3.1 Characteristics of inter-provincial carbon transfer
From 2012 to 2017, China's economically developed provinces (autonomous regions and municipalities) were the major carbon emissions transfer-in and out region of most provinces, and provinces with a weak base of economic development and lagging behind in science and technology were the smallest carbon transfer-out regions. Figure 1 shows the relationship between inter-provincial carbon transfer, the darker the color means, the larger the amount of inter-provincial carbon transfer. In 2012, Jiangsu was the largest carbon emissions transfer-in region for most provinces,
followed by Guangdong. In terms of the direction of carbon emissions transfer out, Jiangsu was also the largest carbon emissions transfer-out region for most provinces. In 2017, the carbon emissions transferred in from various provinces were concentrated from Guangdong, Hebei, Inner Mongolia and Henan. The largest carbon emissions transferred in region for individual provinces was Jiangsu, Zhejiang, Inner Mongolia and Henan. In terms of the direction of carbon emissions transfer out, Jiangsu remains the largest carbon emissions transfer-out region of most provinces, followed by Guangdong. Jiangsu and Guangdong are the top two carbon emissions transfer-out regions for the vast majority of provinces. With growing economic strength and expanding industrial scale, Guangdong has closer trade links with other provinces, and more and more products are produced and processed through other provinces. From 2012 to 2017, the smallest carbon emissions transfer-out region of each province has always been Hainan as well as the western provinces of Qinghai and Ningxia, indicating that the development gap between the above regions and other provinces has always been large.
3.2 Characteristics of inter-regional carbon transfer
The MR was always the region that transferred the most carbon emissions out, while the JJ transferred the least. The region with the highest amount of carbon emissions transferred in shifted from the MR to the EC. The region with the lowest carbon emissions transferred in shifted from the SC to the JJ. As can be seen in Fig. 2, the MR had the highest total amount of carbon emissions transferred in and out in 2012. Carbon emissions transferred out of all regions to the MR were the largest, with the amount of carbon emissions transferred in of the MR reaching as high as 521.61 millions of tons. NE, NC, EC, SC, and MR all had the lowest carbon emissions transferred out to the SC. JJ, NW, SW transferred the least amount of carbon emissions to the JJ. The MR was also the largest source of carbon emissions for other regions, with a total of 537.132 million tons of carbon emissions transferred out. In 2017, the largest amount of carbon emissions transferred in was in the EC, and the largest carbon emissions transferred out was still in the MR. Except for NC, EC, and SC, other regions transferred out the most carbon emissions to the EC, and the EC carbon emissions transferred in was as high as 585.52 million tons. Except for NC and MR, carbon emissions transferred out of other regions to the JJ were the least, with only 176.59 million tons of carbon emissions transferred in of the JJ. Same as in 2012, the smallest carbon emissions transferred in of all 8 regions are from the JJ. Compared with 2012, carbon emissions transferred in of all other regions except for NE, NC, and NW were increased. In addition, carbon emissions transferred out of the JJ and the SC declined, while all other regions increased to varying degrees. It can be seen that with the continuous development of the economy and society, the carbon emissions generated by the economically developed regions themselves were decreasing, and the carbon emissions transferred through the value chain were increasing. This also shows that the economically developed regions were continuously transferring and relocating the high-carbon emission production chains to the less-developed regions. This had led to a continued increase in direct carbon emissions of less economically developed regions.
3.3 Characteristics of net carbon transfer
China's carbon emissions as a whole show the characteristics of transferring from regions with a higher economic development level to lower regions, and from the east to the west. In 2012, the net transfer-out regions were mainly distributed in the regions with rich natural resources in Central and Western China, as well as economically developed Shanghai, Zhejiang, and Guangdong (Fig. 3a). In 2017, the net transfer-out regions were mainly distributed in the NE, NW, SW, and some provinces in the MR with rich natural resources. As can be seen in Fig. 4, Beijing was a complete net transfer-in region, that is, there were more transfers in than transfers out with all provinces. Tianjin, Jiangsu, Fujian, Jiangxi, Shandong, Henan, Hunan, Sichuan, and Shaanxi were also net transfer-in regions from 2012 to 2017. Fujian had the largest increase in net carbon transfer (Fig. 4). However, the net transfer amount of Tianjin, Jiangxi, Hunan, Sichuan, and Shaanxi decreased (Fig. 4).
Ningxia was a complete net transfer out region, that is, there were more transfers out than transfers in with all the provinces. Shanxi, Inner Mongolia, Liaoning, Anhui, Guizhou, Yunnan, Gansu, and Xinjiang were also consistently net carbon transfer-out regions. Except for Shanxi and Anhui, the net transfer amount of other regions had increased, and the growth rate was high. Comparison of Fig. 3 shows that the net transfer direction of other provinces had changed from 2012 to 2017. Hebei, Jilin, Heilongjiang, Guangxi, Hainan, Chongqing, and Qinghai all changed from net transfer-in regions to net transfer-out regions, while Shanghai, Zhejiang, Hubei, and Guangdong changed from net transfer-out regions to net transfer-in regions. Although the net transfer direction of these regions had changed, the net transfer amount of most regions had increased to varying degrees.
Carbon emissions from the JJ were mainly transferred out to the Northern China in close proximity, but the carbon emissions transferred out to the Southern China were also increasing (Fig. 1). In 2012, the net transfer amount between Beijing and Hebei was the largest (Fig. 5a). The net transfer amount between Tianjin and Jilin was the largest (Fig. 5a). By 2017, the net carbon transfer between Beijing and Henan was the largest (Fig. 5b). Tianjin had the largest net carbon transfer with Inner Mongolia (Fig. 5b).
As an old industrial base in China, the NE is one of the net transfer-out regions of carbon emissions. As can be seen from Fig. 5, in 2012, Liaoning had the highest net transfers with Inner Mongolia, Guangdong and Shanxi, all of which Liaoning served as a net transfer-in region. By 2017, the largest net transfers were with Jiangsu, Guangdong and Shanghai, and Liaoning was a net transfer-out region. This also shows that the carbon emissions transferred in Liaoning through the value chain decreased and the carbon emissions transferred out increased. Net carbon transfers between Heilongjiang and Jilin were the highest in 2012–2017. Jilin was a net transfer-in region in 2012 and Heilongjiang was the net transfer-in region in 2017. Jilin's net carbon transfers increased with most provinces. In the future, Jilin will still continue to be a net transfer-out region to take over some of the production links from the provinces. Heilongjiang, on the other hand, had seen a decrease in net carbon transfer with most provinces. This indicates that the industrial structure of Heilongjiang was adjusting, and the reliance of the region on its heavy industry products was decreasing.
In the NC of Hebei and Shandong, industries had been transformed and the direction of net carbon transfer had changed significantly. As can also be seen in Fig. 5, the largest net transfer was between Shandong and Guangdong in 2012, with Shandong being the net transfer-in region. In 2017, Guangdong then became a net transfer-in region for Shandong. This indicates that Guangdong needed Shandong to complete more energy-consuming industrial segments and emit more carbon emissions. While in 2017, the net transfer between Shandong and Inner Mongolia was the largest, and Shandong had been acting as a net transfer-in region. This illustrates the increased dependence of Shandong's industrial development on resource-based provinces such as Inner Mongolia. In 2012, Hebei had the largest net transfer of carbon emissions out to Beijing, followed by a net transfer in from Guangdong. In 2017, the largest net transfer was with Shanxi, followed by Shandong, Beijing and Guangdong, and all of Hebei became a net transfer-out region.
The three provinces in the EC had larger net carbon transfers to the NE, SW and NW. The above regions can provide the EC with the production and processing of various agricultural products, fossil energy and other raw materials. It also can be seen in Fig. 5, Jiangsu had the largest net carbon transfer with Guangdong, Shanghai and Jiangxi in 2012. However, by 2017, net carbon transfers with all three of these regions had declined, and the net carbon transfers with Xinjiang, Inner Mongolia, and Liaoning were the largest. In 2012, Shanghai had the largest net carbon transfer with Jiangsu and Shandong, and Shanghai is all a net transfer-out region. In 2017, net carbon transfers with Inner Mongolia and Henan were the largest, and Shanghai is already a net transfer-in region. In 2012, the net carbon transfer between Zhejiang and Hebei, Guangdong and Shandong was the largest, and Zhejiang was the net transfer-out region of Hebei and Shandong. In 2017, the net carbon transfer between Zhejiang and Guangdong, Jiangsu and Inner Mongolia was the largest, and Zhejiang was also a net transfer-out region for Jiangsu and Guangdong. It indicates that the industrial structure and the direction of industrial transfer in Shanghai and Zhejiang had changed considerably during 2012–2017. Economic development was becoming more closely linked to other provinces and cities, and the dependence on other provinces was also increasing.
The net carbon transfer of Guangdong, Fujian and Hainan in the SC was mainly from the Northern China. The net carbon transfers between 3 provinces in the SC and Jiangsu and Hebei, Shandong and Inner Mongolia in Northern China were significantly larger in 2012 (Fig. 5a). Of these, the largest net transfers in Fujian and Hainan were from Guangdong. By the time 2017 rolled around, the direction of the net transfer among the 3 regions had changed considerably. As can be seen in Fig. 5b, Fujian remained the largest net carbon transfer with Guangdong and Inner Mongolia. Hainan had larger net transfers mainly with Henan and the Yangtze River Delta region. Guangdong, on the other hand, had the largest net carbon transfer inwards from Xinjiang. Guangdong is a major textile producer and exporter, and has become the world's third-largest apparel export base. Xinjiang has a well-developed cotton cultivation and processing industry, and is an important production and supply location for the upstream and midstream segments of Guangdong's textile and apparel industry.
The direction of net carbon transfers varies considerably across provinces in the MR. The net carbon transfer between Shanxi and Jiangsu was the largest in 2012 (Fig. 5a), but has since declined. The net carbon transfers between Shanxi and Guangdong and Hebei increased the most. Thus by 2017, the largest net carbon transfer was with the above two provinces (Fig. 5b). Shanxi is rich in mineral resources and is dominated by energy and metallurgical industries. This indicates that Guangdong and Hebei needed more energy and related products for their development during 2012–2017, while Jiangsu's demand decreased. The net carbon transfer characteristics of Hubei and Hunan provinces were more similar. The net carbon transfer with Shanxi and Inner Mongolia was larger in 2012, and with Beijing, Hebei and Inner Mongolia in 2017 (Fig. 5). Meanwhile, the net carbon transfer characteristics of Henan and Anhui were also more similar. The regions with larger net carbon transfers in both places were concentrated in Beijing, Inner Mongolia, Shanghai, Shandong, Guangdong and Xinjiang, and the direction of transfer was basically the same. Changes in the net carbon transfer volume in Jiangxi were generally characterized by a substantial decrease with the NC, SW, and part of the MR, and a substantial increase with other provinces.
The continuing transfer of industry chain links from all regions of China to the NW had resulted in an increasing net carbon transfer between provinces in the NW and most of the country. In particular, the incremental net carbon transfer with China's economically developed Yangtze River Delta and Pearl River Delta regions were both particularly large. And the NW basically served as a net carbon transfer-out region (Fig. 5). It also can be seen in Fig. 5, the larger net carbon transfers of provinces in the NW were also basically with the JJ, NC, EC, and Guangdong, which have higher levels of economic development in China.
As in the NW, the net carbon transfers between provinces in the SW and most provinces in China had increased. This is especially the case with the economically developed regions such as Beijing, the Yangtze River Delta and the Pearl River Delta, and regions with more abundant energy resources but a more homogenous industrial structure, such as the NE, NW, Inner Mongolia and Jiangxi. And except with the NW, the provinces in the NW basically were as a net transfer-out region (Fig. 5). Once again, the net carbon transfer in China was roughly characterized by a transfer from east to west and from southeast to northwest.
3.4 Analysis of influencing factors
Carbon emissions are affected by many factors, including the natural environment, society and the economy. Because there are too many influencing factors to be considered, it is difficult for multiple regression to avoid the problem of indicator covariance, so factor analysis can be used to realize the downscaling and simplification of indicators, and better analyze the influencing factors of inter-provincial carbon emission transfer. Factor analysis summarizes independent influencing factors from many indicators, i.e., the common factors, which should reflect as much information as possible about the original variables. By calculating the correlation coefficient matrix, eigenroots and eigenvectors, and variance contribution ratio, the number of common factors and the number of original variables they represent can be judged. Multiple regression models are constructed through the common factors. In this paper, 9 indicators such as regional GDP per capita, percentage of secondary industry, total import and export of goods, number of patents for inventions granted, resident population, disposable income per capita, consumption levels of residents, energy consumption per unit GDP, primary energy production are selected as the original influencing factor indicators for factor analysis, which are shown in Table 1.
In KMO test, the probability is 0.000 is less than the level of significance and the original hypothesis is rejected, which is significantly different from the unit matrix.The KMO is 0.742, which indicates that it is suitable for factor analysis. The factor analysis process usually uses principal component analysis to select principal components with eigenvalues greater than 1 as common factors. In this paper, a total of three common factors were extracted (Fig. 6). From the data in Table 2, it can be seen that the variance of the common factor of 7 of the 9 influencing factors is greater than 80%. The three common factors extracted together explained 82.505% of the information. Therefore the information loss of the original influence factor indicators is small, and the overall effect of the common factor extraction is ideal.
Table 1
Impact factor indicators and meanings
Indicator | Meaning |
GDP per capita (I1) | Economic development level |
Percentage of secondary industry (I2) | Industrial Structure |
Total import and export of goods (I3) | Trade Import/Export Demand |
Number of patents for inventions granted (I4) | Technical level |
Resident population (I5) | Population Concentration |
Disposable income per capita (I6) | Living standard of the population |
Consumption levels of residents (I7) | Resident consumption level |
Energy consumption per unit GDP (I8) | Energy consumption |
Primary energy production (I9) | Local resource availability |
In KMO test, the probability is 0.000 is less than the level of significance and the original hypothesis is rejected, which is significantly different from the unit matrix.The KMO is 0.742, which indicates that it is suitable for factor analysis. The factor analysis process usually uses principal component analysis to select principal components with eigenvalues greater than 1 as common factors. In this paper, a total of three common factors were extracted (Fig. 6). From the data in Table 2, it can be seen that the variance of the common factor of 7 of the 9 influencing factors is greater than 80%. The three common factors extracted together explained 82.505% of the information. Therefore the information loss of the original influence factor indicators is small, and the overall effect of the common factor extraction is ideal.
Table 2
| Initial | Extracted |
I1 score | 1.000 | .901 |
I2 score | 1.000 | .658 |
I3 score | 1.000 | .756 |
I4 score | 1.000 | .851 |
I5 score | 1.000 | .863 |
I6 score | 1.000 | .960 |
I7 score | 1.000 | .968 |
I8 score | 1.000 | .649 |
I9 score | 1.000 | .821 |
Appropriate rotation of the loading matrix makes the common factor of the original influencing factors more obvious and more conducive to the interpretation of the actual problem. As can be seen from Table 3, GDP per capita, disposable income per capita, and consumption levels of residents have the greatest contribution to the common factor 1 (F1), which is above 90%, so the F1 can be interpreted as the level of economic development. Resident population contributes the most to the common factor 2 (F2) with 92.1%, so the F2 can be interpreted as population agglomeration and social needs. Primary energy production contributes the most to the common factor 3 (F3) with 90.2%, so the F3 is interpreted as energy demand.
Table 3
| Component |
F1 | F2 | F3 |
I1 score | .942 | .100 | − .058 |
I2 score | − .525 | .395 | .475 |
I3 score | .450 | .741 | − .060 |
I4 score | .740 | .530 | − .149 |
I5 score | − .107 | .921 | − .064 |
I6 score | .960 | .046 | − .188 |
I7 score | .970 | .067 | − .150 |
I8 score | − .444 | − .415 | .528 |
I9 score | − .042 | − .076 | .902 |
Extraction method: principal component analysis. Rotation method: kaiser normalized maximum variance method. |
a. The rotation has converged after 5 iterations. |
Finally, the standardized original matrix of influence factor indicators is multiplied by the component score coefficients (Table 4) to calculate the common factor scores. The expressions for each common factor are shown in Eqs. (8)- (10) below:
$$\text{F}1=0.286{\text{I}}^{{\prime }}1-0.107{\text{I}}^{{\prime }}2+0.062{\text{I}}^{{\prime }}3+0.158{\text{I}}^{{\prime }}4-0.139{\text{I}}^{{\prime }}5+0.269{\text{I}}^{{\prime }}6+0.278{\text{I}}^{{\prime }}7+0.011{\text{I}}^{{\prime }}8+0.191{\text{I}}^{{\prime }}9$$
8
$$\text{F}2=-0.041{\text{I}}^{{\prime }}1+0.286{\text{I}}^{{\prime }}2+0.353{\text{I}}^{{\prime }}3+0.206{\text{I}}^{{\prime }}4+0.504{\text{I}}^{{\prime }}5-0.078{\text{I}}^{{\prime }}6-0.065{\text{I}}^{{\prime }}7-0.156{\text{I}}^{{\prime }}8+0.004{\text{I}}^{{\prime }}9$$
9
$$\text{F}3=0.160{\text{I}}^{{\prime }}1+0.321{\text{I}}^{{\prime }}2+0.079{\text{I}}^{{\prime }}3+0.055{\text{I}}^{{\prime }}4-0.038{\text{I}}^{{\prime }}5+0.046{\text{I}}^{{\prime }}6+0.083{\text{I}}^{{\prime }}7+0.348{\text{I}}^{{\prime }}8+0.781{\text{I}}^{{\prime }}9$$
10
where \({\text{I}}^{{\prime }}\)i denotes the standardized value of the raw impact factor indicator.
Table 4
Matrix of component score coefficients
| Component |
F1 | F2 | F3 |
I1 score | .286 | − .041 | .160 |
I2 score | − .107 | .286 | .321 |
I3 score | .062 | .353 | .079 |
I4 score | .158 | .206 | .055 |
I5 score | − .139 | .504 | − .038 |
I6 score | .269 | − .078 | .046 |
I7 score | .278 | − .065 | .083 |
I8 score | .011 | − .156 | .348 |
I9 score | .191 | .004 | .781 |
Extraction method: principal component analysis. Rotation method: kaiser normalized maximum variance method. Component Score. |
There is no linear relationship between the common factors (Table 5). Therefore, the regression coefficients of each influencing factor can be derived by regressing the three common factors on the net carbon transfer, so as to further analyze the relationship between each influencing factor and the changes in the net carbon transfer of inter-provincial. The F-value of the regression result is 9.952 and the p-value is 0.000 < 0.05 (Table 6), indicating that the model is constructed in a meaningful way and at least one of the independent variables will have an effect on the dependent variable. As can be seen from the p-values, all three public factors have a significant effect on the net carbon transfer. According to the regression results, the regression equation of the net carbon transfer with the three public factors can be obtained as:
$$\text{Y}=0.539\text{F}1+0.273\text{F}2-0.359\text{F}3+0.27$$
$$=0.086{\text{I}}^{{\prime }}1-0.095{\text{I}}^{{\prime }}2+0.101{\text{I}}^{{\prime }}3+0.122{\text{I}}^{{\prime }}4+0.076{\text{I}}^{{\prime }}5 + 0.107{\text{I}}^{{\prime }}6+0.102{\text{I}}^{{\prime }}7-0.162{\text{I}}^{{\prime }}8-0.176{\text{I}}^{{\prime }}9$$
11
Table 5
Component score covariance matrix
Component | 1 | 2 | 3 |
1 | 1.000 | .000 | .000 |
2 | .000 | 1.000 | .000 |
3 | .000 | .000 | 1.000 |
Extraction method: principal component analysis. Rotation method: kaiser normalized maximum variance method. Component Score. |
Table 6
The net carbon transfer | Coef. | St.Err. | t-value | p-value | [95% Conf Interval] | Sig |
F1 | .539 | .115 | 4.69 | 0 | .309 | .769 | *** |
F2 | .273 | .103 | 2.65 | .011 | .066 | .48 | ** |
F3 | − .359 | .104 | -3.45 | .001 | − .569 | − .15 | *** |
2012b | 0 | . | . | . | . | . | |
2017 | − .541 | .232 | -2.33 | .023 | -1.005 | − .077 | ** |
Constant | .27 | .154 | 1.75 | .085 | − .039 | .58 | * |
Mean dependent var | -0.000 | SD dependent var | 1.000 | |
R-squared | 0.420 | Number of obs | 60 | |
F-test | 9.952 | Prob > F | 0.000 | |
Akaike crit. (AIC) | 146.594 | Bayesian crit. (BIC) | 157.065 | |
*** p < .01, ** p < .05, * p < .1 |
From Eq. (11), it can be seen that among the nine influencing factors, the percentage of secondary industry, energy consumption per unit GDP and primary energy production are negatively correlated with the value of net carbon transfer i.e., the larger the value is, the smaller the net carbon transfer is, i.e., carbon emissions transferred out is greater than transferred in. Other factors are positively correlated with the net carbon transfer, i.e., the larger the value is, the greater the net carbon transfer is, i.e., carbon emissions transferred in is greater than transferred out.
The top three indicators of the correlation coefficient are primary energy production, energy consumption per unit GDP and number of patents for inventions granted, which have the greatest impact on the inter-provincial carbon transfer in China. Therefore, the local energy supply capacity, energy consumption level, and science and technology level of a certain region were most important for the difference between the carbon emissions transferred in and out of that region. It is the key to determine whether the region was a net transfer-in or net transfer-out region. China's resource-rich provinces, autonomous regions and municipalities had taken on more high-energy-consuming links in the industrial value chain, and it is most critical for such regions to reduce emissions by improving their independent innovation capabilities and production technology levels and reducing energy consumption levels. Provinces, autonomous regions and municipalities with high technological levels, whose resource holdings could not meet local development, transfer high-energy-consuming segments to other regions, so that their carbon emissions were transferred in more than they are transferred out. The coefficient for resident population is the smallest, suggesting that population size did not have a significant effect on changes in the net carbon transfer.
In addition, disposable income per capita, consumption levels of residents and total import and export of goods all had a large positive effect on the change in net carbon transfer. It indicates that the greater the social demand in a certain region, the greater the net carbon transfer, i.e., the more carbon emissions were transferred. Therefore, the income of the population should be safeguarded and raised to stimulate consumption and expand domestic demand. Neither the percentage of secondary industry nor GDP per capita had a significant effect on the net carbon transfer. The industrial structure and economic development level of a region were not the key factors determining the difference between the carbon emissions transfer in and out of the region based on the value chain.