Influencing factors and decoupling analysis of carbon emissions in China’s manufacturing industry

The manufacturing industry directly reflects national productivity, and it is also an industry with high energy consumption and severe carbon emissions. This study decomposes the influential factors on carbon emissions in China’s manufacturing industry from 1995 to 2018 into industry value added, energy consumption, fixed asset investment, carbon productivity, energy structure, energy intensity, investment carbon intensity, and investment efficiency by Generalized Divisia Index Model. The decoupling analysis of carbon emissions and industry value added is carried out to investigate the states of the manufacturing industry under the pressure of “low carbon” and “economy.” Results show that first, fixed asset investment is the driving force of carbon emissions, followed by industry value added; investment carbon intensity, carbon productivity, investment efficiency, and energy intensity are the mitigating factors; simultaneously, the impacts of energy consumption and energy structure are fluctuating. Second, the decoupling of manufacturing has improved, especially in the light industry. Third, the decoupling of carbon emissions and economic development is mainly dominated by the decoupling of energy consumption and industry added value. Therefore, reducing the proportion of coal consumption and optimizing the energy structure are significant ways to promote the low-carbon development of the manufacturing industry.


Introduction
The greenhouse effect caused by carbon emissions seriously threatens the development of human beings. Effectively solving climate and environmental problems becomes the primary standard for measuring social development quality (Nwaka et al. 2020;Ahmad et al. 2021a, b). Human activities are vital for the production of carbon emissions. Therefore, the initial research is basically from the perspective of socioeconomic. Decomposition analysis is an analytical framework for studying the characteristics of greenhouse gas emissions, which increasingly apply in the study of environment and economy (Song et al. 2011;Yue et al. 2013). Generally speaking, greenhouse gas emissions are determined by the technological level, affluence, energy structure, economic structure, population size, etc. The research scopes of decomposition analysis are broad, primarily including countries, regions, and industries (Chen et al. 2018;Chai et al. 2019;Meng and Zhou 2020).
With the rapid development of industrialization, China achieved the "Made in China," which brought about economic leaps. Meanwhile, energy plays a significant role in the economy, and the consumption of energy based on fossil fuels results in massive carbon emissions. China became the largest carbon emitter globally (Choi and Oh 2014;Tan and Lin 2018). Data from the China Statistical Yearbook showed that in 2017 and 2018, the manufacturing industry's industrial added value accounted for 28.11% and 27.84% of the gross domestic product (GDP). Energy consumption accounted for 54.65% and 61.88% of the total energy consumption, respectively. At present, China's economic growth is still dominated by the manufacturing industry (Zhang et al. 2016). Therefore, exploring the driving factors and decoupling status of the manufacturing industry at different times are the basis for verifying the effects of carbon reduction policies.
The existing research of low-carbon study depends mainly on the effective utilization of resources rather than on absolute reductions in resource consumption and carbon emissions (Wang et al. 2012;Sinha et al. 2020). Our study not only analyzes the carbon emissions caused by energy consumption, energy intensity, energy structure, and economic activity; the contributions of investment carbon intensity, investment efficiency are investigated simultaneously. Considering the technological heterogeneity, we explore the influential factors and decoupling status of the light industry and the heavy industry. In China, the manufacturing industry is the central pillar industry of the economy, inevitably accompanied by high carbon emissions (Yang et al. 2018). Exploring the relationship between economic development and synchronous carbon emissions change has been the primary national sustainable development issue. For further exploring the decoupling of carbon emissions and industry value added, we decompose the decoupling between carbon emissions and energy consumption and that between energy consumption and industry value added.
The main contributions of this study are as follows: Firstly, we estimate the carbon emissions of the manufacturing industry, the light industry, and the heavy industry from 1995 to 2018 in China. Secondly, the Generalized Divisia Index Model is established to study the absolute and relative influential factors on carbon emissions of the manufacturing industry. Meanwhile, the influential factors and decoupling status of the light industry and the heavy industry are analyzed, respectively. Thirdly, we distinguish the decoupling of carbon emissions and energy consumption and the decoupling of energy consumption and industry value added. This study aims to investigate the driving factors of carbon emissions in the manufacturing industry, analyze the manufacturing industry's decoupling states, and explore the path of reducing carbon emissions.
The remaining contents are as follows: The "Literature review" section presents the literature review of the decoupling among carbon emissions, energy consumption, and economic growth of China, the development and application of decomposition models, and decoupling analysis; the "Methodology and data" section presents the main models and data sources. The "Results and discussions" section is the results and discussions. The "Conclusions and policy implications" section puts forward the conclusions, policies, and limitations of the study.

Literature review
The study of carbon emissions, energy consumption, and economic growth in China Research on the relationship between economic development and carbon emissions is a hot issue. In order to reduce the pressure of carbon emission reduction and ensure sound and rapid economic growth, it is essential to explore the decoupling status between carbon emission and economic development. Song et al. (2019) discussed the relationship between economic development and carbon emissions based on the decoupling model and environmental Kuznets curve (EKC). Dong et al. (2020) studied the decoupling state of economic growth and carbon emissions in six industrial sectors. Research showed that the paths to achieving peak carbon emissions targets were adjusting industrial structure and energy structures. Energy consumption also plays a vital role in economic growth. Zhang et al. (2018) explored China's coal consumption factors and analyzed the decoupling relationship between coal consumption and economic growth. The study indicated that the industrial sector was the largest consumer of coal, and the energy intensity effect played a leading role in reducing coal consumption. Wen and Zhang (2019) focused on the decoupling of carbon emissions from the old industrial base in Liaoning. The results indicated that economic growth, investment structure, and energy structure were the main contributors to the increase in industrial carbon emissions. However, energy intensity and energy technology had restrained the growth of carbon emissions. Many studies begin to explore the decoupling of carbon emissions-energy-economic activity. Ahmad et al. (2021a, b) investigated the relationship between energy investment, air pollution, and sustainable development of China. The study revealed that it was necessary to transform the energy structure of the industry to decouple air pollution from economic growth. The influencing factors of carbon emissions from energy consumption mainly include energy intensity, population, economic activity, energy structure, economic structure, and industrial structure. However, the production activities of the manufacturing industry are accompanied by large-scale investment in fixed assets, so it is essential to consider the impact of carbon productivity, investment efficiency, and investment carbon intensity. Simultaneously, the existing literature identifies the decoupling relationship between carbon emissions and economic growth in the country, provinces, cities, and critical industries. However, few studies explore the decoupling of carbon emissions, energy consumption, and industry added value in China's manufacturing industry.

Comparison of the decomposition method
Decomposition methods conduct quantitative research on the contribution of influencing factors to the changes in energy or environment. Currently, the commonly used energy identities are IPAT identity and Kaya identity. Proposed by Ehrlich and Holdren (1970), the IPAT identity reflected the impact of the environment with population, per capita wealth, and technological level. Kaya identity was proposed by Kaya (1989), and it reflected the factors that lead to the change of carbon emissions (Pui and Othman 2019;Ortega-Ruiz et al. 2020).
To verify whether the decomposed variables better reflect the mechanism of influencing factors than the original variables, Hwang et al. (2020) conducted a multivariate cluster analysis to study the CO 2 generated by fossil fuels with the IPAT/Kaya identity. The results showed that decomposition variables were more helpful in identifying the relevant drivers. Based on the IPAT identity, Waggoner and Ausubel (2002) further decomposed the technology level T into the technology consumed on per unit of GDP and the environmental impact on per unit of technology, which is called the ImPACT identity. However, York et al. (2002) believed that both the IPAT and ImPACT models have some limitations, reflecting the effect of independent variables on the linear change of dependent variables. To make up for the deficiencies of the IPAT model and analyze the influence factors on the nonlinear environmental dependent variables, York further established the STIRPAT method based on IPAT identity. Subsequently, many scholars have conducted extensive research on the STIRPAT model (Xu and Lin 2016;Nasrollahi et al. 2020;Ma et al. 2020). The IPAT model, the ImPACT model, and the STIRPAT model have similar conceptual foundations while different purposes. York et al. (2003) discussed the relationship between the three formulas and improved the STIRPAT model by establishing the concept of ecological elasticity. The results showed that the STIRPAT model with ecological resilience explained the driving force of environmental impacts more accurately. The model provided a scientific basis for ecological change and identified factors that may be most sensitive to policy. Nowadays, the extended model basing on the concept of the IPAT model promoted the decomposition method. Introducing the industrial structure and urbanization level into the IPAT model, Li et al. (2011) adopted the Path-STIRPAT model to study the driving forces of carbon emissions in China. The results believed that the most significant impact on carbon emissions was per capita GDP, followed by industrial structure, population, urbanization, water, and technological level. To further figured out the role of energy in the environment, Wen and Li (2019) introduced energy into the IPAT model and used the IPAT-E model to explore the influencing factors of regional carbon emissions in China.
The decomposition methods are roughly divided into three categories: structural decomposition analysis (SDA), production-theoretical decomposition analysis (PDA), and index decomposition analysis (IDA). SDA method is a structural decomposition method combined with an input-output model (Hirotaka 2020). PDA model is derived from data envelopment analysis (DEA) technology. Based on the relationship between input and output, linear programming techniques are used to analyze the efficiency of the decisionmaking unit (Wang and Feng 2018a). IDA method is a simple index decomposition analysis and mainly includes the Laspeyres index and Divisia index decomposition method. The Divisia index decomposition method mainly includes Arithmetic Mean Divisia Index (AMDI) and Log Mean Divisia Index (LMDI). The LMDI decomposition method is further divided into LMDI decomposition and LMDI I (II) decomposition, which has multiplicative decomposition and additive decomposition simultaneously (Wang and Feng 2018b). Table 1 listed the literature on the decomposition method and affecting factors of carbon emissions in chronological order. Alexander (2014) found that the existing decomposition methods were limitations in interdependence and absolute changes of the affecting factors, making factors have mutual dependence in form. Alexander proposed the Generalized Divisia Index Method (GDIM) and applied the method to study the influencing factors of carbon emissions from 1980 to 2012 in China. At present, the study of the GDIM is still in the initial stage. Li et al. (2019) applied the GDIM to decompose the affecting factors of the construction industry and predicted the peak of carbon emissions combined with the scenario analysis method in China. Using the GDIM model, Yang and Shan (2019) studied the driving force of industrial sulfur dioxide emissions in Jiangsu and assessed the contribution rates of carbon emissions factors. It identified the driving factors of regional sulfur dioxide emissions and provided a basis for formulating more reasonable emission reduction policies. Wang et al. (2018) first adopted the GDIM to analyze the influence factors of carbon emissions on the transportation industry in China. And the improved Tapio model was used to explore the decoupling elasticity of the transportation industry. To identify the drivers of carbon emissions in the mining industry and five sub-sectors in China, Shao et al. (2016) used the GDIM to decompose the affecting factors. Meanwhile, the scenario analysis method was established to explore the feasibility of energy mitigation methods and policy suggestions. To avoid the limitation of continuous multiplicative on the LMDI method, Fang et al. (2020) analyzed the influence of three quantitative factors and five related factors on electricity consumption with the GDIM. At the same time, it revealed the mechanism of electricity consumption.
The reviews of decomposition methods showed that the IPAT and the mIPAT models have a deficiency in analyzing the change of nonlinear influencing factors. The STIRPAT model and LMDI model reflect the influence of nonlinear factors; however, these models cannot distinguish the impact of absolute factors and relative factors. When it comes to industrial carbon emissions decomposition factors, capital increments are more plastic than capital stocks. Meanwhile, fixed asset investment in the incremental sense directly impacts carbon emission reduction in the manufacturing industry. Furthermore, the factors related to fixed asset investment can effectively provide a basis for reduction policies (Shao et al. 2017). The GDIM model overcomes the shortcomings of the above decomposition methods and can be combined with the decoupling model to explore the relationship between industrial economic development and carbon emissions.

Literature review on the decoupling model
Decoupling theory is widely utilized to measure the relationship between economic growth, material consumption, and environmental protection. The asynchronous relationship is mainly derived from the response of the government basing on environmental pressure under economic developments. Currently, there are two decoupling models: the OECD decoupling model and the Tapio decoupling model. The OECD decoupling model utilized the ratio of environmental pressure to GDP at the end of the period and the initial period to present the decoupling state, which effectively identifies the correlation between economic development and environmental pollution; however, it cannot distinguish the decoupling status. Tapio (2005) established the Tapio decoupling model, which overcomes the defects of OECD decoupling.
Basing on the LMDI and the Tapio index, Wang and Yang (2015) quantitatively analyzed the decoupling index of industrial growth and environmental pressure in the Beijing-Tianjin-Hebei region. The results showed that economic growth was the main factor leading to industrial decoupling. Energy structure and energy intensity have an essential impact on the process of industrial decoupling. From the perspective of the industry sector, Andreoni and Galmarini (2012) and Lu et al. (2015) utilized a decomposition model to analyze the decoupling relationship of carbon emissions. The distinctions were that Andreoni and Galmarini (2012) divided the study into two periods and considered factors such as carbon intensity, energy intensity, structural change, and economic activity. The study analyzed the decoupling state of agricultural, thermal production, water, gas, transportation, and service sectors. In comparison, Lu et al. (2015) divided the industry into three main sectors and 38 sub-sectors and studied the decoupling relationship between carbon emissions intensity and economic growth. The most important was that five manufacturing industries had achieved a low-carbon economy in varying degrees. Combining the improved Laspeyres index method with the decoupling model, Diakoulaki and Mandaraka (2007) and Ren and Hu (2012) studied the decoupling relationship between industry growth and carbon emissions at the national and industry levels. The former took advantage of an improved Laspeyres model to decompose the affecting factors into output, energy intensity, structure, fuel structure, and utility structure. It mainly evaluated the actual efforts and effectiveness of countries in economic development and the environment. The latter divided the factors into industrial scale, energy structure, energy intensity, and public utility structure. And it is believed that the growth of the industry was a significant factor in carbon emissions. To further explore the degree of the economy dependent on energy input, Bithas and Kalimeris (2013) tried to re-estimate the decoupling effect of energy-economic growth, incorporating the energy/capita GDP ratio into the decoupling model. Results denoted that the energy/capita GDP ratio is closer to the energy attributes than the energy/gross domestic product ratio. Zhang and Da (2015) decomposed Chinese carbon emissions and carbon intensity into the energy source and the industrial structure by the LMDI method, respectively. Then introduced the decoupling index to analyze the decoupling relationship between carbon emissions and economic growth. From the national perspective, de Freitas and Kaneko (2011) and Roinioti and Koroneos (2017) investigated the decoupling state in Brazil and Greece, respectively. The difference was that the former employed the LMDI decomposition model, while the latter utilized the full decomposition technique developed by JW Sun. Compared with the LMDI model, the complete decomposition technique developed by JW Sun effectively deals with zero values in the data set. Decomposing the decoupling state between carbon emissions and economic growth into multi-dimensional decoupling has become a hot topic. It can reflect the decoupling status and the dynamic change path of the decoupling index. Zhang et al. (2020) believed that the current decoupling theory could not distinguish the decoupling state of countries with different levels of economic development. Therefore, the study established a two-dimensional decoupling model of economic growth and energy footprint by combining the Tapio index with the economic development index for the first time. Song et al. (2020) discussed the decoupling status and dynamic path of CO 2 emissions from 2000 to 2016 in China. A twodimensional decoupling model and a decoupling analysis framework were constructed with the GDP per capita as the horizontal axis and the Tapio decoupling index as the vertical axis. Based on the LMDI model, Wang et al. (2017) decomposed the decoupling of CO 2 emissions-GDP into CO 2 emissions-fossil energy consumption decoupling, fossil energy consumption-total energy consumption, and total energy consumption-GDP decoupling. These studies provide references to explore the decoupling between carbon emissions and economics.
Notwithstanding, the decomposition model analyzes the driving factors of carbon emissions; however, it cannot precisely and objectively measure the government's efforts to carbon reduction. Therefore, a decoupling model basing on the GDIM of the manufacturing industry is necessary.

Methodology and data
The Generalized Divisia Index Model (GDIM) Basing on the principle of Kaya identity, the GDIM decomposes the multi-dimensional factors of carbon emissions and makes up for the shortcoming of the interdependence of factor selection in the existing decomposition methods. At the same time, the absolute and relative factors are investigated to avoid double counting. The expressions of GDIM are as follows: where TC stands for carbon emissions; E is energy consumption; IVA represents the industry value added; FAI is fixed asset investment; EC (EC = TC/E) represents the energy structure; CP (CP = TC/IVA) denotes carbon productivity; ICI (ICI = TC/FAI) is investment carbon intensity; EI (EI = E/IVA) indicates energy intensity; and IE (IE = IVA/FAI) means investment efficiency. Furthermore, Eq. 1 can be transformed as: The functionTC(X)represents the contribution of factor X to carbon emissions. Combined with the above formula, the Jacobian matrix is constructed as: The changes in carbon emissions can be decomposed into the sum of the contributions of affecting factors as: where L represents the time span; I is the identity matrix; "+" means generalized inverse matrix; ∇TC = (CP IAV 0 0 0 0 0 0) T ; and if the columns of φ X in the Jacobian are linearly Therefore, the influencing factors on carbon emissions of the manufacturing industry can be decomposed intoΔIVA, ΔE, ΔFAI, ΔCP, ΔEC, ΔICI, ΔIE, andΔEI, whereΔIVA, ΔE, andΔFAIare the absolute influence factors on carbon emissions;ΔCP, ΔEC, ΔICI, ΔIE, andΔEIare the relative influence factors.ΔIVAreflects the effect of output scale; ΔErepresents the effect of energy consumption; ΔFAIis the effect of investment scale; ΔCPis the effect of carbon productivity;ΔECis energy structure effect; ΔICIdenotes the impact of investment carbon intensity; ΔEIstands for energy intensity effect; ΔIEis the effect of investment efficiency.

The decoupling model
The decoupling status classified by the Tapio model is more accurate and not limited by time. According to the elasticity value, the decoupling states can be divided into weak decoupling, strong decoupling, weak negative decoupling, strong negative decoupling, growth negative decoupling, growth connection, recession decoupling, and decline connection. Based on the Tapio model, the decoupling model of carbon emissions in the manufacturing industry is as follow: In order to further study the relationship between carbon emissions and economic development, we decomposed the decoupling of φ(TC, IVA) into φ(TC, E) and φ(E, IVA) (Gao et al. 2021 ).
whereφ(TC, IVA),φ(TC, E), andφ(E, IVA)are decoupling of carbon emission and industry value added, decoupling of carbon emission and energy consumption, and that between energy consumption and industry value added, respectively.ΔTC,ΔE, andΔIVAseparately stand for the change of carbon emissions, energy consumption, and the industry value added;ε IVA , ε E , ε FAI , ε CP , ε EC , ε EI , ε ICI , andε IE are the decoupling elastic value of IVA, E, FAI, CP, EC, EI, ICI, and IE. The decoupling states and elasticity levels are shown in Table 2.

Data source
(1) Name of the manufacturing industry. Due to the inconsistent statistical caliber of industry names from 1995 to 2018, this study mainly refers to the China Statistical Yearbook to unify the data source of industry classification caliber. The carbon emissions of the light industry and the heavy industry are calculated by the sub-industries. The total manufacturing carbon emissions are calculated by the carbon emissions of 17 light industries and 11 heavy industries. The classified names for light industry and heavy industry are shown in Table 6 in the Appendix.   Table 7 in the Appendix.

Results and discussions
The results of carbon emissions As shown in Fig. 1 The trend of carbon emissions in the heavy industry is roughly the same as that of the entire manufacturing industry. In contrast, the carbon emissions of the light industry change slightly and account for less than 20% of the manufacturing industry. Obviously, the heavy industry is a significant contributor to the carbon emissions of the manufacturing industry, which should be paid more attention to. After China entered the World Trade Organization in 2001, carbon emissions continue to rise as the market expands and exports increase. Meanwhile, with the expansion of functions in Chinese urban, the demand for infrastructures also goes upward. Notwithstanding that the manufacturing industry's total economic output continues to grow, the carbon emissions problem has gradually emerged. In 2014, the Chinese economy entered a new normal period, which brings an opportunity to develop a green and low-carbon economy. The industrial structure has been adjusted and optimized, and the growth rate of total carbon emissions in the manufacturing industry slows down. Generally, energy consumption is a rigid demand for the development of the manufacturing industry. With the acceleration of industrialization and urbanization, the carbon emissions of the manufacturing industry are severe, especially that of heavy industry. Therefore, it is necessary to analyze the driving factors of carbon emissions in the manufacturing industry to provide a basis for the reduction strategies.

The results of GDIM decomposition
Decomposition results of China's manufacturing industry This study decomposes the carbon emissions of the manufacturing industry in China from 1995 to 2018 based on Eq. 1 to Eq. 4. The carbon emissions of 28 sub-industries and manufacturing industry in China from 1995 to 2018 are shown in Fig. 4 in the Appendix.
By taking into account the technological heterogeneity of the light industry and the heavy industry, the influencing factors of carbon emissions are decomposed, respectively. It can be seen from Fig. 2 that fixed asset investment and industry value added are the driving factors. Investment carbon intensity, carbon productivity, investment efficiency, and energy intensity are the restraining factors. However, energy consumption and energy structure present inconsistent effects at different stages.
Fixed asset investment is the most vital factor in increasing manufacturing carbon emissions, and industry value added is the main factor. The driving effects of fixed asset investment and industry value added are not fully manifested during the "Ninth Five-Year Plan" period; however, the driving effects are particularly obvious during the "Twelfth Five-Year Plan" period. Furthermore, during the "Thirteenth Five-Year Plan" period, carbon emissions driven by fixed asset investment and industry value added are decreased. Since the reform and opening-up, China's economy has been in a stage of extensive growth. The reform of the property rights system and the government management system in the manufacturing sector is still lagging behind. Investment and construction in fixed assets are still at the primary stage, and the output scale needs to be improved. Therefore, during the "Ninth Five-Year Plan" period (1995)(1996)(1997)(1998)(1999)(2000), the government explicitly proposed transforming the economic growth pattern from extensive to intensive.
During the "Tenth Five-Year Plan" period (2000)(2001)(2002)(2003)(2004)(2005), China's social productivity and foreign economic relations have experienced significant changes. China entered the World Trade Organization and became the "world factory." Export volume increased sharply. The investment in fixed assets in the manufacturing industry further expanded, leading to increased carbon emissions caused by fixed asset investment and industry value added.
During the "Eleventh Five-Year Plan" period (2005)(2006)(2007)(2008)(2009)(2010), the continuous improvement of economic and Increasing investment in fixed assets can solve employment, promote income growth, and maintain social stability. However, it will cause serious environmental issues with too much attention paid to scale expansion and neglect carbon emissions. After the international financial crisis in 2008, the phenomenon of overcapacity in China changed from overcapacity in local industries to an overall surplus. Therefore, China first time put forward the constraint target of energy conservation and emission reduction during the "Eleventh Five-Year Plan" period. During the Twelfth Five-Year Plan (2010-2015) period, China proposed to enhance the core competitiveness of the manufacturing industry. In 2015, China committed in the Paris Agreement: Carbon dioxide emissions will reach a peak around 2030 and strive to reach the peak as soon as possible, and carbon dioxide emissions per unit of GDP will be reduced by 60-65% compared with 2005. Therefore, at the beginning of the "Thirteenth Five-Year Plan" (2015-2018), the increase in carbon emissions caused by fixed asset investment and industry value added has been significantly reduced. It showed that China's manufacturing industry had achieved certain results in cleaner production and green transformation under the new economic normal background.
Investment carbon intensity is the most important reason for reducing carbon emissions, while carbon productivity is an essential factor. The following are investment efficiency and energy intensity. The promotion effect of investment carbon intensity is particularly obvious during the "Eleventh Five-Year Plan" and "Twelfth Five-Year Plan" periods. Investment, consumption, and exports are the "troika" that promotes economic growth, and the Chinese market is mainly in an investment-driven economic growth model. During the "Eleventh Five-Year Plan" period, the government emphasized the resource and environmental pressures on sustainable development caused by blind investment and lowlevel expansion. Therefore, in the "Twelfth Five-Year Plan," the emphasis is on promoting economic growth to rely on consumption, investment, and export-coordinated shift. Thus, the changes in the pattern of economic growth make investment carbon intensity and carbon productivity be an essential reason for reducing carbon emissions.
Investment efficiency and energy intensity have relatively weak restraint effects on manufacturing carbon emissions. It indicates that the strategy of energy conservation and emission reduction during the "Eleventh Five-Year Plan" period has a particular impact on the manufacturing industry. The results of investment efficiency show that fixed asset investment in the manufacturing industry had formed a certain production capacity and achieved a certain level of production technology.
The effects of energy consumption and energy structure are varied in the light industry and the heavy industry. During the "Ten Five-Year Plan" and "Eleventh Five-Year Plan," energy consumption has a promoting effect on the carbon emissions of the manufacturing industry. In contrast, during the "Nine Five-Year Plan," "Twelfth Five-Year Plan," and "Thirteenth Five-Year Plan," energy consumption inhibits carbon emissions. Furthermore, during the "Thirteenth Five-Year Plan" period, the reduction effect of energy consumption is significantly higher than that of the "Nine Five-Year Plan" and "Twelfth Five-Year Plan" periods. It indicates that the impact of energy consumption has the potential for emission reduction. It is worth noting that in the "Twelfth Five-Year Plan," energy consumption has a decreasing impact on the carbon emissions of the light industry while an increasing effect on the heavy industry. The heavy industry is mainly characterized by a certain industrial scale; therefore, the production of the heavy industry is more driven by energy.
Energy structure exhibits an inhibitory effect in the "Nine Five-Year Plan" period, and then it shows a promoting effect in the "Ten Five-Year Plan," "Eleventh Five-Year Plan," "Twelfth Five-Year Plan," and "Thirteenth Five-Year Plan." The difference is that in the early period of the "Thirteenth Five-Year Plan," energy structure has a slight inhibition effect on the light industry carbon emissions. In 2011, several industry policies were successively introduced, such as "Development Plan for Industrial Transformation and Upgrading during the Twelfth Five-Year Plan Period," "Development Plan for the Petroleum and Chemical Industry during the Twelfth Five-Year Plan Period," and the "Twelfth Five-Year" development plan for sub-sectors such as Pesticides, Rubber, and Paper Chemicals. The policies gradually facilitate the development of a low-carbon economy in the light industry.

The cumulative contribution on influencing factors of carbon emission in China's manufacturing industry
As shown in Fig. 3, the cumulative contribution rate presents a fluctuating trend from 1995 to 2018. The cumulative carbon emissions change from negative to positive after 2002 and show an upward trend until 2014. However, at the beginning of the "Thirteenth Five-Year Plan" period, cumulative carbon emissions begin to decline. The reasons may be that the manufacturing industry developed rapidly due to the comparative advantages of resource endowments and factor costs. Meanwhile, the manufacturing industry gradually forms a pattern of marketization and globalization, which led to a large number of carbon emissions. With the increasing emphasis on the environment, the growth of carbon emissions in the manufacturing industry slows since 2015. Fixed asset investment, industry value added, and energy consumption are the driving factors of carbon emissions in the manufacturing industry. Energy structure is the driving factor, and it makes the change range of carbon emissions accumulation small. Investment carbon intensity and carbon productivity are the main inhibited factors. Meanwhile, energy intensity and investment efficiency have a weak inhibition effect. It indicates that the inhibition effect of energy intensity and investment efficiency in carbon emissions has not been fully utilized, and the effects need to be further strengthened.
China is a developing country; however, the above analysis shows that investment and economic growth are the main reasons for the increase in carbon emissions. Therefore, the mitigation strategies of the manufacturing industry could be formulated from the aspects of improving investment efficiency and carbon productivity. The government should encourage enterprises to eliminate outdated production capacity and optimize energy structure, and meanwhile, guide enterprises to pay more attention to the investment and application of energy-saving technologies and equipment. The cumulative contributions of carbon emissions in the light industry and the heavy industry from 1995 to 2018 are shown in Fig. 5 and Fig. 6 in the Appendix. Table 3 shows the decoupling states of China's manufacturing industry from 1995 to 2018. It demonstrates that due to the distinguishing characteristics of the light industry and heavy industry, the decoupling status of the manufacturing industry is fluctuating.

Decoupling of carbon emissions and industry value added in China's manufacturing industry
The manufacturing industry mainly experiences weak decoupling, expansive coupling, expansive negative decoupling, and strong decoupling. During the Ninth Five-Year Plan period, the manufacturing sector changes from a weak decoupling state to a strong decoupling state. The reasons may be that the economic model changes from a traditional planned economy to a socialist market economy. Under the wave of "deindustrialization," China vigorously implemented the reform and opening-up policy. For the first time, it put forward the concept of "expanding domestic demand" and increased investment in infrastructure construction. At the beginning of the "Ten Five-Year Plan," "Eleventh Five-Year Plan," and "Twelfth Five-Year Plan," the manufacturing industry is mainly in the state of expansive coupling. With the rise of e-commerce and online shopping, large-scale manufacturing capacity and industrial clusters formed in coastal areas, which makes China became the world's largest exporter in 2009.
In the late "Twelfth Five-Year Plan" period, the manufacturing industry begins to show a strong decoupling state and continue to the beginning of the "Thirteenth Five-Year Plan." The outline of the "Twelfth Five-Year Plan" policy points out the main line of development of "traditional manufacturing transfer and upgrade" and puts forward the goal of transforming and upgrading the manufacturing industry. It is not only conducive to improving the production efficiency of the manufacturing industry and improving the relationship between economic development and carbon emissions but also conducive to the green development of the manufacturing industry. The light industry experiences strong decoupling, weak decoupling, expansive negative decoupling, and expansive coupling. During the "Twelfth Five-Year Plan" period, the light industry is mainly in a state of strong decoupling. One reason may be the "Food Industry 'Twelfth Five-Year' Development Plan" organized by the National Development and Reform Commission and the Ministry of Industry and Information Technology during 2010-2015. The plan put forward higher energy conservation and emission reduction than the "Eleventh Five-Year Plan" period. Therefore, the relationship between the industry added value and carbon emissions gradually improves.
The decoupling status of the heavy industry is unsatisfactory, which experiences weak decoupling, strong decoupling, and expansive negative decoupling. The possible reason may be that the heavy industry provides the material foundation for developing the national economy. The government pays more attention to infrastructure construction, which leads to the significant carbon emissions of heavy industry. From the "Thirteenth Five-Year Plan," the decoupling status of the heavy industry shows a slight improvement, which presents the strong decoupling in this stage.

Decoupling contribution of influencing factors
From the decoupling contribution of the influencing factors in Table 4, industry value added and fixed asset investment hinder the decoupling of manufacturing carbon emissions. From 1995 to 2018, the industry value added decoupling factor is always greater than 0. The carbon emissions generated by the industry value added continued to increase, only a slight decline in 2010, and then shows an upward trend. It illustrates that with industrialization, economic growth would lead to an increase in carbon emissions. The fixed asset investment decoupling factor is less than zero in 1997, 1999, and 2015, while the fixed asset investment decoupling factor is greater than 0 during the rest of the period. The 1997 Asian financial crisis impacted global finance, causing many large Asian companies to lose money. Bankruptcy, unemployment, social and economic depression, and manufacturing industries are also affected. There is a short-term decline in carbon emissions from fixed asset investment. After the financial crisis, the carbon emissions generated by fixed asset investment fluctuates, the overall trend remained rising. Carbon productivity, investment carbon intensity, investment efficiency, and energy intensity promote the decoupling effect of manufacturing carbon emissions. The carbon productivity decoupling factor is only less than 0 from 2003 to 2005. Carbon productivity has a reciprocal relationship with "carbon emission intensity per unit of GDP," emphasizing emission output efficiency. Judging whether an industry is low carbon is mainly to see whether it can effectively use resources, rather than just focusing on the absolute reduction in resource consumption and greenhouse gas emissions.
The decoupling value of investment carbon intensity and investment efficiency continues to be less than zero from 1995 to 2018, which shows that investment carbon intensity and investment efficiency are beneficial to suppress the increase in carbon emissions. Therefore, it is necessary to reasonably improve the investment carbon intensity, energy intensity, and investment efficiency in energy saving and emission reduction. The effect of energy consumption and energy structure fluctuates, sometimes promoting carbon emissions and sometimes suppressing decoupling. The development of the manufacturing industry is based on high energy consumption and is inevitably accompanied by high carbon emissions. In the short term, fossil fuels are the primary source of energy. Therefore, with the adjustment of national economic policies, the improvement of the energy structure still requires longterm plans. The decoupling contribution of the influencing factors in the light industry and the heavy industry is shown in Table 8 and Table 9 in the Appendix.

Decoupling of carbon emission and energy consumption, energy consumption, and industry value added in China's manufacturing industry
Achieving low-carbon economic goals while maintaining economic development is a severe challenge for China's manufacturing industry. According to the results of "The cumulative contribution on influencing factors of carbon emission in China's manufacturing industry," the impact of energy consumption and energy structure in different periods is not consistent, and the impact degree is relatively weak. Therefore, the goal of advancing the energy consumption revolution, optimizing the energy structure, and strictly controlling the excessive growth of energy consumption is a long way to go. According to the results of Table 3 and Table 5, the decoupling state of carbon emissions and industry added value is similar to that of energy consumption and industry added value. The results show that the decoupling state of carbon emissions and economic development is mainly  During "Nine Five-Year Plan" and "Thirteenth Fiveyear Plan," the decoupling of carbon emissions and energy consumption mainly included expansive coupling, expansive negative decoupling, recessive decoupling, and recessive coupling. The strong decoupling only appeared in 2008-2009. The 2008 US financial crisis may have a more significant impact on the manufacturing industry in the real economy, especially labor-intensive sectors. However, China hosted the 29th Summer Olympic Games in 2008, which expanded the domestic demand of the manufacturing industry. During this period, the "Green Olympics" was the central concept and, therefore, impacted energy consumption.
From "Ten Five-Year Plan" to "Eleventh Five-Year Plan," the decoupling state of energy consumption and industry added value is a step backward. Since the 1980s, China's economic development has entered a period of accelerated industrialization. During this period, the process of industrialization was accompanied by the market-oriented reform of the economic system, and energy consumption also increased substantially.
Subsequently, opening up to the outside world continued to accelerate, providing opportunities for the development of manufacturing (Liu et al. 2021). At the beginning of the "Thirteenth Five-year Plan," the decoupling of energy consumption from the industry's added value is improved, and the strong decoupling continued to appear. China's manufacturing industry is turning to a new path of industrialization to achieve economic development and the goal of low energy consumption. Therefore, in order to improve the decoupling of carbon emissions from the added value of the industry, it is necessary to develop clean energy and improve t h e e n e r g y c o n s u m p t i o n s t r u c t u r e o f C h i n a ' s manufacturing industry. Decoupling of carbon emission and energy consumption and that of energy consumption and industry value added in the light industry and the heavy industry are present in Table 10 and Table 11 in the Appendix.

Conclusions and policy implications
This study first estimates the manufacturing industry's carbon emissions from 1995 to 2018 in China. It indicates that China's manufacturing industry has the potential to reduce carbon emissions, especially the heavy industry. Moreover, the influencing factors on carbon emissions of the manufacturing industry, light industry, and heavy industry are elaborated in detail. Meanwhile, we study the decoupling states and sources of the discrepancy. Finally, the decoupling of carbon emissions and industry added value is decomposed into the decoupling of carbon emissions and energy consumption, energy consumption, and industry added value in China's manufacturing industry. Based on the above analysis, the main conclusions are as follows: First, fixed asset investment and industry value added are the main reasons for promoting carbon emissions. Investment carbon intensity, carbon productivity, investment efficiency, and energy intensity are essential factors in reducing carbon emissions. Energy consumption and energy structure have different effects at different stages. Second, from 1995 to 2018, the light industry's decoupling state is relatively sound, while that of the heavy industry is partially improved during the "Thirteenth Five-Year Plan." Simultaneously, emphasis should be placed on fixed asset investment in hindering the decoupling of the manufacturing industry. And further strengthen the contribution of carbon productivity, investment carbon intensity, energy intensity, and investment efficiency on decoupling. Third, the decoupling of carbon emissions and economic development is similar to energy consumption and industry added value. Meanwhile, the decoupling of heavy industry needs to be improved. Therefore, properly handling energy issues is significant to the future development of the manufacturing industry, especially for the heavy industry with high energy consumption. Based on the above research, the policy inspirations on energy conservation and emission reduction in China's manufacturing industry mainly include the following: Firstly, the model of economic growth in China is in urgent need of transformation. Relying on investment to stimulate economic development has the hidden danger of overcapacity, which reduces investment returns and the national economy. Furthermore, it will lead to an increase in energy supply, carbon emissions, and environmental pressure. Fixed asset investment and industry value added have a more obvious linkage effect. Therefore, to overcome the disadvantages of the investment-led growth model and improve investment quality, the government should avoid low-benefit production and blind investment. Meanwhile, the economic growth model driven by investment should gradually shift to the consumption-driven economic development model to achieve sustainable development.
Secondly, pay full attention to the inhibitory effects of investment carbon intensity, carbon productivity, investment efficiency, and energy intensity on manufacturing carbon emissions. At present, energy structure, investment efficiency, and energy intensity have not fully exerted inhibitory effects on carbon emissions. Notably, energy structure has the most unsatisfactory emission reduction effect. However, China cannot eliminate the production mode dominated by fossil energy in the short term. Therefore, the energy market's development and improvement should be promoted and establish a reasonable market mechanism of "energy use right." In the long run, optimizing the energy structure is an effective means to reduce the carbon emissions of the manufacturing industry.
Finally, correctly understand the "quantity" and "quality" issues in economic development, focusing on the quality of growth, resource utilization efficiency, and environmental protection. Simultaneously, appropriate subsidies should be adopted to encourage high carbon emissions industries to utilize advanced equipment and improve energy utilization efficiency to achieve sound and rapid economic development of the manufacturing industry.
We study the influencing factors and decoupling analysis of carbon emissions in China's manufacturing industry. However, there are some limitations. This study focuses on the relationship between China's manufacturing carbon emissions and industrial economic development and the influencing factors. However, this study does not analyze the influencing factors that affect manufacturing energy consumption and the critical point between different decoupling statuses. Meanwhile, structural shifts concerning sectoral shift in the Chinese economy over can be further study. Due to the space limitation, Table 6, Table 7, Table 8, Table 9, Table 10,  Table 11, Fig. 4, Fig. 5, and Fig. 6 are shown in the Appendix.      S7 S8 S9 S10 S11 S12 S15 S16 S17 S27 S28  Fig. 6 The cumulative contribution of the heavy industry from 1995 to 2018