes have a markable negative impact on local and neighborhood CO₂ intensity by examining the relationship between technological innovation and CO₂ emissions, and capital-saving technological change in low-CO₂ cities has a stronger suppression of local CO₂ intensity than in non-low-CO₂-emitting cities ((Li et al.) [26]; (Zhang H and Haiqian K. [27])). Therefore, this paper proposes Hypothesis 3:

Hypothesis 3.

Carbon emissions in the Yangtze River (main stream) basin of China have the characteristics of positive spatial spillover effect.

3. Research Design

3.1. Model

To test the regional impact of CO₂ emissions in the Yangtze River basin, the following econometric model is constructed:

Spatial Autoregressive model

$$lnC{O}_{2it}={\alpha }_0+\rho {{\displaystyle \sum }}_{i\ne j}{w}_{ij}\times lnC{O}_{2it}+{\beta }_{1}contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}$$

(1)

In addition, the more comprehensive spatial Durbin model is used as the baseline model for the empirical study, which is set-up as follows:

$$lnC{O}_{2it}={\alpha }_0+\rho {\displaystyle \sum }_{i\ne j}{w}_{ij}\times lnC{O}_{2it}+{\beta }_{1}contro{l}_{it}+{\beta }_2{\displaystyle \sum }_{i\ne j}{w}_{ij}\times contro{l}_{it}+{\mu }_i+{\eta }_t+{\epsilon }_{it}$$

(2)

where $lnC{O}_{2it}$ represents the CO₂ emissions of the ith region in year t; $w_{ij}$ represents an N × N symmetric spatial weight matrix, and the inverse distance matrix is used in this paper; $w_{ij}\times lnC{O}_{2it}$ represents the spatial lag term of CO₂ emissions, i.e., the weighted average of CO₂ emissions of neighboring regions; $contro{l}_{it}$ represents a set of control variables; ${\mu }_i$ represents the province fixed effect; ${\eta }_t$ represents the year fixed effect; ${\epsilon }_{it}$ represents the residual term. $\rho$ represents the spatial autocorrelation coefficient, if it is significantly greater than 0, it indicates that there is a positive spatial autocorrelation effect of carbon emissions, i.e., carbon emissions in neighboring areas have a significant positive effect on local carbon emissions, and vice versa; ${\beta }_1, {\beta }_2$ is the coefficient to be estimated, if ${\beta }_1$ significantly greater than 0, it indicates that the control variables have a positive effect on local carbon emissions, and vice versa, it has a negative effect; if ${\beta }_2$ significantly greater than 0, it indicates that the relevant The control variables have positive effect on local carbon emission, and vice versa, they have negative effect.

$lnC{O}_{2it}$——the CO₂ emissions of the ith region in year t

$w_{ij}$——an N × N symmetric spatial weight matrix, and the inverse distance matrix is used in this paper

$w_{ij}\times lnC{O}_{2it}$—— the spatial lag term of CO₂ emissions, i.e., the weighted average of CO₂ emissions of neighboring regions

$contro{l}_{it}$——a set of control variables

${\mu }_i$——the city fixed effect

${\eta }_t$——the year fixed effect

${\epsilon }_{it}$——the residual term

$\rho$——the spatial autocorrelation coefficient (if $\rho >$0, it indicates that there is a positive spatial autocorrelation effect of carbon emissions, i.e., carbon emissions in neighboring areas have a significant positive effect on local carbon emissions, and vice versa).

${\beta }_1,{\beta }_2$——the coefficient to be estimated (if ${\beta }_1>$ 0, it indicates that the control variables have a positive effect on local carbon emissions, and vice versa, it has a negative effect; if ${\beta }_2<$ 0, it indicates that the relevant The control variables have positive effect on local carbon emission, and vice versa, they have negative effect).

3.2. Variables and Data Description

3.2.1. Study Area

The study area is delineated according to the natural catchment area defined by the Yangtze River Protection Commission under the Ministry of Water Resources, and the integrity of each administrative region along the main stream of Yangtze River, which spans nine provinces and two municipalities directly under the central government from Qinghai to Shanghai. Weighing the scientific nature of the data study and the relevance of subsequent policies, 74 sample sites at the prefecture (city and state) level were selected for this paper. The list of sample cities is shown in

Table 1. Division of administrative districts in Yangtze River (main stream).

Province (11)	City (74)
Qinghai	Yushu Tibetan Autonomous Prefecture, Haixi Mongol and Tibetan Autonomous Prefecture, Golog Tibetan Autonomous Prefecture (3)
Sichuan	Panzhihua, Liangshan Yi Autonomous Prefecture, Ganzi Tibetan Autonomous Prefecture, Ngawa Tibetan and Qiang Autonomous Prefecture, Ya’an, Leshan, Meishan, Chengdu, Deyang, Mianyang, Guangyuan, Bazhong, Nanchong, Guangan, Lushi, Yibin, Ziyang, Suining, Zigong, Neijiang (20)
Tibet	Changdu (1)
Yunnan	Diqing Tibetan Autonomous Prefecture, Lijiang, Dali bai Autonomous Prefecture, Chuxiong Yi Autonomous Prefecture, Zhaotong, Kunming, Qujing (7)
Chongqing	Chongqing (1)
Hubei	Enshi, Yichang, Jingzhou, Wuhan, Huanggang, Ezhou, Huangshi, Xiaogan, Jingmen, Xiangyang, Shiyan, Xianning, Suining (13)
Hunan	Huaihua, Zhangjiajie, Changde, Yiyang, Yueyang, Changsha, Xiangtan, Zhuzhou, Hengyang, Yongzhou (10)
Jiangxi	Jiujiang, Nanchang, Xinyu, Ji’an, Ganzhou (5)
Anhui	Anqing, Chizhou, Tongling, Wuhu, Ma’anshan (5)
Jiangsu	Nanjing, Zhenjiang, Yangzhou, Taizhou, Changzhou, Wuxi, Suzhou, Nantong (8)
Shanghai	Shanghai (1)

: Table 1 contains a list of the 11 provinces and 74 prefecture-level cities involved in the Yangtze River main stream basin.

3.2.2. Variable Specification

Dependent variable: The dependent variable in this paper is the absolute amount of carbon dioxide emissions, which is treated logarithmically to reduce the influence of heteroscedasticity.

Independent variable: Referring to the study by Zhao, GDP share of secondary industry, retail sales of consumer goods per capita, and population density are selected as explanatory variables to study their impact on CO₂ (Zhao et al.) [28]. In order to reduce heteroscedasticity, logarithmic processing is carried out on per capita GDP.

3.2.3. Data Sources and Variable Descriptions

In this paper, CO₂ emissions are used as the dependent variables, and with reference to existing research results, gross regional product, the share of secondary industry in gross regional product, total retail sales of social consumer goods per capita, and population density are used as explanatory variables. The data are mainly from the US Department of Energy’s Carbon Data Analysis Center, China Urban Statistical Yearbook, and regional statistical yearbooks. Individual missing data are mainly filled in by the interpolation method. Variable descriptions information is shown in

Table 2. Variable description and descriptive statistics.

Variable Type	Variable Name	Variable Meaning	Mean	Standard Deviation	Minimum	Median	Maximum	Number of Obs.
Dependent Variable	lnCO2	Carbon dioxide emissions (logged $\times$ 1000)	2507.95	1227.53	−1602.33	2533.47	5635.26	730
Independent Variable	lngdp	gross domestic product (logged)	16.46	1.18	11.79	16.40	19.76	730
	lngdp2	The logarithm of gross domestic product squared	272.39	38.51	138.89	269.08	390.45	730
	sipg	The ratio of the secondary industry in the gross regional product	48.54	9.67	14.74	48.48	86.44	730
	conspc	Total retail sales of consumer goods per capita	19,007.56	16,947.35	0.39	13,469.15	110,000.00	730
	popd	Density of population	0.04	0.04	0.00	0.04	0.23	730

:

4. Empirical Results and Analysis

4.1. Spatial Distribution Characteristics of Carbon Emissions in Yangtze River Basin

4.1.1. Time-Series Characteristics and Basin Differences of CO₂ Emissions in the Yangtze River Basin

Figure 1 shows the mean value and coefficient of variation of CO₂ emissions in the main stream of Yangtze River basin from 2010 to 2019. From the figure, it can be seen that the mean value of CO₂ emissions in the Yangtze River basin basically shows a fluctuating upward trend, among which there is an obvious upward trend from 2010 to 2013, a decline from 2014 to 2016, and then an upward trend; the reason for this phenomenon may be related to China’s economic development policy and macro context. China has experienced rapid economic growth from the late 1970s, creating a miracle of sustained high economic growth in the world. Along with the rapid economic development and industrialization comes an increase in CO₂ emissions, especially in the main stream of Yangtze River basin, which accounts for a large part of China’s economy, so the average value of CO₂ emissions showed a significant upward trend from 2010 to 2013. However, as China’s economy entered the new normal, the momentum of the early rapid economic growth slowed down, the economic structure faced transformation and upgrading, the Chinese government began to put more emphasis on environmental protection, and CO₂ emissions declined from 2013 to 2016. However, after 2016, the average value of CO₂ emissions in the Yangtze River basin showed an upward trend again, which may be related to China’s economic stimulus policies and the growth of residents’ income. In response to the pressure of slowing economic growth during the economic transition period, the Chinese government introduced a series of policies to stimulate economic development, including preferential policies on vehicle purchase tax, a move that greatly stimulated a significant increase in vehicle sales. According to the CAAM, China’s vehicle sales increased by 13.65% year-on-year in 2016, and the increase in vehicle sales directly contributed to the increase in CO₂ emissions. In addition to this, carbon emissions have also increased to a certain extent with the significant rise in Chinese residents’ income, higher consumption levels, and increased demand. The coefficient of variation of CO₂ emissions in the main stream Yangtze River basin basically shows a smooth moving trend and reached a peak of 1.6485 in 2016, indicating a relatively smooth dispersion of CO₂ emissions within there between 2010 and 2019.

Figure 2 shows the average CO₂ emissions upstream, intermediate, and downstream of the production chain. The figure shows that CO₂ emissions in the Yangtze River basin from 2010 to 2019 are in descending order in the downstream, midstream, and upstream regions. According to the average value, the CO₂ emissions from the lower Yangtze River region account for 67.30%, which is more than half of the total emissions, while the overall difference between the middle and upper reaches is not conspicuous, accounting for 17.45% and 15.24%, respectively. On average, the average CO₂ emissions growth in the upper and middle basins in 2019 compared to 2010 is 2.59 million tons, 2.96 million tons, and 11.41 million tons, respectively, with the largest growth in CO₂ emissions in the lower Yangtze River region, and the larger the economic volume, the higher the CO₂ emissions. The difference is reflected as related to the level of economic development of each region, and form the data of China City Statistical Yearbook. The total GDP of the Yangtze River (main stream) basin in 2020 is 30.9 trillion yuan, while the total GDP of the lower Yangtze River region reaches 14.23 trillion yuan, accounting for 46.04% of the whole basin, which inevitably corresponds to the high CO₂ emissions brought by the rapid economic development.

4.1.2. Spatial Correlation of CO₂ Emissions in Yangtze River Basin

Spatial autocorrelation measures the degree of spatial clustering of variables; therefore, in this paper, the global Moran index of CO₂ emissions in municipal areas of the main stream of Yangtze River basin in China is calculated using the inverse distance matrix to measure spatial autocorrelation. Equation (3) is the formula for the Moran index I, where ${\mathrm{S}}^2=\frac{{{\displaystyle \sum }}_{\mathrm{n}-1}^{\mathrm{n}}{({\mathrm{x}}_{\mathrm{i}}-\overline{\mathrm{x}})}^2}{\mathrm{n}}$ is the sample variance, $w_{ij}$ is the spatial weight matrix, and ${\displaystyle \sum }_{\mathrm{n}=1}^{\mathrm{n}}{\displaystyle \sum }_{\mathrm{j}=1}^{\mathrm{n}}$ is the sum of all spatial weights.

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

(3)

Usually, −1 < I < 1. If I > 0, it indicates H-H,L-L clustering characteristics (the positive correlation; if Moran’s index I→0, it exhibits a random distribution characteristic in space and there is no spatial correlation; if it <0, it indicates the L-H,H-L connection (negative correlation). To increase robustness, this paper uses a bivariate test and the computational results are shown in Figure 3. As can be seen from Figure 3, the global Moreland index of CO₂ emissions in the Yangtze River basin in 2019 was 0.33, which was remarkable at the 1% level, implying a very notable positive spatial correlation of CO₂ emissions, i.e., in the Yangtze River basin (main river), CO₂ emissions show a clustering of H-H and L-L. The main reason for this regional concentration is that the eastern provinces and cities of the lower Yangtze River, such as Jiangsu, Zhejiang, and Shanghai, which are early areas of reform and opening-up, tend to be more closely linked economically and have highly developed industries due to the long-standing cooperation and division of labor and, therefore, have high levels of concentration. The upper reaches of the Yangtze River are mostly primary industries with a weak industrial base and tend to have lower carbon emissions than the national average, so their concentration is low. The above conclusion proves Hypothesis 1.

4.1.3. Spatial Agglomeration Feature Analysis

Figure 4 further presents the results of the hot spot analysis of CO₂ emissions in the main stream of Yangtze River basin in 2019. From Figure 4, we can see that the CO₂ emissions in the main stream of Yangtze River basin show obvious spatial clustering characteristics, with the hot spot areas mainly concentrated in the downstream areas and the cold spot areas mainly concentrated in the upstream areas, which indicate that the CO₂ emissions in the downstream areas are higher and more concentrated, while the CO₂ emissions in the upstream areas are lower. The spatial clustering characteristics of CO₂ emissions again indicate that CO₂ emissions in the Yangtze River basin show a remarkable positive correlation during the study period. This is consistent with the previous conclusion, again testing Hypothesis 1.

4.2. Empirical Results

4.2.1. Basic Regression Results

First, to test the spatial impact of CO₂ emissions in the main stream of Yangtze River, Equations (1) and (2) were estimated, and the outcomes are presented in

Table 3. Spatial model estimation results.

	(1)	(2)
	lnCO2	lnCO2
lngdp	2.49 *	2.46 *
	(1.88)	(1.83)
lngdp2	−0.10 **	−0.10 **
	(−2.07)	(−2.11)
Sipg	0.02 *	0.01
	(1.85)	(0.71)
conspc	0.00	0.00 **
	(1.44)	(2.20)
Popd	17.09	17.52
	(1.13)	(1.11)
W × lngdp		32.99 ***
		(3.17)
W × lngdp2		−1.40 ***
		(−3.80)
W × sipg		0.09 **
		(2.01)
W × conspc		0.00 ***
		(2.79)
W × popd		248.59 *
		(1.96)
rho	0.72 ***	0.64 ***
	(10.65)	(7.91)
sigma2_e	0.55 ***	0.53 ***
	(18.85)	(18.96)
N	730	730
r2_w	0.60	0.71
ll	−825.04	−810.05
City Fix	Yes	Yes
Year Fix	Yes	Yes

Note: The t statistic is in parentheses, where *, **, and *** are significant at the level of 10%, 5%, and 1%, respectively.

.

According to column (1) of the table above, the spatial autocorrelation coefficient is 0.72 and significant at the 1% level, illustrating that there is a significant positive spatial correlation of carbon emissions. The coefficient of the squared gross regional product (lngdp2) is −0.10 and significant at the 5% level, indicating an inverted U-shaped relationship between regional economic development level and CO₂ emissions, a finding that is consistent with the Kuznets curve for the environment, i.e., with economic development, the level of pollution first increases and then decreases. The coefficient for the share of secondary industry (sipg) is 0.02 and is significant at the 10% level, indicating a significant positive relationship between the share of secondary industry in the region and the region’s CO₂ emissions, i.e., the larger the share of secondary industry in a region and the larger the energy consumption, the higher the region’s CO₂ emissions. In contrast, neither the total number of retail sales of consumer goods per capita (conspc) nor the population density (popd) has a significant effect on CO₂ emissions, confirming the validity of hypothesis 2.

After taking into consideration the spatial terms of the explanatory variables, the spatial autocorrelation coefficient is 0.64 based on the regression results present in column (2), which is also significant at the 1% level, once again verifying the existence of a significant positive spatial correlation of CO₂. The coefficient of squared GDP of neighboring regions (W × lngdp2) is −1.40 and significant at the 1% level, verifying that the economic development of neighboring regions and the impact of neighboring regions on CO₂ emissions (lnCO₂) also show an inverted U-shaped curve, i.e., neighboring regions’ economic development and per capita income have first a positive and then a negative impact on CO₂ emissions in the region, which may be due to the fact that, based on the theory of “voting with one’s feet”, the neighboring regions competed to improve the economic development level at the initial stage in order to enhance the competitiveness of the region, and the CO₂ emission level increased accordingly. After a certain level of economic development, the increasing environmental pollution problems force local governments to promote environmental protection and change their development philosophy. The spatial term of the proportion of secondary industry’s (W × sipg) coefficient is 0.09 and significant at the 5% level, which implies that there is a significant positive correlation between the secondary industry in the neighboring regions and the CO₂ emissions in the region, that is, secondary industry‘s proportion in the neighboring regions not only affects the CO₂ emissions in the region, but also has a significant positive spillover effect on the region. This phenomenon may be due to the fact that the development of secondary industry requires the continuous production and output of energy in the neighboring regions, and the industrial agglomeration effect and the mobility of air pollutants pull the CO₂ emissions of the neighboring regions. Per capita consumption (W × conspc) and population density (W × popd) in neighboring regions are significant at the 1% and 10% levels, respectively, indicating that there is also a significant spatial spillover effect of consumption level and population density in neighboring regions on CO₂ emissions (lnCO₂), and with the continuous expansion of population size and per capita consumption level, it will greatly increase the demand for commercial products such as real estate, cars and electrical appliances. This production and consumption activity will not be limited to the local economy, but will also “spread” to the production and consumption in the surrounding areas, which will lead to a large increase in energy and CO₂ emissions. This conclusion verifies hypothesis 3.

4.2.2. Direct and Indirect Effects

Furthermore, we calculated the direct and indirect effects of the model, and the specific funding is reflected in

Table 4. (Direct/Indirect) Effects of spatial models.

	lnCO2
	SAR Model			SDM Model
	Direct Effects	Indirect Effects	Total Effects	Direct Effects	Indirect Effects	Total Effects
lngdp	2.66 *	7.16	9.82	4.04 ***	98.70 ***	102.74 ***
	(1.86)	(1.38)	(1.55)	(2.60)	(2.73)	(2.79)
lngdp2	−0.10 **	−0.28	−0.38 *	−0.17 ***	−4.18 ***	−4.35 ***
	(−2.06)	(−1.47)	(−1.67)	(−2.98)	(−2.99)	(−3.05)
sipg	0.02**	0.05	0.07 *	0.01	0.26 *	0.27**
	(2.04)	(1.45)	(1.65)	(1.26)	(1.93)	(1.99)
conspc	0.00	0.00	0.00	0.00 ***	0.00**	0.00**
	(1.51)	(1.26)	(1.36)	(2.89)	(2.46)	(2.54)
popd	17.81	46.88	64.69	28.89	732.06 *	760.95 *
	(1.17)	(1.01)	(1.07)	(1.58)	(1.84)	(1.86)
N	730	730	730	730	730	730
r2_w	0.60	0.60	0.60	0.71	0.71	0.71
ll	−825.04	−825.04	−825.04	−810.05	−810.05	−810.05

Note: The t statistic is in parentheses, where *, **, and *** are significant at the level of 10%, 5%, and 1%, respectively.

.

In Table 4, the SAR model regression consequences reflect that the direct-effect coefficient of the squared regional gross product (lngdp2) is −0.10 and significant at the 5% level, demonstrating that the economic level of the region has an inverse U-shaped relationship with CO₂ emissions. The coefficient for the regional secondary industry share (sipg) is 0.02 and significant at the 5% level, indicating that the regional secondary industry share is positively related to the region’s CO₂ emissions. The indirect effects of regional gross product squared (lngdp2) and regional secondary industry share (sipg) both pass the significance test, again demonstrating that there are regional spillover effects of economic level and industrial development on CO₂ emissions.

The regression results of the SDM model show that both the direct and indirect effects of regional gross domestic product (lngdp2) pass the significance test of 1%, which is consistent with the conclusion of the SAR model. The indirect effect of the secondary industry share (sipg) is significant at the 10% level, indicating that the proportion of secondary industry in neighboring regions has a significant and active impact on the region’s CO₂ emissions. Both the direct and indirect effect of the total volume of retail sales of social consumer goods (conspc) per capita passed the significance at the 1% level, demonstrating that it has a direct pull and positive impact on CO₂ emissions in neighboring areas. Population density (popd) also has a positive multiplier effect on CO₂ emissions in neighboring areas. This conclusion confirms Hypothesis 3 again.

5. Further Discussion

5.1. Robustness Test

5.1.1. Replace the Space Weight Matrix

In the previous analysis, the inverse distance matrix is adopted by us to calculate the spatial weight of each prefecture level city, and regression analysis is also carried out. However, there are many calculation methods of the spatial weight matrix and different spatial weight matrices grant different weight values to adjacent areas, which may cause changes in the conclusion. In order to improve the robustness of the conclusion, we use pairs of adjacency matrices (1) and (2) between prefecture-level cities simultaneously in a robustness test. From the regression results in

Table 5. Robustness test—Change in the spatial weight matrix.

	(1)	(2)
	lnCO2	lnCO2
lngdp	1.49	2.24
	(0.58)	(0.84)
lngdp2	−0.05	−0.07
	(−0.67)	(−0.93)
sipg	0.01	0.00
	(1.24)	(0.01)
conspc	0.00	0.00
	(0.74)	(1.21)
popd	17.43	12.00
	(1.12)	(0.75)
W × lngdp		0.15
		(0.04)
W × lngdp2		−0.05
		(−0.41)
W × sipg		0.05 **
		(2.37)
W × conspc		−0.00
		(−0.55)
W × popd		55.04 *
		(1.85)
rho	0.25 ***	0.23 ***
	(5.50)	(5.13)
sigma2_e	0.55 ***	0.54 ***
	(17.61)	(17.63)
N	630	630
r2_w	0.25	0.59
ll	−710.62	−704.80
City Fix	Yes	Yes
Year Fix	Yes	Yes

Note: The t statistic is in parentheses, where *, **, and *** are significant at the level of 10%, 5%, and 1%, respectively.

, we can discover that the spatial autocorrelation coefficient is still positively significant. Meanwhile, the coefficients of other explanatory variables are no longer statistically significant, which may have something to do with the fact that the adjacency matrix is less accurate than the inverse distance matrix in measuring interregional correlation.

5.1.2. Change the Explained Variable

In the previous analysis, we use the logarithm of regional CO₂ emissions as the explanatory variable for regression. However, the method to measure regional CO₂ emissions intensity in absolute terms for analysis may be affected by many factors, including the level of economic development and the industrialization process of the region. Although we have controlled the regional GDP and the proportion of the secondary industry in the explanatory variables, and excluded the interference of these factors to some extent, in order to go deeper into the analysis from the perspective of the relative scale of CO₂ emissions at the same time, and improve the robustness of the previous conclusions, we choose the logarithm of CO₂ dioxide emissions of the industrial enterprises above the unit size as the explanatory variable for re-regression analysis, trying to explore whether the relative scale of CO₂ emissions can still have significant positive spatial correlation (see

Table 6. Robustness test—replace the dependent variable.

	(1)	(2)	(3)	(4)
	lnCO2_1	lnCO2_1	lnCO2_1	lnCO2_1
lngdp	1.54 ***	1.53 ***	−0.92 **	−0.92 **
	(5.33)	(5.13)	(−2.34)	(−2.27)
lngdp2	−0.06 ***	−0.06 ***	0.00	0.00
	(−6.11)	(−5.94)	(0.17)	(0.15)
sipg	−0.00	−0.00	0.00	0.00
	(−1.46)	(−1.34)	(0.30)	(0.97)
conspc	0.00 ***	0.00 ***	0.00 ***	0.00 ***
	(6.49)	(6.38)	(3.74)	(2.88)
popd	1.15	1.24	0.52	1.64
	(0.35)	(0.35)	(0.22)	(0.67)
W × lngdp		−0.59		0.18
		(−0.26)		(0.28)
W × lngdp2		−0.01		−0.01
		(−0.12)		(−0.37)
W × sipg		0.00		−0.00
		(0.10)		(−1.38)
W × conspc		−0.00		0.00 *
		(−0.30)		(1.67)
W × popd		16.87		−7.58 *
		(0.60)		(−1.66)
rho	0.55 ***	0.52 ***	0.31 ***	0.29 ***
	(5.35)	(4.81)	(8.30)	(7.22)
sigma2_e	0.03 ***	0.03 ***	0.01 ***	0.01 ***
	(18.92)	(18.93)	(17.58)	(17.59)
N	730	730	630	630
r2_w	0.08	0.07	0.18	0.18
ll	292.55	294.07	470.82	475.37
City Fix	Yes	Yes	Yes	Yes
Year Fix	Yes	Yes	Yes	Yes

Note: The t statistic is in parentheses, where *, **, and *** are significant at the level of 10%, 5%, and 1%, respectively.

for regression results). In Table 6, Columns (1) to (2) present the regression results where the inverse distance matrix is the spatial weight matrix, columns (3) to (4) present the regression results where the neighborhood matrix is the spatial weight matrix, columns (1) and (3) present the regression results of the SAR model, and columns (2) and (4) present the regression results of the SDM model. The regression results show that the spatial autocorrelation coefficient for CO₂ is still positively significant at the 1% level, which again confirms the positive spatial correlation of CO₂ emissions.

In

Table 7. The direct/indirect effects of robustness test.

	lnCO2
	Inverse Distance Matrix			Adjacency Matrix
	DE	IndE	TE	DE	IndE	TE
lngdp	1.55 ***	0.59	2.13	−0.91 **	−0.06	−0.97
	(4.87)	(0.12)	(0.43)	(−2.15)	(−0.08)	(−1.00)
lngdp2	−0.06 ***	−0.09	−0.16	0.00	−0.01	−0.01
	(−5.67)	(−0.53)	(−0.87)	(0.05)	(−0.41)	(−0.32)
sipg	−0.00	−0.00	−0.00	0.00	−0.01	−0.00
	(−1.32)	(−0.10)	(−0.22)	(0.95)	(−1.42)	(−0.86)
conspc	0.00 ***	0.00	0.00	0.00 ***	0.00 **	0.00 ***
	(6.45)	(0.31)	(0.77)	(3.31)	(2.43)	(3.51)
popd	1.77	37.63	39.40	0.96	−9.22	−8.26
	(0.47)	(0.59)	(0.60)	(0.39)	(−1.50)	(−1.13)
N	730	730	730	630	630	630
r2_w	0.07	0.07	0.07	0.18	0.18	0.18
ll	294.07	294.07	294.07	475.37	475.37	475.37

Note: The t statistic is in parentheses, where **, and *** are significant at the level of 5%, and 1%, respectively.

, both the direct and indirect effects of the above SDM model are calculated. It proves that the economic development level still has a significant impact on CO₂ emissions, with no very clear spatial spillover effect, while the direct effect of the per capita total retail sales of consumer goods is still significant at the level of 1%. When using the adjacency matrix, the indirect effect of the per capita total retail sales of consumer goods is also positively significant, and basically consistent with the previous conclusions.

5.2. Dynamic Effect Test

Further, in order to explore the dynamic effect on the time of CO₂ emissions, we created a system GMM estimation model, and the specific model form is as follows:

$$dlnC{O}_2{}_{it}=\alpha 0+{\beta }_{1}dlnC{O}_2{}_{i,t-1}+{\beta }_{2}dlnC{O}_2{}_{i,t-2}+{\beta }_{3}Contro{l}_{it}+{\epsilon }_{it}$$

(4)

Among them, the first-order difference representing the logarithm value of CO₂ emissions stands for a group of control variables, as well as the residual item. The results based on model (3) are displayed in

Table 8. Dynamic effect test.

	dlnCO2
L.dlnCO2	0.66 ***
	(14.26)
L2.dlnCO2	−0.00
	(−0.01)
lngdp	183.56 ***
	(2.76)
lngdp2	−5.66 ***
	(−2.79)
sipg	−0.97 ***
	(−4.74)
conspc	0.00 ***
	(3.55)
popd	−261.91 *
	(−1.72)
_cons	−1439.38 ***
	(−2.63)
AR(1) in first differences	P > z = 0.000
AR(2) in first differences	P > z = 0.112
Hansen test	Prob > chi2 = 0.107
N	511

Note: The t statistic is in parentheses, where * and *** are significant at the level of 10% and 1%, respectively.

.

In Table 8, the consequences represent that the first-order lag of CO₂ emissions has a remarkably positive impact, illustrating that the carbon emissions of the previous year have a positive impact on the CO₂ emissions of this year, but the second-order lag is no longer significant, indicating that this dynamic effect has only been maintained for one year. This dynamic effect of CO₂ emissions proves that the regions not only have spatial agglomeration in CO₂ emissions, but also have continuity in time, with the performance of a path dependence. As a result, the dual effects of space and time lead to the growth trend of CO₂ emissions.

6. Conclusions and Policy Recommendations

With the continuous development and reform of economy, CO₂ emissions reduction has become one of the crucial factors that the Chinese government needs to consider when making decisions. By studying the sample city in the Yangtze River (main stream) basin, this paper draws the following four important conclusions: First, the overall CO₂ emissions between 2010 and 2019 in the Yangtze River (main stream) basin present a fluctuating upward trend, while the growth rate is low, and the overall CO₂ emissions have not reached the peak. Second, the spatial agglomeration of CO₂ emissions in the main stream of Yangtze River basin is magnificent, the relative “high” and “low” agglomeration areas are formed in different basins, and there is a significant positive spatial spillover effect, which proves that the CO₂ emissions reduction should break through the inherent administrative division. The administrative division of CO₂ emissions reduction has to be considered in an integrated manner to formulate and implement CO₂ emissions reduction policies. Third, the impact of economic development on CO₂ emissions conforms to the Kuznets curve, that is, economic development in this region will inhibit CO₂ emissions after reaching the inflection point. Fourth, the secondary industry will remarkably increase the emission of CO₂, and the spatial spillover effect is apparent. The secondary industry has the largest demand and consumption of energy among three major industries, which generates the most CO₂ emissions, so the larger the proportion of secondary industry, the larger the CO₂ emissions, and the spatial spillover effect is significant, indicating that the optimization, upgrading, and reasonable layout of secondary industry are not only important for CO₂ emissions reduction in the region, but also equally important for neighboring regions.

The innovations of this paper are mentioned in the introduction part. From the existing essays, the breakthroughs of this paper are as follows: First, the literature samples used in this paper are totally different from the existing ones. Previous studies on CO₂ emissions in the Yangtze River basin mainly took provincial panel data as samples ((Wang et al.) [29]; (Zhao X and Zhang X) [30]), and the panel data of prefecture-level cities are used for analysis. The second point is to propel the existing literature analysis processes. In this paper, the connection between economic growth and CO₂ emissions in the main stream of Yangtze River basin is analyzed with the nonlinear method, while most previous scholars used a linear method ((Tang D et al.) [31]; (Tang Z et al.) [32]).

Based on the above conclusions, the following policy recommendations are introduced: First, from the spatial characteristics of CO₂ emissions in the Yangtze River (main stream) basin, we should actively carry out the concept of green, low-carbon, and sustainable development, reduce the dependence on high-carbon energy sources in the economic development of the basin, and improve the efficiency of resource utilization, while continuing to maintain the stability of the regional economy. Second, we should strengthen the supervision of backward production capacity and promote industrial energy to transform and upgrade. On the one hand, for the transfer of CO₂ emissions between regions and the transfer of high-energy-consumption industries that cannot produce positive economic growth effects, relevant national policies and regulations should be strictly enforced and backward production capacity should be eliminated in a timely manner. On the other hand, we should actively carry out the innovation of clean technologies, reduce the dependence of industries on traditional energy resources, and introduce the policies to promote the benign flow of energy. Third, we should strengthen regional cooperation efforts to promote low-carbon development in a concerted manner. As a greenhouse gas, CO₂ has a significant spatial spillover effect, and no region can do it alone, so promoting CO₂ emissions reduction is in progress, and we should break through the administrative boundaries and achieve cross-regional cooperation and shared governance. At the same time, the principle of reciprocity of rights and responsibilities should be respected, and the CO₂ emissions reduction tasks of each area should not be simply divided, but also should take into account the economic development levels of different regions, so that each region can truly integrate into the low-carbon sustainable development strategy of synergy.