The Analysis of Carbon Emission’s Characteristics and Dynamic Evolution Based on the Strategy of Unbalanced Regional Economic Development in China

Abstract

Analyzing the evolution law of carbon emissions is particularly important for the designation of policies on energy conservation and emission reduction. Based on the regional division of China, this paper uses a spatial panel model to find the causes of the differences in carbon emission, and the non-parametric model, logarithmic mean Divisia index (LMDI) model and the extended STIRPAT model to analyze the relevant influencing factors in detail. From the studies in this paper, there come the following conclusions: (1) The environmental Kuznets curve (EKC) in the eastern region resembles the national EKC, demonstrating the same “N” pattern. However, the “upside-down U” pattern in the middle and western regions not only confirms the assumption of EKC in some Chinese regions but also demonstrates the effective restraint in high energy consumption and high emission levels when narrowing down the gaps between the central and western regions and the eastern regions. (2) In addition, good education can effectively suppress the increase in carbon emission, and every 1% increase in the proportion of educated people (college and above) results in emission reduction, respectively, by 0.22%, 0.51% and 0.44% in the eastern, central and western regions of China. (3) Significantly, the effect of tertiary industry structure on carbon emissions is positive, reflecting the trend of “deterioration” of China’s industrial structure over long time scales. This study functions positively in understanding the evolutionary pattern of regional carbon emissions and proposing differentiated policies on emission reduction.

1. Introduction

Climate change has long been a significant international concern [1,2]. The Paris Agreement requires all parties to put forward their best efforts to prevent the temperature increase to 1.5 degrees Celsius above the temperature level before the Industrial Revolution [3,4]. The 23rd session of the Conference of the Parties (COP 23) to the U.N. Convention on Climate Change (UNFCCC) was held in Bonn Germany, pointing out in its report that China is “on its way to complete its 2030 reduction goals ahead of schedule”.

As the world’s largest carbon emitter and the second largest economy [5], China is still in the middle of its urbanization and industrialization, facing tremendous international pressures from public opinions, and is required to take its responsibilities and obligations through reducing carbon emissions [6]. China has increased its goals of reducing carbon emission per unit value of GDP in 2030 by 60–65% over the same period in 2005 [7], and pledged at the United Nation’s Climate Conference to strive to reach “peak carbon” by 2030 and achieve “carbon neutrality” by 2060, which not only marked a new normal for economic development in China but also fully reflected its determination to build a community of human destiny as a reliable power [8].

After the Reform and Opening up, China, based on some geographical, historical and political considerations and the fundamental principles of market economy, worked out and practiced its three-level unbalanced development plan, prioritizing the economic development in eastern coastal provinces. In 2000, China launched its Western Development Program, considering the historical and enlarging gaps in economic levels between eastern and western regions. After that, the Rise of Central China Program was proposed in 2004 to make the regional economic development coordinated, through which China established its well-known eastern, central and western development strategies. For a long time, the unbalanced development in China has become a fundamental reality, resulting in an increased economic gap that held China back from its social goals of being overall well-off. To solve this problem, coordinated development, a reality-driven solution, is proposed for the unbalanced and unsustainable development. In the Government Work Report 2018, the improvements of both the Yangtze River Economic Zone and the “Belt and Road” Initiative are intended to coordinate the eastern, central and western development and narrow down the urban–rural economy gap. However, the unbalanced development in China will still be a problem in the long run, reflected in the relations among urban–rural development, economic development, social development, and technology level. If China does not pay sufficient attention to its overall development, the buckets effect will soon appear and grow and a series of social conflicts and environmental problems will aggravate.

Though still in the middle of urbanization and industrialization, China faces extreme international pressure and increasing calls for taking liability for emission reduction. Therefore, the unbalanced development strategy has revealed significant differences in economic level, industrial structure and natural resources in China, a country with vast territory. To maintain stable and fair economic growth, the Chinese government not only focused on scientific emission reduction but also ensured the environmental carrying capacity is not broken. As a result, this paper aims to solve the following issues: (1) In order to study the carbon emission differences in China through the latest ten years, is it reasonable to divide China into such three regions as eastern, central and western by referring to the Chinese regional development strategy? (2) What are the primary causes for the differences in regional emission levels, and what is the collective evolution like under other influencing factors?

Specifically, this paper comprises the following sections: Section 2 Literature Review, Section 3 Models and Data, and Section 4 Three-Level Regional Division and Causes of Carbon Emission Difference. Finally, the conclusions.

2. Literature Review

It is found in the systematic literature that most of the literature on carbon emission differences are focused on four levels, namely, international, national, provincial and local. (1) At the international level, many research achievements have been realized by scholars of international emission measurement, emission quota trading and allocation, and by impacts of related policies on carbon emission [9,10]. Den Elzen et al. [11] studied the goals and costs of developing countries in emission reduction based on their 2020 emission goals. Cutlip and Fath [12] compared the per capita carbon emission in the past 200 years and the accumulated emission level by taking six countries as examples. Wang et al. [13] studied the urbanization in BRICS countries on their carbon emissions. Li et al. [14] decomposed and studied the influential factors on carbon emission in the power generation sectors in 11 countries between 1990 and 2013. Dai et al. [15] studied the impacts of the U.S. withdrawal from the Paris Agreement on carbon emission space and emission costs in the E.U., Japan and China. (2) On a national level, more and more researchers have set their eyes on the changes in overall carbon emission at a selected time sequence and the impacts on carbon emission by a single or several factors. Pan et al. [16] established a multi-region input/output model according to the input/output data sheet and the newly regional division in China. After carefully considering the economic development strategies, they studied the relationship between economic growth and carbon emission. Wu et al. [17] mainly studied and analyzed the correlation between economic output and carbon emission between 2005 and 2015 in the Chinese construction industry. Zhang et al. [18] discussed the impacts of GDP, trade structure, foreign exchange rates and foreign investments on Chinese carbon emissions between 1982 and 2016. (3) At the provincial (State) level, the focuses of kinds of research in the existing literature have shifted to the description of the cross-regional difference in carbon emission. Jorgenson et al. [19] used the Gini coefficient model to study the correlation among carbon emission in U.S. states and two other income indicators between 1997 and 2012. Cohen et al. [20] analyzed the relations between the gross domestic product (GDP) and carbon emission, believing developed provinces tended to have smaller Kuznets elasticity than less developed provinces. (4) At local levels, studies must consider the cultural and economic differences. Increasingly, researchers focus on in-depth studies on a single region (economic zone, province or city) to come up with well-targeted policies on energy conservation and emission reduction. Li et al. [21] appraised the performance of the carbon emission authority in Beijing and carried out a structural analysis from the aspects of both demands and supplies. Long and Yoshida [22] studied the direct and consequential carbon emission caused by the Tokyo Energy Shift by analyzing the multi-regional input/output sheet.

Through the systematic studies of the related literature on carbon emission in general consideration of the study methods sorted by Zhou et al. [23], this paper classified the methods into four groups: (1) Based on EKC theories, the relationship between the environmental degradation and per capita income is simulated so as to study the routes of carbon emission increase. Narayan et al. [24] established a new correlation analysis test on the assumption of Kuznets to study the correlation between economic growth and carbon emission in 181 countries. Ulucak and Bilgili [25] classified 45 countries into three groups by income levels and utilized EKC to study the relationship between ecological footprint and the high/middle/low-income groups. (2) Based on the decomposition of emission impact factors for qualitative and quantitative analysis, such as the structural decomposition analysis (SDA) and index decomposition analysis (IDA) [23,26]. Ang [27] took the lead in solving the problem of null values in data processing using the LMDI method, in light of a comprehensive comparison of several IDA methods, by which the LMDI method can be extensively used in various applications. To study the influential factors on China’s industrial carbon emission, Lin and Long [28] used the LMDI model. The Tapio decoupling model and the LMDI decomposition model were used by Wang et al. [13] to compare and analyze carbon emissions in China and the U.S. between 2000 and 2014. (3) According to a panel model and spatial panel model, the correlation between carbon emissions and various influential factors is studied. Jiang et al. [29] used measuring models in their studies to study the impacts of natural resources and concluded that fossil energy production in a region significantly impacted the local carbon emission efficiency. Li et al. [30] used the spatial lag model in the studies to study the impacts of urbanization in the Yangtze River Delta between 2000 and 2010 on the carbon emission efficiency. Cai et al. [31] clarified the spatial and industrial difference in the space for carbon emission in the Beijing-Tianjin-Hebei zone after clearly defining boundaries of cities by internationally recognized methods. (4) A series of mathematical models have been established to present changes of any single factor in the system or multiple economies that may have impacts on the overall economy or other economies [32,33]. Mao et al. [34] used the S.D. model in the studies of GDP and emission growth between 2008 and 2020 in three scenarios so as to provide a valuable basis for optimizing local industrial structure. Lin and Jia [35] used the CGE model to study the impacts of six plans of emission quota allocation on energy, economy and environment. Xiao et al. [36] used the CGE model in analyzing the impacts of energy efficiency, energy structure and economic structure on carbon emission.

It may be easily found through careful studies of the mentioned literature that in regional emission studies, especially at the geographical level, carbon emission is usually zoned by provinces [1,37]. However, on considering many historical, geographical and cultural factors, Chinese provinces are geographically established. Some scholars classified Chinese provinces into several groups with the methods of k-means clustering, system clustering, met frontier DEA clustering, and fuzzy PSO clustering [38,39,40,41]. In order to demonstrate the necessity of zoning, more researchers began to take economic zones, national development strategies and regional development strategies as the fundamental basis for their zoning plans. However, only a few researchers studied the scientificity and reasonableness of such zoning strategies.

Under this condition, the three-level development strategy in China was taken as the study basis for regional emission differences. In addition to the arguments on the scientificity and reasonableness of the zoning strategy, this paper uses the spatial panel model to study the direct and consequential impacts of GDP, industrial structure, and energy structure on the carbon emission and carries out in-depth studies on the dynamic evolution of various factors and the carbon emission in various regions. The possible contributions are: (1) By building a coordinated evaluation system for energy, economy and environment, a prerequisite for scientific division of regions and research is provided. (2) Theoretical support is provided for the government to formulate differentiated energy conservation and emission reduction policies through the difference analysis on influential factors of regional carbon emission. (3) Based on in-depth and systematic research on the influential factors of carbon emissions, the theoretical achievements in this field is further enriched.

3. Model and Data

3.1. Carbon Emission Measurement

In this paper, the end-user consumption method is used to calculate carbon emission volume, meaning the end-user consumption as indicated in the energy balance sheet is used to calculate the carbon emission levels while the carbon emission in the form of energy loss is ignored in the processing and transformation to avoid double counting. IPCC (2006) proposed a specific method for calculation; however, the carbon volume produced in the combustion of fossil fuels may be obtained by estimation and summary of carbon volume produced in energetic consumption, with the calculation formula to be expressed as:

(1)$C{E}^e=\frac{44}{12}\times \left({\displaystyle \sum _i{\displaystyle \sum _j{A}_{ij}\times {\alpha }_j\times e_j\times {\beta }_j\times o}}\right)$

where, subscript i represents the industrial sectors, such as agriculture, manufacturing and construction industries; subscript j represents various types of energy, such as coking coal, crude oil and natural gas; A_ij the volume of fossil energy j consumed by industry i, and α the TCE conversion coefficient of j fossil energy; e_j the net calorific value of energy j; β_j the carbon emission coefficient of j fossil energy; o the energy oxygenation efficiency. As a result, it is assumed in this paper that all types of energy can be fully oxidized. The actual calculation process includes the carbon emission in power generation CE^e and heat production CE^k.

3.2. The Coordination Evaluation and Assessment Model of Energy, Economy and Environment (3E)

There are underlying interactions among energy, economic development and environmental protection, and these primary principles and connections are studied with a coordinated Energy–Economy–Environment (3E) development framework. The “coordinated development” of the 3E system aims to stress the positive low–high, simple–complex and unordered–ordered evolution inside the subsystems or between the inherent elements of the 3E system. The degree of coordination is a good indicator between various system factors. The function of coordination may be expressed as:

(2)$M=f(E_e,E_c,E_v)$

where, M represents the degree of coordination of 3E; E_e, E_c, and E_v, respectively, the degrees of coordination of energy, environment and economy. A conventional 3E coordination model may be expressed as:

(3)$M=\sqrt[3]{E_e,E_c,E_v}$

The value range of coordination degree is usually between [0, 1], which is better as the value approaches 1, and vice versa. By referring to the methods proposed by Liu et al., the coordination degree is further classified [42].

Concerning the methods proposed by Yi et al. [43], an entropy weight based fuzzy matter element 3E system coordination evaluation and assessment model is established on the basis of the entropy weight and fuzzy matter–element theories. Fuzzy matter–element, fundamental to all matters, is composed of the matter M_m, the character C_n and fuzzy characteristics value x_mn. If matter M_m has n characteristics and corresponding fuzzy matter–elements, then R is the fuzzy matter–element in n dimensions, and the combination of m matter–elements in n dimensions forms the complex matter–element R = f (M, C, x), which may be expressed as:

(4)$R_{mn}= \left[\begin{array}{lllll}& M_1& M_2& \cdots & M_m\\ C_1& x_{11}& x_{21}& \cdots & x_{m1}\\ C_2& x_{12}& x_{22}& \cdots & x_{m2}\\ ⋮& & & & \\ C_n& x_{1n}& x_{2n}& \cdots & x_{mn}\end{array}\right]$

(5)$v_{ij}=\frac{x_{ij}-\min \left|x_{ij}\right|}{\max \left|x_{ij}\right|-\min \left|x_{ij}\right|}$

(6)$v_{ij}=\frac{\max \left|x_{ij}\right|-x_{ij}}{\max \left|x_{ij}\right|-\min \left|x_{ij}\right|}$

where, x_ij represents the value of the jth characteristics of matter i. Positive and negative indicators are converted into the optimal degree of membership in accordance with Formulas (5) and (6). The so-called optimal degree of membership means the membership degree of fuzzy quantities in correspondence to the evaluation indicators of any single indicator that is subordinate to a standard matter. The optimal membership of fuzzy matter–element $R_{mn}^{*}$ obtained by the conversion is:

(7)$R_{mn}^{*}{}_{}= \left[\begin{array}{lllll}& M_1& M_2& \cdots & M_m\\ C_1& v_{11}& v_{21}& \cdots & v_{m1}\\ C_2& v_{12}& v_{22}& \cdots & v_{m2}\\ ⋮& & & & \\ C_n& v_{1n}& v_{2n}& \cdots & v_{mn}\end{array}\right]$

where, R_on represents the standard element of the optimal membership degrees of the maximum and minimum values of various characteristics in Formula (4). Then, the difference square operation is conducted by the items in R_on and by the correspondent items to obtain the difference that is the squared complex matter–element.

(8)$\Delta v_{ij}={(v_{oj}-v_{ij})}^2$

(9)$\Delta R= \left[\begin{array}{lllll}& M_1& M_2& \cdots & M_m\\ C_1& {\Delta }_{11}& {\Delta }_{21}& \cdots & {\Delta }_{m1}\\ C_2& {\Delta }_{12}& {\Delta }_{22}& \cdots & {\Delta }_{m2}\\ ⋮& & & & \\ C_n& {\Delta }_{1n}& {\Delta }_{2n}& \cdots & {\Delta }_{mn}\end{array}\right]$

Then, the weight of region j under the indicator i, P_ij, is calculated, i.e., the finite discrete probability distribution.

(10)$P_{ij}=\frac{T_{ij}}{{\displaystyle \sum _{i=1}^m{T}_{ij}}}$

where, T_ij represent the standardized characteristic value of indicator i in region j. The information entropy H_i at ith indicator is calculated, and the entropy weight w_i of the i indicator is determined by the entropy method:

(11)$P_{ij}=\frac{T_{ij}}{{\displaystyle \sum _{i=1}^m{T}_{ij}}}$

(12)$H_i=-\frac{1}{\ln m}{\displaystyle \sum _{i=1}^m(p_{ij}\times \ln p_{ij})}$

(13)$w_i=\frac{1-H_i}{{\displaystyle \sum _{i=1}^m(1-H_i)}}$

Proximity describes the approach of various plans to the standard deviation with the more significant value indicating a closer approach, and is used to measure the approach degree of two items. The development of various subsystems is calculated by Euclidean distance:

(14)$\rho H_j=1-\sqrt{{\displaystyle \sum _{i=1}^m{w}_i{\Delta }_{ij}}}$

Assuming that all subsystems of energy, economy and environment are equally important, then the comprehensive coordination degree of the 3E system may be obtained by using Formula (15):

(15)$E=(\rho H_e+\rho H_c+\rho H_v)/3$

3.3. Spatial Econometric Models

The concept of spatial econometrics was proposed by Paelinck and Klaassen [44], who defined the settings of spatial interdependence, the asymmetry of spatial relationships, and the importance of explanatory spatial variables. After that, Cliff and Ord [45] extended the spatial autoregression model and discovered many other models, parameter estimation and testing techniques. The integrated spatial factors become not only more effective in econometric modeling but also gradually a complete set of econometrics systems, which is now favored by a wide range of scholars [46].

3.4. Data Source

• (1). The data of China’s energy consumption in 2007–2020 are from the China Energy Statistics Yearbook. The rest of China’s data are mainly from the China Statistical Yearbook, China Labor Statistics Yearbook, China Science and Technology Statistics Yearbook, China Population and Employment Statistics Yearbook, China Foreign Economic Statistics Yearbook and statistical yearbooks of corresponding years in 30 provinces, municipalities, and autonomous regions.

• (2). The conversion that is coefficient of various types of energy consumed is determined by the industry in tons of coal equivalent (TCE) and the physical quantities are calculated by referring to the Energy Balance Sheet. The industry’s carbon emission, that is the coefficient of energy consumed, is determined by referring to the IPCC Guidelines for National Greenhouse Gas Inventories 2006 (Appendix A). The carbon emission coefficient of various types of fossil energy is released on the IPCC Guideline (2006). Despite those changes in the carbon emission coefficient in recent years, they are insignificant and, therefore, can be ignored owing to the macro changes studied in this paper.

• (3). The regional development strategy of China is divided into three zones in this paper: ① Eastern: Beijing, Fujian, Guangdong, Hainan, Hebei, Jiangsu, Liaoning, Shandong, Shanghai, Tianjin, Zhejiang, and Guangxi; ② Central: Heilongjiang, Anhui, Henan, Jilin, Hubei, Hunan, Jiangxi, Inner Mongolia, and Shanxi; ③ Western: Gansu, Guizhou, Ningxia, Qinghai, Shaanxi, Sichuan, Xinjiang, Yunnan, and Chongqing.

4. Analysis and Discussion

4.1. Basis for Zoning

From a general point of view, carbon emission is subject to the systematic framework of energy, economy, and environment. Studies on the regional difference in carbon emission levels are not only to realize the commitments on emission reduction but also to solve the underlying conflicts between energy, economy and environment subsystems. Accordingly, this paper hopes to build a 3E coordination evaluation and assessment system and to analyze China’s unbalanced regional development and structural characteristics, on the basis of which, scientificity and reasonableness of zoning plans in eastern, central and western regions can be studied.

• (1). The establishment of 3E coordination evaluation and assessment system

Through systematic studies on the related literature and comprehensive considerations of the statistical indicators in China in the aspects of energy, economy and environment, a 3E coordination evaluation and assessment system, composed of 3 destination layers, 10 principle layers, and 30 specific indicators, has been established. In accordance with the 3E coordination evaluation and assessment system established on Table 1 and Formulas (4)–(12), the difference squared complex matter element ΔR is obtained, so that the degrees of coordination (Table 2) of 3E subsystems between 2007 and 2020 in various provinces are calculated by considering the weight indicators of these subsystems through these years.

• (2). 3E coordination analysis

The degrees of coordination in over one-third of Chinese provinces between 2007 and 2020 are unbalanced, which, for energy, is relatively high, indicating that local governments have attached great importance to energy development. For over one decade in the past, Chinese governments have devoted to pushing the reform of energy supply, stressing the construction of energy storage and transportation facilities, facilitating the efficient and clean transformation of energy, promoting the in-depth reform of energy systems, and improving the energy technologies. The differences in energy coordination degrees in all provinces, except Qinghai, Ningxia, and Hainan, are insignificant, suggesting indirectly that diversified energy structure has been greatly improved in eastern provinces after the adjustment of energy structure and that the construction of energy-related facilities and infrastructure has been ameliorated, resulting in narrowed space for further adjustments. Despite being merely sectional data, the fact that the lowest degree of coordination is found in the energy subsystem reflected that the deterioration of environmental conditions in China was not suppressed from the roots and that some regional environmental ecologies were still severely damaged in their pursuit of economic growth.

Through screening, the 3E coordination degrees in 30 Chinese provinces may be found to be apparently clustering distribution, which are primarily good and moderate (above 0.7) in eastern provinces, but basic and mild (0.5–0.7) in central provinces, and poor (0.5 and lower) in western provinces, with Ningxia being the worst. For such remarkable geographical distribution, the root causes are that, firstly, eastern provinces received the industries transferred from international markets to China and established relatively complete laws and regulations on environmental protection while pushing the development of high-tech industries. Secondly, industries that have high energy consumption and high pollution are transported to the western provinces, despite improvement of the technology levels in the central and western provinces assisted by the east–west industrial transfer, resulting in more investments in production factors. At last, China, in recent years, has been taking strict controls over coal consumption in eastern provinces, forcing the industrial transfer from eastern provinces to western provinces and the purchase of electric power and coal gas from western provinces, which may likely be the final causes in the unsatisfactory environment coordination degree in western provinces.

4.2. Regional Carbon Emissions Status

The Theil coefficient, variation coefficient, Gini coefficient and loss deviation coefficient are adopted to describe the current status of regional carbon emissions.

• (1). Theil index

The Theil index may measure the cross-regional difference in emission intensity, and the most significant advantage lies in the possible decomposition of absolute difference in various regions. In other words, the cross-provincial difference in the national dimension in emission intensity may be decomposed as the summary of cross-provincial differences within the regions. When the Theil index approaches 0, the difference becomes smaller; or when the Theil index goes up, the difference increases. The degree of difference may be expressed as follows:

(16)$T={\displaystyle \sum _{i=1}^n(C{E}_i/CE)\times \log (\frac{C{E}_i/GD{P}_i}{CE/GDP})}$

It may be concluded from Table 3 that the absolute difference in China in 2020 was 0.0349 and that 86.81% of this is caused by the difference inside the regions, which is much higher than the cross-regional difference, showing that the total national difference is primarily induced by differences inside the regions other than between them. In terms of contribution rate, most of the differences are from the eastern and central provinces, which in 2020 is, respectively, 46.92% and 30.38%.

• (2). Variation coefficient and Gini coefficient

The variation coefficient reflects the uniformity, stability and consistency inside physical objects. A large variation coefficient means a significant difference in emission level, while a small one an insignificant difference (Formula (17)).

(17)$CV=S^{\prime }/\overline{X}$

(18)$G=\frac{S_0}{S}=2(0.5-S_1)=1-2S_1$

The following conclusion may be drawn by analyzing the data in 2020 (Table 4) that the variation coefficient in eastern provinces with total emission level is 66.17%, which is slightly higher than the national average (63.43%), indicating that there is a remarkable difference and un-uniformed distribution in emission levels in eastern provinces. The coefficient is only 36.22%, the smallest in the central region, meaning that the emission is relatively distributed uniformly. In view of the per capita carbon emission, the variation coefficient is only 34.87% in eastern provinces, which is much lower than the national average, showing that the difference in per capita carbon emission is slight in the eastern provinces. Because of the difference in carbon emission, the eastern provinces still have the lowest emission difference, followed by the central and western provinces. However, the highest deviation of national emission is 49.23%. When comparing these three regions, it is easy to find significant differences in intensity.

From 2007 to 2020, the Gini coefficient of carbon emission in China was maintained between 0.3002 and 0.4221. However, in 2020, when the carbon emission in all regions is ranked from large to small, a Lorenz curve (Figure 1) may be established with the cumulative population as its abscissa axis and the cumulative carbon emission as its ordinate axis. In 2020, the Gini coefficient was 0.2216 in the eastern region, 0.4676 in the central region and 0.3734 in the western region, with the national Gini coefficient being 0.4221. A significant cross-regional difference can be found, that is, the largest difference in the central region, then in the central region and the smallest in western regions. The primary causes for such phenomena are that the carbon emission in Inner Mongolia in 2020 was 478 million, accounting for 15.64% of the total emission in the region. However, the population in Inner Mongolia was only 25.05 million, accounting for 5.52% of the regional in total. A small population with high carbon emissions shows badly unbalanced regional carbon emission.

• (3). Loss deviation coefficient

By referring to the spatial Gini coefficient, the Loss/Profit Deviation (LPD) coefficient is calculated in economic benefits and environmental loss so as to measure the “external economic benefits and internal environmental loss” caused by carbon emission in the region. The LPD coefficient may be expressed as:

(19)$LDP=\frac{EC}{EDC}=\frac{\frac{GD{P}_i}{GDP}/\frac{C{E}_i}{CE}}{E{I}_i/EI}$

From Table 5, the LPD coefficients of all eastern, central and western regions from 2007 to 2020 can be found to be below 1, but were increasing year by year, which shows that the unbalanced regional development is narrowing despite carbon emission LPD to certain extents in all three regions, especially in the eastern region where the LPD coefficient was approaching 1 after 2019. In western regions, the LPD coefficient was 0.43 in 2020, indicating the realization of economic profits in this region by means of 0.43 carbon emission weight over the national level. However, the environmental damage coefficient is 1.89 on the other hand, indicating the 1.89 times severity of environmental damage caused by carbon emission, which is greater than the national level. It has something to do with such factors as the regional economic development level and emission reduction technologies, reflecting, to a certain extent, the inequity in the regional development.

4.3. Causes for the Cross-Regional Difference in Carbon Emission Level

• (1). Spatial dependence characteristics of carbon emission

As is shown in Table 6, there is a significant spatial positive autocorrelation in the carbon emission in China; however, the regional heterogeneity failed to present. Therefore, Moran’s I scatter plot is used to further analyze the heterogeneity characteristics of regional carbon emission. However, on considering the time intervals, data in the years 2007, 2011, 2016 and 2020 are used to draw the Moran’s I scatter plot (Figure 2). It may be found from the plot that most Chinese provinces fall in the 1st, 2nd, and 3rd quadrants and that a spatial clustering pattern can be found in relation to carbon emission.

• (2). Test findings of spatial econometric models

The significant spatial linkage between 30 Chinese provinces in the fields of carbon emission has been demonstrated in the Moran’s I spatial correlation test. Through systematic studies and the existing literature, such factors as the social development level, energy consumption, industrial structure, population, energy efficiency, and openness are selected as the explanatory variables. The stability and multicollinearity of the selected variables need to be tested before the establishment of a time series model. After the completion of the tests, the variable “openness” is removed due to multicollinearity. Then, some models, including SAR, SEM and SDM, are used to fit the samples. The non-observed effects are no longer controlled but classified into four groups: the first with no fixed effects, the second with spatial fixed effects, the third with temporal fixed effects, and the fourth with both spatial and temporal fixed effects.

By analyzing the test findings, as shown in Table 7 and Table 8, as well as the specific effect setting of the three major panel models, it is found that, although the SEM model with spatial fixed effect has satisfactory fitting results, all the seven variables are abandoned due to the failing of significant tests. In studying the SME model with dual fixed effects, although R² is 0.9781, the result of Corrected-R² is merely 0.4167, surviving only energy consumption and industrial structure. Therefore, it could neither explain well the issues discussed in this paper. As for the SAR model with temporal effects, the value of Corrected-R² is 0.9618, with a satisfactory fitting result. Theoretically, on the steady-state level, the temporal effect represents the influence of contextual variables. The subjects studied in this paper have large periods, so the results are significantly influenced by the temporal contextual variables. Through general consideration of the facts given above, the SAR model with temporal effects is selected for studying regional emission differences in this paper, and further, the general effects are broken down and studied in detail.

After selecting the data model, it may be found in Table 9 that the direct effect, consequential effect and general effect of LnGDP are significant at a coefficient of 10%, of which the regression coefficients of the direct effects, consequential effects and general effects of LnGDP are, respectively, 0.4831, 0.0416 and 0.5247, meaning the increase of carbon emission by 0.52% for every 1% GDP growth. The consequential effects show a positive spillover effect on neighboring provinces produced by the economic growth. In other words, the fast growth of the Chinese economy remains the main contributor to the carbon emission increase, which is inevitable. The positive spillover of consequential effects has, on the one hand, justified the theories of taking significant cities and well-developed regions as the “drivers” of economic growth and, on the other hand, shown that developed areas with good spatial patterns will work positively in driving the economic growth in adjacent areas. In the future, the regression coefficient of consequential effects may become lower as the local governments work out new major regional development strategies on the basis of local conditions.

The direct, consequential and general effects of LnUR are insignificant and positive, showing that the urbanization in China does not reflect the actual energy consumption and that the changes in urbanization level have a less satisfactory effect on carbon emission.

Both LnEG and LnES are significant below 1%, of which the elastic coefficient of energy consumption is 1.6489, meaning that each 1% energy consumption will result in a 1.65% emission increase. In terms of energy structure, the general effect is also significant, reaching 0.1102, which means that each 1% economic growth will increase carbon emissions by 0.11%.

The direct and consequential effects of both LnEG and LnES are, respectively, 1.4132 and 0.1418, and 0.1253 and 0.0072. The positive and significant effects show that the region’s energy consumption and energy structure may exert positive spillover effects on neighboring areas. However, the former plays a much more significant role than the latter, indicating that the development of industrialization has been a driving factor for the increasingly rigid demands for fossil energies. Furthermore, what should not be neglected is that, in some regions, the proportion of coals and fuel oils may increase abruptly, for which the root causes may include, without limitation to the special structure of resources, the dependence on fossil energies due to extensive development models, the twisted price systems that result in seriously underestimated scarcity and external impacts of such resources, and, as well, the lack of breakthroughs in the fields of renewable energy technologies, infrastructure construction and national policies.

The direct and general effects of LnIS are significantly below 10%. However, the consequential effect failed the 10% significant test. The general effect (0.1385) shows that each 1% increase of the industrial weight will increase carbon emissions by 0.14%. This paper studies the weight of the tertiary industry in the economy. Since having significant and positive influence on the carbon emissions, it reflected that the industrial structure in the Chinese economy is “worsening”, especially, the weights of industries with high energy consumption are “increasing other than decreasing”. In addition, it also reveals the fact that under the macroeconomic policies, local governments may blindly expand investments and introduce more projects so as to pursue the expansion in volume and size in new industries and high-tech industries. At the same time, they may turn blind to the low efficiency in the local industrial structure. As a consequence, energy consumption may be finally intensified by the extensive economic expansion.

To some extent, an unbalanced industrial structure may be caused, in which too many investments are made in industries with huge profits while the fundamental industries are deserted. Though the consequential effect is positive, it is insignificant, showing that no spillover effect was produced by the industrial structure. The main causes for such phenomena lie in the dynamic evolution of industrial structure, the narrowing cross-regional difference in industrial structure, and the gradually insignificant geographical features. As the industries tend to be accomplished in the region, relatively complete industrial systems have been established, leading to the uniform distribution of major industries and product production, and lowered centralization.

The direct effect of LnP is significant at 5%; however, the consequential effect and general effect are at 10%. The direct, consequential and general effects of the population are, respectively, 0.4319, 0.0878 and 0.5197, suggesting each 1% yearly growth will result in 0.52% increase in carbon emission. Despite that, the spillover effect is significant and positive, remaining much lower than its direct effect. In the spatial lag model, population growth contributes significantly to the carbon emission increase in China. As for the causes for this phenomenon, firstly, the wealth increase of residents expands people’s consumption. People’s desire for better houses and durable commodities increased in the context of smaller families and increased households, which further amplified the energy consumption in production and routine lives. Secondly, with the popularization of education, citizens’ quality keeps increasing, and people’s awareness and capabilities in reshaping the world have increased dramatically. Furthermore, with the continuous improvements in technological levels and the content diversification of demand structures, people have developed more high-quality products and are more willing to upgrade their higher-grade products and services. All these factors work together and increase energy consumption. Thirdly, the urbanization level keeps increasing quickly, causing the continuous expansion of cities. The infrastructural construction has also driven the energy consumption in several industries, such as the cement and steel industries. In addition, the continuously expanding urbanization will take up more farms and forests, resulting in more carbon emissions brought by the changes in land use.

The direct and general effects of LnEI are significant and negative at 5%, while the consequential effect is at 10%. The direct, consequential and general effects of energy intensity are −0.5970, −0.0376, and −0.6346. The most effective emission reduction approach so far is to increase energy efficiency. As the energy consumption ratio exceeds the GDP, dropping energy intensity indicates the increase of energy efficiency. However, technological improvement is the most effective means of improving the energy efficiency. In the context of continuously increasing energy efficiency and increased difficulties in adjusting industrial and energetic structure, to increase the energy utilization rate through technological advancement has become one of the most important means. Nevertheless, the potential for emission reduction has been dramatically exploited, and the high efficiency of policies on energy conservation relies largely on top-down administrative measures, requiring significant capital investments and eventually leading to significantly increased costs for energy conservation and emission reduction, and, what is more, challenging the economic development. As such, there is a minimal space for reducing the energy intensity. It may be concluded from Table 10 that each 1% reduction of energy intensity will result in carbon emission reduction by 0.63%. In the spatial lag model, the consequential effect of energy intensity is significant at 5%. In contrast, the existence of technological cooperation between different regions under the principles of shared benefits is reflected by the significant and negative spatial spillover effect on reduced carbon emission by technological advancement. At the same time, they play the best of their respective geographical advantages. Furthermore, it also proved that educational resources or research institutes are the basis for enterprises from two neighboring regions producing similar products when resources are shared. In this way, cooperation may be established among the industries, education and research cycles, as the common goal of seeking solutions to overcome the technical barriers is shared.

4.4. Studies on Influential Factors

The spatial correlation exists among such factors as economic growth, energy consumption, energy structure, population, energy efficiency, and carbon emission. This paper, in order to demonstrate the influences of such factors on carbon emission in the context of regional differences, establishes non-parametric models, decoupling models, LMDI models, ridge regression models, and threshold regression models.

• (1). Economic level

The per capita carbon emission data and per capita GDP data in 30 Chinese provinces between 2007 and 2020 are selected and analyzed by utilizing non-parametric fixed-effect models and non-parametric random effect models. The analysis may be expressed as follows:

(20)$y_{it}=m_t(x_{it} )+u_{it}\begin{array}{cc}& i=1,2,\cdots ,N\begin{array}{c} t=1,2\end{array}\end{array},\cdots ,N\begin{array}{c}\end{array}{u}_{it}\sim (0,{\sigma }^2{I}_n)$

where, y is the explained variable; x the explanatory variable; m(x) an unknown function; u_it a random error, reflecting the omitted observable and non-observable influences and errors. Subscripts i and t represent, respectively, regions and years. When the explanatory variable is a determining variable, E(y_it) = m_t(x_it); and when the explanatory variable is a determining variable, Ey_it|Ex_it = m_t(x_it).

The Gaussian function is selected as the kernel, with the optimal window-width h = a × (N × T)^−1/7, and if a = 1.2, window width h = 0.5. The value of a influences insignificantly on both non-parametric models, which will not influence the estimates of the non-parametric model. Values between the maximum and minimum are divided into 17 equal groups, and then are estimated point-by-point by the non-parametric fixed effect model and non-parametric random effect model (Table 11). When smooth curves link up the estimated values, the EKC of per capita emission is formed, and Figure 3 is formed when the per capita carbon emission and the per capita carbon emission scatter plot is mapped on this basis.

Based on an observation of the figure, the curves of both the per capita carbon emission and per capita GDP in China between 2007 and 2020 are easily to be found to be basically in an “N” pattern. When the per capita GDP reached CNY 78.4 thousand, the per capita carbon emission decreased as the GDP grew. However, the second turning point appeared when the per capita GDP reached CNY 91.7 thousand, and the per capita carbon emissions increased, meaning the carbon emission in China is not wholly in agreement with the environmental Kuznets hypothesis. Before reaching the peak point on the left of the “N”-shaped EKC curve, it rises more abruptly than the drops on the right side of the curve, showing that it may be a rather prolonged process; although environmental pressure may reduce the premises of economic growth. Figure 3 indicated that most points fall on the left side of the “N”-shaped curve, suggesting that there is not only a very huge difference in per capita GDP between the regions, but also an enlarged wealth gap in the three-levelled economic development pattern in China.

When the data of all Chinese provinces are put together for non-parametric analysis, a general understanding of the EKC curve of carbon emission in China is shaped. However, there is, as stated above, a huge difference in the regional economic development in China, owing to which it is not enough to focus the studies merely on the local situations from a national view. Through more careful examination, it may be found that there is a huge difference in the EKC curve of per capita carbon emission. In a similar “N” pattern, the per capita carbon emission EKC in eastern provinces resembles the national EKC. When per capita GDP, respectively, reached CNY 79.9 thousand and CNY 91.4 thousand, the two turning points appeared. The “upside-down U” EKC pattern agrees with not only the national environmental EKC curve but also the appearance of the turning points in the two regions, suggesting that, in order to narrow down the economic gap with developed eastern regions, the central and western regions are in the middle of industrialization and urbanization. However, the high energy consumption and emission volume have been suppressed in the contexts effectively where the central government highly values the efficiency and consistency of policies on the low-emission economy in the central and western regions. Although these regions have gathered many conventional industries and resource-dependent cities, rigid fossil energy demands still remained. Therefore, the central and western regions received many industries transferred from the eastern region. From another perspective, some outdated production capacity has been eliminated, helping to push up the energy efficiency and the industrial strategy shifts gradually from a resource-oriented to a market-oriented economy.

• (2). Energy structure

A model of carbon emission factor decomposition is established based on the LMDI method so as to analyze the influences of energy structure and industrial structure on carbon emission levels in all three regions in China. In Formula (18), ES^t_ij = E^t_ij/E^t_i represents the proportions of j energies in the period t; and S^t_i = GDP^t_i/GDP^t represents the GDP contributed by and the proportion of GDP of industry i in the period t.

(21)$C{E}^t={\displaystyle \sum _i{\displaystyle \sum _{j}C{E}_{ij}^t=}}{\displaystyle \sum _i{\displaystyle \sum _j\frac{C{E}_{ij}^t}{C{E}_{ij}^t}\times \left(\frac{E_{ij}^t}{E_i^t}\right)}}\times \left(\frac{E_{ij}^t}{GD{P}_i^t}\right)\times \left(\frac{GD{P}_i^t}{GD{P}_{}^t}\right)\times GD{P}_{}^t$

(22)$\Delta C{E}_{}^t=C{E}_{}^t-C{E}_{}^0=C{I}_{\mathrm{effect}}+E{S}_{\mathrm{effect}}+E{I}_{\mathrm{effect}}+S_{\mathrm{effect}}+GD{P}_{\mathrm{effect}}$

Because of the conventions of China in taking five years as an entire planning period, this paper divides the period from 2007 to 2020 into three segments. With the help of related statistical data, it may be found that the energy structure keeps developing economy in China. The proportion of coal energy has fallen from 72.5% in 2007 to 56.8% in 2020, while in the same period that of new energy has risen from 7.4% to 24.3%. There is a stable positive correlation between the proportion of non-coal energies and the economic growth in China, showing that the economic growth will not be in conflict with energy structure optimization in terms of emission reduction.

It can be concluded from Table 12 that, in the process of economic development, energy structure is not always at the “optimal level”, so that its role in emission reduction does not always exist. Li et al. [47] argued that since the declining quantity limited energy consumption, the problem of low usage of clean energy has not changed so that the implementation of the policy of de-capacity and de-energy consumption in the country was not as satisfactory as it should be. For example, between 2007 and 2011, impacted by the outdated energy structure, carbon emissions were increased by 3.26 million tons. In the periods from 2012 to 2016 and from 2017 to 2020, the energy structure in the central region resulted in emission increases by 22.31 million tce and 0.14 million tce, respectively. From 2007 to 2011, consumed by the secondary industry in the western region, the proportion of coal energies increased from 64.83% to 68.42%, caused mainly by: ① Although the energy structure keeps optimizing as the society and economy develops, the supply of hydropower and natural gas is seriously unbalanced in the western region. ② According to statistical data, there were 1824 coal supplying companies in Northwest China in the “11th Five-Year Plan” period, among which 63.8% were rural coal production enterprises with production capacity below 100 thousand tons. The repeated appearance of such enterprises shaped the common characteristics of enterprises in the western region: small size, low industrial concentration, and backward technology and management levels. From 2012 to 2016, the changes in energy structure in the central region resulted in an emission increase of 22.33 million tce, largely because of “West-East Natural Gas Transmission Project” during the “12th Five-Year Plan” period, resulting in a significant increase of the gas energies in the central region. In 2016, the consumption of gas energies of the secondary and the tertiary industries increased, respectively, by 0.88% and 3.1%, over the same in 2012, increasing carbon emission to 7.07 million tce and 12.78 million tce.

Similarly, Table 12 shows that the changes in industrial structures in all three regions have led to an increased emission of 6.81 million tce, 22.19 million tce and 16.73 tce in the period from 2007 to 2011. However, the changes of industrial structure resulted in emission reduction by 171.11 million tce, 101.40 million tce and 52.13 million tce in all three regions from 2017 to 2020. The leading causes of this phenomenon lie in the dropping of industrial weights of the secondary industry (−4.01%, −5.22%, and −4.41%). In the three regions, the changes in energy structure in the tertiary industry also resulted in a carbon emission increase of 87.30 million tce, 78.41 million tce and 38.44 million tce.

• (3). Industrial structure

The influence of industrial structure on the carbon emission level is visualized by Formula (23), which, however, still fails to prove that such changes are a part of a long-term trend. Specifically, it cannot demonstrate that the industrial structure changes will foster industrial optimization and structural adjustment, under the circumstance that China can suppress carbon emission increase as it is attaching increasing importance to carbon emission problems in its economic development. This paper adopts the β convergence theory of Barro [48] to explain and justify the “convergence” of economic growth, and the absolute β convergence theory to examine the trends of influences of industrial structures. Despite the fact that the latter implies assumptions that are entirely the same as other factors, it is sufficiently intuitive and simple when examining the trends of changes and the degree of “convergence”. Therefore, the absolute β convergence model is adopted in this paper and set as follows:

(23)$\ln (\frac{S_{i,t+1}}{S_i})=\alpha +\beta \left(S_{i,t}\right)+{\epsilon }_{i,t}$

where, S represents the influence caused by industrial structure on carbon emission; α the intercept; ε_i,t the disturbance. In 30 Chinese provinces between 2007 and 2020, the panel data are selected. S is obtained by Formula (23) and summarized in Table 13.

The secondary industry, mainly manufacturing, is an industry with highly concentrated carbon emissions compared to the primary and tertiary industries, whose larger contribution to the local economy means more pressure from increasing carbon emissions. It may be found in Table 14 that the β coefficient is always negative at a 1% significance level in the eastern region, the central region, the western region and even at the national level, showing that absolute β convergence exists in the industrial structure factors. Specifically, an optimized industrial structure slows down the increase of carbon emissions. Alternatively, on carbon emission increase, a declining trend may be found in the influence of industrial structure. It is found through observations that the convergence rate caused by industrial structure adjustment in the eastern and western regions is slower than the national level, indicating the slow adjustments to industries with high levels of emissions.

In contrast, adjustments to industries with low carbon emissions are relatively weak and behind schedule, implying that industries with high emission levels may likely be aggregated. In the central region, the β coefficient is higher than the national level, denoting the priority of industries with low emission levels and the suppression of the development of high-emission industries in the region. By considering the β convergence test findings as summarized in Table 14, the changes of the regional industrial structure may be found to have significantly influenced the increase in carbon emission. However, in response to the regional industrial structural adjustment, a significant difference has been produced due to the cross-regional industrial transfer and the difference in regional industrial policies. Zhu and Zhang [49] argued that the restriction of cross-regional industries can, on the one hand, strengthen the vertical and horizontal links of industries and form a spatial structure with reasonable division of labor and complementary functions, and, on the other, achieve energy conservation and emission reduction.

• (4). Population

In this paper, the STIRPAT model (Formula (25), developed by Dietz et al. (1994)) is adopted to study the non-proportional changes of the population influences on carbon emission.

(24)$I=P\times A\times T$

(25)$I=a{P}^b\times A^c\times T^d\times e$

where, I represents environmental pressure; P population; A the degree of affluence; T technology level; a, b, c and d are the parameters, and e indicates errors.

In order to further study the influence of population changes on carbon emission levels, the population factor is further divided into such subgroups as population scale (Ps), age structure (Pc), quantities of households (Ph), and average education level (Pe). The STIRPAT model is extended, and the logarithm on both sides of the model may be obtained:

(26)$\ln I=\ln a+b\ln Ps+c\ln Pc+d\ln Ph+e\ln Pe+e_i$

In this study, the population scale is represented by the actual population in the region, the age structure by the proportion of young and middle-aged (younger than 65 years old) population, the quantity of households by the actual number of households in the region, and the average education level by the proportion of the population with college and higher education background. Many multicollinearity problems, as shown in Table 13, are discovered through careful examination of the characteristics, matrix correlation coefficient, VIP and correlation coefficient matrix in all three regions. Therefore, the ridge regression model is adopted to solve these problems.

As shown in the findings of ridge regression models in all three regions (Table 15), the model is not completely significant. Because of the growth of the population scale, age structure and quantity of households, the carbon emission increased. However, the education level generated a significantly inhibitory effect on carbon emissions, and was verified in part of the literature [50], which may be found through studies of the weights of these factors, with age structure being the most important influencing factor. Every 1% increase in age structure in the eastern, central and western regions will increase emissions, respectively, by 5.24%, 12.33% and 10.06%. According to statistical data, the proportion of the population younger than 65 years old had dropped by 2.35%, 3.12% and 4.50%, respectively, in the eastern, central and western regions between 2007 and 2020. The shrinkage of the work-age population, as it suppresses economic growth, reduces the carbon emission. However, the ageing population will result in decreased savings which, in turn, leads to an increased per capita capital and consumption level and pushes up the carbon emission. Furthermore, the population scale tends to influence insignificantly on carbon emission in the western region, but the positive pulling effect caused by population growth in the eastern region is much less insignificant than the influences of population structure. In carbon emission, every 1% change in population scale merely causes a 2.23% increase, i.e., the consumption capacity and population structure, other than the population scale, are the dominant factors.

The sizes of families and population jointly determine the number of households, and every 1% increase in household quantity in all three regions will increase carbon emissions by approximately 1.52~2.21%. However, the family size is the eastern, central and western regions has reduced from 3.67 persons, 5.80 persons and 3.74 persons in 2007 to 3.01 persons, 4.74 persons and 3.26 persons in 2020. The difference in the changes of family size shows the higher per capita resource utilization rates in the central and western regions than the eastern region owing to the fact that large families are more advantageous for sharing objects and services than smaller ones. However, the reduced family size and increased population indicate the requirement of more residential apartments, which will significantly expand the demand for land and production materials. What is more noteworthy is that the residential area per family member, despite the reduced family size, dramatically increased to four times the size compared with 30 years ago.

Education is an effective means of improving personal quality. Many researches have proven that the consumption concepts, consumption level and patterns are all closely related to the education background. As such, the family carbon emission would be inevitably influenced by the education levels of family members. In Table 15, the improvement of education level has found not to increase the carbon emission, meaning that people with better education have more desire for better living standards, so that the overall consumption level has been increased. However, due to the better educated and skilled managers in the production process, the advancement of their consumption concepts, the increased awareness of environmental protection and improved productivity have worked together to reduce the CO₂ emission. As a result, every 1% increase in education level will reduce CO₂ emission by 0.24%, 0.53 and 0.47% in the eastern, central and western regions.

• (5). Energy efficiency

Chiu’s [51] studies presented that there is a non-linear relationship between energy efficiency and carbon emission level. Therefore, this paper, in order to understand the mentioned relationship, adopts the threshold panel model developed by Hansen to study the “inherent grouping” characteristics of data and reveal the relationship between energy efficiency and carbon emission level at several stages. The static threshold panel model may be expressed as:

(27)$C_{it}=c_i+\alpha X_{it}+{\beta }_1{Y}_{it}I(q_{it}\le \gamma )+{\beta }_2{Y}_{it}I(q_{it}>\gamma )+{\epsilon }_{it}$

The double threshold model is:

(28)$C_{it}=c_i+\alpha X_{it}+{\beta }_1{Y}_{it}I(q_{it}\le {\gamma }_1)+{\beta }_2{Y}_{it}I({\gamma }_1<q_{it}\le {\gamma }_2)+{\beta }_3{Y}_{it}I(q_{it}>{\gamma }_2)+{\epsilon }_{it}$

where, C_it is the explanatory variable; c_i represents the individual effect of regional difference; X_it is the control variable; q_it the threshold variable; I the exponential function, and the result is valued as 1 if all conditions are met, or as 0 if the conditions are not met; β₁ and β₂ are the coefficients of the threshold to be estimated; ε_it is independently and identically distributed, with the mean value being 0 and variance being the stochastic disturbance of σ².

Based on panel data in 30 Chinese provinces between 2007 and 2020, the model is established, of which energy efficiency (EI), economic growth (GDP), energy structure (ES), industrial structure (IS), openness (OP), and technology investment (TI) are used as the threshold variables, and with which a single threshold only exists at 5% GDP growth. The p value is 0.00218 in a single threshold test. A single threshold exists at a 10% significant level for technology investments, of which the p-value is 0.0610. Figure 4 and Figure 5 reported the relationship between threshold parameters and the likelihood ratio. From the above-mentioned figures, the likelihood ratio is found to be zero when the threshold values for economic growth and technology investment are, respectively, 8.9310 and 0.7123. Furthermore, when the threshold values are within the confidence intervals [8.9550, 8.9800] and [0.5740, 0.6990], the critical value is 7.35 at a 5% likelihood ratio. Based on these findings, it may be concluded that it is true of the estimated threshold values for economic growth and technology investment.

Based on the threshold test, the results of the specific regression analysis are shown in Table 16, with economic growth and technology investment as the thresholds of the panel model. Energy efficiency, with continuous economic growth, industrial structure optimization and technological improvement, will gradually suppress the increase in carbon emissions. However, among relationship coefficients of energy efficiency-carbon emissions of different groups, there are significant distinctions. When the economic growth index falls below 8.9550, specifically when the per capita GDP is less than CNY 7746.52, every 1% increase in energy efficiency will result in CO₂ reduction by 0.66%. When the per capita GDP falls between CNY 7746.52 and CNY 7942.63, the interaction factor will be 0.7376. This suggests that the influence of energy efficiency on carbon emission is not monotonically decreasing but subject to a threshold or turning point. Shao [52] also found that, although energy efficiency has a significantly inhibitory effect on carbon emissions, the effect can be significantly various from period to period. If all provinces are classified into two groups as the higher and the lower, then most Chinese provinces fall in the lower group in 2007, except Beijing, Fujian, Guangdong, Heilongjiang, Jiangsu, Shandong, Shanghai, Tianjin, and Zhejiang, which are in the higher group. Most of the provinces in the higher group are in the eastern region, where more elasticity exists under the influence of energy efficiency on carbon emission, on account of the impact of energy development by the trends of the national economy. China, in such contexts of sharp problems in energy structure, energy production and outdated energy utilization, had successfully built its new energy industry system to suit the socialist economic systems in recent years, and established and improved the macroeconomic control system on energy development, which stresses the roles of economic and legal actions with administrative actions as assistance. Furthermore,, while striving to expand the overseas oil/gas production bases, China further optimized its domestic energy structure, enacted the Law of Energy Conservation, and encouraged the investments in the demonstration and promotion of science and technology.

To a large extent, technology investment is represented by the proportion of R&D investments over GDP. When the T.I. parameter is smaller than 0.574 (i.e., the weight of technology investment is less than 1.77%), the interaction coefficient of energy efficiency on carbon emission is −0.4781, which is slightly lower than the linear regression coefficient (−0.5824). When technology investment is between 1.77% and 2.01%, the suppression effect of energy efficiency on CO₂ emission increases significantly to −0.6234, showing a significant “technology investment threshold effect” between energy efficiency and carbon emission. Supposing that the 30 provinces are studied and re-classified into higher/lower groups in accordance with the same standards with five years as a period, comparison is focused on the values, respectively, in 2007, 2012, 2017 and 2020. It may be found in Table 17 that, although most provinces fall in the lower group, in 2020 only nine provinces were left in the higher group, and the technology investments in these provinces, as is shown by statistical data, account for 70.94% of the national total of 1.31% of Chinese GDP. For regions, most provinces in the higher group are in the eastern region, indirectly supporting the opinion of “significant and negative spillover effect by technology investment on carbon emission” as raised in the foregoing sections of this paper. The study of Zou et al. [53] also showed that the efficiency of pollution control in the western region is lower than that in the eastern region, fluctuating in a wave-like manner from 2012 to 2018. In the central and western provinces in the lower group, the main reasons for the small technology investments are temporary measures for the economic development strategies in these provinces by infrastructure construction on large scales and the inability to support the long-term development in these provinces. The national government first stressed in its regional development strategy that the improvement of infrastructure goes first, then followed by industrial development and optimization, and finally, by the technological innovation in view of the regional industrial advantages. However, natural conditions and the economic and cultural basis in some central and western regions, especially in the under-developed regions, have determined the less rewarding investment in these regions. As such, the investments of private sectors and foreign investors can barely increase government-oriented investments.

5. Conclusions and Policy Suggestions

Based on the regional classification in China, this paper explores the causes of spatial differences in carbon emissions and analyzes the mechanism of each influential factor in detail with empirical evidences. The scientific classification of regions, unlike previous studies, is based on the coordination between energy, economy and environment. Meanwhile, several relatively independent models are used to analyze the influential factors of carbon emissions. Accordingly, this study aims to provide a theoretical reference for governments to formulate policies and reduce trial-and-error costs for the implementation of emission reduction programs.

5.1. Conclusions

• (1). The environmental subsystem has the lowest degree of coordination, reflecting that the deteriorating environmental conditions are not yet controlled from the root and that some regions are still damaging the environment in pursuit of regional economic development. Based on the study of the 3E coordination degree, it is easy to find that the provinces have demonstrated apparent regional (eastern, central and western) features, with most of the poorly coordinated provinces located in the western region. The primary causes lie in outdated economic growth patterns, unbalanced distribution of production factors, backward technology levels and low knowledge popularization, resulting in regional environmental deterioration.

• (2). The regression of the spatial econometric model shows that: ① The direct and general effects of the industrial structure are significant at 10%, suggesting the emergence of some trends of deterioration in industrial structure, reflected explicitly by the rising rather than dropping of weights of industries with high energy consumption. However, the indirect effects are positive but insignificant, showing that the cross-regional difference in industrial structure is narrowing down in the evolution of industrial structures and that the regional characteristics are not that significant. On the basis of increasingly integral industrial categories in each region, a more complete system ad a balanced spatial distribution of major industries and product production, as well as a decline in concentration, has gradually been formed. ② Population growth will influence the increase of CO₂ emission, caused by the increased population, improved quality and expanded demands, which, specifically, are reflected by the expanded consumption desires, upgraded requirements on quality and the expanding city size.

• (3). The increase in population, age structure and quantity of households also function as positive contributors to the increase in CO₂ emission. Education level plays a significantly inhibitory role in this process. Although the improvement of education level stimulates people’s desire for a better life, it also leads to the overall improvement of living standards. However, advanced concepts on consumption have increased the awareness of environmental protection. Furthermore, the knowledge and skills of managers in production activities have indirectly contributed to the improvement of productivity and the reduction of carbon dioxide emissions.

5.2. Policy Suggestions

The irreversible and unpredictable natures of environmental issues have revealed that the tray-and-fail methods cannot solve environmental problems, which will bring about unpredictable risks and hazards if people expect to solve the environmental issues merely by market mechanism, laws and regulations or inactive adaption. Energy conservation and emission reduction require more actions to be purposefully taken in several aspects, such as energy, economy, and living conditions, to improve the environment.

• (1). Firstly, China should hold on to its strategy of developing green energy so as to build up an administration mechanism that suits and matches the energy strategies. Furthermore, modes of highly efficient utilization of energy, research systems that facilitate clean and renewable energy production, and advantageous policies for the environment should be created to maintain a stable and safe energy supply. Secondly, actions should be taken to strengthen target management by virtue of total quantity control. For this purpose, yearly goals of energy conservation and consumption reduction should be established stage-by-stage. The intensity of energy market openness should be determined in view of environment carrying capacity, and a comprehensive energy–environment decision-making mechanism be established to ensure the coordinated development of energy and the environment. Thirdly, efforts should be made to push forward the supply-side reform so as to promote the adjustment of energy consumption structure. Furthermore, active actions should be made to develop renewable energies and replace conventional energies with clean energy. Investments in related technological development should be further strengthened, and some privileged policies be enacted to introduce advanced technologies globally to realize diversified energy structure.

• (2). As for the demand, China should work actively to adjust and optimize its existing industrial structure to facilitate reasonable energy utilization in good order. In the eastern region, based on the advantages in capital and technologies, China should seek opportunities to build up its post-industrialization and emerging industrial systems. In the central region, the conditional industrial transfer may be encouraged to combine the improvement of energy efficiency and industrial application. In the western region, the second-mover advantages should be made best use of so as to work out reasonable industrial layout, strengthen source control and improve industrialization quality. It is noteworthy to mention that a cross-regional coordination mechanism should be established in response to the needs for cross-regional industrial flows, especially in industries with high CO₂ emission levels, and, moreover, overall planning should be done to transfer industries with high CO₂ emission levels. Apart from these, to promote coordinated development and environment and quality improvements across various regions, regions should work out policies that are tailored to their specific conditions.

• (3). Based on the roles in society, such as consumers, schools, families, enterprises, and governments, special efforts should be made to foster the awareness of environmental protection. For example, consumers are encouraged to develop sound environmental protection and to resist bad habits; schools to carry out education programs on the environment and foster positive school cultures; and families to establish advanced and environment-friendly consumption concepts. As for enterprises, green product awareness should be built up so as to transform the operating strategies to actively take their social responsibilities and strengthen the innovation of green technologies. Governments should lead and guide the environment-friendly conditions for consumers, and establish green quality inspection systems and green consumption mechanisms, and strengthen the construction of green infrastructures.

• (4). For western provinces with great potentiality, more elasticity, together with more emission quotas, should be given to the emission reduction policies. Despite the outdated infrastructure, poor worker quality, unbalanced population and resources, and restrictions from conventional industrial paths that resulted in the massive pressure on energy conservation and emission reduction, the top priority in the regional development strategies is still maintained by the West Development Strategy. In the excellent background of the smooth development of industrialization, the “Belt and Road” strategy has revealed new directions and blueprints for economic development in western regions. It is suggested that Chinese governments should, while actively guiding their regional development, give sufficient preferences of materials and time to the western regions. More importance should be attached to the western regions when facing the largest challenges in realizing a well-off society. To achieve this goal, a low-emission and recyclable economy should be stressed to promote their industrialization, urbanization and modern development with local characteristics.

There are, however, several limitations in this study. The Chinese government’s current approach to reducing emissions is dominated by the “double carbon” target, which is not discussed much. As a result, the policy recommendations lack some direction. In the following study, the achievement of the carbon peak target will be the focus and the impact on the economy and social welfare based on total control will be discussed.

Author Contributions

Conceptualization, Q.G.; methodology, Q.G. and Z.L.; software, Q.G. and Z.L.; data curation, A.Z. and X.B.; draft preparation and writing, Q.G. and M.L.; writing—review and editing, Q.G. and Z.L.; visualization, Q.G.; supervision, Z.L. and A.Z.; funding acquisition, Q.G. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data that support the conclusion of this research are available on request from the corresponding authors.

Acknowledgments

We sincerely acknowledge the support from the “Doctoral promotion program of Suzhou Agricultural Vocational and Technical College (Grant No. BS2109)”, “the 333 High-level Talents Training Project in Jiangsu Province” and “Suqian University Talent Introduction Research Startup Fund Support (Grant No. CK004215)”. The authors gratefully thank the anonymous reviewers and editor for their valuable opinions and professional comments.

Conflicts of Interest

The authors declare no conflict of interest.

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Figures and Tables

Figure 1. Lorenz curve of different regions. A straight line of 45 degrees is a perfectly equal line; the higher the deviation from the 45 degree line, the greater the Gini coefficient.

Figure 2. Moran’s I of carbon emissions level in different regions. (a–d) represent 2007, 2011, 2016 and 2020, respectively.

Figure 3. Nonparametric fixed effects and random effects pointwise estimates of per capita carbon emissions EKC.

Figure 4. Estimation value and confidence interval of threshold parameter GDP.

Figure 5. Estimation value and confidence interval of threshold parameter TI.

Table 1

Provincial frame of 3E coordination degree evaluation system.

First Class Indicator	Criterion	Second Class Indicator	Indicator Property
Energy system	Energy scale	Investment in fixed assets in the energy industry (Q1)	+
		Energy generation capacity (Q2)	+
		Investment ration of fixed assets in energy industry (Q3)	+
	Energy consumption	Energy consumption (Q4)	−
		Coal consumption (Q5)	−
		Gas consumption (Q6)	−
		Per capita energy consumption (Q7)	−
		Energy consumption elasticity coefficient (Q8)	−
	Energy efficiency	Energy intensity (Q9)	−
Economic system	Economic scale	GDP per capita (Q10)	+
		Investment in social fixed assets (Q11)	+
		Import–export volume (Q12)	+
	Economic structure	Proportion of second industry increment in GDP (Q13)	−
		Proportion of tertiary industry increment in GDP (Q14)	+
	Economic benefit	Total retail of sales of per capita consumer goods (Q15)	+
		Per capita disposable income (Q16)	+
	Economic quality	GDP growth rate (Q17)	+
		Revenue growth rate (Q18)	+
		Resident income growth rate (Q19)	+
		Growth rate of investment in fixed assets (Q20)	+
		Industrial of output growth rate (Q21)	+
Environmental system	Environmental quality	Per capita park green area (Q22)	+
		Forest cover rate (Q23)	+
	Environmental pollution	Industrial solid waste (Q24)	−
		Industrial dust (Q25)	−
		Carbon emissions (Q26)	−
	Environmental protection	Industrial waste disposal rate (Q27)	+
		Investment in industrial pollution control (Q28)	+
		Investment in environmental pollution control (Q29)	+
		Proportion of investment in environmental pollution control (Q30)	+

Table 2

Results of 3E coordination degree evaluation in different provinces (2007–2020).

Province	Energy Coordination Degree	Economic Coordination Degree	Environment Coordination Degree	3E Coordination Degree	Coordination State
Guangdong	0.7652	0.8299	0.7041	0.7683	Well coordination
Zhejiang	0.6269	0.8027	0.7240	0.7148	Well coordination
Shandong	0.6125	0.8104	0.7352	0.7271	Well coordination
Jiangsu	0.6689	0.7501	0.7437	0.7213	Well coordination
Tianjin	0.7712	0.6848	0.6852	0.7147	Well coordination
Fujian	0.6477	0.7841	0.7082	0.7132	Well coordination
Beijing	0.6739	0.7501	0.6859	0.7073	Well coordination
Shanghai	0.6457	0.7302	0.7144	0.7018	Well coordination
Chongqing	0.6748	0.7803	0.6367	0.6894	Basic maladjustment
Hubei	0.6301	0.8412	0.5774	0.6801	Basic maladjustment
Anhui	0.7403	0.5622	0.6401	0.6508	Basic maladjustment
Hunan	0.6818	0.5723	0.6713	0.6494	Basic maladjustment
Inner Mongolia	0.7271	0.6943	0.5249	0.6482	Basic maladjustment
Heilongjiang	0.6133	0.5051	0.7126	0.6331	Basic maladjustment
Guangxi	0.6780	0.7371	0.4938	0.6291	Basic maladjustment
Jilin	0.6216	0.5493	0.6654	0.6121	Basic maladjustment
Shanxi	0.6910	0.5845	0.5376	0.5957	Mild maladjustment
Jiangxi	0.7857	0.4979	0.5482	0.5914	Mild maladjustment
Shaanxi	0.7436	0.5521	0.4893	0.5845	Mild maladjustment
Liaoning	0.6914	0.6636	0.4012	0.5716	Mild maladjustment
Henan	0.6358	0.4718	0.5352	0.5638	Mild maladjustment
Hebei	0.6233	0.4692	0.5831	0.5699	Mild maladjustment
Sichuan	0.7007	0.4132	0.5451	0.5570	Mild maladjustment
Yunnan	0.6579	0.5693	0.3960	0.5491	Mild maladjustment
Hainan	0.5833	0.4512	0.5067	0.5399	Mild maladjustment
Guizhou	0.6132	0.4479	0.5145	0.5382	Mild maladjustment
Hainan	0.5833	0.4512	0.5067	0.5399	Mild maladjustment
Gansu	0.6669	0.4985	0.3992	0.5236	Mild maladjustment
Xinjiang	0.6210	0.4107	0.4114	0.4911	Major maladjustment
Qinghai	0.3481	0.4591	0.4553	0.3981	Major maladjustment
Ningxia	0.3749	0.3672	0.3935	0.3827	Major maladjustment

Table 3

Overall national disparity.

Time	Overall National Differences (T)	Inter-Regional Differences (TBR)		Intra-Regional Differences (TWR)
		Values	Contribution Rate	Values			Contribution Rate
				East Region	Central Region	West Region	East Region	Central Region	West Region
2007	0.0275	0.0046	16.61%	0.0121	0.0058	0.0051	46.54%	22.31%	19.62%
2008	0.0251	0.0041	16.44%	0.0115	0.0059	0.0035	44.23%	22.69%	13.46%
2009	0.0257	0.0045	17.34%	0.0116	0.0062	0.0035	44.62%	23.85%	13.46%
2010	0.0251	0.0033	13.25%	0.0126	0.0056	0.0036	48.46%	21.54%	13.85%
2011	0.0268	0.0035	13.03%	0.0133	0.006	0.0039	51.15%	23.08%	15.00%
2012	0.0288	0.0039	13.71%	0.0139	0.0064	0.0045	53.46%	24.62%	17.31%
2013	0.0317	0.0043	13.68%	0.0131	0.0082	0.0061	50.38%	31.54%	23.46%
2014	0.0305	0.0044	14.50%	0.0128	0.0081	0.0052	49.23%	31.15%	20.00%
2015	0.0313	0.0047	14.38%	0.0118	0.0074	0.0033	45.38%	28.46%	12.69%
2016	0.0322	0.0038	13.15%	0.0127	0.0063	0.0031	48.85%	24.23%	11.92%
2017	0.0309	0.0036	13.07%	0.0134	0.0059	0.0041	51.54%	22.69%	15.77%
2018	0.0321	0.0041	13.91%	0.0137	0.0061	0.0044	52.69%	23.46%	16.92%
2019	0.0337	0.0044	13.62%	0.0129	0.0083	0.0062	49.62%	31.92%	23.85%
2020	0.0349	0.0045	13.19%	0.0122	0.0079	0.0059	46.92%	30.38%	22.69%

Table 4

Coefficient of variation of different regions.

Regions	Carbon Emission Variation Coefficient	Coefficient of Variation of per Capita Carbon Emission	Coefficient of Variation of Carbon Emission Intensity
Nationwide	63.43%	47.10%	49.23%
Eastern China	66.17%	34.87%	37.29%
Central China	36.22%	60.19%	40.49%
Western China	45.59%	54.33%	42.74%

Table 5

Deviation coefficient of profit and loss of different regions (2007—2020).

Time	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
Eastern China EC	1.17	1.17	1.19	1.19	1.16	1.16	1.14	1.13	1.14	1.12	1.13	1.13	1.14	1.14
Eastern China EDC	2.64	2.61	2.45	2.32	2.38	2.25	2.09	1.88	1.81	1.68	1.55	1.45	1.36	1.33
Eastern China LDP	0.45	0.46	0.49	0.52	0.49	0.52	0.55	0.61	0.64	0.67	0.73	0.79	0.85	0.87
Central China EC	0.85	0.85	0.84	0.85	0.86	0.84	0.85	0.89	0.89	0.91	0.91	0.92	0.9	0.91
Central China EDC	3.62	3.63	3.48	3.26	3.2	3.12	2.8	2.41	2.34	2.07	1.92	1.79	1.73	1.66
Central China LDP	0.24	0.24	0.25	0.27	0.28	0.28	0.31	0.37	0.39	0.45	0.48	0.52	0.53	0.56
Western China EC	0.8	0.8	0.76	0.75	0.79	0.83	0.86	0.83	0.82	0.83	0.83	0.81	0.82	0.81
Western China EDC	3.84	3.82	3.85	3.68	3.49	3.15	2.79	2.58	2.54	2.28	2.1	2.03	1.88	1.89
Western China LDP	0.22	0.22	0.21	0.21	0.23	0.27	0.31	0.33	0.33	0.37	0.4	0.41	0.44	0.43

Table 6

Moran’s I (2007–2020).

Time	Moran’s I	p Values	Z Values	Time	Moran’s I	p Values	Z Values
2007	0.296750	0.009	2.6919	2014	0.274120	0.011	2.9013
2008	0.303145	0.003	3.0308	2015	0.274059	0.008	2.8004
2009	0.253658	0.018	2.2928	2016	0.261229	0.021	2.2197
2010	0.254956	0.022	2.5713	2017	0.259345	0.010	2.2674
2011	0.285425	0.010	2.7891	2018	0.229661	0.021	2.3521
2012	0.309343	0.007	3.0251	2019	0.236845	0.011	2.4521
2013	0.305852	0.009	3.0167	2020	0.209347	0.014	2.2314

Table 7

LM test and robust LM test without spatial panel models.

Variables	Unfixed Effect	Spatial Fixed Effect	Time Fixed Effect	Double Fixation Effect
LnGDP	0.024013(0.759231)	0.011397(0.074382)	0.492311 **(1.898370)	0.077638(0.294184)
LnUR	0.137228 **(2.413623)	0.167891(0.963302)	0.048231 **(0.578302)	0.003147(0.016182)
LnEG	0.948724 ***(54.330901)	0.875313 ***(0.085213)	1.508718 ***(5.982374)	0.886412 ***(0.073591)
LnES	0.153379 ***(5.213401)	0.074432(1.286675)	0.147243 ***(4.716293)	0.073417(1.193884)
LnIS	0.123602 **(1.874413)	−0.079316(−0.868634)	0.078362(1.313449)	0.180752 **(2.147083)
LnP	0.023152(1.236739)	0.004823(−0.50132)	0.493361 **(2.318647)	−0.304318(−0.793216)
LnEI	−0.044208 *(−1.308731)	−0.136592(−1.314572)	−0.647382 **(−2.156182)	−0.069818(−0.203787)
C	−0.581376(−1.786231)
σ 2	0.0312	0.0287	0.0432	0.0158
R 2	0.9513	0.7659	0.9348	0.2836
Logl	190.2847	253.1713	201.5346	268.1129
LM text noSpatial lag	6.40261p = 0.008	1.0633p = 0.283	11.5981p = 0.000	1.1277p = 0.084
Robust LM text no spatial lag	8.5913p = 0.004	2.7212p = 0.089	14.5370p = 0.000	0.0762p = 0.086
LM text nospatial error	0.0749p = 0.763	0.0037p = 0.061	1.8852p = 0.183	1.0836p = 0.284
Robust LM text no spatial error	0.2931p = 0.439	1.8372p = 0.068	6.0558p = 0.015	0.0223p = 0.871

Note: t-statistics in parentheses; *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 8

Wald test and LR test of spatial panel.

Test	Spatial Fixed Effect	Time Fixed Effect	Double Fixation Effect
Wald test spatial lag	56.8102(0.00000)	66.3319(0.52364)	34.5718(0.0149)
LR test spatial lag	77.3329(0.00000)	64.8421(0.368819)	37.5821(0.0449)
Wald test spatial error	25.6385(0.4269)	70.0872(0.00000)	42.3813(0.00169)
LR test spatial error	44.8210(0.26583)	31.4798(0.00103)	38.6847(0.00228)

Table 9

Setting results of four effects of SAR panel model.

Variables	Unfixed Effect	Spatial Fixed Effect	Time Fixed Effect	Double Fixation Effect
LnGDP	0.001823(0.045732)	0.007151(0.045452)	0.519121 ***(1.783632)	−0.070503(−0.237428)
LnUR	0.139869 *(2.58153)	0.152173(0.957813)	0.018362(0.212584)	−0.0163461(0.124423)
LnEG	0.958717 *(58.2335468)	0.914841 *(6.125914)	1.521312 **(0.198114)	0.884682 *(2.979816)
LnES	0.133419 *(3.928346)	0.082712(1.421851)	0.113829 **(3.302852)	0.081183(1.247853)
LnIS	0.129421 **(2.245718)	−0.074281(−0.721971)	0.117218 ***(1.723541)	−0.258321 **(−2.175321)
LnP	0.011437(1.324815)	0.238264(−1.473587)	0.553921 **(2.263420)	−0.242594(−1.125251)
LnEI	−0.035383(−0.998462)	−0.072835(−1.245331)	−0.591623 **(−2.385211)	−0.044353(−0.181453)
W × dep.var.	0.064854 *(2.662913)	−0.070361(−1.269343)	0.080973(4.457573)	−0.097987(−1.543103)
σ 2	0.0235	0.0182	0.0213	0.017
R 2	0.9655	0.9539	0.9632	0.9841
Logl	213.4015	260.2193	239.1145	274.3019

Note: t-statistics in parentheses; *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 10

Direct and indirect effects of explanatory variables.

Variables	Direct Effect	T Values	p Values	Indirect Effect	T Values	p Values	Total Effect	T Values	p Values
LnGDP	0.4831	1.9932	0.0221	0.0416	1.8392	0.0847	0.5247	1.8336	0.0470
LnUR	0.0311	0.2579	0.7112	0.0021	0.3832	0.9159	0.0323	0.2761	0.7432
LnEG	1.4132	5.1343	0.0019	0.1408	4.2803	0.0024	1.5540	5.7423	0.0086
LnES	0.1253	4.1784	0.0031	0.0072	3.7902	0.0032	0.1325	4.1238	0.0014
LnIS	0.1048	1.6832	0.0715	0.0337	1.5022	0.1085	0.1385	1.6962	0.0958
LnP	0.4319	2.2479	0.0338	0.0878	1.8382	0.0597	0.5193	2.2403	0.0287
LnEI	−0.5970	−2.2413	0.0273	−0.0376	−2.1874	0.0519	−0.6346	−2.2374	0.0316

Table 11

Various kernel function expressions.

	Fixed Effect		Random Effect
	m(x)	β(x)	m(x)	β(x)
1	2.8307	0.2539	2.7315	0.2553
2	3.9195	0.0615	3.7184	0.0601
3	4.7832	0.2582	4.5341	0.2483
4	5.9188	0.1832	5.9233	0.1834
5	5.2413	0.1901	5.2636	0.1853
6	7.9013	−0.0021	7.9348	−0.0026
7	12.1992	−0.8431	12.1307	−0.7838
8	9.3959	−0.0894	9.2945	−0.0868
9	5.9174	0.4913	5.8141	0.4746
10	10.1431	0.0512	9.9447	0.0521
11	5.8491	0.6993	6.1592	0.6993
12	14.4643	−0.4310	15.2142	−0.3327
13	3.1823	0.9387	3.2701	0.9841
14	3.9971	0.9264	2.9148	0.8215
15	1.8128	0.8919	1.7120	0.9314
16	1.9235	0.7593	1.8591	0.7849
17	4.3475	0.6974	4.2547	0.6937

Table 12

Impact factors of carbon emissions in different regions.

Time Periods	Variables	East Region	Central China	West Region
2007–2011	CI	1448.37	−154.14	5641.70
	ES	−7224.01	−2481.95	328.96
	EI	−24,647.70	−12,012.00	−14,599.90
	S	6809.12	2219.22	1673.08
	L	141,620.00	69,504.30	36,196.00
2012–2016	CI	3320.79	3564.05	−239.43
	ES	−1412.46	2233.45	−2300.97
	EI	−99,419.30	−96,461.80	−45,086.70
	S	−7352.16	8922.25	5039.26
	L	201,128.00	144,540.00	78,401.40
2017–2020	CI	59,672.10	21,511.60	−46,599.80
	ES	−9471.15	16.72	−6653.94
	EI	−89,990.40	−52,319.60	40,284.10
	S	−17,111.10	−10,140.30	−5213.56
	L	98,628.30	59,887.20	49,173.80

Table 13

Correlation and collinearity test of independent variable and dependent variable.

Regions	Variables	Correlation Matrix					VIF	F	R 2	Characteristic Root
		C	Lnps	Lnpc	Lnph	Lnpe
East region	C	0.00	0.00	0.00	0.00	0.00		110.60	0.977	4.971
	Lnps	0.00	0.00	0.01	0.00	0.01	295.367			0.019
	Lnpc	0.00	0.00	0.23	0.00	0.00	4.395			0.004
	Lnph	0.01	0.00	0.39	0.06	0.05	118.023			0.000
	Lnpe	0.97	1.00	0.34	0.92	0.94	75.192			0.002
Central China	C	0.00	0.00	0.00	0.00	0.00		138.41	0.954	4.591
	Lnps	0.00	0.00	0.03	0.00	0.14	7.781			0.019
	Lnpc	0.00	0.00	0.17	0.00	0.71	6.884			0.003
	Lnph	0.04	0.01	0.78	0.96	0.08	1.913			0.000
	Lnpe	0.96	0.99	0.14	0.05	0.09	1.281			0.000
West region	C	0.00	0.00	0.00	0.00	0.00		202.148	0.981	4.665
	Lnps	0.00	0.00	0.04	0.00	0.16	1.522			0.025
	Lnpc	0.00	0.00	0.18	0.00	0.75	8.547			0.007
	Lnph	0.05	0.01	0.74	0.85	0.00	8.923			0.000
	Lnpe	0.95	0.99	0.05	0.17	0.12	1.904			0.000

Table 14

Convergence test results of carbon dioxide growth impact of industrial structure factor.

		Eastern China	Central China	Western China	Nationwide
β	coefficient	−1.0656	−1.1221	−1.0177	−1.0767
	t	−11.59	−9.72	−7.68	−17.17
	Sig.	0.0000	0.0000	0.0000	0.0000
Constant	coefficient	5.3277	5.3461	4.6593	5.1864
	t	11.57	9.5900	7.68	17.08
	Sig.	0.0000	0.0000	0.0000	0.0000
R2		0.6808	0.6117	0.6036	0.6761
F	coefficient	134.4	94.53	58.92	294.91
	p	0.0000	0.0000	0.0000	0.0000

Table 15

Estimation results of ridge regression.

Regions	Variables	B var	T	Sig.
Eastern China	C	−25.7251	−8.1974	0.0000
	Lnps	2.2318	13.6182	0.0000
	Lnpc	5.2381	1.9313	0.0703
	Lnph	1.5293	9.4852	0.0000
	Lnpe	−0.2434	−3.6803	0.0584
Central China	C	−9.8639	−7.4595	0.0215
	Lnps	0.2694	0.1176	0.4738
	Lnpc	12.3349	4.3325	0.0000
	Lnph	2.1648	4.8215	0.0000
	Lnpe	−0.5384	−4.7633	0.0000
Western China	C	−17.3843	−6.8419	0.0283
	Lnps	0.9812	0.7975	0.7346
	Lnpc	10.0674	7.3923	0.0021
	Lnph	2.2114	8.9563	0.0000
	Lnpe	−0. 4729	−9.8204	0.0000

Table 16

Estimation result of panel threshold model.

Variables	Linear Regression Individual Fixed Effect	Threshold Regression
		GDP	TI
GDP	0.8779 ***	\	0.8417 ***
	−21.83		−21.62
ES	0.0434 *	0.0394 *	0.0639 *
	−0.71	−0.52	−1.13
IS	−0.089	−0.0921	0.0051
	(−0.19)	(−0.86)	−0.47
OP	0.0034	0.0139	0.004
	−0.34	−0.88	−0.22
TI	−0.0162 **	−0.0153 *	\
	(−0.44)	(−0.47)
0 (EI)	−0.5713 ***	−0.6487 ***	−0.4781 ***
	(−7.53)	(−8.42)	(−5.71)
1 (EI)		−0.7481 **	−0.6234 *
		(−9.15)	(−8.62)
TH-1		8.97	0.672
Lower		8.955	0.574
Upper		8.98	0.699
P		0.0221	0.0607
F		−26.49	−28.41

Note: t-statistics in parentheses; *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 17

High and low location divisions in different provinces (2007, 2012, 2017, 2020).

Time	High Location	Low Location
2007	Beijing, Shanghai	Other provinces
2012	Beijing, Shanghai, Tianjin	Other provinces
2017	Beijing, Shanghai, Tianjin, Guangdong, Jiangsu, Shandong, Zhejiang	Other provinces
2020	Anhui, Beijing, Guangdong, Hubei, Jiangsu, Shandong, Shanghai, Tianjin, Zhejiang	Other provinces

Appendix A

Table A1

Standard coal conversion coefficient and CO₂ emission index of primary types of energy.

Types of Energy	Standard Coal Conversion Coefficient	CO2 Emission Index	Types of Energy	Standard Coal Conversion Coefficient	CO2 Emission Index
Raw coal	0.712 kgce/kg	2.69 CO2kg/kg	Gasoline	1.471 kgce/kg	2.76 CO2kg/L
Fuel coal	0.899 kgce/kg	2.53 CO2kg/kg	Kerosene	1.471 kgce/kg	2.56 CO2kg/L
Coking coal	0.971 kgce/kg	2.69 CO2kg/kg	Diesel oil	1.457 kgce/kg	2.73 CO2kg/L
Coke oven gas	0.614 kgce/m3	0.93 CO2kg/m3	Fuel oil	1.457 kgce/kg	2.98 CO2kg/L
Other coke products	1.153 kgce/kg	3.35 CO2kg/kg	Natural gas	1.330 kgce/m3	2.09 CO2kg/m3
Crude oil	1.428 kgce/kg	2.76 CO2kg/L	Coke oven gas	0.613 kgce/m3	0.93 CO2kg/m3

References

1. Anwar, A.; Malik, S. Cogitating the role of technological innovation and institutional quality on environmental degradation in G-7 countries. Int. J. Green Econ.; 2021; 15, pp. 213-232. [DOI: https://dx.doi.org/10.1504/IJGE.2021.120871]

2. Chien, F.; Anwar, A.; Hsu, C.C.; Sharif, A.; Razzaq, A.; Sinha, A. The role of information and communication technology in encountering environmental degradation: Proposing an SDG framework for the BRICS countries. Technol. Soc.; 2021; 65, 101587. [DOI: https://dx.doi.org/10.1016/j.techsoc.2021.101587]

3. Mengis, N.; Partanen, A.L.; Jalbert, J.; Matthews, H.D. 1.5 °C carbon budget dependent on carbon cycle uncertainty and future non-CO₂ forcing. Nature; 2018; 8, 5831. [DOI: https://dx.doi.org/10.1038/s41598-018-24241-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29643459]

4. Naqi, A.; Jang, J.G. Recent progress in green cement technology utilizing low-carbon emission fuels and raw materials: A review. Sustainability; 2019; 11, 537. [DOI: https://dx.doi.org/10.3390/su11020537]

5. Sun, W.; Huang, C.C. Predictions of carbon emission intensity based on factor analysis and an improved extreme learning machine from the perspective of carbon emission efficiency. J. Clean. Prod.; 2022; 338, 130414. [DOI: https://dx.doi.org/10.1016/j.jclepro.2022.130414]

6. Zhu, Q.; Li, X.; Li, F.; Wu, J.; Zhou, D. Energy and environmental efficiency of China’s transportation sectors under the constraints of energy consumption and environmental pollutions. Energy Econ.; 2020; 89, 104817. [DOI: https://dx.doi.org/10.1016/j.eneco.2020.104817]

7. Xi, J. Xi Jinping’s Speech at the Climate Ambition Summit. 2020; Available online: http://www.gov.cn/xinwen/2020-12/13/content_5569138.htm (accessed on 25 March 2021).

8. Xi, J. Building a Community of Life between People and Nature. People’s Daily; 2021; 004267. [DOI: https://dx.doi.org/10.28655/n.cnki.nrmrb.2021.004267]

9. Safi, A.; Chen, Y.; Wahab, S.; Ali, S.; Yi, X.; Imran, M. Financial instability and consumption-based carbon emission in E-7 countries: The role of trade and economic growth. Sustain. Prod. Consum.; 2021; 27, pp. 383-391. [DOI: https://dx.doi.org/10.1016/j.spc.2020.10.034]

10. Liu, L.; Anwar, A.; Irmak, E.; Pelit, I. Asymmetric linkages between public-private partnership, environmental innovation, and transport emissions. Econ. Res. Ekon. Istraživanja; 2022; pp. 1-22. [DOI: https://dx.doi.org/10.1080/1331677X.2022.2049979]

11. Den Elzen, M.G.; Beltran, A.M.; Hof, A.F.; van Ruijven, B.; van Vliet, J. Reduction targets and abatement costs of developing countries resulting from global and developed countries’ reduction targets by 2050. Mitig. Adapt. Strateg. Glob. Chang.; 2013; 18, pp. 491-512. [DOI: https://dx.doi.org/10.1007/s11027-012-9371-9]

12. Cutlip, L.; Fath, B.D. Relationship between carbon emissions and economic development: Case study of six countries. Environ. Dev. Sustain.; 2012; 14, pp. 433-453. [DOI: https://dx.doi.org/10.1007/s10668-011-9328-2]

13. Wang, Q.; Zhao, M.M.; Li, R.R.; Su, M. Decomposition and decoupling analysis of carbon emission from economic growth: A comparative study of China and United States. J. Clean. Prod.; 2018; 197, pp. 178-184. [DOI: https://dx.doi.org/10.1016/j.jclepro.2018.05.285]

14. Li, J.B.; Huang, X.J.; Kwan, M.P.; Yang, H.; Chuai, X.W. The effect of urbanization on carbon dioxide emissions efficiency in the Yangtze River Delta, China. J. Clean. Prod.; 2018; 188, pp. 38-48. [DOI: https://dx.doi.org/10.1016/j.jclepro.2018.03.198]

15. Dai, H.C.; Zhang, H.B.; Wang, W.T. The impacts of U.S. withdrawal from the Paris Agreement on carbon emission space and mitigation cost of China, E.U., and Japan under the constraints of global carbon emission space. Adv. Clim. Chang. Res.; 2017; 8, pp. 226-234. [DOI: https://dx.doi.org/10.1016/j.accre.2017.09.003]

16. Pan, W.; Pan, W.; Shi, Y.; Liu, S.; He, B.; Hu, C.; Tu, H.; Xiong, J.; Yu, D. China’s inter-regional carbon emissions: A input-output analysis under considering national economic strategy. J. Clean. Prod.; 2018; 197, pp. 794-803. [DOI: https://dx.doi.org/10.1016/j.jclepro.2018.06.207]

17. Wu, Y.; Chau, K.W.; Lu, W.S.; Shen, L.Y.; Shuai, C.Y.; Chen, J.D. Decoupling relationship between economic output and carbon emission in the Chinese construction industry. Environ. Impact Assess. Rev.; 2017; 71, 6060. [DOI: https://dx.doi.org/10.1016/j.eiar.2018.04.001]

18. Zhang, Y.; Zhang, S.F. The impacts of GDP, trade structure, exchange rate and FDI inflows on China’s carbon emission. Energy Policy; 2018; 120, pp. 347-353. [DOI: https://dx.doi.org/10.1016/j.enpol.2018.05.056]

19. Jorgenson, A.; Schor, J.; Huang, X.R. Income inequality and carbon emission in the United States: A state-level analysis, 1997–2012. Ecol. Econ.; 2017; 134, pp. 40-48. [DOI: https://dx.doi.org/10.1016/j.ecolecon.2016.12.016]

20. Cohen, G.; Jalles, J.T.; Loungani, P.; Marto, R.; Wang, G.W. Decoupling of emissions and GDP: Evidence from aggregate and provincial Chinese data. Energy Econ.; 2019; 77, pp. 105-118. [DOI: https://dx.doi.org/10.1016/j.eneco.2018.03.030]

21. Li, J.S.; Zhou, H.W.; Meng, J.; Yang, Q.; Chen, B.; Zhang, Y.Y. Carbon emissions and their drivers for a typical urban economy from multiple perspectives: A case analysis for Beijing city. Appl. Energy; 2018; 226, pp. 1076-1086. [DOI: https://dx.doi.org/10.1016/j.apenergy.2018.06.004]

22. Long, Y.; Yoshida, Y. Quantifying city-scale emission responsibility based on input-output analysis-inside from Tokyo, Japan. Appl. Energy; 2018; 218, pp. 349-360. [DOI: https://dx.doi.org/10.1016/j.apenergy.2018.02.167]

23. Zhou, D.Q.; Zhou, X.Y.; Xu, Q.; Wu, F.; Wang, Q.W.; Zha, D.L. Regional embodied carbon emission and their transfer characteristics in China. J. Clean. Prod.; 2018; 197, pp. 178-184. [DOI: https://dx.doi.org/10.1016/j.strueco.2018.05.008]

24. Narayan, P.K.; Saboori, B.; Soleymani, A. Economic growth and carbon emissions. Econ. Model.; 2016; 53, pp. 388-397. [DOI: https://dx.doi.org/10.1016/j.econmod.2015.10.027]

25. Ulucak, R.; Bilgili, F. A reinvestigation of EKC model by ecological footprint measurement for high, middle and low income countries. J. Clean. Prod.; 2018; 188, pp. 144-157. [DOI: https://dx.doi.org/10.1016/j.jclepro.2018.03.191]

26. Dong, F.; Yu, B.L.; Hadachin, T.; Dai, Y.J.; Long, R.Y. Drivers of carbon emission intensity changes in China. Resour. Conserv. Recycl.; 2018; 129, pp. 187-201. [DOI: https://dx.doi.org/10.1016/j.resconrec.2017.10.035]

27. Ang, B.W.; Zhang, F.Q.; Choi, K. Factorizing changes in energy and environmental indicators through decomposition. Energy; 1998; 23, pp. 489-495. [DOI: https://dx.doi.org/10.1016/S0360-5442(98)00016-4]

28. Lin, B.Q.; Long, H.Y. Emissions reduction in China’s chemical industry-Based LMDI. Renew. Sustain. Energy Rev.; 2018; 35, pp. 1348-1355.

29. Jiang, Y.; Bai, H.T.; Feng, X.Y.; Luo, W.; Huang, Y.Y.; Xu, H. How do geographical factors affect energy-related carbon emissions? A Chinese panel analysis. Ecol. Indic.; 2018; 93, pp. 1226-1235. [DOI: https://dx.doi.org/10.1016/j.ecolind.2018.05.080]

30. Lin, Y.; Yang, Y.T.; Zhang, X.; Tang, K. Whether China’s industrial sectors make efforts to reduce CO₂ emissions from production? -A decomposed decoupling analysis. Energy; 2018; 160, pp. 796-809.

31. Cai, B.F.; Li, W.X.; Dhakal, S.; Wang, J.H. Source data supported high resolution carbon emissions inventory for urban areas of the Beijing-Tianjin-Hebei region: Spatial patterns, decomposition and policy implications. J. Environ. Manag.; 2018; 206, pp. 786-799. [DOI: https://dx.doi.org/10.1016/j.jenvman.2017.11.038]

32. Anwar, A.; Sinha, A.; Sharif, A.; Siddique, M.; Irshad, S.; Anwar, W.; Malik, S. The nexus between urbanization, renewable energy consumption, financial development, and CO₂ emissions: Evidence from selected Asian countries. Environ. Dev. Sustain.; 2022; 24, pp. 6556-6576. [DOI: https://dx.doi.org/10.1007/s10668-021-01716-2]

33. Sun, Y.; Anwar, A.; Razzaq, A.; Liang, X.; Siddique, M. Asymmetric role of renewable energy, green innovation, and globalization in deriving environmental sustainability: Evidence from top-10 polluted countries. Renew. Energy; 2022; 185, pp. 280-290. [DOI: https://dx.doi.org/10.1016/j.renene.2021.12.038]

34. Mao, G.; Dai, X.; Wang, Y.; Guo, J.; Cheng, X.; Fang, D.; Song, X.; He, Y.; Zhao, P. Reducing carbon emissions in China: Industrial structural upgrade based on system dynamics. Energy Strategy Rev.; 2013; 2, pp. 199-204. [DOI: https://dx.doi.org/10.1016/j.esr.2013.07.004]

35. Lin, B.Q.; Jia, Z.J. Impact of quota decline scheme of emission trading in China: A dynamic recursive CGE model. Energy; 2018; 149, pp. 190-203. [DOI: https://dx.doi.org/10.1016/j.energy.2018.02.039]

36. Xiao, B.W.; Niu, D.X.; Wu, H. Exploring the impact of determining factors behind CO₂ emissions in China: A CGE appraisal. Sci. Total Environ.; 2017; 518, pp. 559-572. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2016.12.164]

37. Sharif, A.; Raza, S.A.; Ozturk, I.; Afshan, S. The dynamic relationship of renewable and nonrenewable energy consumption with carbon emission: A global study with the application of heterogeneous panel estimations. Renew. Energy; 2019; 133, pp. 685-691. [DOI: https://dx.doi.org/10.1016/j.renene.2018.10.052]

38. Grübler, A.; Nakicenovic, N. International Burden Sharing in Greenhouse Gas Reduction; International Institute for Applied Systems Analysis: Laxenburg, Austria, 1994.

39. Yu, S.W.; Wei, Y.M.; Fan, J.L.; Zhang, X.; Wang, K. Exploring the regional characteristics of inter-provincial CO₂ emissions in China: An improved fuzzy clustering analysis based on particle swarm optimization. Appl. Energy; 2012; 92, pp. 522-562. [DOI: https://dx.doi.org/10.1016/j.apenergy.2011.11.068]

40. Zhou, X.; Zhou, M.H.; Zhang, M. Contrastive Analyses of the Influence Factors of Inter-provincial Carbon Emission Induced by Industry Energy in China. Nat. Hazards; 2016; 3, pp. 1405-1433. [DOI: https://dx.doi.org/10.1007/s11069-015-2096-9]

41. Li, Z.Z.; Li RY, M.; Malik, M.Y.; Murshed, M.; Khan, Z.; Umar, M. Determinants of carbon emission in China: How good is green investment?. Sustain. Prod. Consum.; 2021; 27, pp. 392-401. [DOI: https://dx.doi.org/10.1016/j.spc.2020.11.008]

42. Liu, Y.Q.; Zhang, J.C.; Zhu, Z.S.; Zhao, G.H. Impacts of the 3E (economy, energy and environment) coordinated development on energy mix in China: The multi-objective optimisation perspective. Struct. Chang. Econ. Dyn.; 2019; 50, pp. 56-64. [DOI: https://dx.doi.org/10.1016/j.strueco.2019.05.005]

43. Yi, Q.; Feng, J.; Wu, Y.L.; Li, W.Y. 3E (energy, environmental, and economy) evaluation and assessment to an innovative dual-gas polygene ration system. Energy; 2014; 66, pp. 285-294. [DOI: https://dx.doi.org/10.1016/j.energy.2014.01.053]

44. Fujita, M. Spatial Development Planning: A Dynamic Convex Programming Approach; North-Holland Publishing Company: Amsterdam, The Netherlands, 1979; pp. 501-503.

45. Cliff, A.D. Evaluating the percentage points of a spatial autocorrelation coefficient. Geogr. Anal.; 1971; 3, pp. 51-62. [DOI: https://dx.doi.org/10.1111/j.1538-4632.1971.tb00347.x]

46. Wang, X.W.; Wang, S.; Zhang, Y.S. The impact of environmental regulation and carbon emissions on green technology innovation from the perspective of spatial interaction: Empirical evidence from urban agglomeration in China. Sustainability; 2022; 14, 5381. [DOI: https://dx.doi.org/10.3390/su14095381]

47. Li, X.; Wang, J.M.; Zhang, M.; Ouyang, J.M.; Shi, W.T. Regional differences in carbon emission of China’s industries and its decomposition effects. J. Clean. Prod.; 2020; 270, 122528. [DOI: https://dx.doi.org/10.1016/j.jclepro.2020.122528]

48. Barro, R.B. Technological diffusion, convergence, and growth. J. Econ. Growth; 1997; 2, pp. 1-27. [DOI: https://dx.doi.org/10.1023/A:1009746629269]

49. Zhu, B.; Zhang, T.L. The impact of cross-region industrial structure optimization on economy, carbon emissions and energy consumption: A case of the Yangtze River Delta. Sci. Total Environ.; 2021; 778, 146089. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2021.146089]

50. Tao, A.X. Research on the realization path of carbon emission reduction in Zhejiang Province. Energy Rep.; 2022; 8, pp. 501-506. [DOI: https://dx.doi.org/10.1016/j.egyr.2022.03.103]

51. Chiu, Y.B. Carbon dioxide, income and energy: Evidence from a non-linear model. Energy Econ.; 2017; 61, pp. 279-288. [DOI: https://dx.doi.org/10.1016/j.eneco.2016.11.022]

52. Shao, M.X.; Xue, M.G. Decomposition analysis of carbon emissions: Considering China’s energy efficiency. Energy Rep.; 2022; 8, pp. 630-635. [DOI: https://dx.doi.org/10.1016/j.egyr.2022.05.114]

53. Zou, W.J.; Zhang, L.Q.; Xu, J.Y.; Xie, Y.F.; Chen, H.X. Spatial–Temporal Evolution Characteristics and Influencing Factors of Industrial Pollution Control Efficiency in China. Sustainability; 2022; 14, 5152. [DOI: https://dx.doi.org/10.3390/su14095152]