Differentiation Analysis on Carbon Emission Efficiency and Its Factors at Different Industrialization Stages: Evidence from Mainland China

Industrial production is currently the main source of global carbon emissions. There are obvious differences in regional carbon emission efficiencies (CEE) at different industrial stages. We investigate CEE and explore its factors in mainland China at different industrialization stages from 2008-2020 using the super-SBM model with an undesirable output and the STIRPAT model. There is significant spatial heterogeneity in regional CEE, with gaps gradually widening. CEE’s spatial heterogeneity in mid-industrialized provinces is narrowing, while in late-industrialized and post-industrialized provinces, it is widening. CEE’s factors also differ in provinces at different industrialization stages. At the mid-industrialization stage, the industrial structure (IS) is the dominant factor, while population urbanization (PU) is dominant at the late-industrialization stage, and both PU and IS are dominant at the post-industrialization stage. Based on CEE’s characteristics at different industrialization stages, we propose suggestions for green development.


Introduction
Carbon emissions have become a major constraint to regional development, limiting regions whose increases are necessitated by COVID-19 epidemic normalization, economic recovery, and enormous energy demands. Global CO 2 emissions exceeded 33.8 billion tons in 2021, demonstrating a growth rate of 5.00%. As the largest global carbon emitter, China's carbon emissions reached 10.5 billion tons in 2021. In terms of global CO 2 emissions in 2021 (Global CO 2 emission in 2021: Asia Pacific accounts for more than half of carbon emissions, China tops the world in carbon emissions: https://www.sohu.com/a/578797859_120950203 (accessed on 11 October 2022)). As a developing country, China still relies on extensive energy consumption. As such, China's carbon emissions are still continuously increasing. For sustainable development, China strives to reach "carbon peaking" by 2030 and "carbon neutrality" by 2060. At the 20th National Congress of the Communist Party of China, it was proposed that China should actively and steadily promote "carbon peaking" and "carbon neutrality" and actively respond to the global governance of climate change.
Carbon emission efficiency (CEE) measures regional contributions (e.g., to climate change) and reflects the ratio of the production relationship between the minimum CO 2 emissions created and the maximum economic output [1]. In terms of research scales, countries [2][3][4], regions [5][6][7], and some particular units [8] are some of the main objects of previous studies, while different industries (e.g., industrial structure, transportation industry, and environmental regulation) are among the main directions [9,10]. However, due to differentiated resource endowments, interregional economies are uneven. As a result, regional carbon emissions may also differ, yet there are fewer studies [11] on cross-sectional comparisons [12] of the same geographical units, which would help to provide tailored suggestions for regional characteristics. A single index measurement can reflect the local CEE situation [13][14][15][16] but often ignores the relationships and influences of other indices,
We completed a panel regression analysis with Stata16. To eliminate the possible influence of heteroskedasticity on upper models, we took the logarithms of all variables. The factors' regression analysis was conducted using the random effects model (RE) and fixed effects model (FE). The Hausman test showed that the fixed effects model (FE) was more appropriate for our research of nationwide and mid-industrialization and lateindustrialization stages. A high goodness-of-fit coefficient formed in the fixed effects model (FE), two-way fixed-effects model (FE-tw), and two-way fixed model (FE-tw). Thus, a two-way fixed model (FE-tw) was selected. At the post-industrialization stage, the random effects model (RE) was more appropriate for our research. To prevent "pseudo-regression", the panel series were tested for stationarity using an HT test. The results (Table 2) showed that all panel data passed the significance test. We measured the local industrialization level using five indicators, namely: economic development, industrial structure, manufacturing structure, spatial structure, and employment structure ( Table 3). The evaluation values, numbers, and weights of individual indicators were counted and calculated. Thus, a composite index of local industrialization level was calculated using the additive synthesis method.
The results show that all 30 provinces were at the mid-industrialization stage and beyond in 2020 (Table 4). Hainan, Heilongjiang, Guangxi, and the other three provinces were at the mid-industrialization stage. Hebei, Inner Mongolia, Jilin, and 18 other provinces were at the late-industrialization stage. Fujian, Shandong, Hubei, Guangdong, Chongqing, and Zhejiang ranked at the top at the late-industrialization stage, where industrialization was significantly improved. Beijing, Tianjin, and Shanghai were already at the postindustrialization stage because of their strong economy and long industrialization history.

Super-SBM Model of Undesirable Output
The traditional DEA model [36] measures efficiency from a single perspective and ignores the possible slackness of input and output, which can cause specific measurement errors. The non-radial, non-angle SBM model solves some problems of the traditional DEA model [37]. However, this model usually results in the efficiency value of multiple decisionmaking units being 1, making it impossible to further distinguish each unit's efficiency. The super-efficient DEA and SBM models form the super-efficient SBM model [38], which can handle situations where multiple decision units have an efficiency value of 1 and can distinguish the efficiency of effective decision units. We measured CEE in China's provinces using the Super-SBM model with an undesirable output [39] via Matlab2020b. The CEE index system consists of three input indicators, namely: capital, labor, and energy. The output indicators include desired and undesirable output factors (Table 5). Notes: The standard value of GDP per capita was calculated based on the GDP deflator. A, I, and S denote the proportion of the value-added of primary, secondary, and tertiary industries in GDP. Manufacturing value added/total merchandise production value added was mainly the ratio of industrial value added to value added by primary and secondary industries. Data processing and stage judgments were compared to three decimal places for data that were at critical values. Table 4. Division of industrialization stages in 2020.  The coefficient of variation (CV), Gini coefficient (G), and Thiel index (T) were chosen to explore CEE's overall differences nationwide and at different industrialization stages. In the Gini coefficient, CEE in each province needs to be ranked in descending order and then calculated. The formulas are as follows.

Mid-industrialization
denotes the square root of the arithmetic mean of CEE in province i and China's average; and Z denotes the mean of the overall CEE.

Econometric Model
IPAT is an equation for studying environmental factors [40]. Based on IPAT, the STIRPAT model was proposed to lift the model's limitations due to homogeneous linear variation. The STIRPAT model is as follows: where I, P, A, and T denote environmental impact, population, affluence, and technology, respectively; a is the model coefficient; b, c, and d are indices to be taken into account; and e is the error term.
Heteroskedasticity may have an impact on the model results. Therefore, we transformed the STIRPAT model into a logarithmic form.
Additionally, factors such as industrial structure (IS), foreign investment (FI), energy consumption (EC), and population urbanization (PU) were added to the model on top of the basic model consistently.
where CEE denotes carbon emission efficiency; ST denotes technology; ED denotes GDP, PC denotes population capacity; PU denotes population urbanization; IS denotes industrial structure; EC denotes energy consumption; FI denotes foreign investment; α denotes the intercept term; ε is a random error term; and I and t denote the observation province and observation year, respectively. µ is the coefficient of the explanatory variable.

Spatial Evolution
We classified CEE in mainland China's 30 provinces into high-efficiency, mediumefficiency, and low-efficiency zones via K-means cluster analysis ( Figure 2). Overall, China's CEE was in constant change, with significant provincial differences. The highefficiency zone showed a geographical characteristic of a "concentration-decentralization-

Spatial Evolution
We classified CEE in mainland China's 30 provinces into high-efficiency, mediumefficiency, and low-efficiency zones via K-means cluster analysis ( Figure 2). Overall, China's CEE was in constant change, with significant provincial differences. The high-efficiency zone showed a geographical characteristic of a "concentration-decentralization-contraction" phase. The medium-efficiency zone was mainly concentrated in the central region and gradually clustered to the southeast. The low-efficiency zone spread from the west to the northeast and southwest. In 2008, the high-efficiency zone was mainly concentrated in China's eastern coastal region. The medium-efficiency zone was scattered. The lowefficiency zone was mainly concentrated in the middle-upper reaches of the Yellow River and the upper reaches of the Yangtze River. In 2014, the high-efficiency zone was gradually dispersed. The medium-efficiency zone was concentrated in the Yangtze River basin and distributed closer to the southeast. The low-efficiency zone extended to the northeast. In 2020, only Beijing was still in the high-efficiency zone. The medium-efficiency zone was concentrated in the Yangtze River basin and closer to the southeast. The low-efficiency zone extended to the northeast, from which Yunnan and Guizhou gradually withdrew. contraction" phase. The medium-efficiency zone was mainly concentrated in the central region and gradually clustered to the southeast. The low-efficiency zone spread from the west to the northeast and southwest. In 2008, the high-efficiency zone was mainly concentrated in China's eastern coastal region. The medium-efficiency zone was scattered. The low-efficiency zone was mainly concentrated in the middle-upper reaches of the Yellow River and the upper reaches of the Yangtze River. In 2014, the high-efficiency zone was gradually dispersed. The medium-efficiency zone was concentrated in the Yangtze River basin and distributed closer to the southeast. The low-efficiency zone extended to the northeast. In 2020, only Beijing was still in the high-efficiency zone. The medium-efficiency zone was concentrated in the Yangtze River basin and closer to the southeast. The low-efficiency zone extended to the northeast, from which Yunnan and Guizhou gradually withdrew. The trends in CV, G, and T were generally consistent at the same industrialization stage. The opposite is true at different industrialization stages. Meanwhile, there were some differences in provincial CEE at different industrialization stages (Figure 3). At the

Analysis of the Overall Results
The panel regression results showed that an increase in economic development (ED), industrial structure (IS), and foreign investment (FI) contributed to the improvement in CEE. Every 1% increase in economic development (ED) increased CEE by 0.252%, while every 1% increase in industrial structure (IS) increased CEE by 0.712%. A 1% increase in foreign investment (FI) contributed to a 0.040% increase in CEE. In contrast, population capacity (PC) and energy consumption (EC) had a significant negative impact on CEE and were not conducive to improving CEE. At the 1% significance level, every 1% increase in population capacity (PC) decreased CEE by 0.210%. At the 10% significance level, every 1% increase in energy consumption (EC) decreased CEE by 0.023% (Table 6).   The trends in CV, G, and T were generally consistent at the same industrialization stage. The opposite is true at different industrialization stages. Meanwhile, there were some differences in provincial CEE at different industrialization stages (Figure 3). At the mid-industrialization stage, China's provinces were more homogeneous, and internal differentiation gradually narrowed at the mid-industrialization stage, T ∈ [0.018, 0.056]. The variability of provincial CEE showed a narrowing and then widening characteristic at the late-industrialization stage, T ∈ [0.041, 0.090]. At the post-industrialization stage, CEE's differentiation was smaller, while the regional differentiation increased over time. The highest value of T was 0.097 in 2020, while the lowest value was 0.002 in 2008. Although they were at the same industrialization stage, there were large differences in economic and energy aspects.

Analysis of the Overall Results
The panel regression results showed that an increase in economic development (ED), industrial structure (IS), and foreign investment (FI) contributed to the improvement in CEE. Every 1% increase in economic development (ED) increased CEE by 0.252%, while every 1% increase in industrial structure (IS) increased CEE by 0.712%. A 1% increase in foreign investment (FI) contributed to a 0.040% increase in CEE. In contrast, population capacity (PC) and energy consumption (EC) had a significant negative impact on CEE and were not conducive to improving CEE. At the 1% significance level, every 1% increase in population capacity (PC) decreased CEE by 0.210%. At the 10% significance level, every 1% increase in energy consumption (EC) decreased CEE by 0.023% (Table 6).
Currently, China's economy is rapidly developing. When the economy reaches a certain level, regional development will transfer from pursuing development speed to development quality. In China in 2021, CO 2 emissions of 10 4 RMB GDP decreased by 3.8% compared to 2020. From 2017 to 2021, China's CO 2 emissions of 10 4 RMB GDP declined continuously and exceeded the climate action target in 2020, ahead of schedule aiming for 2019. The expansion of population capacity (PC) puts pressure on urban transport infrastructure. Traffic congestion and heavy transport consume a lot of energy and increase carbon emissions, which are not conducive to CEE. An excessive increase in population capacity (PC) can also damage the local environment. Coal is dominant in China's primary energy consumption. By 2021, the share of non-fossil energy in China's primary energy consumption reached 16.60%, and China decreased energy consumption per unit of GDP by 26.40% compared to 2012, saving 1.4 billion tons of standard coal. Relying on extensive coal consumption for economic development is not conducive to green and low-carbon development, which reflects the importance and necessity of transforming energy consumption (EC).  Note: "***", "**", "*" denote 1%, 5%, and 10% significance levels; values in brackets are t-statistics or Z-values, "-" are items not involved.

Analysis of Zoning Results
At the mid-industrialization stage, economic development (ED), population urbanization (PU), and industrial structure (IS) all passed the significance test and had a significant positive relationship with CEE (Table 7). At the 1% significance level, every 1% increase in economic development (ED) increased CEE by 0.146%, every 1% increase in population urbanization (PU) increased CEE by 0.735%, and every 1% increase in industrial structure (IS) increased CEE by 0.748%. Population capacity (PC) is not conducive to improving CEE. Every 1% increase in population capacity (PC) decreased CEE by 0.163% at the 5% significance test. Among all positive factors, industrial structure (IS) is the dominant factor of CEE at the mid-industrialization stage.
At the late-industrialization stage, technology (ST) passed the 5% significance test and showed a significant negative correlation with CEE (Table 7). Every 1% increase in technology (ST) decreased CEE by 0.107%. Economic development (ED), population urbanization (PU), and industrial structure (IS) all passed the significance test and contributed to improving CEE. Every 1% increase in economic development (ED) increased CEE by 0.263%, while every 1% increase in population urbanization (PU) increased CEE by 1.245%. At the late-industrialization stage, population urbanization (PU) was CEE's main factor.
At the post-industrialization stage, all variables, except foreign investment (FI), passed the significance test and showed a significant relationship with CEE (Table 7). Among these, technology (ST), economic development (ED), and industrial structure (IS) contributed to the improvement in CEE. Every 1% increase in technology (ST) increased CEE by 0.262%, every 1% increase in economic development (ED) increased CEE by 0.165%, and every 1% increase in industrial structure (IS) increased CEE by 1.024%. Population capacity (PC), population urbanization (PU), and energy consumption (EC) hindered the improvement in CEE. Every 1% increase in population capacity (PC) decreased CEE by 0.985%, every 1% increase in population urbanization (PU) decreased CEE by 3.744%, and every 1% increase in energy consumption (EC) decreased CEE by 0.189%. Among all significant positive factors, industrial structure (IS) was the most dominant, while among all significant negative factors, population urbanization (PU) was dominant.   Note: "***", "**", and "*" denote 1%, 5%, and 10% significance levels; values in brackets are t-statistics or Z-values, and "-" are unrelated items.
However, at different industrialization stages, there are many similarities regarding CEE's factors. Both economic development (ED) and industrial structure (IS) are significant. Population capacity (PC) has a significant negative impact at the mid-and post-industrialization stages, which significantly differ. Technology (ST) has a significant negative impact on CEE in the provinces at the late-industrialization stage but has a significant positive effect at the post-industrialization stage. Population urbanization (PU) passes the significance test at all different industrialization stages. Population urbanization (PU) has a positive impact and enhances CEE at the mid-and late-industrialization stages, which is detrimental to CEE at the post-industrialization stage.
There is a difference in the impact of technology (ST) on CEE in provinces at different industrialization stages. At the late-industrialization stage, every 1% increase in technology (ST) decreased CEE by 0.107% but increased CEE by 0.262% at the post-industrialization stage. Economy, industry, and space differ among provinces at the late-industrialization stage. Urban construction and economic development still require a significant amount of capital investment. Technology (ST) requires much more capital and human resources to drive progress and improvement. Capital investment far outweighs economic output generated by technology (ST), while R&D requires more energy consumption (EC), all of which will increase provincial financial pressure to the detriment of local development at the late-industrialization stage. Meanwhile, pursuing economic development (ED) requires the intensive use of energy consumption (EC). Therefore, an increase in technology (ST) requires more energy consumption (EC), decreasing CEE.
Population urbanization (PU) significantly positively affected CEE at the mid-and late-industrialization stages. A 1% increase in population urbanization (PU) increased CEE by 0.735% at the mid-industrialization stage, and increased CEE by 1.245% at the late-industrialization stage. At the post-industrialization stage, population urbanization (PU) had a significant negative effect on CEE. Every 1% increase in population urbanization (PU) will decrease CEE by 3.744%. As population urbanization (PU) rises, human capital sustains local economic development. Meanwhile, human capital drives sustainable economic development, transitioning towards low-carbon development and clean production. It also leads to changes in urban residents' consumption, shifting towards low-carbon, environmental protection, and cleanliness. The shift in urban residents' consumption has increased energy use efficiency, reduced carbon emissions, and promoted CEE. Further increases in population urbanization (PU) will burden urban construction. Urban residents directly or indirectly increase energy consumption (EC) and carbon emissions, which is not conducive to improving CEE.

Robustness Test
There may be issues such as omitted variables in the model, resulting in endogeneity issues affecting the model's results and the stability of the regression results. Based on previous research [41], we chose five representative quantile points of 10%, 25%, 50%, 75%, and 90% to complete a panel quantile regression model ( Table 8). The results showed that the nature and significance of each variable's effects were generally consistent with the overall regression results, indicating that our study's results are reliable and stable.

Problem and Recommendations
(1) At the mid-industrialization stage, population migration leads to a decrease in the population urbanization rate, which is not conducive to CEE. At this stage, every 1% increase in population urbanization increases CEE by 0.967%, and an increase in urbanization increases CEE. The scale of migration across provinces in 2020 was 125 million persons, an increase of 38.96 million persons or 45.37% over 2010. At the mid-industrialization stage, China's provinces face a population exodus, which is not conducive to CEE. Provinces need to attract the population back and bring in high-level talent at the mid-industrialization stage. Human capital can accelerate the upgrading and transformation within local enterprises through the skills and proficiency of the educated population [42]. It can apply advanced technology and production methods to local practices, which can help to increase the efficiency of enterprises' production and operation, improve the regional input-output situation, and thus improve CEE [43].
(2) Enhancement in technology innovation inhibits improvement in CEE, which is significantly related to the input and output situation in technical innovation. At the lateindustrialization stage, every 1% increase in technology decreases CEE by 0.107%. China's innovation input index was 219.0 in 2021, while its innovation output index was 353.6. Environmental protection and low-carbon development should receive attention in some provinces at the late-industrialization stage, and technical innovation should be improved in some provinces at the late-industrialization stage [44]. Enterprises and research institutes should increase capital investment in R&D and improve the efficiency of transforming technical achievements [12].
(3) At the post-industrialization stage, some provinces advance China in terms of socioeconomic development. At the post-industrialization stage, a 1% increase in population urbanization decreases CEE by 3.744%. Population urbanization is not conducive to CEE. China is home to a high amount of talent, capital, and immigrants. The population urbanization rate reached 64.72% in China in 2021 and was above 80.00% in Beijing, Tianjin, and Shanghai. Urban scale structure and urban development should be optimized [45]. Environmental awareness and guiding residents to develop green consumption and low-carbon lifestyles should be strengthened [46]. Based on provincial characteristics, China needs to develop high-tech, low-carbon green industries to improve urban functions, reduce carbon emissions, and thus improve local CEE [47].
(4) It is necessary to develop targeted efficiency improvement strategies according to local conditions. Local governments should guide enterprises to shift to a low-carbon model in their production and operations [20]. In pursuing low-carbon development, it is essential to provide capital, technical, and human resources assistance to provinces with low CEE [48]. Provinces should accelerate regional communication, fully use the technology of energy conservation and emission reduction and accelerate regional industrial upgrading based on local resources [49]. China should strengthen its environmental laws and institutions and enhance the enforcement of environmental policies [50]. Foreign investment can bring advanced production technology experience, enhance international cooperation, and take advantage of the "technology spillover" effect of foreign enterprises in energy conservation and emission reduction to raise the environmental awareness of domestic enterprises [51]. Additionally, air pollution, such as PM 2.5 [52], is an important external factor for CEE. Interregional exchange and learning can improve CEE and reduce regional differences.

Research Perspectives
In this paper, we investigated CEE's spatial heterogeneity and its factors at different industrialization stages. The large variability among industrialization stages proves our study is of theoretical and practical value. The CEE analysis paradigm based on differentiated industrialization stages is our greatest theoretical innovation and contribution to research. Based on the differentiated characteristics and existing problems, we propose suggestions for human capital, technical innovation, and urbanization construction, which can help promote efficient and green development. However, there are still certain shortcomings or limitations to our study. We used provinces as the basic study unit and only explored industrialization stages. Additionally, provincial CEE does not reflect its specific actual situation in a comprehensive and detailed manner. Some of China's provinces (e.g., Inner Mongolia) span a wide range of regions, with significant disparities in the economy, carbon emissions, energy consumption, population, etc. The data on carbon emissions are relatively old. Furthermore, there are still some missing data from recent years. In this paper, we predict carbon emissions using the model. However, in terms of actual measurements, carbon emissions are affected by other factors. Therefore, the prediction does not reflect the actual local situation effectively. These are the research aspects that we will strive to improve in the future.

Conclusions
By constructing panel regression models at different industrialization stages, we explore CEE's various factors. China's CEE shows a significant fluctuation, with provincial CEE increasing at the post-industrialization stage and decreasing at other industrialization stages. CEE is characterized by CEE post-industrialization > CEE China's average > CEE late industrialization > CEE mid-industrialization . Provincial CEE's differentiation gradually decreases at the mid-industrialization stage, which keeps increasing at the late industrialization and post-industrialization stages. Among CEE's factors, industrial structure (IS), economic development (ED), population capacity (PC), and population urbanization (PD) are the main factors. There are differences in the influence of factors at different industrialization stages. Industrial structure (IS) is the most important factor of CEE, especially during the mid-industrialization. Population urbanization (PU) is the most significant factor, while whose impact is opposite at the late-and post-industrialization stages.

Data Availability Statement:
The data used to support the findings of this study are available from the corresponding author upon request.